STREET GREENING QUALITY DETECTION METHOD BASED ON PHYSIOLOGICAL ACTIVATION RECOGNITION

Abstract

A street greening quality detection method based on physiological activation recognition is provided. The street greening quality detection method includes establishing a greening quality factor index system, and obtaining and uniformly processing street greening images; collecting raw data, and performing reclassification and differential wave processing on the raw data to obtain valid physiological data that can be used for activation feature recognition of greening quality factors; calculating physiological activation feature parameters, training the physiological activation feature parameters by transfer learning fusion to determine importance of physiological activation features, and recognizing weighted average greening activation indexes of the greening quality factors; analyzing weighted average greening activation index data of the greening quality factors to form a street greening quality detection model; and inputting annotated street samples to be analyzed into the street greening quality detection model to obtain annotated results of street greening quality grading detection target data.

Claims

1. A street greening quality detection method based on a physiological activation recognition, comprising the following steps: establishing a greening quality factor index system according to high-frequency street landscape characteristics, and obtaining and uniformly processing street greening images for greening stimulus physiological experiments; collecting EEG, ECG, EDA and EMG raw data stimulated by the street greening images, and performing reclassification and differential wave processing on the raw data according to greening quality factor indexes to obtain valid physiological data that can be used for activation feature extraction of greening quality factors; calculating EEG, ECG, EDA and EMG physiological activation feature parameters of the greening quality factors according to the obtained valid physiological data, training the physiological activation feature parameters by transfer learning fusion to determine importance of physiological activation features, and recognizing weighted average greening activation indexes of the greening quality factors; analyzing weighted average greening activation index data of the greening quality factors to form a street greening quality detection model for contrast detection of street greening quality; and inputting annotated street samples to be analyzed into the street greening quality detection model to obtain annotated results of street greening quality grading detection target data.

2. The street greening quality detection method based on the physiological activation recognition according to claim 1, wherein the process of establishing the greening quality factor index system comprises the following steps: collecting statistics on frequencies of street greening structures, plant attributes, visual landscape and the like, selecting high-frequency street greening quality factors, and sorting out constituent elements, typical features and environmental connotations of street greening by a theoretical analysis method, to establish the street greening quality factor index system, wherein the street greening quality factor index system comprises primary element dimensions, secondary variable factors and tertiary factor change form indexes; acquiring data of built environment street view images, determining street greening scene class images in the street view images by location scene recognition technology, performing feature sampling on single greening quality variable factors of the greening scene class images by image element semantic segmentation technology, and determining clear street greening target images by a square gradient function; and randomly selecting m street greening images from the street greening target images, and performing three phase randomizations on the m street greening images to obtain 3*m random phase images and form an experimental stimulus image library; dividing all experimental stimulus images in the experimental stimulus image library into n groups through inter-group experiments, and playing the experimental stimulus images at a random inter-group and same frequency in a laboratory environment to obtain EEG, ECG, EDA, EMG and trigger signals of corresponding data segments of the images in real time.

3. The street greening quality detection method based on the physiological activation recognition according to claim 2, wherein primary element dimension indexes of the greening quality factor index system comprise greening structure, plant texture, line of sight relationship, and landscape characteristics; the secondary variable factor indexes are extensions of primary greening quality elements; and the tertiary factor change form indexes are manifestations of variable factors, and factor features of greening quality are sampled through the built environment street view images.

4. The street greening quality detection method based on the physiological activation recognition according to claim 1, wherein the process of collecting the EEG, ECG, EDA and EMG raw data stimulated by the street greening images, and performing the reclassification and differential wave processing on the raw data according to the greening quality factor indexes comprises the following steps: capturing raw data of each street greening target image when stimulated, and classifying raw data segments representing a same greening quality variable factor into one class according to markers recorded by trigger signals, wherein each class of data segments reflects EEG, ECG, EDA and EMG changes of subjects under an influence of a variable factor; performing baseline correction, bandpass filtering, average reference processing, ICA, noise reduction and artifact removal on the raw data segments, and correcting signal offsets by EMD, so as to solve average amplitudes and differential waves of non-stimulus state electrical signals caused by greening quality factor stimulus states; and analyzing amplitudes and phase images of differential waveforms by Hanning windowing, fast Fourier transform and wavelet transform, and extracting ? and ? frequency bands of five EEG differential wave leads PZ, P4, P5, O1, OZ, and O2, low-frequency and high-frequency bands of ECG differential waves at R-R intervals, high-frequency bands of EMG differential waves after full-wave rectification, and a normalized conductivity GSR of EDA differential waves within an exposure time window of the street greening target images, so as to calculate power spectral densities of the EEG, ECG and EMG frequency bands and a first-order difference of the EDA conductivity, to obtain valid physiological data for physiological activation feature recognition of the greening quality factors.

5. The street greening quality detection method based on the physiological activation recognition according to claim 1, wherein the process of calculating the EEG, ECG, EDA and EMG physiological activation feature parameters of the greening quality factors according to the obtained valid physiological data, training the physiological activation feature parameters by the transfer learning fusion to determine the importance of the physiological activation features, and recognizing the weighted average greening activation indexes of the greening quality factors comprises the following steps: superposing and averaging physiological data of each class of greening quality factors according to the obtained valid physiological data, calculating the EEG, ECG, EDA and EMG physiological activation feature parameters of the greening quality factors respectively, and normalizing the calculated EEG, ECG, EDA and EMG physiological activation feature parameters of the greening quality factors; obtaining a physiological activation feature vector A.sup.(m)={a.sub.x.sup.(m)}, (x=1, 2, . . . , N, m=1, 2, . . . , J) of the greening quality factors from the normalized EEG, ECG, EDA and EMG physiological activation feature parameters of the greening quality factors, wherein a.sub.i.sup.(m) represents an m.sup.th physiological activation feature of an x.sup.th class of greening quality factor objects; constructing a physiological activation feature importance determination matrix B={b.sub.ij}, wherein by represents an importance degree ratio of an i.sup.th activation feature dimension to an j.sup.th activation feature dimension; consequently, obtaining a weight vector w*=[w.sub.1, w.sub.2 . . . , w.sub.j] of each feature; and fusing the physiological activation features by transfer learning TLDA, using 70% of the samples as a source domain dataset and remaining 30% as a target domain dataset, obtaining marker activation values of source domain street greening target images, performing sparse self-encoding on the physiological activation feature vector A.sup.(m) and the marker activation value Y of the greening quality factors, determining the number of neurons q (q<m) in a self-encoder, introducing A.sup.(m) into a neural network, and assigning physiological activation feature weights after neural network training to obtain an ensemble vector E of fused features and a weighted average greening activation index O corresponding to E, as follows: $O=\frac{1}{4}\underset{i=1}{\overset{m}{.Math.}}\frac{{?}_i}{\underset{j=1}{\overset{m}{.Math.}}{?}_j}{o}_k$ wherein ?.sub.i represents a weight of each source domain, namely, a modulus of a similarity vector in the ensemble vector E, and o.sub.k represents a predicted activation degree of a k.sup.th class of greening quality factors.

6. The street greening quality detection method based on the physiological activation recognition according to claim 5, wherein the process of calculating the EEG, ECG, EDA and EMG physiological activation feature parameters of the greening quality factors comprises the following steps: calculating an EEG activation feature parameter A.sub.EEG of a greening factor object by the following formula: $A_{\mathrm{EEG}}=\frac{P_{?,x}}{P_{?,x}}$ wherein P.sub.?,x and P.sub.?,x represent relative average power of the five leads PZ, P4, P5, O1, OZ, and O2 with respect to ? and ? bands of a currently calculated greening quality factor object x; calculating an ECG activation feature parameter A.sub.ECG of the greening factor object by the following formula: $A_{\mathrm{ECG}}=\frac{P_{\mathrm{LF},x}}{P_{\mathrm{HF},x}}$ wherein P.sub.LF,x represents a power value of an ECG low-frequency component of the greening quality factor x, and P.sub.HF,x represents a power value of an ECG high-frequency component of the greening quality factor x; calculating an EDA activation feature parameter A.sub.EDA of the greening factor object by the following formula: $A_{\mathrm{EDA}}=\frac{1}{2}\left(t_{\mathrm{peak}}-t_{\mathrm{onset}}\right)*\left(s_{\mathrm{peak}}-s_{\mathrm{onset}}\right)$ wherein t.sub.peak and t.sub.onset represent a peak and beginning of ?GSR rise time during stress response, s.sub.peak and s.sub.onset represent a peak and beginning of a ?GSR amplitude value during stress response, and A.sub.EDA represents activation of 10% higher than a baseline during GSR stress response; calculating an EMG activation feature parameter A.sub.EMG of the greening factor object by the following formula: $A_{\mathrm{EMG}}=\frac{1}{2}{?}_0^{+?}{P}_{\mathrm{EMG},x}\mathrm{df}$ wherein P.sub.EMG,x represents a power spectral density function of EMG signals of the currently calculated greening quality factor object x, and f represents a frequency of the EMG signals; and normalizing the physiological activation feature parameters of each greening quality factor and introducing the following calculation formula: $A_{\mathrm{Normalized}}=\left(\frac{A\left(i\right)-A_{\min }}{A_{\max }-A_{\min }}\right)?100$ wherein A(i) represents an i.sup.th activation feature parameter, A.sub.min and A.sub.max represent a Normalized minimum value and a maximum value of the activation feature parameters, and A.sub.Normalized represents normalized activation feature parameters.

7. The street greening quality detection method based on the physiological activation recognition according to claim 1, wherein the process of analyzing the weighted average greening activation index data of the greening quality factors to form the street greening quality detection model comprises the following steps: obtaining a weighted average greening activation index of each class of greening quality factors, checking the weighted average greening activation index data of the greening quality factors by KMO measure of sampling adequacy and Bartlett's test of sphericity, wherein when KMO value >0.6 and sphericity test adjoint probability P value?0.01, it is considered that factor variables are strongly correlated and are suitable for further analysis on greening factor objects; calculating a cumulative variance contribution rate M.sub.K of latent principal components of an initial greening quality variable matrix X={x.sub.ij}, (i=1, 2, 3, . . . , m; j=1, 2, 3, . . . , n), and selecting latent principal components of greening quality at M.sub.K?80% as follows: $M_K=\underset{i=1}{\overset{k}{.Math.}}\frac{{?}_{\mathrm{ij}}}{\sqrt{{?}_j}}$ wherein x.sub.ij represents a j.sup.th greening quality variable factor of an i.sup.th sample; ?.sub.ij/?{square root over (?.sub.j)} represents a k.sup.th latent principal component of the greening quality, ?.sub.ij represents factor load of a j.sup.th latent principal component of an i.sup.th variable factor, and ?.sub.ij represents a characteristic root of the j.sup.th latent principal component; extracting first k latent principal components to detect the street greening quality, and calculating a weight w.sub.i of a single greening quality variable factor according to the related coefficient matrices and the variance contribution rate as follows: $w_i^{}=.Math.\underset{i=1}{\overset{n}{.Math.}}\left(\frac{{?}_{\mathrm{ij}}}{\sqrt{{?}_j}}*{?}_j\right).Math./{?}_j$ wherein ?.sub.j represents a variance contribution rate corresponding to the j.sup.th latent principal component of the greening quality, and the larger the w.sub.i, the greater the importance of the greening quality variable factor; and forming the street greening quality detection model according to weight coefficients of the greening quality variable factors, which is used for contrast detection of the street greening quality: $G={?}_1{x}_1+{?}_2{x}_2+{?}_3{x}_3+.Math.+{?}_j{x}_j$ wherein ?.sub.1 represents an influence coefficient of the i.sup.th factor, and x.sub.j represents greening activation index data of a j.sup.th re-extracted factor.

8. The street greening quality detection method based on the physiological activation recognition according to claim 1, wherein the process of inputting the annotated street samples to be analyzed into the street greening quality detection model to obtain the annotated results of the street greening quality grading detection target data comprises the following steps: collecting physiological data of J subjects with respect to I street greening images of N street samples to obtain an initial greening quality variable matrix Z={z.sub.ij}, (i=1, 2, 3, . . . , M; j=1, 2, 3, . . . , N) of M greening quality variable factors of the N street samples, and annotating the EEG, ECG, EDA and EMG physiological activation feature parameters of the N street samples in classes according to the classes of the greening quality variable factor indexes; establishing a greening activation relationship fusion model among the EEG, ECG, EDA and EMG activation feature parameters, generating a fused greening activation index of the J subjects with respect to the street greening quality variable factors, setting confidence of the greening activation index data within a [0, 1] interval, and annotating the variable factors of the street samples with activation degrees; presetting greening quality detection conditions, dividing the street greening quality into four levels G.sub.1, G.sub.2, G.sub.3, and G.sub.4, and assigning hierarchical values to the element dimensions of the street samples from high to low, to rank the greening quality of the street samples; $C_i=\{\begin{array}{c}4,x_{\mathrm{ij}}?\left[\stackrel{?}{x}+0.5?,x_{\max }\right]\\ 3,x_{\mathrm{ij}}?[\stackrel{?}{x},\stackrel{?}{x}+0.5?)\\ 2,x_{\mathrm{ij}}?[\stackrel{?}{x}-0.5?,\stackrel{?}{x})\\ 1,x_{\mathrm{ij}}?[x_{\min },\stackrel{?}{x}-0.5?)\end{array}$ wherein x.sub.ij represents greening activation annotation data of an j.sup.th element dimension at an i.sup.th location, x represents an average value of greening activation degrees of all samples, ? represents a standard deviation of the greening activation degrees of all samples, and C.sub.i represents a level of assigned greening quality of an i.sup.th element dimension; and inputting the annotated street samples into the street greening quality detection model, obtaining activation degrees of the element dimensions of the street samples one by one through the calculated greening variable factors and element dimensions, on this basis, reassigning the greening quality of each element dimension with reference to the greening quality grading detection conditions, and performing weighted superposition on dimension weights of the greening quality elements to form and annotate dimensionless greening quality values of the street samples; $Y=\underset{t=1}{\overset{n}{.Math.}}{w}_t^{*}{C}_t^{}$ wherein Y represents the greening quality values of the street samples, w*.sub.t represents a weight of a i.sup.th element dimension, and C.sub.t represents level assignment of greening quality at the t.sup.th element dimension.

9. A device, comprising: one or more processors; and a memory, configured to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors are enabled to implement the street greening quality detection method based on the physiological activation recognition according to claim 1.

10. A storage medium comprising computer executable instructions, wherein the computer executable instructions are used to perform the street greening quality detection method based on the physiological activation recognition according to claim 1 when executed by a computer processor.

11. The device according to claim 9, wherein in the street greening quality detection method, the process of establishing the greening quality factor index system comprises the following steps: collecting statistics on frequencies of street greening structures, plant attributes, visual landscape and the like, selecting high-frequency street greening quality factors, and sorting out constituent elements, typical features and environmental connotations of street greening by a theoretical analysis method, to establish the street greening quality factor index system, wherein the street greening quality factor index system comprises primary element dimensions, secondary variable factors and tertiary factor change form indexes; acquiring data of built environment street view images, determining street greening scene class images in the street view images by location scene recognition technology, performing feature sampling on single greening quality variable factors of the greening scene class images by image element semantic segmentation technology, and determining clear street greening target images by a square gradient function; and randomly selecting m street greening images from the street greening target images, and performing three phase randomizations on the m street greening images to obtain 3*m random phase images and form an experimental stimulus image library; dividing all experimental stimulus images in the experimental stimulus image library into n groups through inter-group experiments, and playing the experimental stimulus images at a random inter-group and same frequency in a laboratory environment to obtain EEG, ECG, EDA, EMG and trigger signals of corresponding data segments of the images in real time.

12. The device according to claim 11, wherein in the street greening quality detection method, primary element dimension indexes of the greening quality factor index system comprise greening structure, plant texture, line of sight relationship, and landscape characteristics; the secondary variable factor indexes are extensions of primary greening quality elements; and the tertiary factor change form indexes are manifestations of variable factors, and factor features of greening quality are sampled through the built environment street view images.

13. The device according to claim 9, wherein in the street greening quality detection method, the process of collecting the EEG, ECG, EDA and EMG raw data stimulated by the street greening images, and performing the reclassification and differential wave processing on the raw data according to the greening quality factor indexes comprises the following steps: capturing raw data of each street greening target image when stimulated, and classifying raw data segments representing a same greening quality variable factor into one class according to markers recorded by trigger signals, wherein each class of data segments reflects EEG, ECG, EDA and EMG changes of subjects under an influence of a variable factor; performing baseline correction, bandpass filtering, average reference processing, ICA, noise reduction and artifact removal on the raw data segments, and correcting signal offsets by EMD, so as to solve average amplitudes and differential waves of non-stimulus state electrical signals caused by greening quality factor stimulus states; and analyzing amplitudes and phase images of differential waveforms by Hanning windowing, fast Fourier transform and wavelet transform, and extracting ? and ? frequency bands of five EEG differential wave leads PZ, P4, P5, O1, OZ, and O2, low-frequency and high-frequency bands of ECG differential waves at R-R intervals, high-frequency bands of EMG differential waves after full-wave rectification, and a normalized conductivity GSR of EDA differential waves within an exposure time window of the street greening target images, so as to calculate power spectral densities of the EEG, ECG and EMG frequency bands and a first-order difference of the EDA conductivity, to obtain valid physiological data for physiological activation feature recognition of the greening quality factors.

14. The device according to claim 9, wherein in the street greening quality detection method, the process of calculating the EEG, ECG, EDA and EMG physiological activation feature parameters of the greening quality factors according to the obtained valid physiological data, training the physiological activation feature parameters by the transfer learning fusion to determine the importance of the physiological activation features, and recognizing the weighted average greening activation indexes of the greening quality factors comprises the following steps: superposing and averaging physiological data of each class of greening quality factors according to the obtained valid physiological data, calculating the EEG, ECG, EDA and EMG physiological activation feature parameters of the greening quality factors respectively, and normalizing the calculated EEG, ECG, EDA and EMG physiological activation feature parameters of the greening quality factors; obtaining a physiological activation feature vector A.sup.(m)={a.sub.x.sup.(m)}, (x=1, 2, . . . , N, m=1, 2, . . . , J) of the greening quality factors from the normalized EEG, ECG, EDA and EMG physiological activation feature parameters of the greening quality factors, wherein a.sub.i.sup.(m) represents an m.sup.th physiological activation feature of an x.sup.th class of greening quality factor objects; constructing a physiological activation feature importance determination matrix B={b.sub.ij}, wherein b.sub.ij represents an importance degree ratio of an i.sup.th activation feature dimension to an j.sup.th activation feature dimension; consequently, obtaining a weight vector w*=[w.sub.1, w.sub.2, . . . , w.sub.j] of each feature; and fusing the physiological activation features by transfer learning TLDA, using 70% of the samples as a source domain dataset and remaining 30% as a target domain dataset, obtaining marker activation values of source domain street greening target images, performing sparse self-encoding on the physiological activation feature vector A.sup.(m) and the marker activation value Y of the greening quality factors, determining the number of neurons q (q<m) in a self-encoder, introducing A.sup.(m) into a neural network, and assigning physiological activation feature weights after neural network training to obtain an ensemble vector E of fused features and a weighted average greening activation index O corresponding to E, as follows: $O=\frac{1}{4}\underset{i=1}{\overset{m}{.Math.}}\frac{{?}_i}{\underset{{j=1}^{}}{\overset{m}{.Math.}}{?}_j}{o}_k$ wherein ?.sub.i represents a weight of each source domain, namely, a modulus of a similarity vector in the ensemble vector E, and o.sub.k represents a predicted activation degree of a k.sup.th class of greening quality factors.

15. The device according to claim 14, wherein in the street greening quality detection method, the process of calculating the EEG, ECG, EDA and EMG physiological activation feature parameters of the greening quality factors comprises the following steps: calculating an EEG activation feature parameter A.sub.EEG of a greening factor object by the following formula: $A_{\mathrm{EEG}}=\frac{P_{?,x}}{P_{?,x}}$ wherein P.sub.?,x and P.sub.?,x represent relative average power of the five leads PZ, P4, P5, O1, OZ, and O2 with respect to ? and ? bands of a currently calculated greening quality factor object x; calculating an ECG activation feature parameter A.sub.ECG of the greening factor object by the following formula: $A_{\mathrm{ECG}}=\frac{P_{\mathrm{LF},x}}{P_{\mathrm{HF},x}}$ wherein P.sub.LF,x represents a power value of an ECG low-frequency component of the greening quality factor x, and P.sub.HF,x represents a power value of an ECG high-frequency component of the greening quality factor x; calculating an EDA activation feature parameter A.sub.EDA of the greening factor object by the following formula: $A_{\mathrm{EDA}}=\frac{1}{2}\left(t_{peak}-t_{onset}\right)*\left(s_{peak}-s_{onset}\right)$ wherein t.sub.peak and t.sub.onset represent a peak and beginning of ?GSR rise time during stress response, s.sub.peak and s.sub.onset represent a peak and beginning of a ?GSR amplitude value during stress response, and A.sub.EDA represents activation of 10% higher than a baseline during GSR stress response; calculating an EMG activation feature parameter A.sub.EMG of the greening factor object by the following formula: $A_{\mathrm{EMG}}=\frac{1}{2}{?}_0^{+?}{P}_{\mathrm{EMG},x}\mathrm{df}$ wherein P.sub.EMG,x represents a power spectral density function of EMG signals of the currently calculated greening quality factor object x, and f represents a frequency of the EMG signals; and normalizing the physiological activation feature parameters of each greening quality factor and introducing the following calculation formula: $A_{\mathrm{Normalized}}=\left(\frac{A\left(i\right)-A_{\min }}{A_{\max }-A_{\min }}\right)?100$ wherein A(i) represents an i.sup.th activation feature parameter, A.sub.min and A.sub.max represent a minimum value and a maximum value of the activation feature parameters, and A.sub.Normalized represents normalized activation feature parameters.

16. The device according to claim 9, wherein in the street greening quality detection method, the process of analyzing the weighted average greening activation index data of the greening quality factors to form the street greening quality detection model comprises the following steps: obtaining a weighted average greening activation index of each class of greening quality factors, checking the weighted average greening activation index data of the greening quality factors by KMO measure of sampling adequacy and Bartlett's test of sphericity, wherein when KMO value >0.6 and sphericity test adjoint probability P value?0.01, it is considered that factor variables are strongly correlated and are suitable for further analysis on greening factor objects; calculating a cumulative variance contribution rate M.sub.K of latent principal components of an initial greening quality variable matrix X={x.sub.ij}, (i=1, 2, 3, . . . , m; j=1, 2, 3, . . . , n), and selecting latent principal components of greening quality at M.sub.K?80% as follows: $M_K=\underset{i=1}{\overset{k}{.Math.}}\frac{{?}_{\mathrm{ij}}}{\sqrt{{?}_j}}$ wherein x.sub.ij represents a j.sup.th greening quality variable factor of an i.sup.th sample; ?.sub.ij/?{square root over (?.sub.j)} represents a k.sup.th latent principal component of the greening quality, ?.sub.ij represents factor load of a j.sup.th latent principal component of an i.sup.th variable factor, and ?.sub.ij represents a characteristic root of the j.sup.th latent principal component; extracting first k latent principal components to detect the street greening quality, and calculating a weight w.sub.i of a single greening quality variable factor according to the related coefficient matrices and the variance contribution rate as follows: $w_i^{}=.Math.\underset{i=1}{\overset{n}{.Math.}}\left(\frac{{?}_{\mathrm{ij}}}{\sqrt{{?}_j}}*{?}_j\right).Math./{?}_j$ wherein ?.sub.j represents a variance contribution rate corresponding to the j.sup.th latent principal component of the greening quality, and the larger the w.sub.i, the greater the importance of the greening quality variable factor; and forming the street greening quality detection model according to weight coefficients of the greening quality variable factors, which is used for contrast detection of the street greening quality: $G={?}_1{x}_1+{?}_2{x}_2+{?}_3{x}_3+.Math.+{?}_j{x}_j$ wherein ?.sub.1 represents an influence coefficient of the i.sup.th factor, and x.sub.j represents greening activation index data of a j.sup.th re-extracted factor.

17. The device according to claim 9, wherein in the street greening quality detection method, the process of inputting the annotated street samples to be analyzed into the street greening quality detection model to obtain the annotated results of the street greening quality grading detection target data comprises the following steps: collecting physiological data of J subjects with respect to I street greening images of N street samples to obtain an initial greening quality variable matrix Z={z.sub.ij}, (i=1, 2, 3, . . . , M; j=1, 2, 3, . . . , N) of M greening quality variable factors of the N street samples, and annotating the EEG, ECG, EDA and EMG physiological activation feature parameters of the N street samples in classes according to the classes of the greening quality variable factor indexes; establishing a greening activation relationship fusion model among the EEG, ECG, EDA and EMG activation feature parameters, generating a fused greening activation index of the J subjects with respect to the street greening quality variable factors, setting confidence of the greening activation index data within a [0, 1] interval, and annotating the variable factors of the street samples with activation degrees; presetting greening quality detection conditions, dividing the street greening quality into four levels G.sub.1, G.sub.2, G.sub.3, and G.sub.4, and assigning hierarchical values to element dimensions of the street samples from high to low, to rank the greening quality of the street samples; $C_i=\{\begin{array}{c}4,x_{\mathrm{ij}}?\left[\stackrel{?}{x}+0.5?,x_{\max }\right]\\ 3,x_{\mathrm{ij}}?[\stackrel{?}{x},\stackrel{?}{x}+0.5?)\\ 2,x_{\mathrm{ij}}?[\stackrel{?}{x}-0.5?,\stackrel{?}{x})\\ 1,x_{\mathrm{ij}}?[x_{\min },\stackrel{?}{x}-0.5?)\end{array}$ wherein x.sub.ij represents greening activation annotation data of an j.sup.th element dimension at an i.sup.th location, x represents an average value of greening activation degrees of all samples, ? represents a standard deviation of the greening activation degrees of all samples, and C.sub.i represents a level of assigned greening quality of an i.sup.th element dimension; and inputting the annotated street samples into the street greening quality detection model, obtaining activation degrees of the element dimensions of the street samples one by one through the calculated greening variable factors and element dimensions, on this basis, reassigning the greening quality of each element dimension with reference to the greening quality grading detection conditions, and performing weighted superposition on dimension weights of the greening quality elements to form and annotate dimensionless greening quality values of the street samples; $Y=\underset{t=1}{\overset{n}{.Math.}}{w}_t^{*}{C}_t^{}$ wherein Y represents the greening quality values of the street samples, w*.sub.t represents a weight of a i.sup.th element dimension, and C.sub.t represents level assignment of greening quality at the t.sup.th element dimension.

18. The storage medium according to claim 10, wherein in the street greening quality detection method, the process of establishing the greening quality factor index system comprises the following steps: collecting statistics on frequencies of street greening structures, plant attributes, visual landscape and the like, selecting high-frequency street greening quality factors, and sorting out constituent elements, typical features and environmental connotations of street greening by a theoretical analysis method, to establish the street greening quality factor index system, wherein the street greening quality factor index system comprises primary element dimensions, secondary variable factors and tertiary factor change form indexes; acquiring data of built environment street view images, determining street greening scene class images in the street view images by location scene recognition technology, performing feature sampling on single greening quality variable factors of the greening scene class images by image element semantic segmentation technology, and determining clear street greening target images by a square gradient function; and randomly selecting m street greening images from the street greening target images, and performing three phase randomizations on the m street greening images to obtain 3*m random phase images and form an experimental stimulus image library; dividing all experimental stimulus images in the experimental stimulus image library into n groups through inter-group experiments, and playing the experimental stimulus images at a random inter-group and same frequency in a laboratory environment to obtain EEG, ECG, EDA, EMG and trigger signals of corresponding data segments of the images in real time.

19. The storage medium according to claim 18, wherein in the street greening quality detection method, primary element dimension indexes of the greening quality factor index system comprise greening structure, plant texture, line of sight relationship, and landscape characteristics; the secondary variable factor indexes are extensions of primary greening quality elements; and the tertiary factor change form indexes are manifestations of variable factors, and factor features of greening quality are sampled through the built environment street view images.

20. The storage medium according to claim 10, wherein in the street greening quality detection method, the process of collecting the EEG, ECG, EDA and EMG raw data stimulated by the street greening images, and performing the reclassification and differential wave processing on the raw data according to the greening quality factor indexes comprises the following steps: capturing raw data of each street greening target image when stimulated, and classifying raw data segments representing a same greening quality variable factor into one class according to markers recorded by trigger signals, wherein each class of data segments reflects EEG, ECG, EDA and EMG changes of subjects under an influence of a variable factor; performing baseline correction, bandpass filtering, average reference processing, ICA, noise reduction and artifact removal on the raw data segments, and correcting signal offsets by EMD, so as to solve average amplitudes and differential waves of non-stimulus state electrical signals caused by greening quality factor stimulus states; and analyzing amplitudes and phase images of differential waveforms by Hanning windowing, fast Fourier transform and wavelet transform, and extracting ? and ? frequency bands of five EEG differential wave leads PZ, P4, P5, O1, OZ, and O2, low-frequency and high-frequency bands of ECG differential waves at R-R intervals, high-frequency bands of EMG differential waves after full-wave rectification, and a normalized conductivity GSR of EDA differential waves within an exposure time window of the street greening target images, so as to calculate power spectral densities of the EEG, ECG and EMG frequency bands and a first-order difference of the EDA conductivity, to obtain valid physiological data for physiological activation feature recognition of the greening quality factors.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to explain the embodiments of the present invention or the technical solutions in the prior art more clearly, the accompanying drawings required for the description of the embodiments or the prior art will be introduced simply. Apparently, those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0061] FIG. 1 is a flowchart of a street greening quality detection method based on physiological activation recognition of the present invention;

[0062] FIGS. 2A-2D show a feedback graph of superimposed average physiological activation features of subjects according to an embodiment of the present invention;

[0063] FIG. 3 is a diagram of a fused structure of physiological activation features according to an embodiment of the present invention;

[0064] FIGS. 4A-4B show a diagram showing a relationship between latent principal components of greening quality according to an embodiment of the present invention; and

[0065] FIG. 5 is a distribution diagram of influence weights and comprehensive score coefficients of greening quality factors according to an embodiment of the present invention.

DETAILED DESCRIPTION OF THE EMBODIMENTS

[0066] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the invention. Apparently, the described embodiments are only some of the embodiments of the present invention, no all of them. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without any creative effort shall fall within the protection scope of the present invention.

[0067] As shown in FIG. 1, a street greening quality detection method based on physiological activation recognition includes the following steps: [0068] (1) A greening quality factor index system is established according to high-frequency street landscape characteristics, and street greening images are obtained and uniformly processed for greening stimulus physiological experiments: [0069] (1-1) Statistics on frequencies of street greening structures, plant attributes, visual landscape and the like in existing literature are collected, high-frequency street greening quality factors are selected, and constituent elements, typical features and environmental connotations of street greening are sorted out by a theoretical analysis method, to establish the street greening quality factor index system, including primary element dimensions, secondary variable factors and specific tertiary factor change form indexes, as shown in Table. 1. The primary element dimensions include greening structure, plant texture, line of sight relationship, and landscape characteristic. The secondary variable factor indexes are extensions of the primary greening quality elements, including 18 street greening variable factors such as green form, greening composition, and green looking ratio. The tertiary indexes are manifestations of the variable factors, and factor features of greening quality are sampled through the built environment street view images.

TABLE-US-00001 TABLE 1 Street greening quality factor index system Element Variable dimension factor Specific change forms of indexes Greening Green pattern (GP) Regular Slightly natural Natural structure Green class (GC) ?2 classes 3-5 classes >5 classes Green combination Shrub Arbor Arbor and shrub (GP) Green form (GF) Linear, group, cluster, coupled, and isolated planting View method (GM) Flower mirror Flower bed Three- dimensional greening Plant Green color (GO) Monochromatic Contrast two Composite colors texture system colors Green situation (GS) Poor Ordinary Good Plant trait (BF) Branches Contour Branches and leaves Plant luster (BL) Poor Ordinary Good Seasonal ratio (SR) ?30% (30, 60]% (60100]% Line of sight Green visual ratio (5, 30]% (30, 60]% (60100]% relationship (GV) Intervisibility (VA) ?10% (10, 40]% (40100]% Green density (DC) ?30% (30, 60]% (60100]% Green height (GH) ?0.6 m (0.6, 1.5]m.sup. >1.5 m Landscape Oddity (ODD) Low price Moderate price High price characteristics Awareness (ARE) Poor Ordinary Good Ancient wood ratio ?10% (10, 40]% (40100]% (PAW) Tree age (GAT) ?10 years (10, 30) years (30, 100] years [0070] (1-2) Data of built environment street view images are acquired, street greening scene class images in the street view images are determined by location scene recognition technology, feature sampling is performed on single greening quality variable factors of the greening scene class images by image element semantic segmentation technology, and clear street greening target images are determined by a square gradient function; and [0071] (1-3) m street greening images are randomly selected from the experimental materials, and three phase randomizations are performed on the m street greening images to obtain 3*m random phase images and form an experimental stimulus image library. All experimental stimulus images are divided into n groups through inter-group experiments, and played at a random inter-group and same frequency in a laboratory environment to obtain EEG, ECG, EDA, EMG and trigger signals of corresponding data segments of the images in real time.

[0072] In this embodiment, a Place365-CNN model dataset and a ResNet152-Hybrid1365 scene classification model are selected as street greening scene detection tools. The images sorted in the top three places in detection labels and related to the semantics of greening elements are determined as street greening scene class images. Feature sampling is performed on variable factors of the street greening scene class images by using an ADE20K-CNN dataset and a Cascade DilatedNet semantic segmentation model. The images with a green visual ratio >5% are used as the street greening target images for next analysis and processing.

[0073] 80 typical street greening target images with 3-5 different change forms are selected for each greening quality factor, and 3*80=240 random street greening phase images after three phase randomizations are obtained to form the experimental stimulus image library. The experimental stimulus images are divided into 2 groups for inter-group experiments, with each image cycled three times and flashed for 3000 ms at a frequency of 10 Hz and with a resting state of 3000 ms between the images. When the experimental stimulus images are played, markers are sent to a physiological oscilloscope to record and collect raw data of EEG, ECG, EDA, EMG and oscilloscope triggered signal changes in real time during playback of each image. A total of 320 raw data segments are captured. [0074] (2) EEG, ECG, EDA and EMG raw data stimulated by the street greening images are collected, and reclassification and differential wave processing are performed on the raw data according to greening quality factor indexes to obtain valid physiological data that can be used for activation feature extraction of greening quality factors; [0075] (2-1) Raw data of each street greening target image when stimulated are captured, and raw data segments representing a same greening quality variable factor are classified into one class according to markers recorded by the trigger signals, where each class of data segments reflects EEG, ECG, EDA and EMG changes of subjects under the influence of a variable factor. Baseline correction, bandpass filtering, average reference processing, independent component analysis (ICA), noise reduction and artifact removal are performed on the raw data segments, and signal offsets are corrected by empirical mode decomposition (EMD), so as to solve average amplitudes and differential waves of non-stimulus state electrical signals caused by greening quality factor stimulus states; and [0076] (2-2) The amplitudes and phase images of differential waveforms are analyzed by Hanning windowing, fast Fourier transform and wavelet transform, and (8-12 Hz) ? and (14-30 Hz) ? frequency bands of five EEG differential wave leads PZ, P4, P5, O1, OZ, and O2, low-frequency (LF: 0.04-0.15 Hz) and high-frequency (LF: 0.15-0.4 Hz) bands of ECG differential waves at R-R intervals, high-frequency (MF: 50-150 Hz) bands of EMG differential waves after full-wave rectification, and a normalized conductivity GSR of EDA differential waves within an exposure time window of the street greening target images are extracted, so as to calculate power spectral densities of the EEG, ECG and EMG frequency bands and a first-order difference of the EDA conductivity, to obtain valid physiological data for physiological activation feature recognition of the greening quality factors.

[0077] In this example, experimental data of a total of 65 subjects are collected to obtain 61 groups of valid data. The sampling frequency of the signals is 500 Hz. The baseline correction, noise reduction, artifact removal, filtering, independent component analysis (ICA), and signal offset correction pre-processing of the raw data are completed on a Matlab platform by using software packages such as ECGLab, LedaLab, and HRVAS. A resistance value of each lead electrode in EEG is 10 k? or below. A size of the Hanning window is set to 25 ms, and the wavelet transform is Daubechies db2. EEG, ECG, EDA and EMG signal change data from first 2000 ms to last 5000 ms of street greening target image stimulus are captured, and then differential wave analysis is performed on the reclassified signal change data according to the 18 classes of greening quality factors. [0078] (3) EEG, ECG, EDA and EMG physiological activation feature parameters of the greening quality factors are calculated according to the obtained valid physiological data, the physiological activation feature parameters are trained by transfer learning fusion to determine importance of physiological activation features, and weighted average greening activation indexes of the greening quality factors are recognized; [0079] (3-1) Physiological data of each class of greening quality factors are superposed and averaged according to the valid physiological data obtained in step (2-2) after the street greening factors are treated, and the EEG, ECG, EDA and EMG physiological activation feature parameters (as shown in FIGS. 2A-2D) of the greening quality factors are calculated respectively. Specific steps are as follows: [0080] (3-1-1) An EEG activation feature parameter A.sub.EEG of a greening factor object is calculated by the following formula:

[00012]$A_{\mathrm{EEG}}=\frac{P_{?,x}}{P_{?,x}}$ [0081] where P.sub.?,x and P.sub.?,x represent relative average power of the five leads PZ, P4, P5, O1, OZ, and O2 with respect to the ? and ? bands of the currently calculated greening quality factor object x, and the larger the value of A.sub.EEG, the higher the degree of activation of a brain visual area by street greening; [0082] (3-1-2) An ECG activation feature parameter A.sub.ECG of the greening factor object is calculated by the following formula:

[00013]$A_{\mathrm{ECG}}=\frac{P_{\mathrm{LF},x}}{P_{HF,x}}$ [0083] where P.sub.LF,x represents a power value of an ECG low-frequency component of the greening quality factor x, P.sub.HF,x represents a power value of an ECG high-frequency component of the greening quality factor x, and the larger the value of A.sub.ECG, the more active the sympathetic nerve, that is, the higher the degree of activation by street greening; [0084] (3-1-3) An EDA activation feature parameter A.sub.EDA of the greening factor object is calculated by the following formula:

[00014]$A_{\mathrm{EDA}}=\frac{1}{2}\left(t_{peak}-t_{onset}\right)*\left(s_{peak}-s_{onset}\right)$ [0085] where t.sub.peak and t.sub.onset represent a peak and beginning of ?GSR rise time during stress response, s.sub.peak and s.sub.onset represent a peak and beginning of a ?GSR amplitude value during stress response, A.sub.EDA represents activation of 10% higher than a baseline during GSR stress response, and the larger the value of A.sub.EDA, the higher the activation energy for street greening; [0086] (3-1-4) An EMG activation feature parameter A.sub.EMG of the greening factor object is calculated by the following formula:

[00015]$A_{\mathrm{EMG}}=\frac{1}{2}{?}_0^{+?}{P}_{\mathrm{EMG},x}\mathrm{df}$ [0087] where P.sub.EMG,x represents a power spectral density function of EMG signals of the currently calculated greening quality factor object x, f represents a frequency of the EMG signals, and the larger the value of A.sub.EMG, the higher the activation energy for street greening;

[0088] In order to eliminate individual differences among the subjects, the physiological activation features of each greening quality factor are normalized and the following calculation formula is introduced:

[00016]$A_{\mathrm{Normalized}}=\left(\frac{A\left(i\right)-A_{\min }}{A_{\max }-A_{\min }}\right)?100$ [0089] where A(i) represents the i.sup.th activation feature parameter, A.sub.min and A.sub.max represent a minimum value and a maximum value of the activation feature parameters, and Normalized represents normalized activation feature parameters; [0090] (3-2) A physiological activation feature vector A.sup.(m)={a.sub.x.sup.(m)}, (x=1, 2, . . . , N, m=1, 2, . . . , J) of the greening quality factors is obtained from the normalized EEG, ECG, EDA and EMG physiological activation feature parameters of the greening quality factors, where a.sub.i.sup.(m) represents an m.sup.th physiological activation feature of an x.sup.th class of greening quality factor objects. A physiological activation feature importance determination matrix B={b.sub.ij} is constructed, where b.sub.ij represents an importance degree ratio of an i.sup.th activation feature dimension to an j.sup.th activation feature dimension; and consequently, a weight vector w*=[w.sub.1, w.sub.2, . . . , w.sub.j] of each feature is obtained; [0091] (3-3) The physiological activation features are fused by transfer learning TLDA (as shown in FIG. 3), 70% of the samples are used as a source domain dataset, remaining 30% are used as a target domain dataset, marker activation values of source domain street greening target images are obtained, sparse self-encoding is performed on the physiological activation feature vector A.sup.(m) and the marker activation value Y of the greening quality factors, the number of neurons q (q<m) in a self-encoder is determined, A.sup.(m) is introduced into a neural network, and physiological activation feature weights are assigned after neural network training to obtain an ensemble vector E of fused features and a weighted average greening activation index O corresponding to E, as follows:

[00017]$O=\frac{1}{4}\underset{i=1}{\overset{m}{.Math.}}\frac{{?}_i}{\underset{j=1}{\overset{m}{.Math.}}{?}_j}{o}_k$ [0092] where ?.sub.i represents a weight of each source domain, namely, a modulus of a similarity vector in the ensemble vector E, and o.sub.k represents a predicted activation degree of a k.sup.th class of greening quality factors.

[0093] In this example, EEG signals of the five leads PZ, P4, P5, O1, OZ, and O2 are collected, so EEG has five features. Besides, 2 EDA features, 1 EEG feature, and 1 EMG feature are added, and a total of 9 physiological activation features are collected. For the setting of the activation degree Y of the source domain marker, the environmental activation degree of an SAM scale is used as a determination index. Four transfer learning models are trained according to the EEG, ECG, EDA and EMG physiological activation features. A sigmoid activation function is used for the models, weight parameters of the model are optimized by gradient descent SGD, and fusion models are evaluated by means of balanced F score F1-score and accuracy (as shown in Table 2). After training, the accuracy of the EEG+ECG+EDA+EMG model is the highest, with physiological activation feature weights of 44.2%, 35.47%, 12.16%, and 8.17%.

TABLE-US-00002 TABLE 2 Contrast of fusion results of physiological activation features Fused features Accuracy (%) F1-score EEG 86.47 0.8634 EEG + heart rate 91.04 0.9043 EEG + heart rate + EDA 91.89 0.9125 EEG + heart rate + EDA + EMG 92.47 0.9323 [0094] (4) Weighted average greening activation index data of the greening quality factors are analyzed to form a street greening quality detection model for contrast detection of street greening quality [0095] (4-1) A weighted average greening activation index of each class of greening quality factors is obtained, the weighted average greening activation index data of the greening quality factors are checked by KMO measure of sampling adequacy and Bartlett's test of sphericity, where when KMO value >0.6 and sphericity test adjoint probability P value?0.01, it is considered that factor variables are strongly correlated and are suitable for further analysis on the greening factor objects; [0096] (4-2) A cumulative variance contribution rate M.sub.K of latent principal components of an initial greening quality variable matrix X={x.sub.ij}, (i=1, 2, 3, . . . , m; j=1, 2, 3, . . . , n) is calculated, and latent principal components of greening quality at M.sub.K?80% are selected as follows:

[00018]$M_K=\underset{i=1}{\overset{k}{.Math.}}\frac{{?}_{\mathrm{ij}}}{\sqrt{{?}_j}}$ [0097] where x.sub.ij represents a j.sup.th greening quality variable factor of an i.sup.th sample; ?.sub.ij/?{square root over (?.sub.j)} represents a k.sup.th latent principal component of the greening quality, ?.sub.ij represents factor load of a j.sup.th latent principal component of an i.sup.th variable factor, and ?.sub.ij represents a characteristic root of the j.sup.th latent principal component; [0098] (4-3) The first k latent principal components are extracted to detect the street greening quality, and a weight w.sub.i of a single greening quality variable factor is calculated according to the related coefficient matrices and the variance contribution rate as follows:

[00019]$w_i^{{}_{}}=.Math.\underset{i=1}{\overset{n}{.Math.}}\left(\frac{{?}_{\mathrm{ij}}}{\sqrt{{?}_j}}*{?}_j\right).Math./{?}_j$ [0099] where ?.sub.j represents a variance contribution rate corresponding to the j.sup.th latent principal component of the greening quality, and the larger the w.sub.i, the greater the importance of the greening quality variable factor; and [0100] (4-4) The street greening quality detection model for contrast detection of the street greening quality is formed according to the factor weight coefficients. The model reflects main factors that affect the greening quality and their contribution rates to the greening quality, specifically:

[00020]$G={?}_1{x}_1+{?}_2{x}_2+{?}_3{x}_3+.Math.+{?}_j{x}_j$ [0101] where ?.sub.1 represents an influence coefficient of the i.sup.th factor, and x.sub.j represents greening activation index data of a j.sup.th re-extracted factor.

[0102] In this example, the initial greening quality variable matrix is constructed with an overall activation degree of a street greening environment as a dependent variable and the greening quality factor greening activation data as independent variables. After KMO and Bartlett's test of sphericity, the KMO sampling adequacy of the variable matrix is 0.610>0.6, and the adjoint probability P value of the Bartlett's test is 0.000?0.01, so the two satisfy conditions and can be further analyzed. Latent principal components are extracted from the greening quality factors to obtain 5 new variables that are independent of each other and include initial factor information, the obtained explanatory total variance is 80.782%, which is more than 80% (as shown in Table 3), and a diagram showing a relationship between latent principal components of greening quality is obtained accordingly (as shown in FIGS. 4A-4B). A latent component factor coefficients (as shown in Table 4) and influence weight (as shown in FIG. 5) of each initial greening quality variable factor are obtained according to step (4-3), and the influence weights of all the greening quality factors are distributed between 5% and 7%, with the view method (GM) having the greatest impact on the greening quality.

TABLE-US-00003 TABLE 3 Explanation on total variance of latent principal components of greening quality Component Initial eigenvalue Sum of squares of extracted load Serial Variance Cumula- Variance Cumula- number Total percentage tive/% Total percentage tive/% 1 5.132 28.513 28.513 5.132 28.513 28.513 2 4.066 22.586 51.1 4.066 22.586 51.1 3 2.446 13.591 64.691 2.446 13.591 64.691 4 1.662 9.232 73.922 1.662 9.232 73.922 5 1.235 6.86 80.782 1.235 6.86 80.782

TABLE-US-00004 TABLE 4 Matrix of latent principal component score coefficients of greening quality Latent principal component Variable factor 1 2 3 4 5 Green pattern (GP) 0.429 0.811 0.238 0.013 ?0.09 Green class (GC) 0.325 0.813 0.39 ?0.092 ?0.159 Green combination (GP) 0.209 0.816 0.093 ?0.154 ?0.064 Green form (GF) 0.576 0.421 ?0.36 0.455 0.167 View method (GM) 0.27 0.252 0.167 ?0.224 0.75 Green color (GO) ?0.06 0.6 0.607 ?0.009 ?0.294 Green situation (GS) 0.399 0.266 ?0.381 0.427 ?0.376 Plant trait (BF) 0.235 0.576 ?0.482 ?0.343 0.163 Plant luster (BL) 0.724 ?0.26 ?0.058 ?0.241 0.224 Seasonal ratio (SR) 0.738 ?0.301 0.114 ?0.365 0.064 Green visual ratio (GV) 0.774 ?0.446 0.056 ?0.129 ?0.141 Intervisibility (VA) 0.832 ?0.285 0.327 ?0.054 ?0.206 Green density (DC) 0.708 ?0.045 0.428 0.21 0.027 Green height (GH) 0.81 ?0.449 0.185 ?0.116 ?0.082 Oddity (ODD) 0.693 0.27 ?0.548 0.257 0.087 Awareness (ARE) 0.082 0.143 0.301 0.673 0.426 Ancient wood 0.056 ?0.555 0.429 0.528 0.043 ratio (PAW) Tree age (GAT) 0.412 ?0.19 ?0.637 0.083 ?0.194 [0103] (5) Annotated street samples to be analyzed are input into the street greening quality detection model to obtain annotated results of street greening quality grading detection target data. [0104] (5-1) Physiological data of J subjects with respect to I street greening images of N street samples are collected to obtain an initial greening quality variable matrix Z={z.sub.ij}, (i=1, 2, 3, . . . , M; j=1, 2, 3, . . . , N) of M greening quality variable factors of the N street samples, and the EEG, ECG, EDA and EMG physiological activation feature parameters of the N street samples are annotated in classes according to the classes of the greening quality variable factor indexes; [0105] (5-2) A greening activation relationship fusion model among the physiological activation feature indexes is established through (3-3), and a fused greening activation index of the J subjects with respect to the street greening quality variable factors is generated, which are conducive to accurate recognition on the degrees of activation of the subjects with respect to different street greening within a sampling time point. Activation confidence is set within a [0, 1] interval, and the variable factors of the street samples are annotated with the activation degrees; [0106] (5-3) Greening quality detection conditions are preset, the street greening quality is divided into four levels G.sub.1, G.sub.2, G.sub.3, and G.sub.1, and hierarchical values are assigned to the element dimensions of the street samples from high to low, to rank the greening quality of the street samples;

[00021]$C_i=\{\begin{array}{c}4,x_{\mathrm{ij}}?\left[\stackrel{\_}{x}+0.5?,x_{\max }\right]\\ 3,x_{\mathrm{ij}}?[\stackrel{\_}{x},\stackrel{\_}{x}+0.5?)\\ 2,x_{\mathrm{ij}}?[\stackrel{\_}{x}-0.5?,\stackrel{\_}{x})\\ 1,x_{\mathrm{ij}}?[x_{\min },\stackrel{\_}{x}-0.5?)\end{array}$ [0107] where x.sub.ij represents greening activation annotation data of an j.sup.th element dimension at an i.sup.th location, x represents an average value of greening activation degrees of all samples, ? represents a standard deviation of the greening activation degrees of all samples, and C.sub.i represents a level of assigned greening quality of the i.sup.th element dimension; and [0108] (5-4) The annotated street samples are input into the street greening quality detection model, activation degrees of the element dimensions of the street samples are obtained one by one through the calculated greening variable factors and element dimensions, on this basis, the greening quality of each element dimension is reassigned with reference to the greening quality grading detection conditions, and weighted superposition is performed on the dimension weights of the greening quality elements to form and annotate dimensionless greening quality values of the street samples;

[00022]$Y=\underset{t=1}{\overset{n}{.Math.}}{w}_t^{*}{C}_t^{{}_{}}$ [0109] where Y represents the greening quality values of the street samples, w*.sub.t represents a weight of a t.sup.th element dimension, and C.sub.t represents level assignment of greening quality at the i.sup.th element dimension.

[0110] In this example, EEG, ECG, EDA, and EMG physiological data of 60 subjects with respect to greening images from 4 street sample locations are collected. According to the greening quality factor index system in (1-1), a plurality of greening activation feature parameters are fused to obtain greening activation index data of 18 variable factors and 4 element dimensions of the street samples. The greening quality values of the element dimension indexes of the samples are compared, counted, and annotated through latent principal component analysis, weighted superposition and greening quality grading detection, feasible improvement measures are provided for areas with abnormal indexes, and the greening quality statuses among the obtained street samples are ranked and compared (as shown in Table 5). The annotation results in this example are compared with the results obtained by testing with the previous greening quality detection model (4-4), showing that the matching rate reaches 86%.

TABLE-US-00005 TABLE 5 Degrees of activation and environmental quality of some street samples Wuhan Zhongshan Shanghai Xinhua Guangzhou Shamian Nanjing Zhongshan Avenue Road Street Road Element Degree of Degree of Degree of Degree of dimension activation Assignment activation Assignment activation Assignment activation Assignment Greening 32.17 2 73.27 4 56.79 4 61.41 4 structure Plant texture 45.64 3 52.14 3 66.14 3 27.67 2 Line of sight 33.45 2 53.45 3 42.74 3 56.78 3 relationship Landscape 23.74 1 32.78 2 47.67 3 39.95 3 characteristics Quality value 2.0674 3.0748 3.2831 3.0074 Rank 4 2 1 3

[0111] Based on the same inventive concept, the present invention further provides a computer device. The computer device includes: one or more processors; and a memory, configured to store one or more computer programs. The programs include program instructions, and the processor is configured to execute the program instructions stored in the memory. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The processor is a computing core and a control core of a terminal, and is configured to implement one or more instructions, specifically to load and execute one or more instructions in a computer storage medium to implement the foregoing method.

[0112] It should be further explained that, based on the same inventive concept, the present invention further provides a computer storage medium storing a computer program that is executed by a processor to perform the foregoing method. The storage medium may be one of or any combination of more computer-readable media. The computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium may be, but is not limited to a system, apparatus, or device of electricity, magnetism, light, electricity, magnetism, infrared, or semiconductor, or any combination of the above. More specific examples of the computer-readable storage medium (non-exhaustive list) include: an electrical connections with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, the computer-readable storage medium may be any tangible medium including or storing a program, and the program may be used by or in combination with an instruction execution system, apparatus, or device.

[0113] In the description of this specification, the description with reference to the term one embodiment, example, specific example, or the like means that a specific feature, structure, material, or characteristic described in conjunction with the embodiment or example is included in at least one embodiment or example of the present disclosure. In this specification, the schematic expression of the foregoing term does not necessarily refer to the same embodiment or example. Moreover, the described specific feature, structure, material, or characteristic can be combined in any one or more embodiments or examples in a suitable manner.

[0114] The above shows and describes the basic principles, main features, and advantages of the present disclosure. Those skilled in the art should be understood that the present disclosure is not limited to the foregoing embodiments. The descriptions in the foregoing embodiments and specification only illustrate the principles of the preset disclosure. The present disclosure has various changes and improvements without departing from the spirit and scope of the present disclosure, and these changes and improvements fall within the scope of protection of the present disclosure.