A method for searching a path by using a 3D reconstructed map includes: receiving 3D point-cloud map information and 3D material map information; clustering the 3D point-cloud map information with a clustering algorithm to obtain clustering information, and identifying material attributes of objects in the 3D point-cloud map information with a material neural network model to obtain material attribute information; fusing the those map information based on their coordinate information, thereby outputting fused map information; identifying obstacle areas and non-obstacle areas in the fused map information based on an obstacle neural network model, the clustering information, and the material attribute information; and generating 3D path information according to the non-obstacle areas. Since the 3D path information is generated based on those map information, the obstacle areas and flight spaces are effectively determined to generate an accurate flight path.

Methods and systems are disclosed for correlating synthetic LiDAR data to a real-world domain for use in training an model for use by autonomous vehicle when operating in an environment. To do this, the system will obtain a data set of synthetic LiDAR data, along with images of a real-world environment. The system will transfer the synthetic LiDAR data to a two-dimensional representation, use the two-dimensional representation and the images to train a model that a vehicle can use to operate in a real-world environment.

A display system includes a controller and a head-mounted display. The head-mounted display includes a display, a head support coupled to the display for supporting the display on a head of a user to be viewed by the user, and sensors coupled to the head support for sensing an environment from the head-mounted display unit in low light. The sensors include one or more of an infrared sensor for sensing the environment with infrared electromagnetic radiation, or a depth sensor for sensing distances to objects of the environment, and also include an ultrasonic sensor for sensing the environment with ultrasonic sound waves. The controller determines graphical content according to the sensing of the environment with the one or more of the infrared sensor or the depth sensor and with the ultrasonic sensor, and operates the display to provide the graphical content concurrent with the sensing of the environment.

The invention concerns a device for the holographic and hyperspectral measurement and analysis (2) of a sample (3), comprising; —an acquisition means (2) for acquiring a diffracted image (11) of an image of the sample (3); and interference patterns (12) of a reference light signal (R) and the light signal (O) having passed through the sample (3) to be measured and analysed; and—a means for illuminating the sample (3) focused on the sample (3); and—a means for reconstructing and analysing (1) the hyperspectral holographic image comprising a deep convolutional neural network generating an image for analysis and detection of particularities in the sample.

The present invention discloses an image fusion method based on Fourier spectrum extraction, which comprises: step 1, inputting a plurality of to-be-fused images to a processor of a computer by an input unit of the computer, and performing the following steps by the processor of the computer: performing Fourier transform on images at different focus positions, extracting a frequency component corresponding to an image with the maximum frequency amplitude in the images at different focus positions in a transformed frequency domain space, taking the frequency component as a frequency component of a fused image at the corresponding spatial frequency, traversing each frequency to generate a frequency domain component of the fused image, finally performing inverse Fourier transform on the frequency domain component of the fused image to obtain the fused image; and step 2, outputting the fused image obtained by the processor by an output unit of the computer.

An object texture measurement device includes a visual detection unit that detects a visual texture of an object, a tactile detection unit that detects a tactile texture of the object, and an acquisition unit that acquires a detection result of each of the visual detection unit and the tactile detection unit while relatively moving each of the visual detection unit and the tactile detection unit with respect to the object.

A method and a product for AI recognizing of embolism based on VRDS 4D medical image, the method is applied to a medical imaging apparatus, and the method includes the following steps: determining a bitmap (BMP) data source according to a plurality of scanned images of a target site of a target user, wherein the target site includes an embolism formed on a wall of a target blood vessel; generating target medical image data according to the BMP data source; performing 4D medical imaging according to the target medical image data and determining a feature attribute of the embolism according to an imaging result, wherein the feature attribute includes at least one of the following: density, crawling direction, correspondence with a site of cancer focus and edge characteristics; and determining a type of the embolism according to the features and outputting the type.

An image processing method includes: analyzing, by a white balance gain estimation circuit, a plurality of exposure frames with different exposures to generate a plurality of sets of white balance gain settings; combining the plurality of sets of white balance gain settings to generate a white balance gain table; applying the white balance gain table to the plurality of exposure frames to generate a plurality of white balance compensated frames; and generating a high dynamic range (HDR) frame by performing pixel fusion according to pixel data derived from the plurality of white balance compensated frames.

On-floor obstacle detection using an RGB-D camera is disclosed. An obstacle on a floor is detected by receiving an image including depth channel data and RGB channel data through the RGB-D camera, estimating a ground plane corresponding to the floor based on the depth channel data, obtaining a foreground of the image corresponding to the ground plane based on the depth channel data, performing a distribution modeling on the foreground of the image based on the RGB channel data to obtain a 2D location of the obstacle, and transforming the 2D location of the obstacle into a 3D location of the obstacle based on the depth channel data.

A method and system for training an image classification model is disclosed. An aspect is to separate training processes of a feature value extraction model and an image classification model and train the feature value extraction model on a representative feature value suitable for image classification into a specific label value (e.g., “Peak”), thereby improving accuracy and performance of a classification model for a ground-penetrating radar (GPR) image that is captured by a GPR and is not easy for feature value extraction.