137 results about "Reflectance map" patented technology

The invention discloses a weak light image enhancement method fusing noise adaptive diffusion and illumination perception, and the method comprises the steps: firstly designing a reflection-illumination decomposition module RID based on a Retinex theory, and extracting a reflectivity component and an illumination component of an input weak light image; then designing an ADRD (adaptive diffusion reflection denoising) module, performing denoising enhancement on the reflectivity component, and outputting an enhanced reflectivity image; introducing a noise estimation network, and dynamically adjusting the de-noising amplitude of the diffusion model according to the noise intensity of the region in the reflectivity graph; designing an illumination perception enhancement module IEM, fusing an attention mechanism and a local enhancement block, enhancing the illumination image, and outputting an enhanced illumination component; and building a complete neural network framework, carrying out joint training on the provided module, finally testing the stored neural network model, and outputting an enhanced image. According to the method, the problems of detail loss, noise residue and uneven illumination of the enhanced image in the prior art are solved.

A laser radar-camera online calibration method and system based on a reflectance map includes: data preprocessing, generating data that meets the input of the neural network part; inputting the preprocessed data into the trained neural network; the neural network will output the edge point features of the reference image and the reflectance map, that is, the reference image and the reflectance map descriptor; by comparing the similarity of the reference image and the reflectance map descriptor, find the matching relationship between the reference image feature and the reflectance map feature point; use the mapping relationship between the reflectance map and the 3D point cloud to find the matching relationship between the reference image and the point cloud; use the EPnP algorithm to estimate the transformation matrix from the given 2D-3D point pair to complete the calibration process.

A visual defect detection method, device, medium and product relate to the field of visual detection, and the method comprises the following steps: firstly, recording brightness changes of pixel points in different illumination directions by constructing an illumination response matrix, and establishing a complete mapping of surface reflection characteristics; secondly, performing joint calculation by using the light source direction matrix and the illumination response matrix to obtain normal vector data and reflectivity data, and representing spatial geometric features and material optical features of the surface respectively; thirdly, calculating a gradient feature map, a curvature feature map and a deformation feature map based on the normal vector data, generating a reflectivity map by using the reflectivity data, and forming a multi-dimensional feature representation system; and finally, in a defect detection stage, a strategy of combining gradient feature pre-screening with multi-feature verification is adopted, and final judgment is carried out by calculating a comprehensive defect feature value, so that the technical problem of insufficient micro defect detection precision of a visual detection method in the prior art is solved.

The invention discloses a method for clearly imaging a mask under an optical objective, and relates to the technical field of optical detection and image processing, and the method comprises the following steps: S1, constructing a light intensity distribution model based on the boundary condition of the field of view of the optical objective in combination with the reflection characteristic and incident angle change rule of a metal edge material, and determining the coverage range of metal edge reflection crosstalk, generating a mask boundary interference prediction map; s2, according to the mask boundary interference prediction map, performing region division on the micro-reflectivity image, extracting brightness gradient characteristics of reflectivity lifting in an interference region, and generating a brightness gradient parameter set required by image filtering; the method is based on light intensity modeling, fusion direction filtering, pixel stripping, gradient reconstruction and credibility weighted fusion, realizes closed-loop control from interference prediction to pixel restoration, has high resolution and adaptivity, can accurately strip edge reflection artifacts and restore a real image structure, remarkably reduces misjudgment and rework rate, and is suitable for large-scale popularization and application. And the mask yield and the stability of the detection system are improved.

The invention discloses a multi-source data fusion image sequence generation method and system based on space-time diffusion, and belongs to the technical field of image sequence generation, and the method comprises the steps: taking an existing satellite remote sensing image and an existing radar echo reflectivity image as input, taking an existing lightning drop point space distribution diagram as output, and training a synchronous mapping network; training a prediction network through the existing satellite remote sensing image and the existing radar echo reflectivity image to obtain a future satellite remote sensing image and a future radar echo reflectivity image; inputting a future satellite remote sensing image and a future radar echo reflectivity image into the trained synchronous mapping network to obtain a first prediction map of a future lightning drop point spatial distribution map; inputting the existing lightning drop point spatial distribution map into the space-time diffusion model to obtain a second prediction map of the future lightning drop point spatial distribution map; and the first prediction image and the second prediction image are spliced to obtain a final prediction image sequence of the future lightning drop point spatial distribution map, so that the prediction precision and robustness are remarkably improved.

The invention discloses an aviation surface reflectance image production method and a related product, and relates to the technical field of optical remote sensing. Comprising the following steps: acquiring an aviation optical image and a ground feature spectrum corresponding to the aviation optical image; and inputting the aviation optical image and the ground feature spectrum into a reflectivity image generation model to obtain an aviation surface reflectivity image. According to the method, after the aviation optical image and the corresponding ground feature spectrum are obtained, complex operation is not needed, the data are directly input into the reflectivity image generation model, and the aviation surface reflectivity image can be obtained.

The invention belongs to the technical field of digital image processing, and relates to an unpaired low-illumination image enhancement method based on Fourier transform. The invention provides an innovative scheme, an unpaired training strategy is adopted, and dependence on paired data is avoided. The method comprises the following steps: firstly, acquiring a low-illumination image and an unpaired high-illumination image, then extracting illumination and reflectivity components from the low-illumination or high-illumination image by using a pre-training separation network based on a Retinex algorithm, and training an enhanced reflectivity component; then, the illumination components are sent to a cyclic consistency generative adversarial network (CycleGAN) for non-pairing training; and finally, multiplying the enhanced reflectivity by the illumination component to obtain a final enhanced image. Through combination of the pre-trained reflectivity graph and the non-pairing training, the illumination and detail recovery capability of the image is effectively improved, the adaptability and generalization capability of the model are high, a remarkable image enhancement effect can be realized without data pairing, and the method has wide practical application potential.

The invention relates to a defect detection method and device, equipment and a storage medium. Acquiring image data of the surface of the to-be-detected object; determining a normal vector diagram and a reflectivity diagram of the surface of the to-be-detected object according to the image data and light source information of a light source adopted when the image data is acquired; inputting the reflectivity graph into a defect detection model to obtain first defect information of the surface of the to-be-detected object; performing image analysis on the normal vector diagram to obtain second defect information of the surface of the to-be-detected object; and determining target defect information of the surface of the to-be-detected object according to the first defect information and the second defect information. At least three images with different brightness distributions on the surface of the to-be-detected object are obtained, so that high-quality imaging on the surface of the to-be-detected object is realized, and the object surface defect detection accuracy is improved; the defect information of the surface of the to-be-detected object is determined through the detection model and image analysis, so that the determined defect information is more accurate, and the object surface defect detection accuracy is further improved.

The invention provides an underwater single-photon laser radar target reconstruction method based on modal fusion and collaborative variational imaging optimization, and the method comprises the following steps: reconstructing the original echo data of an underwater single-photon laser radar into a superposed photon number sequence corresponding to X * Y pixel points, the superposed photon number sequence corresponding to each pixel point is generated by carrying out photon number superposition based on a time bin lattice on multiple laser pulse echo signals received at the position of the pixel point, and the number M of laser pulses for carrying out photon number superposition on each pixel point is the same; performing depth-reflectivity estimation based on modal fusion on the superposed photon number sequence of each pixel point to obtain a depth map and a reflectivity map of the underwater target; and optimizing the depth map and the reflectivity map of the underwater target based on a collaborative variational imaging optimization strategy. By using the method provided by the invention, the high-quality reconstructed image of the underwater target can be quickly obtained.

The invention discloses a high-speed camera for semiconductor graph-free dark field detection and a detection method, and relates to the technical field of semiconductor detection, and the method comprises the steps: integrating an adjustable laser light source array, a high-performance TDI camera and a pre-scanning sensor, carrying out the brightness information collection of a pre-scanning region, and generating a pre-scanning reflectivity graph; sequentially irradiating a pre-scanning area with different illumination light through an adjustable laser light source array, generating a plurality of scanning images, and extracting features to generate a multi-modal differential feature field; predicting a main defect type in the region through a multi-mode illumination and defect feature collaborative enhancement algorithm to generate a corresponding illumination mode instruction; calculating an optimal TDI parameter through a TDI parameter dynamic modulation algorithm to generate a TDI camera control instruction; and scanning the wafer according to the two instructions, collecting original image data, inputting the original image data into the depth defect classification model, and outputting a defect information graph and corresponding information. The invention aims to solve the problem that semiconductor detection precision and efficiency are difficult to balance.

The invention discloses a thunderstorm and gale identification method and system based on an attention mechanism convolutional neural network. The thunderstorm and gale identification method comprises the following steps: 1, generating a sample set according to historical weather radar single station base data and thunderstorm and gale event occurrence actual measurement records, and splitting the sample set into a training set and a verification set; 2, training the recognition model by adopting a training set, verifying the generalization ability of the recognition model by adopting a verification set, and selecting an optimal model parameter according to the model performance; the input of the identification model is a radar reflectivity graph and a radar radial velocity graph of a to-be-identified area at the same time, and the output is a thunderstorm and gale estimation label; and 3, generating a radar reflectivity chart and a radar radial velocity chart according to radar data acquired by a real-time weather radar single station, and inputting the radar reflectivity chart and the radar radial velocity chart into the recognition model to obtain a thunderstorm and gale recognition result. According to the method, for nonlinearity, randomness and burstiness of thunderstorm and gale event space distribution, multi-scale information and position information can be effectively utilized, and the thunderstorm and gale recognition accuracy is improved.

A system and method are provided for analyzing perfusion parameters of skin. The method includes acquiring a multi-band reflectance image of skin; delineating at least one clinically relevant spatial component of the multi-band reflectance image of the surface, where the at least one clinically relevant spatial component has substantially homogenous optical properties; performing reconstruction on the at least one clinically relevant spatial component using a corresponding at least one tailored reconstruction algorithm, respectively, where the at least one tailored reconstruction algorithm is specific to the at least one clinically relevant spatial component; and outputting estimated perfusion parameters of the skin for the at least one reconstructed clinically relevant spatial component from the at least one tailored reconstruction algorithm.

The invention belongs to the technical field of image processing, and particularly relates to a motor assembly defect detection system based on image recognition, which comprises a highlight area segmentation module, a multi-scale highlight repair module, a reflectivity decomposition module and a defect detection module, and is characterized in that a stator magnetic steel potting image is acquired, a highlight area is segmented, and a highlight mask is obtained; constructing a multi-scale pyramid based on the brightness component and the saturation component of the stator magnetic steel potting image, and generating a mixed weight map according to the highlight intensity and the detail significance under each scale; according to the multi-scale pyramid, the mixed weight map and the highlight mask, carrying out layered restoration on the brightness component to obtain a synthetic substrate image, and according to the synthetic substrate image, carrying out guided filtering and reflectivity decomposition to obtain an inherent reflectivity map; and carrying out encapsulation defect detection according to the inherent reflectivity graph. According to the invention, the interference of highlight reflection on encapsulation defect detection is eliminated, and the accuracy and robustness of encapsulation defect detection are improved.

The invention provides a scratch defect enhancement method based on multi-modal gradient graph fusion, which can enhance graph features, obtain a fused image with obvious scratch defects, and provide a data basis for improving the detection precision of tiny scratches on the metal surface. An eight-direction light source is used for obtaining an image, specular reflection of the surface of a workpiece can be restrained, and the synthesis quality is closer to a theoretical value; a direction-insensitive gradient enhancement method is constructed, gradient enhancement is performed on X and Y directions of a normal vector, scratch defects in the normal vector are effectively extracted, then gradient enhancement is performed on a reflectivity image, the noise of the obtained enhanced image is obviously reduced, but the performance of the scratch effect is weakened, and finally, fusion is performed on the three images, so that the reflectivity of the image is improved. A multi-modal fusion gradient map with obvious scratches and less noise is obtained, and scratch enhancement and noise suppression are realized by fusing multiple gradient enhancement maps.

The invention relates to the technical field of defect detection, in particular to a laser remelting surface defect area segmentation method which comprises the following steps: S1, acquiring a laser remelting surface digital image; s2, performing illumination reflection separation processing on the surface digital image to obtain a reflectivity image, and performing highlight detection and suppression on the reflectivity image to obtain a highlight suppression image; s3, on the highlight suppression image, estimating the surface dominant texture orientation and consistency of the highlight suppression image based on neighborhood gradient statistics; by processing the digital image of the laser remelting surface, the image quality is improved from the source, background texture interference is specifically eliminated, fine defect features are intelligently enhanced, and finally high-precision and robust segmentation of the laser remelting surface defects is realized through global optimization and refined post-processing. The problems of low detection precision and poor reliability caused by surface characteristics in the prior art are effectively solved.

The invention discloses a hydraulic concrete structure underwater surface crack detection method, and relates to the technical field of image sensors, and the method comprises the steps: carrying out the transmissivity mapping and background radiation field separation of underwater light field characterization parameters, generating a target reflectivity image, carrying out the multispectral-polarization data fusion of the target reflectivity image, and generating an enhanced feature image; performing multi-scale convolution kernel space transformation and feature pyramid fusion on the enhanced feature image to generate a crack response diagram, and performing anisotropic non-maximum suppression and morphological skeleton refinement on the crack response diagram to generate crack detection data; and performing geometric topology joint optimization and parameter quantification on the crack detection data, and outputting a structured detection report. According to the invention, by constructing the light field transmission model and synchronously carrying out polarization and depolarization ratio combined noise filtering and geometric distortion correction, adaptive noise filtering is realized, and the measurement precision is improved.

This invention discloses a radar echo extrapolation method that incorporates three-dimensional environmental field data. Radar baseline data is collected; each set of 21 consecutive radar reflectivity maps is converted into a sample, with the first 11 serving as input and the last 10 serving as labels. Based on the sample information, matching ERA5 global analysis data is collected. The radar reflectivity sequence and the corresponding environmental field are used as input to train a 3D convolutional neural network model. Experimental results show that the inclusion of the environmental field improves the extrapolation performance compared to a model without the environmental field, and is more consistent with meteorological interpretation of the extrapolation principle. By incorporating simultaneous three-dimensional environmental field data, the invention achieves accurate constraints and corrections on the extrapolation results.

The invention discloses a terminal welding abnormal defect detection method, and relates to the technical field of image data processing, and the method comprises the following steps: carrying out the image collection through a multi-light-source image collection module, and obtaining a single-light-source image and a uniform-light-source image; calculating through a single light source image to obtain a reflectivity image, and obtaining a reflection grey-scale map through bilateral filtering and nonlinear transformation of an enhanced kernel function; fourier transform is carried out on the reflection grey-scale map, and a frequency domain interest map is obtained through mean filtering and subtraction; performing feature extraction through binary matrix sampling and a conditional probability model to obtain a two-dimensional interest graph; carrying out two times of pixel point screening and maximum value suppression on the reflection grey-scale map to obtain a texture interest map; carrying out HSV color space conversion on the uniform light source image, and calculating through Gaussian filtering and a normal form distance model to obtain a color interest graph; and obtaining a defect feature map through weighted fusion, carrying out threshold segmentation on the defect feature map, marking defects, and completing identification of the terminal welding defects.

The invention provides a progressive Retinex-guided low-light image enhancement method, which comprises the following steps of: acquiring an original RGB low-light image Ilow, extracting an illumination component initial state # imgabs0 # and a reflection component initial state # imgabs1 # of the Ilow through a convolution block, and then performing iterative enhancement on # imgabs2 # # imgabs3 # by using a complementary coding module, a complete illumination component # imgabs4 # and a complete reflection component # imgabs5 # are generated, an illumination adjustment module is used for enhancing the complete illumination component # imgabs6 # of the Ilow, an enhanced illumination result # imgabs7 # is obtained, and a noise removal module is used for conducting denoising operation on the complete reflection component # imgabs8 # of the Ilow; in the reconstruction stage, an obtained illumination result # imgabs9 # and a denoising result # imgabs10 # are multiplied element by element to generate a final enhanced image result # imgabs11 #. According to the method, a multi-scale residual dense network is introduced to remove noise, and particularly unexpected particles are removed in a dark area of a real image. The reflectance map is recovered by associating local and global features, thereby better eliminating noise.

An apparatus is described which includes two or more colour displays arranged to provide piece-wise continuous illumination of a volume, and one or more cameras arranged to image the volume. The apparatus is configured to control the colour displays and the cameras to illuminate the volume with each of two or more illumination conditions. The apparatus is also configured to obtain two or more sets of images, which include sufficient information for calculation of a reflectance map and a photometric normal map of an object or subject positioned within the volume. Each set of images is obtained during illumination of the volume with one or more corresponding illumination conditions. When viewed from the volume, the apparatus only provides direct illumination of the volume from angles within a zone of a hemisphere, which is less than a hemisphere.

The invention relates to a non-reference low-illumination remote sensing image enhancement method based on multi-branch feature fusion. A model of the invention is totally divided into three modules: a denoising module, a multi-dimensional feature fusion enhancement module and a detail recovery module. The de-noising module is mainly responsible for de-noising the image; the multi-dimensional feature fusion enhancement module obtains an enhanced illumination image by designing a semantic information extraction branch, a spatial information extraction branch and a frequency information extraction branch and fully utilizing feature information under three different dimensions; and a detail recovery module designs a space detail recovery branch and a frequency detail recovery branch under the guidance of the enhanced illumination image in the previous stage to respectively extract space information and frequency information of the image, and a finally generated reflectivity image is used as an enhanced image. The model significantly improves the practicability and generalization ability of low-illumination remote sensing image enhancement.

Embodiments herein provide a method and system for a vicarious calibration of optical data from satellite sensors for urban scene flat fields. Identifying test sites automatically in the urban scene helps in vicarious calibration or on-board calibration of the hyperspectral / multispectral image. An internal average relative reflectance is calculated to get a relative reflectance of the image. Band ratios for various pixels is determined to assess flatness of the spectrum. Flat field candidates are identified from the various pixels having average band ratio nearing zero and a morphological technique is applied to determine a flat field. Finally, the image is calibrated vicariously based on the determined flat field as a test site. The on-board calibration of the remote sensing image may lead to a faster way to get the reflectance image of the scene, with the help of the calibration constants.

The invention discloses an image fusion method and system with low light enhancement and overexposure suppression, and belongs to the technical field of image processing, and the method comprises the steps: extracting a reflectivity feature map, a visible light feature map and a dark channel feature map of a visible light image and an infrared feature map of an infrared image in the training process of an image fusion model; obtaining an enhanced reflectivity image according to the reflectivity characteristic pattern; fusing the reflectivity feature map, the visible light feature map and the infrared feature map, and then performing color space conversion to obtain a color fused image; determining an overexposure area and a non-overexposure area of the dark channel feature map; replacing the overexposure area with the features in the infrared feature map, and replacing the non-overexposure area with the features in the enhanced reflectivity image to obtain an overexposure suppressed fusion image; and the fusion image subjected to overexposure suppression participates in model loss calculation. The quality of the color fusion image is improved, and the technical problem that the image fusion quality is affected by overexposure at present is solved.

An electronic device and method for generation of reflectance maps for relightable 3D models is disclosed. The electronic device acquires multi-view image data that includes a set of images of an object and generates a 3D mesh of the object based on the multi-view image data. The electronic device obtains a set of motion-corrected images based on a minimization of a rigid motion associated with the object between images of the set of images and generates texture maps in a UV space based on the set of motion-corrected images and the 3D mesh. The electronic device obtains specular and diffuse reflectance maps based on a separation of specular and diffuse reflectance components from the texture maps, and obtains a relightable 3D model of the object based on the specular and diffuse reflectance maps.

The present application relates to a kind of multi-surface reflection error correction method and equipment for pressure sensitive paint measurement, introduce the process of "pressure sensitive paint luminescence to camera imaging" is simulated by micro light ray tracing technique, based on the known pressure sensitive paint reflectivity, image resolution, camera and model relative position such as parameter, accurate analysis model surface pressure sensitive paint real luminous intensity by back propagation algorithm.Finally, by adjusting the reflection parameter of pressure sensitive paint to zero, directly generate the image without reflection influence under the view angle of camera imaging, so that the corresponding relationship between image pixel and model surface is maintained unchanged while effectively eliminating the interference of multi-surface reflection, avoid additional image registration step.Compared with prior art, the present application has the advantages of high measurement accuracy, wide application scenarios, low cost and the like.

A precipitation scanning aperture RADAR system comprises at least one transmitting satellite (TX) and a plurality of receiving satellites (RXs) flying in a multi-static formation. The TX emits fixed frequency, pseudo-random, non-repeating phase modulated RADAR energy. Range and Doppler information is extracted from reflected signals received by the RXs to form coherent 2D along track / vertical images, which are combined into a 3D reflectivity map using a power-based incoherent modeling algorithm, such as a Gaussian mixture model. Embodiments implement a spotlight mode in which, for each RX, a plurality of the coherent 2D images obtained during successive coherent processing intervals (CPIs) are incoherently combined, and the satellites adjust their orientations between each CPI to compensate for satellite movement. Inter-satellite timing synchronization of 10 ns or better is sufficient for forming the 3D reflectivity map. The range and Doppler information can be extracted by concurrently demodulating at a plurality of reference frequencies.

Disclosed are a monocular high-resolution multispectral imaging system and a control method, relating to the technical field of vegetation index monitoring. The imaging system includes: an image sensor, configured to convert an optical signal of each waveband into a digital signal; a distortionless imaging lens, configured to focus an original image optical signal of a target object on the image sensor to generate a red-green-blue (RGB) original image; a multi-waveband combined narrow-band optical filter, placed between the distortionless imaging lens and the image sensor and configured to constrain a spectral range of an actual receivable optical signal of the image sensor; an incident light calibration module, configured to acquire incident irradiance of ambient light; and a parallel data processing unit, connected to the image sensor and configured to analyze the digital signal in combination with the incident irradiance and generate a reflectivity image for vegetation analysis. It is conducive to solving the problems of high cost as well as insufficient image synchronization and registration precision of existing multispectral sensors.

A system and method for automated timber inventory. The system and method use photogrammetry and an artificial neural network to produce digital maps and models of highly accurate tree canopy delineation, species prediction, tree height modeling, and subsequent merchantable timber values. In one aspect, a method for automated timber inventory involves acquiring a plurality of aerial images for a timber stand of interest, photogrammetrically generating a three-dimensional representation of the timber stand of interest (e.g., an orthomosaic map, a reflectance map, a digital terrain model, a digital surface model, or any combination of the same) from the plurality of aerial images, identifying one or more tree canopies in the three-dimensional representation using an iteratively-improving artificial intelligence convolutional neural network, and generating a digital canopy map of the timber stand of interest using the convolutional neural network, where the digital canopy map includes labels for the tree canopies.

The present invention relates to a self-attention multi-scale encoding and decoding intrinsic image decomposition network, method and application, and belongs to the field of image recognition technology. The network uses a parallel encoding and decoding strategy, that is, two sets of identical but independent encoders and decoders are used to generate images of reflectance maps and illumination maps. Each network includes three parts: an encoder, a decoder and a skip connection. The encoder and decoder are both composed of multiple self-attention modules and downsampling or upsampling modules. The adaptive window self-attention proposed in this method can adaptively adjust the window according to the image content to strengthen the network's attention to different types of features and reduce the resource usage of the model; the channel enhancement self-attention module enhances the effective information in the feature map by expanding the self-attention calculation between different dimensions.

The invention discloses a defect identification method and device based on multi-modal data of a chip preparation process, and the method comprises the steps: obtaining a graphite carrier image after epitaxial growth is completed, and generating a graphite carrier and wafer mapping graph; the method comprises the following steps of: performing multi-modal imaging on a wafer to be detected, acquiring a gray scale image, a reflectivity image and a photoluminescence intensity image, synchronizing according to a pixel level, respectively performing normalization processing, and calculating a physical normalization factor according to a normalization result to obtain a multi-modal image; constructing a wafer grid space database, and storing abnormal areas in the multi-modal image, the electrical test data and the graphite carrier image into the wafer grid space database; and based on the mapped multi-modal image, the electrical test data and the multi-source data of the abnormal region in the graphite carrier image, identifying the defect type according to a preset rule. The problem that in the prior art, due to the fact that data generated in the chip manufacturing process are mutually independent, the defect root cause analysis efficiency is low is solved.