1061 results about "Image correction" patented technology

The invention discloses a real-time single-stage remote sensing image correction target detection method based on YOLOV8, and relates to the technical field of remote sensing image processing. According to the method, a deformable convolution dynamic prediction local geometric distortion parameter is embedded based on a YOLOv8 backbone network, an adaptive deformation field is generated, pixel-level real-time correction is realized, shallow details and high-level semantic features are fused through a bidirectional path aggregation network, and channel attention and a space gating mechanism are combined, so that the real-time correction of the image is realized. The small target detection capability is enhanced, background noise is suppressed, angle prediction is divided into discrete classification and continuous residual error regression tasks through a decoupling type rotation detection head, angle periodic errors are eliminated in combination with a direction sensitive loss function, and the rotation frame positioning precision is improved. And constructing a dynamic multi-task collaborative loss function, introducing gradient distribution consistency constraint to jointly optimize correction and detection tasks, and realizing feature semantic alignment and model self-enhancement through end-to-end closed-loop training. And the rotating target detection precision and the complex scene robustness are obviously improved.

The invention provides a hull surface defect detection system based on machine vision, and relates to the technical field of data processing. The image correction module is used for carrying out illumination equalization processing and geometric distortion correction; the region construction module is used for identifying a defect-free stable region and generating reference region data which comprises a brightness model and a texture model; the candidate generation module is used for detecting a region where texture interruption or abnormal bright spots exist locally to form candidate defect data, and the candidate defect data comprise pixel positions and local contrast parameters; the stability judgment module is used for carrying out projection matching in the multiple frames of images and simultaneously carrying out joint comparison with the brightness model and the texture model of the reference area data to form real defect data and false defect data; the result output module is used for generating a detection result containing defect coordinates, defect contours, image frame numbers and interference sample prompts; the accuracy of hull surface defect detection is improved.

The invention relates to the technical field of environment assessment, in particular to a coastal zone culture area multi-dimensional environment assessment method and a cloud platform. The method comprises the following steps: continuously recording dissolved oxygen concentration, pH value, centigrade temperature, salinity unit, turbidity scattering unit, chlorophyll fluorescence intensity, ammonia nitrogen milligram per liter and nitrite milligram per liter, synchronously calling multi-source remote sensing images covering a culture area and an adjacent water area, extracting remote sensing reflectivity data through image correction and atmospheric correction, and calculating to obtain water color parameters. According to the method, the water quality sensor array is deployed at the key point of the culture area, continuous multi-parameter environmental data acquisition is carried out, multi-source remote sensing images are synchronously integrated, water color information is extracted in real time, and the timeliness and data accuracy of culture water area environmental monitoring are improved; trend decomposition and dynamic baseline construction are carried out based on the data sequence, so that the environmental fluctuation evaluation of the breeding area is more objective and accurate.

The invention relates to the technical field of image processing and computer vision, in particular to a digital human image correction method, electronic equipment and a storage medium. The digital human image correction method comprises the following steps: preprocessing an image; extracting features; deviation analysis; generating a correction strategy; and correcting the image. According to the invention, image correction is realized by using an image processing library, a deep learning framework and the like through related software modules, so that the digital human image quality is improved, and the sense of reality and the visual effect are enhanced.

The invention discloses a monocular and binocular cooperative localization and mapping method and device for underwater refraction compensation, and the method comprises the steps: extracting checkerboard angular points, compensating an underwater light path through employing a two-time vector correction model based on a Snell law, and precisely calibrating the internal and external parameters of a binocular camera. According to the method, an underwater real light path is effectively restored, and calibration errors caused by medium refraction are remarkably reduced. Then epipolar correction is performed on the corrected left and right images, so that the geometric consistency of stereo matching is ensured; and finally, monocular and binocular poses are fused, and seven-degree-of-freedom similarity transformation global alignment is applied, so that the problem of monocular scale uncertainty is effectively solved, and high-precision and high-robustness three-dimensional mapping is realized by means of binocular real scale introduction. Key links such as camera pre-calibration, refraction compensation, image correction, monocular tracking, binocular depth measurement, scale alignment and the like are connected in series to form an integrated process, so that the underwater vision positioning and mapping precision is improved.

The invention relates to the technical field of machine vision, and discloses a side slope anchor rod drilling platform construction quality intelligent detection system based on machine vision. The method comprises the steps that an acquisition module acquires a multi-angle original image with a space-time mark; the correction module performs preprocessing to obtain a standardized image; the extraction module constructs a drilling platform three-dimensional model; the monitoring module detects track deviation by using an improved adaptive Kalman filtering algorithm; the calculation module evaluates the drilling quality; and the analysis module comprehensively analyzes and generates quality grade and early warning information. According to the intelligent detection method for the construction quality of the side slope anchor rod drilling platform based on machine vision, real-time monitoring, dynamic analysis and quality early warning of the construction process are achieved, subjectivity and hysteresis of traditional manual detection are overcome, and the detection precision and efficiency of the anchor rod drilling construction quality are improved.

The invention relates to the technical field of image correction, in particular to a correction method and system for a low-distortion image of a panoramic camera. The method comprises the following steps: based on a second-hand vehicle scene acquisition demand, acquiring a panoramic image and executing basic preprocessing to obtain a preprocessed panoramic image; calling a lens distortion coefficient, analyzing a corresponding distortion parameter in combination with a checkerboard calibration algorithm, and generating a polar region distortion distribution diagram; carrying out segmentation and grid reconstruction on a polar region of the preprocessed panoramic image, and executing distortion elimination and geometric reduction to obtain a preliminary correction image; fine tuning is carried out on the pixel position of the preliminary correction image, color fault and edge dislocation of a multi-lens splicing seam are eliminated, and an optimized correction image is obtained; executing second-hand car scene customized enhancement processing on the optimized correction image to obtain an enhanced correction image; and automatically checking the enhanced correction image, and outputting a low-distortion final correction image. According to the invention, the correction efficiency of the panoramic low-distortion image of the second-hand vehicle can be improved.

The invention belongs to the technical field of image processing, and particularly relates to a cable shielding layer weaving density detection system based on image analysis, which comprises an image correction module, a frequency domain analysis module, a self-adaptive filtering module and a density defect detection module, obtaining the maximum peak point of the centralized complex frequency spectrum; constructing a band elimination filter according to the maximum peak point, filtering the centralized complex frequency spectrum through the band elimination filter to obtain a filtered complex frequency spectrum, executing inverse Fourier transform to obtain a filtered space image, and binarizing the filtered space image through a threshold segmentation algorithm to obtain a pore mask; the porosity is obtained according to the ratio of the non-pore pixel points to the total number of the pixel points; and judging the weaving defect of the shielding layer according to the porosity. According to the method, the strong periodic signals of the weaving textures are automatically identified and filtered in the frequency domain, so that non-periodic defects such as pores are effectively highlighted, the technical problem that periodic textures cover defect signals is solved, and the accuracy of defect detection is improved.

The invention provides an image correction method and system based on a nonlinear distortion distribution elliptical fisheye lens, and a medium. The method comprises the following steps: establishing a nonlinear mixed distortion imaging model based on a mapping relation between image coordinates and a spherical direction; defining a distortion vector, and performing fitting processing on the distortion vector based on a two-dimensional B-spline function to obtain a distortion function; the method comprises the following steps: acquiring images in a five-surface three-dimensional angle calibration space based on an elliptical fisheye lens, analyzing lens imaging parameters, and generating parameter values of a distortion function; analyzing parameter values of the distortion function based on a nonlinear mixed distortion imaging model to generate calibration parameters, and remapping pixel points in the distorted image based on the calibration parameters; performing image reconstruction on the remapped distorted image to obtain a corrected image; improving the sampling density of a target area by constructing a nonlinear mixed distortion imaging model; the system adapts to complex monitoring environments such as galleries and entrances and exits; and spatial distortion is flexibly expressed through a B spline, so that the modeling capability is improved.

The invention discloses a wide-angle camera distorted image correction method and device based on track visual detection. The method comprises the following steps: acquiring a track distorted image; the projection imaging intersection point between the sleeper and the steel rail is adopted to form a mark point; solving the distortion center of the camera by combining a distortion model according to the mapping relation between the three-dimensional world coordinate and the projection coordinate of the three-dimensional world coordinate in the distorted image; solving distortion parameters of a camera by combining a Levenberg-Marquardt algorithm according to the fact that blanking points formed by parallel steel rails in the image must be subjected to collinear constraint; re-projection correction is carried out on the mark points participating in calculation; screening out bad mark points and removing the bad mark points; the distortion center and the distortion parameters are calculated again until all the mark points meet the 3 sigma criterion; and according to the optimal distortion center and the distortion parameters, track image correction is carried out through a distortion model. According to the invention, high-efficiency and accurate distortion correction can be carried out on line without the participation of a special cooperative calibration object.

The invention relates to the technical field of image correction, in particular to an error calibration method and system for a three-dimensional laser scanner, and the method comprises the steps: obtaining three-dimensional point cloud data in each scanning direction through the three-dimensional laser scanner; any three-dimensional point cloud number is subjected to pooling processing, a plurality of boundary feature curves in pooled point cloud data are obtained, the pooled point cloud data are screened to obtain target point cloud data, reference point clouds are further screened out, the positions of feature points in the reference point clouds are adjusted, adjacent feature points are combined to obtain an adjustment trend consistent area, and the adjustment trend consistent area is obtained; and carrying out movement adjustment on data points except the feature points in the reference point clouds by utilizing the overall adjustment condition of the feature points in the adjustment trend consistent region, carrying out position adjustment on all the data points in all the reference point clouds to obtain new reference point clouds, and carrying out registration splicing on all the new reference point clouds. According to the invention, the accuracy of the three-dimensional point cloud data collected by the three-dimensional laser scanner is improved.

The invention relates to the technical field of image processing, and discloses an image correction method and system for binocular diffraction AR intelligent glasses. The method comprises the following steps: calculating a dispersion response parameter of a diffractive optical element in the AR glasses in a visible light wave band; respectively carrying out dispersion distortion detection on the left eye image and the right eye image in the AR glasses to obtain a left eye dispersion distortion feature and a right eye dispersion distortion feature; interpupillary distance change data of a user is obtained, effective aperture offset calculation is carried out on the diffractive optical element, and interpupillary distance compensation parameters are obtained; performing coupling calculation on the left eye dispersion distortion feature, the right eye dispersion distortion feature and the interpupillary distance compensation parameter to obtain a binocular dispersion correction matrix; and performing dispersion compensation on the left eye image and the right eye image to obtain a dispersion correction image, and outputting the dispersion correction image to a display module of the AR glasses. The method can respond to the pupil distance change of the user in real time, automatically adjust dispersion correction parameters, realize accurate and comprehensive dispersion correction, and ensure the stability of dispersion correction performance.

The invention discloses a farmland soil moisture content real-time surveying system and method based on unmanned aerial vehicle multispectral imaging. The method comprises the steps that multispectral sensor parameters are configured, and an unmanned aerial vehicle flight scanning path is planned; a space coordinate system and time synchronization mechanism is established, and high-precision positioning and time synchronization are achieved; collecting and calibrating visible light, near-infrared and short-wave infrared original image data to form a spectral signal set; calculating an image signal-to-noise ratio, screening a low-quality image, and executing noise suppression; extracting timestamp information from the suppressed image, performing spatial positioning, and establishing a spatial corresponding relation database; calculating time synchronization precision and performing image correction; evaluating the spatial resolution according to the distortion coefficient; calculating a comprehensive quality parameter, comparing the comprehensive quality parameter with a threshold value, and obtaining a soil moisture content distribution parameter when a condition is met; a farmland moisture state map is generated, partition marking is carried out according to a multi-level standard, a monitoring result is output, and high-precision and real-time monitoring of the farmland soil moisture content is achieved.

The invention discloses a defect detection method and defect detection equipment for a circular printed matter. The method comprises the following steps: automatically extracting a minimum enclosing rectangle from a sample circular printed matter area to obtain an image correction proportion; fitting a circular contour of the circular printed matter in the sample circular printed matter area after the proportion correction, randomly selecting two matching areas in an image area enclosed by the circular contour, and performing Mark point positioning based on the two matching areas; acquiring a to-be-detected circular printed matter image, performing proportion correction on the to-be-detected circular printed matter image according to the image correction proportion, and selecting two matching areas to perform Mark point positioning; inversely transforming a Mark point positioning result into the to-be-detected circular printed matter image after scale correction; and aligning the to-be-detected circular printed matter image with the sample circular printed matter image, and carrying out defect detection by adopting a difference image method. The automatic detection precision and the detection efficiency of the circular printed matter can be improved.

The invention discloses a deep learning-based mixed-shooting image correction method and system, and relates to the field of computer vision, and the method comprises the steps: carrying out the illumination normalization processing of a mixed-shooting image group, and constructing a Gaussian pyramid image group; inputting the Gaussian pyramid image group into a ResNet-34 network to extract a multi-scale feature map, and generating spatial position and channel statistical information; and converting the spatial position and the channel statistical information into a DNA sequence fragment through a DNA base mapping rule. Illumination normalization is achieved through Retinex decomposition and Gaussian filtering, the exposure difference of different devices is effectively eliminated, the input quality of follow-up processing is guaranteed, traditional feature description is converted into biological sequence comparison through DNA base coding, geometric parameters are optimized in combination with quantum annealing, and the accuracy of the method is improved. The matching precision of a low-texture area and the correction effect of a large-view-angle-difference scene are remarkably improved, and the sub-pixel-level matching precision is achieved based on weighted RANSAC and multi-scale pyramid fusion.

The invention relates to the technical field of image stitching, and discloses a video transmission high-definition image intelligent stitching method and system. The system comprises an image preprocessing module, a feature point matching and screening module, an image optimal splicing seam generation module and a video output module. The method comprises the following steps: firstly, acquiring a video, carrying out dynamic range equalization, carrying out filtering processing by using an adaptive noise reduction method, and carrying out image correction; secondly, extracting feature points by using a self-adaptive corner detection algorithm, matching the feature points based on a multi-dimensional spatial data index to obtain feature point matching pairs, and screening the feature points; searching an overlapping region, and introducing a dynamic search algorithm to generate an optimal image splicing path to obtain an optimal image splicing seam; and finally, dividing the image according to the dynamic grid, and realizing video output by using priority ranking. According to the method, the video images are processed and spliced, the purpose of intelligent image splicing is achieved, and the method is accurate and objective.

The invention discloses a fish tank dynamic interaction display device based on position detection and an interaction implementation method. Belongs to the technical field of intelligent aquarium equipment and digital interaction and collects fish body images in real time, a control screen supports touch and voice interaction, and a background display screen dynamically updates display content according to the position of a fish body. The control processing hardware module integrates a microprocessor, a real-time digital graphic processing chip and an artificial intelligence processing unit, and image correction, multi-fish body detection tracking, coordinate mapping and dynamic background rendering are achieved. The method comprises the steps of image acquisition, geometric correction, position detection, coordinate mapping, dynamic rendering, cross-device linkage and the like, and supports gesture interaction, AI background generation and multi-fish tank linkage functions. Through combination of real-time position detection and a dynamic display technology, immersive interactive experience of fish moving and scene following is constructed, the problem that a traditional fish tank is insufficient in interactivity is solved, and the fish tank is suitable for families, commercial display and other scenes.

The invention discloses an industrial robot grabbing and positioning system based on computer vision, and relates to the technical field of grabbing and positioning, and the system comprises a projection calculation module which is used for determining a roller direction unit vector of a roller based on a calibration plate image, and constructing a projection data pair based on the calibration plate image and the roller direction unit vector; the data acquisition module is used for generating a one-dimensional cosine curve based on the projection data pair and the candidate angular frequency set of the protection window; the image correction module is used for performing coordinate correction on the pixel coordinates of the calibration plate image based on the one-dimensional cosine curve to obtain a corrected image; the key point extraction module is used for carrying out geometric feature extraction on the correction image to obtain a key point set; and the pose optimization module is used for determining the optimal pose of the calibration plate relative to the camera based on the internal reference matrix of the camera and the three-dimensional reference point set and the key point set of the calibration plate. The grabbing stability and reliability of the robot are improved.

The invention discloses an aerial target tracking method based on multi-frame fusion, and belongs to the technical field of target tracking, and the method comprises the steps: carrying out the preprocessing of an image based on an image correction model, and obtaining a first feature set through combining with a pre-trained deep convolutional neural network model; obtaining a first motion track based on the first feature set; if the deviation between the first motion track and the target position in the actual frame image exceeds a preset threshold value, obtaining an auxiliary data set and combining the auxiliary data set with the first feature set to obtain a second feature set; based on the second feature set, adopting a feature enhancement algorithm based on an attention mechanism to obtain a third feature set; and a continuous tracking state of the target is obtained based on the third feature set, and if the target is lost, a target re-identification algorithm based on a regional proposal network is adopted to obtain a recovered target tracking state, and a tracking optimization algorithm based on an adaptive threshold is adopted to carry out optimization. According to the method, complex illumination and shielding can be effectively dealt with, and the accuracy, robustness and continuity of multi-target tracking are improved.

The embodiment of the invention provides a check image correction method, device and equipment. The method comprises the following steps: acquiring an original deformation check image; processing the original deformation check image through a backbone network, a distortion encoder, a correction decoder and an output predictor in an image correction model in sequence to obtain a target pixel displacement field representing a pixel corresponding relation from the original deformation check image to a corrected check image; and according to the target pixel displacement field, performing bilinear difference distortion operation on the original deformed check image to obtain a corrected check image after the original deformed check image is corrected. An original deformed check image is encoded, an encoded feature map is learned, a corresponding pixel displacement field is generated by a correction decoder, and pixels are sampled from the original deformed check image using the predicted pixel displacement field to reconstruct a corrected check image. Automatic and high-quality deformation check image automatic correction is realized, the labor cost is reduced, and the correction accuracy is improved.

The invention relates to the technical field of detection, in particular to a waste landscaping detection system and method.The waste landscaping detection system comprises an image acquisition module, an image pre-fusion module, an image correction module, an image fusion module and a comprehensive feedback module, and the image correction module comprises a geometric correction sub-module and a radiation correction sub-module; according to the waste landscaping detection system and method provided by the invention, a set of complete detection process is constructed from image acquisition, preprocessing, correction, fusion and feedback, and the problem of inaccurate recognition caused by poor image quality in traditional detection is effectively solved through cooperative and efficient work of all modules, by improving the geometric accuracy of the image and optimizing the radiation characteristic, high-quality image data is provided for subsequent image fusion and recognition, the precision, reliability and efficiency of waste detection are improved, and powerful support is provided for waste landscaping treatment.

The invention discloses an underwater target identification and positioning method based on fusion of a physical model and deep learning, and belongs to the technical field of computer vision, and the method comprises the steps: carrying out the transmissivity estimation and physical restoration of a collected underwater monocular image according to an underwater light propagation model, and carrying out the image correction; key feature points are extracted, and high-quality matching point pairs are screened in combination with the joint similarity; further deriving a basic matrix through the high-quality matching point pairs meeting the epipolar geometric constraint relation, solving an essential matrix, and obtaining relative attitude parameters between the cameras in combination with weighted re-projection-LM optimization; and finally, refraction correction triangulation is carried out according to the Snell's law, pixel-level fusion is carried out after scale normalization and space alignment are carried out on the refraction correction triangulation and dense depth output by the MiDaS, three-dimensional space coordinates of the target are inverted, and high-precision recognition and positioning of the underwater target are achieved. The system is light in structure, efficient in calculation, suitable for being integrated on various underwater autonomous or remote control robot platforms and used for tasks such as target recognition, tracking and positioning.

The invention relates to the technical field of electron microscopic imaging, in particular to a correction method and device for geometric distortion in scanning electron microscope measurement, computer equipment, a storage medium and a computer program product. The method comprises the following steps: acquiring a scanning electron microscope image of a standard sample; calculating an orthogonal included angle parameter and a length-width ratio parameter of the scanning electron microscope image; iteratively adjusting the angle correction parameter and the gain correction parameter under a fixed scanning rotation angle until a preset local calibration constraint condition is met; verifying the orthogonal included angle parameter and the length-width ratio parameter under each scanning rotation angle; and when the orthogonality included angle parameter and the length-width ratio parameter do not meet a preset global stability constraint condition, adjusting a relative gain coefficient, and re-executing calibration verification under each scanning rotation angle until a full-angle correction requirement is met. By adopting the method, high precision and stability of imaging correction can be kept in a full-angle range.

A system including: a multiscopic display; camera(s) that is / are positioned such that a display area of the multiscopic display lies at least partially within a field of view of the camera(s); and processor(s) configured to: display an image via the multiscopic display, whilst capturing image(s) of the display area using the camera(s); rectify the captured image(s) of the display area, to represent the display area only; generate a frequency-domain representation of the displayed image and a frequency-domain representation of the captured image(s) of the display area after rectifying; and determine a correct value of parameter(s) of a multiscopic optical element of the multiscopic display, based on the frequency-domain representation of the displayed image and the frequency-domain representation of the captured image(s) of the display area.

The invention discloses a civil engineering foundation pit deformation monitoring system based on image analysis, and relates to the technical field of foundation pit deformation monitoring, and the system collects and enhances a multi-modal image, eliminates interference, corrects image distortion, tracks key points, constructs a panoramic image, generates a point cloud, extracts deformation features, and constructs a deformation map. The method comprises the following steps: identifying a construction disturbance event, constructing a deformation chain and performing causal reasoning, identifying a high-risk area, predicting a deformation trend and outputting an early warning, displaying a monitoring result, collecting user feedback and synchronizing system data. A deformation track is accurately extracted through image correction and key point tracking, a three-dimensional deformation map is constructed in combination with point cloud modeling, causal analysis is carried out based on construction disturbance and a deformation path, a high-risk area is predicted and identified by using a time sequence trend, an early warning is given out, user feedback collection and result synchronization are supported, and monitoring intelligence and response efficiency are improved.

The invention relates to the technical field of image correction of computed tomography equipment, in particular to a cone beam computed tomography equipment correction method, and aims to solve the problems that geometric parameters projected by existing CBCT equipment are inaccurate, and a three-dimensional image subjected to three-dimensional reconstruction has obvious artifacts. In order to achieve the purpose, the method comprises the following steps: acquiring position coordinates of a marking substance in a phantom coordinate system and projection coordinates of the marking substance on a flat panel detector coordinate system when a conical ray originates from a plurality of positions; based on the position coordinates and the projection coordinates, calculating a projection matrix when the conical ray originates from a plurality of positions; calculating an average internal reference matrix of the plurality of internal reference matrixes based on the plurality of projection matrixes; determining a calculation projection coordinate corresponding to the position based on the average internal reference matrix and the position coordinate; and a nonlinear least square optimization algorithm is adopted to minimize the distance error between the projection coordinates and the calculated projection coordinates, and a calibration projection matrix is obtained. Random errors are suppressed by calculating an arithmetic average of the plurality of internal reference matrices.

The invention provides a distortion image correction method based on optical imaging, and the method comprises the steps: training a distortion correction model through the historical data of a source imaging device, and obtaining the optical features and similarity weights of all regions of a source device and a target device; training an image correction model based on the historical data of the source device, and generating correction images and spatial transformation features of each region of the target device through the model in combination with the historical data and the similarity weight of the target device; and then calculating image quality evaluation indexes of each region according to the correction result, the transformation features and the optical features, and summarizing to obtain the distortion correction completion degree of the target equipment. According to the method, the distortion correction precision of the target equipment can be improved by effectively utilizing abundant data of the source equipment, and quantitative verification of the correction effect is realized.

The invention discloses a stay cable vibration recognition method and system based on unmanned aerial vehicle video and image correction. The method comprises the steps that S100, a dynamic response video of a stay cable under natural excitation is shot through an unmanned aerial vehicle; s200, performing semantic segmentation on a stay cable region in the video image to obtain a clear stay cable pixel region; s300, pixel displacement tracking is carried out on the inhaul cable target area and the peripheral fixed reference point area, and time sequence displacement data containing inhaul cable vibration response and unmanned aerial vehicle disturbance components are extracted; s400, geometric correction of the image sequence is realized, and overall jitter caused by attitude change of the unmanned aerial vehicle is eliminated; s500, extracting an intrinsic mode component; identifying a common-mode disturbance component by comparing modal features of the two, and stripping from the inhaul cable response; and S600, carrying out identification and early warning on abnormal vibration behaviors such as loosening, strand breaking and vibration excitation abnormity. And high-precision and automatic monitoring of the vibration state of the stay cable is realized through a technical closed loop from video acquisition to abnormal early warning.

The invention discloses an intelligent pipe surface defect detection method based on image recognition. The method comprises the following steps: firstly, acquiring three-dimensional point cloud data of a to-be-detected pipe fitting, and calculating Gaussian curvature and normal vector of each point; dividing the surface of the pipe fitting into a plurality of geometric homogeneous areas based on Gaussian curvature difference; a multi-view original image is collected in each area, a local equidistant mapping function is constructed by using a normal vector and Gaussian curvature, and the image is corrected and expanded into a two-dimensional plane image; illumination normalization and learnable zooming and translation operations are carried out on the expanded image; inputting the processed image into a defect detection neural network comprising a geometric attention module and a context sensing decoder, and outputting a pixel-level defect probability graph; and finally, projecting a defect result back to a three-dimensional space through an inverse mapping function. The method effectively overcomes imaging distortion and identification difficulty caused by complex geometrical shapes of the pipe fittings, and is suitable for high-precision surface defect detection of various special-shaped pipe fittings.

The invention provides a soft bag detection process monitoring method and system and a soft bag light inspection conveying device, and relates to the technical field of soft bag detection monitoring, and the soft bag detection process monitoring method comprises the steps of image feature optimization, camera triggering time sequence optimization, image correction and soft bag sorting. According to the method, the soft bag images at different visual angles are obtained, whether image feature optimization is executed to correct the difference between the cameras is judged based on the image features of the soft bag images, then the soft bag is delivered to the particle foreign matter detection station, the sudden stop operation is executed, and the sudden stop soft bag image is obtained through the industrial camera after the soft bag is suddenly stopped. Whether camera triggering time sequence optimization is executed or not is judged based on time sequence parameters shot by an industrial camera, then whether geometric distortion correction is executed or not is judged based on consistency parameters of soft bag postures in sudden stop soft bag images, and finally whether emergency response is triggered or not is judged according to soft bag detection time consumption of all soft bags, so that the monitoring reliability of the soft bag detection process is improved, and the detection efficiency is improved. The problem of low monitoring reliability in the soft bag detection process in the prior art is solved.