74 results about "Local variance" patented technology

The invention provides a self-adaptive nonlinear image enhancement method and system for a low-light scene of a mobile terminal, and relates to the technical field of image enhancement, and the method comprises the steps: carrying out the image preprocessing and noise reduction, and carrying out the graying and noise suppression of an input color image through a local variance self-adaptive algorithm; adaptive down-sampling is carried out, and the down-sampling proportion is dynamically adjusted according to the image resolution and the content complexity, so that the processing efficiency is improved; brightness adaptive enhancement is carried out, and the overall brightness of the image is rapidly improved by adopting an Otsu method and a lookup table; contrast nonlinear enhancement: enhancing image details and contrast in combination with a Laplace operator and local mean adjustment; and color restoration: restoring the resolution through bilinear interpolation and performing weighted fusion to realize natural color reconstruction. And finally, a high-quality image of which the brightness, the contrast ratio and the color are remarkably improved is output. According to the invention, the recognition accuracy and processing efficiency of the low-illumination image are improved.

The invention provides an adaptive welding seam detection and three-dimensional reconstruction method based on deep learning and binocular vision. The adaptive welding seam detection and three-dimensional reconstruction method comprises the steps of S1, collecting samples and making a training data set; s2, the picture of the sample to be welded is processed, a feature region is recognized, the image quality of the region to be welded is analyzed and evaluated through wavelet transform and local variance, and the noise level and the contrast ratio are calculated; s3, dynamically generating edge detection parameters and model fitting parameters according to the image quality; s4, using an edge detection algorithm to extract edge point cloud of the welding seam area; s5, performing RANSAC linear fitting, weighted least square fitting and polynomial curve fitting on the edge point cloud in parallel; s6, selecting an optimal fitting result based on an image quality adaptive dynamic scoring model; and S7, carrying out three-dimensional coordinate conversion in combination with the three-dimensional matching model IGEV-Stereo, and outputting a final welding seam three-dimensional coordinate. According to the invention, automatic detection of the position and size of the welding seam can be efficiently and accurately realized.

The invention relates to the technical field of detection, in particular to a complex shell detection method and system based on structured light scanning. The method comprises the following steps: calculating a wrapped phase value and a phase modulation amplitude of each pixel point under each frequency, calculating discrete probability distribution containing candidate stripe series and a corresponding probability for each pixel point based on a multi-frequency wrapped phase difference, generating an initial pixel confidence map, and constructing an energy function; the penalty weight of the smooth item is determined according to the phase modulation amplitude and the wrapped phase gradient local variance, obtaining an initial fringe order distribution diagram, updating discrete probability distribution by using an initial pixel confidence map, re-minimizing an energy function to obtain a corrected fringe order distribution diagram, and reconstructing a final three-dimensional shape. And iteratively optimizing a projector principal point and a radial distortion coefficient in system calibration parameters. According to the scheme, the geometric details of the surface can be reserved, and the processing effect and the measurement precision of the complex curved surface and the high-curvature area are improved.

The invention discloses a metal product defect detection method and system based on image recognition. The method comprises the following steps: firstly, executing reflection disturbance digestion processing on a surface image of a to-be-detected metal product to obtain a reflection digestion image; obtaining surface reference texture features of the defect-free metal product, and generating a reference texture model on the basis of a texture distribution rule, a gray average value and texture continuity; performing defect texture gradient separation on the reflection resolution image based on a reference texture model, positioning an abnormal region, and performing segmentation to obtain a suspected defect texture region; performing boundary pixel reconstruction on the suspected defect texture region to obtain a defect texture reconstruction region; obtaining a target defect texture region through local variance enhancement and neighborhood correlation analysis; and finally, the overexposure area is positioned, gray inverse stretching processing is performed, abnormal pixel group feature clustering is performed on the preprocessed defect area, and then a surface defect detection result is output, so that the precision and accuracy of metal product surface defect detection are improved, and the false detection rate is reduced.

According to the infrared image enhancement method based on style migration and multi-level fusion, a style migration preprocessing step comprises five levels of coding and decoding modules, multi-scale feature fusion is realized through cross-level dense jump connection, an encoder compresses spatial dimensions step by step, and a decoder recovers resolution through deconvolution and cancels a batch normalization layer. Information loss of small target features in the standardization process is avoided, and a triple mixing loss function is adopted for training; in the super-resolution reconstruction step, curvelet decomposition is carried out on a style migration result, curve features such as an arc line are accurately captured by adopting a 36-direction wedge-shaped basis function, a local variance threshold mechanism is adopted for high-frequency sub-band fusion, and a style migration feature weight is given to a high-variance region; low-frequency sub-bands are dynamically weighted according to regional contrast, a residual dense block enhanced SRGAN architecture is adopted in a reconstruction stage, and training stability is improved in cooperation with a spectrum normalization discriminator. Through the mode, the technical problem of effectively improving the quality of the low-quality infrared image is solved.

The invention relates to the technical field of image processing, in particular to a brake disc safety performance detection method and system. When the safety performance of the brake disc is detected, appearance detection, size detection and material performance detection are carried out on the brake disc, and detection is comprehensive; meanwhile, an improved adaptive filtering algorithm is adopted to carry out de-noising processing on the to-be-detected brake disc image; performing multi-scale decomposition on the to-be-detected brake disc image, and decomposing the to-be-detected brake disc image into sub-images of multiple scales; performing local feature extraction on each sub-image, and dynamically allocating weights according to the local features; performing adaptive filtering on the sub-image of each scale according to the weight so as to improve the filtering effect; the filtering intensity is dynamically adjusted according to the local variance, the gradient intensity and the texture information, so that the edges and details of the image can be better reserved, and excessive smoothness is avoided; in combination with global and local information, a self-adaptive weight is distributed for each pixel, the filtering effect is optimized, and the adaptability and robustness of the algorithm are improved.

The invention relates to the technical field of fringe projection three-dimensional reconstruction, in particular to a point cloud enhancement method and system based on polarization structure perception and trend guided recovery. The method comprises the following steps: acquiring a multi-angle polarization image, and generating an object area mask based on a polarization angle consistency mask, a polarization angle gradient mask and a linear polarization degree mask; obtaining an original phase diagram through four-step phase shift and complementary Gray codes, obtaining a missing region in an object according to hole distribution of the original phase diagram, and adding morphological structure indexes and local variance features to recognize a real self-shadow region; aiming at the identified real self-shadow area, establishing a structure guiding recovery model combining a first-order structure gradient guiding item and a second-order structure gradient guiding item to obtain a complete object point cloud; and separating a background and an invalid shadow in the object point cloud through the object area mask to obtain a point cloud which only retains the object area.

The invention relates to an intelligent detection method and system for differential settlement of a super high-rise building foundation, and belongs to the technical field of settlement detection. The method comprises the following steps: collecting settlement monitoring data, and constructing a training data set; calculating local statistics according to the spatial topological relation of the monitoring points, and combining with the adjacent point set to standardize the original data to obtain a standardized settlement value; calculating a curvature mutation factor through second-order difference, superposing a local variance weight factor to amplify the contribution of the mutation region, and splicing the standardized settlement feature vector and the curvature mutation factor to obtain an enhanced feature vector; constructing a differential settlement detection model, and inputting the enhanced feature vector into the model to obtain a differential settlement probability and a local abnormal thermodynamic diagram vector; optimizing the training process of the model through the loss function to obtain a trained model; newly collected real-time settlement monitoring data are processed and then input into the trained model, and an intelligent detection result is obtained. According to the invention, the detection capability of the differential settlement of the foundation can be improved.

The application discloses a textile defect detection method and system based on multi-light source dynamic fusion and a double-branch network, and belongs to the technical field of machine vision and industrial detection. The method acquires multi-view images of a textile under coaxial, low-angle and backlight conditions through an encoder trigger line array camera; a local variance calculation and image fusion are performed by using a Softmax dynamic weight algorithm with a temperature coefficient introduced to generate an enhanced background suppression image; subsequently, a first branch lightweight network containing a depth separable convolution and a CBAM mechanism is used to quickly locate a candidate defect area, and then a second branch network is used to finely classify the cropped candidate area, and a Focal Loss is used to solve the long-tail distribution problem. The application effectively solves the problems of easy loss of small defect features in complex texture textiles and large consumption of video memory, meets the demand of real-time online detection of high-resolution images in industrial fields, and greatly reduces the missed detection rate.

A shrimp feed residue lightweight detection method fusing multi-attention features belongs to the field of image processing, is used for solving the problem of improving small target detection accuracy, and is technically characterized by inputting a multi-scale feature map into a neck network of a YOLO network, calculating a fusion weight value through an EnSimAM attention module of the neck network, and obtaining a fusion feature map according to the fusion weight value; wherein the EnSimAM attention module comprises a global attention branch, a local variance enhancement branch and an edge response enhancement branch; the EnSimAM attention module calculates a fusion weight value based on the following mode; the global attention branch calculates a global attention weight; calculating a local attention weight by a local variance enhancement branch; the edge response enhancement branch calculates an edge attention weight; determining adaptive weight values of the global attention weight, the local attention weight and the edge attention weight; and according to the adaptive weight value, calculating a fusion weight value of the global attention weight, the local attention weight and the edge attention weight.

The present application relates to the technical field of computer vision, and particularly relates to a three-dimensional scene reconstruction rendering method and system. The present application generates a corresponding 3D Gaussian ellipsoid for each sparse point cloud of a scene to be reconstructed; obtains a re-projection area of each 3D Gaussian ellipsoid on a two-dimensional image at different viewing angles; takes each re-projection area as the center, and sets all re-projection areas within a certain pixel distance range as adjacent re-projection areas of the re-projection area; based on the sum of the Gaussian kernel function values from the center of the re-projection area of each 3D Gaussian ellipsoid on the two-dimensional image at different viewing angles to the centers of each adjacent re-projection area of the area, the edge intensity, the local variance, the normalized gradient amplitude, and the absolute value of the dot product of the gradient vector and the normal vector, calculates the split control function value of each re-projection area; compares the split control function value of each re-projection area with the size of a set threshold, and controls the 3D Gaussian ellipsoid to split. The present application improves the three-dimensional scene reconstruction rendering precision.

The invention relates to the technical field of textile fabric detection, in particular to a textile fabric defect detection method and system based on image recognition, and the detection method comprises the following steps: S1, collecting multiple frames of linear array images of a to-be-detected textile fabric; s2, generating a residual vector; s3, generating a model residual error; s4, generating a standardized residual error; s5, identifying a defect candidate signal; s6, when the consistency condition is met, judging that the flaws are real flaws and outputting the flaws; dynamic low-rank factor modeling is constructed for online tracking, time-varying fixed stripe noise is filtered out, complex texture interference is decoupled through space pre-whitening and local variance standardization, and defect signals are accurately captured through sparse constraint and space-time consistency verification. The problems that under narrow-band imaging, fixed stripe noise drifting is difficult to restrain, and false detection and missing detection are serious due to complex texture and noise coupling are solved, and finally the composite beneficial effects of high real-time performance, high detection rate and low false alarm rate are synchronously achieved under the condition of a high-speed production line.

This invention belongs to the field of image processing and computer vision technology, and discloses a lightweight single-image super-resolution reconstruction method based on a directional stripe hybrid network. It captures directional structural information of the image from both the spatial and frequency domains through horizontal stripes, vertical stripes, local window attention of the directional stripe self-attention module, and direction-specific convolution of the directional high-frequency enhancement module. The directional stripe self-attention module effectively captures long-distance dependencies extending along specific directions; the directional high-frequency enhancement module extracts high-frequency components through frequency domain decomposition, combines them with direction-specific convolution, and introduces a dual-guided adaptive enhancement mechanism. It identifies texture-rich regions through local variance maps and structural boundaries through edge maps, generating adaptive weights to guide feature enhancement, accurately enhancing high-frequency edges and textures with different orientations while effectively suppressing noise. The two work together to achieve spatial-frequency domain dual-level directional modeling, significantly improving the reconstruction quality of images rich in directional textures, such as architectural scenes and cartoon images, while maintaining a lightweight design.

The application discloses a video target recognition method based on multi-model hot switching, and belongs to the technical field of computer vision and video processing. The method comprises an arbitration module, a switching control module and a feature adaptation and buffer module. The arbitration module generates a switching preparation signal by extracting multi-dimensional indexes such as optical flow mean, local variance, target density and scene confidence in real time through a lightweight channel independent of main reasoning. The switching control module performs atomic replacement of a computation graph pointer in a vertical blanking period, and realizes millisecond-level hot switching with zero frame loss by combining an asynchronous pre-copy and a chasing mechanism of a double buffer. The feature adaptation and buffer module solves tensor shape mismatch between heterogeneous models through a pre-compiled adaptation layer. The application also provides optimization schemes such as multi-index nonlinear fusion decision, zero-copy memory management, local slice focus reasoning and edge-cloud hierarchical unloading, significantly reduces switching delay, guarantees continuous recognition of a video stream, and is suitable for edge computing scenes with limited resources.

The invention provides a breathing mode classification method and system based on a CNN-LSTM neural network, and the method comprises the steps: collecting body surface point cloud data in a non-shielding state and an arm shielding state in a human body breathing process, and extracting a breathing motion feature on the basis of extracting the breathing motion feature; the invention provides an optimization method for abnormal respiratory movement characteristics of a significant respiratory movement area, solves the interference of environmental noise and newborn clinical characteristics on respiratory signals, and comprises the following steps of: firstly, partitioning a thoracic and abdominal voxel model, and extracting the respiratory movement characteristics of the significant area based on KPCA (Kernel Principal Component Analysis); motion artifacts are removed through a Savitzky-Golay filter, body motion interference is inhibited by using a method of fusing a peak threshold method and a local variance threshold, and baseline drift is removed by using a grey wolf optimization algorithm. According to the method, a CNN-LSTM neural network is constructed to classify four breathing modes of normal, rapid, slow and pause, model parameters and evaluation indexes are determined, effective classification and recognition of the breathing modes are achieved, and the monitoring precision and clinical application value are improved.

The present application relates to the technical field of mechanical hand calibration, and especially to a quick calibration method of a connecting rod type double-arm direct-drive mechanical hand; when detecting joint image data and extracting feature point coordinates of the marker, a dynamic threshold Canny edge detection algorithm is used to detect the joint image data, the detection precision and robustness of the Canny edge detection algorithm to the reflective point marker are improved by combining local image statistical characteristics and a dynamic threshold adjustment strategy, and the algorithm is particularly suitable for scenes with large changes in illumination and uneven contrast; furthermore, an adjustment coefficient is determined based on global gradient, global variance and global average brightness of the image, that is, the influence of the local variance and the local average brightness of each pixel point in the image on the high and low thresholds is constrained by the global characteristics of the image, the sensitivity of the algorithm to different image characteristics can be flexibly controlled, and the dynamic threshold Canny edge detection algorithm is more flexible and universal.

This invention provides a method for predicting aging based on metabolite profiles using nonlocal variance enhancement, relating to the fields of medical data analysis and metabolomics. The method acquires LC-MS and / or GC-MS metabolite profile data and the actual age of the subject to be predicted. It performs missing data checks and intra-class normalization on the data, constructs a class-constrained nonlocal variance enhancement feature extraction model, and combines multi-kernel principal component analysis and linear multi-view fusion to obtain fused metabolite profile features. The fused features are then input into a support vector regression model to output predicted age values. The aging degree is determined based on the difference between the predicted and actual age values, which improves the feature expression ability and interpretability of metabolite profile age prediction.

The invention discloses an image denoising method based on multi-scale feature enhancement and local pixel disturbance. According to the method, features of different scales are extracted by introducing a multi-scale convolution module and cavity convolution, so that the perception capability of complex noise is enhanced. Meanwhile, flat areas are calculated through the local variance of the pseudo clean image, and pixel disturbance is carried out in the areas so as to weaken noise correlation. According to the method, channel attention and space attention mechanisms are combined, and learning of important features is enhanced. In the aspect of loss function design, pixel loss, multi-scale feature loss and local weighting loss are fused. In practical application, a cosine annealing learning rate scheduling and exponential moving average strategy is adopted, and the convergence speed and stability of the model are remarkably improved. According to the method, image details can be effectively reserved, and the method is suitable for denoising tasks of various complex scenes such as natural images and medical images.

This application discloses a method and system for detecting defects in metal products based on image recognition. The method first performs reflection perturbation elimination processing on the surface image of the metal product to be detected to obtain a reflection elimination image; then, it acquires the surface reference texture features of the defect-free metal product, and generates a reference texture model based on texture distribution patterns, gray-level mean, and texture continuity; next, it performs defect texture gradient separation on the reflection elimination image based on the reference texture model, locates abnormal regions, and segments them to obtain suspected defect texture regions; then, it performs boundary pixel reconstruction on the suspected defect texture regions to obtain defect texture reconstruction regions; the target defect texture region is obtained through local variance enhancement and neighborhood correlation analysis; finally, it locates overexposed regions and performs gray-level inverse stretching processing, and outputs the surface defect detection results after clustering abnormal pixel group features of the preprocessed defect regions, thereby improving the accuracy and precision of surface defect detection of metal products and reducing the false detection rate.

The invention discloses a pavement pit slot segmentation method based on adaptive statistical analysis and point cloud geometric features. The method comprises the following steps: converting pavement three-dimensional point cloud data into a depth map; calculating a local variance of the depth map, obtaining a contrast factor according to the local variance, performing CLAHE processing on the depth map, and performing bilinear interpolation on the depth map after the CLAHE processing according to the contrast factor to obtain a contrast enhanced map; binarization processing is carried out on the contrast enhancement image, a binarization result is mapped back to the road surface three-dimensional point cloud, and a target area point cloud is segmented; smoothing the point cloud of the target area, and performing multi-feature combined pre-selected point screening on the smoothed point cloud to obtain a pre-selected point set; and performing plane fitting on the pre-selected point set, calculating the distance from the point cloud of the target area to a fitting plane, and extracting the point cloud of the pavement pit slot based on the distance. The method can rapidly and effectively extract the pit slot on the road surface, and is especially suitable for the road condition with a slope.

The invention discloses a multi-focus image fusion method and system based on Allan variance. The method comprises the following steps: preprocessing a plurality of original multi-focus images with the same scene and different focusing clearness degrees; calculating a pixel mean value in the local window, taking the pixel mean value as a pixel value of a pixel point, and updating the multi-focus grayscale image; obtaining a data sequence in a local window, carrying out multiple groups, carrying out local Allan variance calculation, taking a calculation result as a pixel value of a pixel point, and obtaining a multi-focus feature image under multiple groups; under each group, assignment is carried out according to the pixel values of the multi-focus feature images with different focus definitions at the same position, a plurality of preliminary decision diagrams are generated, optimization and superposition are carried out, and a fusion decision diagram is obtained; and carrying out pixel point fusion on the plurality of original multi-focus images and the fusion decision diagram to generate a multi-focus fusion image. The method can improve the definition of the image and enrich the details of the image.

The invention belongs to the technical field of image analysis, and particularly relates to a waste water circulation treatment monitoring method and system for a sand separator, and the method comprises the steps: collecting a surface image of a waste water pool, enhancing the texture difference between foam and waste water through the technologies of local variance calculation, self-adaptive histogram equalization and the like, and calculating the surface image of the waste water pool; accurately segmenting a foam area by using an active contour model based on an enhanced texture gradient map; the foam coverage rate is obtained by calculating the area surrounded by the contours after convergence, the average brightness and the average texture of a foam area are extracted, and a foam thickness index is obtained through weighted fusion; finally, the system executes a hierarchical cooperative control strategy according to the two key indexes of the foam coverage rate and the thickness index, the defoaming device is intelligently started, stopped or adjusted, and automatic, refined and energy-saving management of foam is achieved.

The invention relates to a self-adaptive gain adjustment fingerprint image enhancement method, which comprises the following steps of: firstly, performing frequency domain decomposition on a fingerprint image to obtain a low-frequency component and a high-frequency component; then calculating a local variance and a local standard deviation of the high-frequency component, and adaptively determining a gain adjustment coefficient according to the standard deviation; performing adaptive enhancement on the high-frequency component by using a gain adjustment coefficient; and finally, recombining the enhanced high-frequency component and the enhanced low-frequency component to obtain an enhanced fingerprint image. Through frequency domain decomposition and adaptive gain adjustment, the definition and contrast of the fingerprint image can be effectively improved, detail features of the fingerprint are highlighted, a good foundation is laid for subsequent fingerprint feature extraction and comparison, and the accuracy of fingerprint identification is improved.

The present disclosure relates to an adaptive semi-global stereo matching parameter setting method, device, electronic equipment and storage medium. The method comprises: based on a left camera image of a binocular camera shooting a target, obtaining the coordinates of a preset left target area and calculating the image proportion; calculating the global variance and the local variance; calculating the local inverse luminosity according to the local variance, generating the minimum disparity number based on a preset correspondence relationship; calculating the joint inverse luminosity according to the image proportion, the global variance and the local variance, generating the sum of absolute differences cost calculation window based on a preset correspondence relationship; configuring the minimum disparity number and the sum of absolute differences cost calculation window to complete the adaptive setting of the semi-global stereo matching parameters. By calculating the minimum disparity number and the sum of absolute differences cost calculation window size of the image, the present disclosure can adapt to various lighting conditions and perform autonomous parameter setting under non-human control.

The application discloses an infrared image non-uniformity prediction and correction method based on wavelet transform, which comprises the following steps: adopting double-density dual-tree complex wavelet transform to perform multi-scale decomposition on an input infrared image; in a high-frequency subband, detecting a blind element position based on a local variance statistical method; in a low-frequency subband, constructing an autoregressive model, predicting non-uniformity of a background region, and performing accurate processing on a prediction result; combining the prediction results of the high-frequency subband and the low-frequency subband to generate a pre-correction coefficient; and adopting the pre-correction coefficient to perform correction processing on an original infrared image to obtain a corrected image. Through the combination of an adaptive fusion strategy, space-time consistency constraint and dynamic gain and bias updating technology, the problems of non-uniformity noise, balance between details and background, device drift compensation and consistency in a dynamic scene in infrared image processing are solved, and the accuracy and stability of infrared image processing are significantly improved.

The present application relates to the technical field of detection, and particularly relates to a complex shell detection method and system based on structured light scanning. The method comprises the following steps: calculating the wrapped phase value and phase modulation amplitude of each pixel point at each frequency, calculating the discrete probability distribution containing the candidate fringe order and the corresponding probability for each pixel point based on the multi-frequency wrapped phase difference, generating an initial pixel confidence map, constructing an energy function, determining the penalty weight of the smoothing term according to the phase modulation amplitude and the local variance of the wrapped phase gradient, obtaining a preliminary fringe order distribution map, updating the discrete probability distribution by using the initial pixel confidence map, re-minimizing the energy function to obtain a corrected fringe order distribution map, reconstructing the final three-dimensional topography, and iteratively optimizing the projector principal point and the radial distortion coefficient in the system calibration parameters. That is, the scheme of the present application can preserve the surface geometric details, improve the effect of processing the complex curved surface and the high-curvature area, and improve the measurement accuracy.

The application belongs to the technical field of image fusion, and particularly relates to a remote sensing image fusion method and system based on local variance image mutual information. The method comprises the following steps: firstly, using a multivariate linear regression component replacement fusion method to perform fusion to obtain an initial fusion image; then, on the basis of the initial fusion image, similarity between an intensity component of a high-resolution multispectral image and an intensity component of the initial fusion image is calculated with the aim of obtaining an optimal Gaussian filter; finally, the original high-resolution panchromatic image is filtered by using the estimated optimal Gaussian filter to obtain a low-resolution panchromatic image, and image fusion is performed on the basis of the low-resolution panchromatic image. The application makes the fused image more conducive to the preservation of spectral characteristics and reduces spatial distortion in the spatial scale.

The invention relates to the technical field of computer systems based on specific calculation models, and discloses a platform spatio-temporal feature fusion method introducing a Transform model, which comprises the following steps: acquiring a platform scene spatio-temporal feature tensor, and calculating similarity distribution to generate an initial attention score matrix; counting the local variance of the score matrix to generate a weight adjustment operator; extracting topological constraint attributes and motion vector parameters of the platform, and mapping to generate a dynamic bias factor; according to the method, a gain adjustment mechanism of semantic uncertainty and physical bias strength is established, the prior intervention strength of the model is automatically enhanced, and the probability of the prior intervention of the model is lowered. Adaptive switching of a data driving mode and a physical guiding mode is achieved, and logic continuity of feature fusion and situation awareness accuracy under complex working conditions are ensured.

The invention belongs to the technical field of computer vision, and particularly relates to an improved RT-DETR-based target detection method in a foggy environment. The invention provides a targeted self-adaptive repair framework which is specially designed for solving the problem of non-uniform degradation caused by fog and is used for positioning degradation first and then performing gating repair. According to the algorithm, firstly, local variance perception is introduced through a variance enhanced convolution block attention mechanism, so that an area degraded due to fog in a feature map is identified; and then decomposing a detection task into two parallel steps of degradation perception and content restoration in a feature degradation enhancement module. In the degradation sensing stage, in order to simulate the physical characteristics of irregular fog form and fuzzy boundary, deformable convolution is adopted, so that the irregular boundary of the degradation area is adaptively fitted. In a content restoration stage, a self-adaptive gating enhancement mechanism is used to take the spatial degradation map generated in the previous step as a pixel-level weight estimation feature correction amount. And finally, performing target detection of the vehicle in the foggy weather by using the repaired features.

The application provides an image processing method based on an improved wavelet threshold function, and belongs to the field of image processing, comprising the following steps: wavelet processing a to-be-processed image to obtain wavelet coefficients; processing the wavelet coefficients by using the improved wavelet threshold function to obtain new wavelet coefficients; and reconstructing to obtain a denoised image according to the new wavelet coefficients, wherein the improved wavelet threshold function involves second-order gradient information of the to-be-processed image, noise standard deviation of the to-be-processed image, local variance and local gradient, and the part of the improved wavelet threshold function processing the wavelet coefficients greater than a threshold value adopts a Gaussian attenuation function, and the part of the improved wavelet threshold function processing the wavelet coefficients less than the threshold value adopts a logarithmic attenuation function. The application can effectively avoid over-denoising or noise residue, and improve image processing quality.