DIAGNOSTIC DICTIONARY REGISTERING DEVICE, DIAGNOSING DEVICE, METHOD, PROGRAM, AND DATA STRUCTURE

Abstract

Even under a different light source, such as outdoor light source, diagnosis of surface states of a diagnosis target object is performed with a high accuracy without measuring spectral distribution information about the light source at the time of measuring a diagnosis target object. A spectral reflectance calculation section (22) calculates, based on pieces of spectral distribution information measured for different surface states of a diagnosis target object and spectral distribution information measured for a reference object with an already-known reflectance, spectral reflectances of the surface states; a reference setting section (23) sets, from a spectral reflectance of a surface state showing a deteriorated state, among the spectral reflectances of the surface states, a wavelength range within which reflectances at the same wavelength are within a predetermined range and the reflectance as a reference wavelength range and a reference reflectance; and a dictionary registration section (24) registers the spectral reflectances of the surface states, the reference wavelength range, the reference reflectance and pieces of spectral distribution information about a plurality of light sources with a dictionary (30).

Claims

1.-8. (canceled)

9. A computer-implemented method for registering and diagnosing objects, the method comprising: receiving, using a plurality of light sources, spectral distribution information of a plurality of surface states of a target object; based at least on the received spectral distribution information and a predetermined spectral distribution information of a reference object, determining a plurality of spectral reflectance values associated with the plurality of surface states of the target object; identifying two or more of the plurality of surface states of the target object in a deteriorating state; determining a range of wavelengths and a reference reflectance value at the range of wavelengths, wherein a plurality of spectral reflectance values of the identified two or more of the plurality of surface states in the deteriorating state are within a predetermined deviation at the range of wavelengths; and registering, as a set of dictionary entries for diagnosing the surface states of a diagnose object, one or more of: the generated plurality of spectral reflectance values of the surface states in the deteriorating state, the determined range of wavelengths, the determined reference reflectance value, and a plurality of spectral distribution information of the plurality of light sources.

10. The computer-implemented method of claim 9, wherein the plurality of light sources are based on a set of predefined light sources for reproducing an outdoor illumination environment.

11. The computer-implemented method of claim 9, the method further comprising: receiving spectral image data of the diagnose object; interactively receiving location information of a deterioration area of the spectral image data; generating spectral distribution information of the deterioration area based on pixel data of the deterioration area of the spectral image; determining degrees of similarity of spectral distribution information in the registered range of wavelengths between the deterioration area and each of the plurality of light sources; based on the determined degrees of similarity, estimating spectral distribution information of one of the plurality of light sources associated with the diagnose object; based on the spectral distribution information of each pixel data of the spectral image and the spectral distribution information of the estimated standard light source, determining surface state at positions associated with each of the pixel data; and providing the determined surface state as a result of a diagnosis of the diagnose object.

12. The computer-implemented method of claim 9, wherein the spectral reflectance values associated with the plurality of surface states of the target object indicate at least a rate of energy, expressed as a ratio of a luminous flux incident of a surface of the target object surface and a reflected luminous flux of each spectrum, that the target object reflects for each wavelength of a light source of spectral distribution, the spectral distribution being a strength of the energy of each wavelength of each of the plurality of light sources.

13. The computer-implemented method of claim 9, wherein the range of wavelengths represent a range of wavelengths where reflectance values of a plurality of surfaces with a deteriorated status are within a predetermined deviation from a predefined reflectance value.

14. The computer-implemented method of claim 9, wherein the spectral distribution information of the plurality of surfaces of the target object are measured using a spectrometer, and wherein the plurality of surfaces include at least one of: polyethylene paint coated over a metal surface, rust fluid, or red rust due to aging.

15. The computer-implemented method of claim 11, the method further comprising: determining a degree of deterioration of the diagnose object based on a ratio of a number of pixels indicating deterioration in a number of pixels in the spectral image data; specifying a status of deterioration of the diagnose object based on the degree of deterioration and a predetermined threshold; and providing the surface status of the deterioration.

16. A system for machine learning, the system comprises: a processor; and a memory storing computer-executable instructions that when executed by the processor cause the system to: receive, using a plurality of light sources, spectral distribution information of a plurality of surface states of a target object; based at least on the received spectral distribution information and a predetermined spectral distribution information of a reference object, determine a plurality of spectral reflectance values associated with the plurality of surface states of the target object; identify two or more of the plurality of surface states of the target object in a deteriorating state; determine a range of wavelengths and a reference reflectance value at the range of wavelengths, wherein a plurality of spectral reflectance values of the identified two or more of the plurality of surface states in the deteriorating state are within a predetermined deviation at the range of wavelengths; and register, as a set of dictionary entries for diagnosing the surface states of a diagnose object, one or more of: the generated plurality of spectral reflectance values of the surface states in the deteriorating state, the determined range of wavelengths, the determined reference reflectance value, and a plurality of spectral distribution information of the plurality of light sources.

17. The system of claim 16, wherein the plurality of light sources are based on a set of predefined light sources for reproducing an outdoor illumination environment.

18. The system of claim 16, the computer-executable instructions when executed further causing the system to: receive spectral image data of the diagnose object; interactively receive location information of a deterioration area of the spectral image data; generate spectral distribution information of the deterioration area based on pixel data of the deterioration area of the spectral image; determine degrees of similarity of spectral distribution information in the registered range of wavelengths between the deterioration area and each of the plurality of light sources; based on the determined degrees of similarity, estimate spectral distribution information of one of the plurality of light sources associated with the diagnose object; based on the spectral distribution information of each pixel data of the spectral image and the spectral distribution information of the estimated standard light source, determine surface state at positions associated with each of the pixel data; and provide the determined surface state as a result of a diagnosis of the diagnose object.

19. The system of claim 16, wherein the spectral reflectance values associated with the plurality of surface states of the target object indicate at least a rate of energy, expressed as a ratio of a luminous flux incident of a surface of the target object surface and a reflected luminous flux of each spectrum, that the target object reflects for each wavelength of a light source of spectral distribution, the spectral distribution being a strength of the energy of each wavelength of each of the plurality of light sources.

20. The system of claim 16, wherein the range of wavelengths represent a range of wavelengths where reflectance values of a plurality of surfaces with a deteriorated status are within a predetermined deviation from a predefined reflectance value.

21. The system of claim 16, wherein the spectral distribution information of the plurality of surfaces of the target object are measured using a spectrometer, and wherein the plurality of surfaces include at least one of: polyethylene paint coated over a metal surface, rust fluid, or red rust due to aging.

22. The system of claim 18, the computer-executable instructions when executed further causing the system to: determine a degree of deterioration of the diagnose object based on a ratio of a number of pixels indicating deterioration in a number of pixels in the spectral image data; specify a status of deterioration of the diagnose object based on the degree of deterioration and a predetermined threshold; and provide the surface status of the deterioration.

23. A computer-readable non-transitory recording medium storing computer-executable instructions that when executed by a processor cause a computer system to: receive, using a plurality of light sources, spectral distribution information of a plurality of surface states of a target object; based at least on the received spectral distribution information and a predetermined spectral distribution information of a reference object, determine a plurality of spectral reflectance values associated with the plurality of surface states of the target object; identify two or more of the plurality of surface states of the target object in a deteriorating state; determine a range of wavelengths and a reference reflectance value at the range of wavelengths, wherein a plurality of spectral reflectance values of the identified two or more of the plurality of surface states in the deteriorating state are within a predetermined deviation at the range of wavelengths; and register, as a set of dictionary entries for diagnosing the surface states of a diagnose object, one or more of: the generated plurality of spectral reflectance values of the surface states in the deteriorating state, the determined range of wavelengths, the determined reference reflectance value, and a plurality of spectral distribution information of the plurality of light sources.

24. The computer-readable non-transitory recording medium of claim 23, wherein the plurality of light sources are based on a set of predefined light sources for reproducing an outdoor illumination environment.

25. The computer-readable non-transitory recording medium of claim 23, the computer-executable instructions when executed further causing the system to: receive spectral image data of the diagnose object; interactively receive location information of a deterioration area of the spectral image data; generate spectral distribution information of the deterioration area based on pixel data of the deterioration area of the spectral image; determine degrees of similarity of spectral distribution information in the registered range of wavelengths between the deterioration area and each of the plurality of light sources; based on the determined degrees of similarity, estimate spectral distribution information of one of the plurality of light sources associated with the diagnose object; based on the spectral distribution information of each pixel data of the spectral image and the spectral distribution information of the estimated standard light source, determine surface state at positions associated with each of the pixel data; and provide the determined surface state as a result of a diagnosis of the diagnose object.

26. The computer-readable non-transitory recording medium of claim 23, wherein the spectral reflectance values associated with the plurality of surface states of the target object indicate at least a rate of energy, expressed as a ratio of a luminous flux incident of a surface of the target object surface and a reflected luminous flux of each spectrum, that the target object reflects for each wavelength of a light source of spectral distribution, the spectral distribution being a strength of the energy of each wavelength of each of the plurality of light sources.

27. The computer-readable non-transitory recording medium of claim 23, wherein the range of wavelengths represent a range of wavelengths where reflectance values of a plurality of surfaces with a deteriorated status are within a predetermined deviation from a predefined reflectance value.

28. The computer-readable non-transitory recording medium of claim 23, wherein the spectral distribution information of the plurality of surfaces of the target object are measured using a spectrometer, and wherein the plurality of surfaces include at least one of: polyethylene paint coated over a metal surface, rust fluid, or red rust due to aging.

Description

BRIEF DESCRIPTION OF DRAWINGS

[0027] FIG. 1 is a functional block diagram of a deterioration diagnosis device according to the present embodiment.

[0028] FIG. 2 is a diagram for illustrating setting of a reference wavelength range and a reference reflectance.

[0029] FIG. 3 is a diagram showing an example of a dictionary.

[0030] FIG. 4 is a flowchart showing an example of a registration processing routine.

[0031] FIG. 5 is a flowchart showing an example of a diagnosis processing routine.

DESCRIPTION OF EMBODIMENT

[0032] An embodiment of the present invention will be described below in detail with reference to drawings.

[0033] A deterioration diagnosis device 10 of the embodiment of the present invention can be configured with a computer that includes a CPU, a RAM and a ROM storing a program for executing a registration processing routine and a diagnosis processing routine to be described later and various kinds of data. Functionally, this deterioration diagnosis device 10 can be represented by a configuration that includes a registration unit 20 and a diagnosis unit 40 as shown in FIG. 1. Note that the registration unit 20 is an example of a dictionary registration device of the present invention, and the diagnosis unit 40 is an example of a diagnosis device of the present invention.

[0034] First, the registration unit 20 will be described. The registration unit 20 can be represented by a configuration that includes a spectral distribution information acquisition section 21, a spectral reflectance calculation section 22, a reference setting section 23 and a dictionary registration section 24.

[0035] The spectral distribution information acquisition section 21 acquires spectral distribution information about each of surface states including a deterioration state of corrosion and the like existing on a surface of a diagnosis target object and original fine states of a coated part and the like. For example, the spectral distribution information about each surface state is measured using a spectrometer such as a hyperspectral camera, for each of samples of surface states of various matters such as polyethylene paint coated on metal to be material for an infrastructure facility, and rust fluid, red rust and brown rust that have occurred due to aging deterioration.

[0036] Further, the spectral distribution information acquisition section 21 acquires spectral distribution information measured for an object with an already-known spectral reflectance using a spectrometer, under the same photographing environment as the time of measuring the spectral distribution information about each surface state. In the present embodiment, a standard white plate is used as the object with an already-known spectral reflectance, and the reflectance is assumed to be 1.00.

[0037] The spectral distribution information acquisition section 21 hands over the acquired spectral distribution information about each surface state and the spectral distribution information about the standard white plate to the spectral reflectance calculation section 22. Note that the spectral distribution information is a brightness value (an amount of radiation or a photometric quantity) for each wavelength.

[0038] The spectral reflectance calculation section 22 calculates a spectral reflectance of each surface state from the spectral distribution information about each surface state and the spectral distribution information about the standard white plate that have been handed over from the spectral distribution information acquisition section 21. Specifically, the spectral reflectance calculation section 22 calculates a spectral reflectance R() for an arbitrary surface state by Formula (1), with a wavelength indicated as , the spectral distribution information about each surface state measured by the spectrometer indicated as C(), and the spectral distribution information about the standard white plate indicated as E().

R()=C()/E()(1)

[0039] The spectral reflectance calculation section 22 hands over the calculated spectral reflectance R() of each surface state to the reference setting section 23 and the dictionary registration section 24.

[0040] Here, FIG. 2 shows an example of the spectral reflectance of each surface state showing a corrosion/deterioration state. As shown in FIG. 2, when a surface state is a corrosion/deterioration state, a wavelength range within which the reflectance is almost constant irrespective of a state of corrosion/deterioration (a degree of progress of corrosion) exists. Therefore, even if the surface states of an observation target object in a corrosion/deterioration state are unknown, it can be assumed that the reflectance of that part is a reflectance within a wavelength range within which the reflectance is almost constant.

[0041] Therefore, from spectral reflectances of surface states of corrosion/deterioration state samples, among the spectral reflectances of the surface states handed over from the spectral reflectance calculation section 22, the reference setting section 23 sets a wavelength range within which a reflectance is almost constant and the reflectance as a reference wavelength range and a reference reflectance R().

[0042] For example, the reference setting section 23 can present a spectral reflectance graph as shown in FIG. 2 to a user, accept a wavelength range and reflectance specified by the user, and set the accepted wavelength range and reflectance as the reference wavelength range and the reference reflectance R().

[0043] Further, for example, the reference setting section 23 may determine a wavelength range within which reflectances at the same wavelength have values within a predetermined range and set the wavelength range as the reference wavelength range . Whether the reflectances at the same wavelength are within the predetermined range or not can be determined, for example, by variance of the reflectances at the same wavelength or by determining whether a difference between the largest and smallest reflectance values at the same wavelength is equal to or smaller than a predetermined value. Further, in this case, the reference setting section 23 can set an average, the largest value, the smallest value or the like of reflectances within the set reference wavelength range as the reference reflectance .

[0044] The reference setting section 23 hands over information about the set reference wavelength range and reference reflectance R() to the dictionary registration section 24.

[0045] With the spectral reflectance of each surface state handed over from the spectral reflectance calculation section 22, the dictionary registration section 24 associates identification information about the surface state and an attribute showing whether the surface state is a deteriorated state or a fine state, and registers the association with a dictionary 30 as a spectral reflectance table 31, for example, as shown in FIG. 3. Note that, in the example of FIG. 3, a sample name of a sample corresponding to a surface state is used as identification information about the surface state. Further, the attribute showing whether a deteriorated state or a fine state is associated with each of the samples showing the surface states in advance, and it is assumed that, at the time of acquiring the spectral distribution information about each surface state, information about the sample name and the attribute is also acquired.

[0046] Further, the dictionary registration section 24 registers the information about the reference wavelength range and the reference reflectance R() handed over from the reference setting section 23 with the dictionary 30 as a reference table 32, for example, as shown in FIG. 3.

[0047] Further, the dictionary registration section 24 associates parameters of a standard light source model corresponding to a light source used at the time of measuring the spectral distribution information acquired by the spectral distribution information acquisition section 21 with pieces of spectral distribution information corresponding to the parameters and registers the association with the dictionary 30 as a light-source spectral distribution information table 33, for example, as shown in FIG. 3. More specifically, the light-source spectral distribution information table 33 can be, for example, a table in which a group of parameters of a standard light source model such as a CIE standard light source D obtained by formulating an outdoor light source are associated with spectral distribution information about the standard light source model calculated using the parameters.

[0048] Next, the diagnosis unit 40 will be described. The diagnosis unit 40 can be represented by a configuration that includes a spectral image acquisition section 41, a light source estimation section 42, a spectral reflectance calculation section 46, a surface state diagnosis section 47 and a result output section 50. Further, the light source estimation section 42 can be represented by a configuration that includes an initial value specification portion 43, a brightness value calculation portion 44 and a similarity determination portion 45. Further, the surface state diagnosis section 47 can be represented by a configuration that includes a pixel attribute identification portion 48 and a deterioration degree calculation portion 49.

[0049] The spectral image acquisition section 41 acquires a spectral image measured by a spectrometer such as a hyperspectral camera for an infrastructure facility or the like which is a diagnosis target object. The spectral image is an image in which each pixel constituting the image has spectral distribution information at a position on the diagnosis target object corresponding to the pixel, as a pixel value. The spectral image acquisition section 41 hands over the acquired spectral image to the initial value specification portion 43 and the spectral reflectance calculation section 46.

[0050] The initial value specification portion 43 presents the spectral image handed over from the spectral image acquisition section 41 to the user and accepts specification of an area showing a corrosion/deterioration position in the spectral image. The initial value specification portion 43 calculates an average of pieces of spectral distribution information which pixels included in the accepted specified area have, as an initial value c() and hands over the initial value c() to the brightness value calculation portion 44.

[0051] The brightness value calculation portion 44 calculates a brightness value e() obtained by dividing a brightness value c() of each wavelength within the reference wavelength range registered with the reference table 32 of the dictionary 30, in the initial value c() of the spectral distribution information about the spectral image handed over from the initial value specification portion 43, by the reference reflectance R() registered with the reference table 32 of the dictionary 30. A formula for calculating the brightness value e() is shown by Formula (2) below.

e()=c()/R()(2)

[0052] The brightness value calculation portion 44 hands over the calculated brightness value e() to the similarity determination portion 45.

[0053] The similarity determination portion 45 calculates a degree of similarity between the brightness value e() handed over from the brightness value calculation portion 44 and a brightness value of spectral distribution information within the reference wavelength range among the pieces of spectral distribution information for the parameters of the standard light source model registered with the light-source spectral distribution information table 33 of the dictionary 30. For calculation of the similarity, for example, a k-nearest neighbor method based on a Euclidean distance or other similarity calculation means can be used. The similarity determination portion 45 judges a parameter of the standard light source model with the highest similarity to the brightness value e() and hands over the parameter of the standard light source model to the spectral reflectance calculation section 46.

[0054] The spectral reflectance calculation section 46 acquires spectral distribution information about the standard light source model associated with the parameter of the standard light source model handed over from the similarity determination portion 45, from the light-source spectral distribution information table 33 of the dictionary 30. By dividing spectral distribution information about each pixel of the spectral image handed over from the spectral image acquisition section 41 by the spectral distribution information about the standard light source model acquired from the dictionary 30, the spectral reflectance calculation section 46 calculates a spectral reflectance of the pixel of the spectral image. The spectral reflectance calculation section 46 hands over the calculated spectral reflectance of each pixel of the spectral image to the pixel attribute identification portion 48.

[0055] The pixel attribute identification portion 48 identifies whether each position on the diagnosis target object corresponding to each pixel of the spectral image handed over from the spectral reflectance calculation section 46 is in a deteriorated state or in a fine state. Specifically, the pixel attribute identification portion 48 calculates a degree of similarity between the spectral reflectance of each pixel of the spectral image handed over from the spectral reflectance calculation section 46 and the spectral reflectance of each surface state registered with the spectral reflectance table 31 of the dictionary 30. For each pixel, the pixel attribute identification portion 48 assigns an attribute associated with a surface state with the highest degree of similarity to the spectral reflectance of the pixel of the spectral image, to the pixel. The pixel attribute identification portion 48 hands over the spectral image for which the attribute of each pixel has been identified, to the deterioration degree calculation portion 49.

[0056] The deterioration degree calculation portion 49 calculates a rate of the number of pixels to which an attribute indicating a deteriorated state is assigned, to an area (the number of pixels) of the whole spectral image, as a degree of deterioration. The deterioration degree calculation portion 49 hands over the calculated degree of deterioration to the result output section 50 as a diagnosis result.

[0057] The result output section 50 outputs the diagnosis result handed over from the deterioration degree calculation portion 49 by displaying the diagnosis result on a display device.

[0058] Next, an operation of the deterioration diagnosis device 10 according to the present embodiment will be described with reference to a flowchart showing a registration processing routine, which is shown in FIG. 4, and a flowchart showing a diagnosis processing routine, which is shown in FIG. 5.

[0059] First, a registration process will be described.

[0060] At step S11 of the registration process shown in FIG. 4, the spectral distribution information acquisition section 21 acquires spectral distribution information about each surface state, and spectral distribution information about a standard white plate measured in the same photographing environment as the time of measuring the spectral distribution information about each surface state. The spectral distribution information acquisition section 21 hands over the acquired spectral distribution information about each surface state and the spectral distribution information about the standard white plate to the spectral reflectance calculation section 22.

[0061] Next, at step S12, the spectral reflectance calculation section 22 calculates a spectral reflectance of each surface state by Formula (1) from the spectral distribution information about each surface state and the spectral distribution information about the standard white plate that have been handed over from the spectral distribution information acquisition section 21. The spectral reflectance calculation section 22 hands over the calculated spectral reflectance of each surface state to the reference setting section 23 and the dictionary registration section 24.

[0062] Next, at step S13, from spectral reflectances of surface states of corrosion/deterioration state samples, among the spectral reflectances of the surface states handed over from the spectral reflectance calculation section 22, the reference setting section 23 sets a wavelength range within which a reflectance is almost constant and the reflectance as a reference wavelength range and a reference reflectance R(). The reference setting section 23 hands over information about the set reference wavelength range and reference reflectance R() to the dictionary registration section 24.

[0063] Next, at step S14, with the spectral reflectance of each surface state handed over from the spectral reflectance calculation section 22, the dictionary registration section 24 associates identification information about the surface state and an attribute showing whether the surface state is a deteriorated state or a fine state, and registers the association with the dictionary 30 as the spectral reflectance table 31, for example, as shown in FIG. 3. Further, the dictionary registration section 24 registers the information about the reference wavelength range and the reference reflectance R() handed over from the reference setting section 23 with the dictionary 30 as the reference table 32, for example, as shown in FIG. 3. Further, the dictionary registration section 24 associates parameters of a standard light source model corresponding to a light source used at the time of measuring the spectral distribution information acquired by the spectral distribution information acquisition section 21 with pieces of spectral distribution information corresponding to the parameters and registers the association with the dictionary 30 as the light-source spectral distribution information table 33, for example, as shown in FIG. 3, and the registration process ends.

[0064] Next, a diagnosis process will be described.

[0065] At step S21 of the diagnosis process shown in FIG. 5, the spectral image acquisition section 41 acquires a spectral image measured by a spectrometer such as a hyperspectral camera for an infrastructure facility or the like which is a diagnosis target object. The spectral image acquisition section 41 hands over the acquired spectral image to the initial value specification portion 43 and the spectral reflectance calculation section 46.

[0066] Next, at step S22, the initial value specification portion 43 presents the spectral image handed over from the spectral image acquisition section 41 to the user and accepts specification of an area showing a corrosion/deterioration position in the spectral image. The initial value specification portion 43 calculates an average of pieces of spectral distribution information which respective pixels included in the accepted specified area have, as an initial value c() and hands over the initial value c() to the brightness value calculation portion 44.

[0067] Next, at step S23, the brightness value calculation portion 44 calculates a brightness value e() obtained by dividing a brightness value c() of each wavelength within the reference wavelength range registered with the reference table 32 of the dictionary 30, in the initial value c() of the spectral distribution information about the spectral image handed over from the initial value specification portion 43, by the reference reflectance R() registered with the reference table 32 of the dictionary 30. The brightness value calculation portion 44 hands over the calculated brightness value e() to the similarity determination portion 45.

[0068] Next, at step S24, the similarity determination portion 45 calculates a degree of similarity between the brightness value e() handed over from the brightness value calculation portion 44 and a brightness value of spectral distribution information within the reference wavelength range among the pieces of spectral distribution information for the parameters of the standard light source model registered with the light-source spectral distribution information table 33 of the dictionary 30. Then, the similarity determination portion 45 judges a parameter of the standard light source model with the highest similarity to the brightness value e() and hands over the parameters of the standard light source model to the spectral reflectance calculation section 46.

[0069] Next, at step S25, the spectral reflectance calculation section 46 acquires spectral distribution information about the standard light source model associated with the parameter of the standard light source model handed over from the similarity determination portion 45, from the light-source spectral distribution information table 33 of the dictionary 30. By dividing spectral distribution information about each pixel of the spectral image handed over from the spectral image acquisition section 41 by the spectral distribution information about the standard light source model acquired from the dictionary 30, the spectral reflectance calculation section 46 calculates a spectral reflectance of the pixel of the spectral image. The spectral reflectance calculation section 46 hands over the calculated spectral reflectance of each pixel of the spectral image to the pixel attribute identification portion 48.

[0070] Next, at step S26, the pixel attribute identification portion 48 calculates a degree of similarity between the spectral reflectance of each pixel of the spectral image handed over from the spectral reflectance calculation section 46 and the spectral reflectance of each surface state registered with the spectral reflectance table 31 of the dictionary 30. Then, for each pixel, the pixel attribute identification portion 48 assigns an attribute associated with a surface state with the highest degree of similarity to the spectral reflectance of the pixel of the spectral image, to the pixel. The pixel attribute identification portion 48 hands over the spectral image for which the attribute of each pixel has been identified, to the deterioration degree calculation portion 49.

[0071] Next, at step S27, the deterioration degree calculation portion 49 calculates a rate of the number of pixels to which an attribute indicating a deteriorated state is assigned, to an area (the number of pixels) of the whole spectral image, as a degree of deterioration. The deterioration degree calculation portion 49 hands over the calculated degree of deterioration to the result output section 50 as a diagnosis result.

[0072] Next, at step S28, the result output section 50 outputs the diagnosis result handed over from the deterioration degree calculation portion 49 by displaying the diagnosis result on the display device, and the diagnosis process ends.

[0073] As described above, according to the deterioration diagnosis device according to the present embodiment, a spectral reflectance of each surface state, a reference wavelength range, within which a reflectance is constant irrespective of a surface state, and a reference reflectance, and spectral distribution information about each parameter of a standard light source model are registered with a dictionary. Then, based on spectral distribution information about a spectral image obtained by measuring a diagnosis target object, and the spectral distribution information about each parameter of the standard light source model, the reference wavelength range and the reference reflectance that have been registered with the dictionary, spectral distribution information about a light source at the time of measuring the spectral image is estimated. Then, by calculating a spectral reflectance of each pixel of the spectral image using the estimated spectral distribution information about the light source, and comparing the spectral reflectance with the spectral reflectance of each surface state registered with the dictionary, a deterioration state of a position on the diagnosis target object corresponding to each pixel is diagnosed. Thereby, it is possible to, even under a different light source, such as outdoor light source, perform diagnosis of surface states of a diagnosis target object with a high accuracy without measuring spectral distribution information about a light source used at the time of measuring a spectral image by a spectrometer.

[0074] The above embodiment is a mere example of the present invention, and it is apparent that the present invention is not limited to the above embodiment. Therefore, addition, omission, replacement and other changes of components may be performed within a range not departing from the technical idea and scope of the present invention.

[0075] For example, the above embodiment has been described on a case where an attribute indicating whether a deteriorated state or a fine state is assigned to each pixel of a spectral image, but the description of the embodiment is non-restrictive. For example, identification information about a surface state (a sample name in the example of the present embodiment) may be assigned. Thereby, a more detailed diagnosis result can be acquired.

REFERENCE SIGNS LIST

[0076] 10 deterioration diagnosis device

[0077] 20 registration unit

[0078] 21 spectral distribution information acquisition section

[0079] 22 spectral reflectance calculation section

[0080] 23 reference setting section

[0081] 24 dictionary registration section

[0082] 30 dictionary

[0083] 31 spectral reflectance table

[0084] 32 reference table

[0085] 33 light-source spectral distribution information table

[0086] 40 diagnosis unit

[0087] 41 spectral image acquisition section

[0088] 42 light source estimation section

[0089] 43 initial value specification portion

[0090] 44 brightness value calculation portion

[0091] 45 similarity determination portion

[0092] 46 spectral reflectance calculation section

[0093] 47 surface state diagnosis section

[0094] 48 pixel attribute identification portion

[0095] 49 deterioration degree calculation portion

[0096] 50 result output section