103 results about "Silhouette edge" patented technology

The invention relates to the technical field of typesetting and drawing, in particular to an automatic typesetting and drawing method and system for lens laser cutting. The method comprises the following steps: collecting an image of a to-be-cut lens, carrying out contour edge detection, generating a contour direction sequence, carrying out polygon fitting on the sequence, screening out effective contour data, constructing a three-dimensional lens model, carrying out classification clustering on the model to extract contour features, carrying out segmentation to obtain a segmented contour model, and carrying out image segmentation on the segmented contour model. The method comprises the steps of obtaining a model, carrying out range limitation and local typesetting processing on the model to form an accurate layout scheme, constructing a cutting path direction chain according to the scheme, carrying out path planning, generating cutting path data, carrying out sequential execution setting on cutting paths based on a preset blank drawing template, and finally obtaining a typesetting template for laser cutting. The automatic typesetting and drawing method for lens laser cutting is more efficient.

The invention relates to the technical field of computer vision, in particular to a jujube tree disease and insect pest recognition method and system combined with a vision technology, and the method comprises the steps: extracting closed edge contours of a disease leaf gray image, and obtaining a contour skeleton of each closed edge contour; based on the graphic features of each contour skeleton, obtaining a to-be-recognized region, analyzing the distribution of gradient amplitudes of the skeleton contour edge and inner pixel points and the gradient amplitude distribution of the skeleton contour outer pixel points, and obtaining an edge transition value of each to-be-recognized region; based on the overall distribution characteristics of the gradient amplitudes of all edge pixel points of the skeleton contour of each to-be-identified area, obtaining an edge energy value of each to-be-identified area; and determining an edge distinguishing value of each to-be-identified area, and obtaining a disease identification result of each to-be-identified area. The invention aims to improve the identification capability of the jujube brown spot and the jujube gray leaf spot and improve the identification precision of disease detection.

The invention relates to the technical field of image processing, in particular to a defect detection method, system and equipment for an optical protective film, and the method comprises the steps: obtaining each frame of film grayscale image, and extracting a film region in each frame of film grayscale image and all contour edges of the film region; determining a matched contour edge corresponding to each contour edge in the film area of each frame of film grayscale image; determining the relative displacement difference and the motion anomaly of each contour edge in the film area of each frame of film grayscale image; determining the distortion possibility degree of each contour edge in the film area of each frame of film grayscale image; and determining the film bending degree and the defect judgment coefficient of each contour edge in the film area of each frame of film grayscale image, and performing defect detection on the surface of the optical protective film. According to the invention, defects and light spots on the surface of the optical protective film can be distinguished more accurately, and the detection precision of the defects on the surface of the optical protective film is improved.

The invention discloses a colloid visual identification method for gluing a rear cover of an air suspension controller, and the method comprises the following specific colloid visual identification steps: collecting an image of colloid on the rear cover of the controller after the colloid is projected by a light source, carrying out the graying processing of the obtained image through an image processing module, carrying out the image segmentation of the image after the graying processing, and carrying out the visual identification of the colloid on the rear cover of the controller. Obtaining a first image; performing sharpening enhancement processing on the image through histogram equalization to obtain a second image, inputting the first image into an image neural network for image denoising enhancement, and marking the output image as a third image; and fusing the second image and the third image through a multi-scale feature fusion technology, carrying out Laplace edge detection on the fused image, and outputting state information of the width and height of the colloid according to the contour edge of the rear cover frame and the contour edge of the colloid shadow. According to the invention, the quality of the colloid image can be effectively improved, parameters of the colloid can be better identified and detected, and the accuracy of identification and detection is improved.

The invention discloses a high-definition image optimization method for an intelligent networked automobile, relates to the technical field of image optimization, and solves the problems of numerical deviation and blurring of the contour edge of an object caused by gray value overlapping of corresponding areas of front and rear frames of images due to object movement. According to the method, the pixel value difference between adjacent frames is fully utilized, gradient calculation and contour area judgment are combined, the dynamic object can be accurately positioned, compared with a traditional method, misjudgment and missed judgment in a complex environment can be reduced, a reliable basis is provided for subsequent optimization, and the sensing accuracy of an intelligent networked automobile to the surrounding dynamic environment is improved; abnormal contour lines can be quickly identified by comparing gradient features of contour areas of similar dynamic objects in continuous frames; according to the method based on gradient column difference analysis, gray scale deviation generated by pixel point overlapping can be sensitively captured, contour blurring caused by foldover is avoided, and the definition and integrity of the edge of a dynamic object in the image are remarkably improved.

The invention relates to the field of image processing, in particular to a workpiece feature recognition method for a CNC machining center, and the method comprises the steps: obtaining the LAB value of each pixel point of a workpiece image to be machined and the gradient value of each pixel point, and traversing the pixel point distribution complexity of each neighborhood window of the pixel points, selecting a neighborhood window corresponding to the maximum pixel point distribution complexity as a target window of the pixel points, and taking the product of the calculated color difference and gradient difference between the pixel points and other pixel points in the target window and the window distribution complexity of the target window as the local saliency of the pixel points; screening all edge points based on the local saliency value to obtain a contour edge image; and performing matching in a workpiece database based on the contour image so as to match a corresponding processing procedure. According to the method, the edge points are screened in combination with the local significance of the pixel points and the optimal significance threshold, the contour edge of the workpiece can be extracted more accurately, and then the accuracy of workpiece feature recognition is improved.

The invention relates to a horse body ruler portable measurement method and system based on a single-view RGB-D camera, and belongs to the technical field of computers, and the method comprises the steps: obtaining a target region based on a segmentation model, and judging whether the target region is in a view range or not; performing attitude confirmation on the target area based on key points output by the key point model; under the condition that the attitude of the target area is confirmed, calculating an initial measurement point according to the confirmed target key points of the preset number of frames; according to the initial measurement point and the contour edge, different attribute measurement points are determined, and the different attribute measurement points comprise a body height measurement point and a body length measurement point; determining a body height value according to the ground point and the body height measuring point of the relative target area, and determining a body length value according to the distance between the chest measuring point and the hip end measuring point on the two-dimensional plane; and estimating the chest circumference length and the tube circumference length according to the chest circumference measuring points and the tube circumference measuring points. According to the method, the automatic measurement of the horse size is realized, the measurement efficiency is improved, and the accuracy is ensured.

The invention provides a method and a system for measuring the size of an object under a complex background based on deep learning. The method comprises the following steps: training to obtain an RCF model under a current scene; shooting a to-be-detected object by using a camera, and collecting an image of the to-be-detected object; sending the to-be-detected object image to the RCF model in the current scene to obtain a coarse contour edge image of the to-be-detected object; processing the coarse contour edge image to obtain sub-pixel coordinate points of the to-be-detected object, and obtaining the image length and the image width of the to-be-detected object according to the sub-pixel coordinate points; and calculating the physical length and width of the to-be-detected object according to the calibration parameters of the camera, the image length and the image width. According to the method, the self-learning ability of the deep neural network is utilized, interference of a complex background is removed, only the edge image of the object to be detected is extracted, the object has sub-pixel precision through binary polynomial interpolation, the measurement precision is improved, and the adaptability of the measurement method is enhanced.

The application relates to the technical field of computer vision, in particular to a jujube tree disease and pest identification method and system combined with visual technology, which comprises the following steps: extracting a closed edge contour of a disease leaf gray image, and obtaining a contour skeleton of each closed edge contour; based on the graphic features of each contour skeleton, obtaining a to-be-identified region, analyzing the gradient amplitude distribution of the skeleton contour edge and the inside pixel points, the gradient amplitude distribution of the skeleton contour outside pixel points, and obtaining an edge transition value of each to-be-identified region; based on the overall distribution features of the gradient amplitude of all edge pixel points of the skeleton contour of each to-be-identified region, obtaining an edge energy value of each to-be-identified region; determining an edge distinguishing value of each to-be-identified region, and obtaining a disease identification result of each to-be-identified region. The application aims to improve the identification ability of jujube brown spot disease and jujube gray spot disease and improve the identification precision of disease detection.

The invention provides a laser radar and camera external parameter calibration method, which is applied to a laser radar and camera combined rigid system, and comprises the following steps: a camera image of a target scene is acquired, the target scene is provided with a calibrator, and the camera image comprises a calibrator image of the calibrator; obtaining the edge contour of the camera image, wherein the edge contour comprises the contour of the calibrator image; constructing a calibrator edge line based on the contour of the calibrator; based on the edge contour and the calibrator edge line, obtaining a Planck coordinate of a fitting edge of the calibrator; obtaining three-dimensional point cloud data of the target scene; based on the three-dimensional point cloud data, extracting an edge point set of the calibrator; and obtaining external parameters of the laser radar and the camera based on the fitting edge and the edge point set. The laser radar and camera external parameter calibration method provided by the invention is beneficial to simplifying an external parameter calibration program and enhancing practicability. The invention further provides a laser radar and camera external parameter calibration system and a storage medium.

The invention provides a multi-module fusion self-supervision denoising method for an industrial CT image, and belongs to the field of digital image processing and industrial nondestructive testing. In order to solve the problems of detail loss, poor denoising effect and the like in an industrial CT image denoising process, the invention provides a self-supervised denoising model which takes U-net as a trunk network and fuses an edge enhancement module, a convolution block attention module and a convolution attention fusion module. Meanwhile, by means of a mixed loss function composed of structural similarity index loss, mean square error loss and ResNet perception loss, the noise level of the industrial CT image can be remarkably reduced, meanwhile, the integrity of the object structure and the contour edge is accurately kept, and the image quality is effectively improved.

The invention relates to a method for determining the layer thickness of an at least partially transparent surface coating on the surface of a substrate, using a measuring device for emitting focused radiation. The invention is characterized in that the distance between the surface of the substrate and the measuring device for emitting focused radiation is changed by a relative movement between the measuring device and the substrate, wherein the image of the radiation is captured in different relative positions, wherein in the case of an image with a clear outline edge, focusing is inferred at the interface between the surface coating and the substrate and the current relative position is stored as the first position, and wherein in the case of an image surrounded by scattering effects, focusing is inferred at the interface between the surface coating and the air and the current relative position is stored as the second position, after which the difference between the first and second positions is used to determine theTaking into account the refractive index of the transparent surface coating, the layer thickness of the at least partially transparent surface coating is determined.

The invention provides a circular curve burr detection method and system based on multi-scale phase consistency and a contour model, and a medium, and the method comprises the steps: constructing a multi-scale Log-Gabor filter group, and carrying out the convolution calculation of an input image, and obtaining a response vector; calculating phase consistency measurement of each pixel point, extracting an initial edge point set, fitting an initial circular contour, performing energy minimization evolution on the initial circular contour based on an active contour model, and fitting a real contour of burrs; constructing a two-dimensional feature vector; carrying out fusion discrimination on the two-dimensional feature vectors based on a linear weighted score function, and outputting a burr detection result; according to the method, phase consistency is used as a core image feature, so that the detection algorithm has natural invariance for uneven illumination, brightness change and contrast difference, and under the scene of failure caused by poor image quality in the prior art, the contour edge can still be stably and completely extracted, and the omission ratio is remarkably reduced.

The invention belongs to the technical field of condition judgment, and particularly relates to an electrode foil corrosion condition judgment method and system, and the method comprises the following steps: S1, obtaining a grayscale image of a to-be-analyzed electrode foil; constructing a structure tensor based on the composite gradient reflecting the local brightness and texture information of the pixel points, determining an anisotropic diffusion coefficient according to the structure tensor, and performing iterative anisotropic diffusion filtering processing on the grayscale image by using the anisotropic diffusion coefficient to obtain a filtered image; and S2, fusing the anisotropic diffusion coefficient determined for each pixel point with the pixel intensity of the filtered image to obtain a high-contrast corrosion significance map. According to the method, the contour edge information of the corrosion area can be kept while the complex texture and noise of the background area of the electrode foil image are smoothed, the contradiction between denoising and edge protection of a traditional filtering method is solved, the result of the corrosion condition is more reliable, and the automation level of electrode foil product quality detection is improved.

The invention relates to the technical field of image recognition, in particular to a wild animal behavior recognition method and system based on visual recognition. The method comprises the following steps: acquiring continuous image frames, analyzing gray direction changes of pixel blocks in adjacent frames, extracting low-photosensitive texture features and forming a texture change track; then fusing the texture trajectory with the geometric offset trend of the animal contour edge points to generate a behavior dynamic fusion feature capable of representing the action direction and intensity; and carrying out continuity calculation on the fusion features, and judging whether the action meets a behavior triggering condition based on a change continuity index, thereby realizing automatic identification and output of the wild animal target behavior. According to the method, the texture change track and the contour geometric deformation direction are fused, stable recognition of continuous actions under the low-light condition is achieved, and the wild animal behavior recognition result is more accurate and reliable and has the quantitative judgment capacity.

The invention provides an image processing method, electronic equipment and a storage medium. The image processing method comprises the steps of performing feature extraction on a standard workpiece image, and constructing a data model based on extracted feature data; matching the data model with a to-be-detected image by adopting an edge feature matching algorithm, and determining positioning information of a workpiece in the to-be-detected image; performing contour extraction on the to-be-detected image according to the positioning information of the workpiece to obtain a mask pattern of the workpiece; and optimizing the mask pattern by using the edge information of the to-be-detected image to obtain a contour edge point set of the workpiece. The method can improve the extraction accuracy of the contour boundary of the workpiece.

The invention discloses a target detection method, which comprises the following steps of: reading a current frame image of a camera and a sequence frame image before the current frame, and respectively processing the sequence frame image and the current frame image to generate an environment background image and a new image; calculating the difference between the new image and the environment background image by adopting a background difference method, binarizing a difference result according to an adaptive difference threshold value to obtain a binary image, and obtaining a contour edge in the binary image by adopting an edge detection algorithm; calculating a minimum rectangular bounding box of the contour edge as a dynamic region bounding box, and obtaining an image containing a dynamic region by adopting a rectangular extraction strategy; inputting the image containing the dynamic region into a target detection model to obtain a target detection result, wherein the target detection model is obtained by training a data set considering meteorological factors in advance; according to the method, the problem that target recognition cannot be carried out due to the fact that the proportion of the target area is too low is avoided, and the model generalization ability is improved by adopting the data set training model considering meteorological factors.

The invention relates to a method, a ray tracing system and a storage medium for performing an intersection test. A ray tracing system and a computer-implemented method for performing an intersection test for a ray bundle with respect to a box are described. Contour edges of the box are identified from a perspective of the ray bundle. For each of the identified contour edges, a component of a vector providing a boundary to the ray bundle is obtained and it is determined whether the vector passes inside or outside the contour edge. A result of the intersection test for the ray bundle with respect to the box is determined using the result of the determination whether the vector passes inside or outside the contour edge for each of the identified contour edges.

This invention relates to the field of image data processing technology and discloses a feature fusion method for multi-target detection and classification of text images. The method includes: acquiring a low-level spatial detail feature tensor and a high-level global semantic feature tensor and aligning the spatial axis resolution to form a feature tensor pair; calculating the channel dot product of the feature tensor pair to obtain a spatial feature modulation matrix; using the element values ​​of the spatial feature modulation matrix as feature flow control weights, combining the feature tensor pairs through a nonlinear gated mapping, and outputting the final multi-target detection and classification fusion feature. This invention establishes a gated constraint boundary through the spatial feature modulation matrix, suppresses high-level feature blurring, strengthens pixel contour edges, eliminates text target feature diffusion and crosstalk, reshapes semantic causal correspondence, and enhances detection recall and classification accuracy.

The present invention discloses a laser radar target edge feature point detection method based on waveform decomposition. The edge detection of the target in the scene is realized through the laser radar echo waveform decomposition method, and the 3D feature point position of the target edge is detected based on the edge judgment result. The present invention innovatively introduces laser echo waveform analysis to determine the target contour edge, realizes rapid detection of the target edge, and thus detects the 3D feature point position of the target. The method does not require full-target three-dimensional imaging detection, but only requires target edge waveform perception. The process is simple, the data volume is small, the calculation speed is fast, and the target edge detection accuracy is high. By introducing threshold adaptive judgment, the adaptive detection of the feature point position of the laser radar calibration plate is realized, the edge detection accuracy is improved, and the detection process of edge feature points is simplified.

The application discloses a cross-modal pedestrian re-identification method and system based on contour map guidance, and belongs to the technical field of pedestrian re-identification. The application adopts an additive and multiplicative collaborative operation mode to weight image features on corresponding contour features. The additive and multiplicative collaborative operation can combine the respective advantages of image features and contour features and save the original image information. A feature energy function is defined to weight the fused features. The feature energy function calculates the dispersion degree of each neuron of the fused features through a mean square deviation. After being weighted by the energy function, the features at the contour edge position will generate greater weights. The features at the contour edge position obtain more attention in subsequent calculations, and the obtained contour guided features and original image features are used for loss calculation together, so that the recognition accuracy of the cross-modal pedestrian re-identification is improved. The problem that the difference between multi-modal images reduces the pedestrian re-identification accuracy in the prior art is solved.

The application relates to a method, system, device and storage medium for image background light illumination homogenization, which comprises the following steps: step S1: inputting an original image, obtaining a rectangular mask, and performing rank filtering on the original image using the rectangular mask to obtain a first image; step S2: performing mean filtering on the first image to eliminate high-frequency gray values in the first image and obtain a second image; and step S3: subtracting the second image from the original image to obtain a final processing image, and scaling the gray value of the final processing image to a preset value. In the application, the original image is filtered using the rectangular mask, which can reduce the difference of the gray values at different positions of the original image; the second image is subtracted from the original image to obtain a third image, which is used to obtain the gray difference value of each pixel point of the second image and the original image, highlight the contour edges in the image, and greatly reduce the problem of uneven image brightness.

The present application provides a laser radar and camera extrinsic calibration method, which is applied to a rigid system of a laser radar and camera combination, and includes: obtaining a camera image of a target scene, the target scene is provided with a calibrator, and the camera image includes a calibrator image of the calibrator; obtaining the edge contour of the camera image, the edge contour includes the contour of the calibrator image; constructing the calibrator edge line based on the contour of the calibrator; obtaining the Planck coordinates of the fitted edge of the calibrator based on the edge contour and the calibrator edge line; obtaining three-dimensional point cloud data of the target scene; extracting the edge point set of the calibrator based on the three-dimensional point cloud data; obtaining the extrinsic parameters of the laser radar and camera based on the fitted edge and the edge point set. The laser radar and camera extrinsic calibration method provided by the present application is conducive to simplifying the extrinsic calibration procedure and enhancing practicality. The present application also provides a laser radar and camera extrinsic calibration system and storage medium.

A multi-point milling path design method for the full surface feature structure of thin-walled spherical shell micro-components is disclosed, relating to the field of micro-component processing technology. This invention can meet the requirements of high-precision, stable, and controllable removal of full surface feature structures of thin-walled spherical shell micro-components under micro-scale constraints. Key technical points: The method includes spatially uniform distribution of feature structures, micro-structure processing sequence planning, and multi-point milling path generation. The thin-walled spherical shell micro-component to be processed is observed offline using an optical microscope to obtain the coordinates of feature points on the contour edge of the micro-component, and the diameter of the micro-component is obtained by fitting. Based on the secondary development platform of UG software, a micro-structure processing sequence planning method driven by the shortest processing path is established to complete the above-mentioned full surface micro-structure processing sequence planning of the micro-component. Based on the point set coordinates optimized by the spherical point set uniform distribution iterative algorithm and the processing sequence driven by the "shortest processing path," combined with the configuration characteristics of ultra-precision shape control machining equipment and the micro-structure processing requirements, the coordinates of several uniformly distributed point sets with the center of the micro-component as the origin are transformed into a machining coordinate system with the tool setting point as the origin, obtaining the original coordinates of the micro-structure point sets and the corresponding feature angles. This invention is used for full-surface milling of thin-walled spherical shell micro-components.

The embodiment of the application provides a wire harness cutting path correction method, which comprises the following steps: S1: collecting a wire harness cutting area image; S2: inputting the image into an improved SegFormer-B2 deep learning model, and the improvement measures comprise: extending three independent convolution output heads in parallel at the end of the original decoder, which are used for synchronous inference and output three target feature maps: a mark point heat map, a wire harness contour edge map and a pixel direction field; S3: fusing the mark point heat map and the wire harness contour edge map, and using the geometric prior provided by the pixel direction field to reconstruct the actual position path of the wire harness; S4: comparing the actual position path with a preset cutting path, calculating the path deviation, and generating corresponding motion control instructions to drive the cutting equipment to correct and cut. The method effectively overcomes the recognition instability problem caused by a complex working scene, significantly improves the precision of wire harness cutting, and improves the production quality and efficiency.

The application provides an industrial CT image coupling artifact multi-scale generative adversarial network correction method, and belongs to the field of digital image processing. The method aims at the problem that image degradation caused by the coupling of various artifacts in industrial CT imaging is difficult to correct effectively, and provides a multi-scale generative adversarial network correction method. The feature pyramid structure is used to help the network capture more comprehensive artifact feature information, and a multi-scale discriminator is used to build a generative adversarial network framework, so that the generated artifact-removed image is clearer and more real. The method has strong coupling artifact correction capability, and can better restore the detail information of the image, thereby ensuring the accuracy of the geometric size and contour edge of the part in the CT image.

This invention discloses a two-branch oocyte target detection method and system for motion blur. Addressing the motion blur problem caused by equipment vibration and oocyte micro-movement during oocyte retrieval, this invention constructs an end-to-end detection model comprising a clear branch and a blurred branch. Each branch includes a lightweight feature recovery network, a multi-band feature fusion module, a frequency structure enhancement module, and an oocyte target detection network. During the training phase, feature domain deblurring loss, feature consistency self-supervised loss, and oocyte detection loss are used for joint training of the two branches, with the clear branch's features supervising the learning of robust feature representations in the blurred branch. During the inference phase, only the blurred branch outputs the detection results. The multi-band feature fusion module repairs degraded features such as contours, edges, and textures through frequency domain decomposition and targeted enhancement, while the frequency structure enhancement module compensates for missing structural information through a cross-attention mechanism. This invention effectively solves the problem of decreased detection accuracy caused by motion blur.

The invention relates to the technical field of robots, in particular to an intelligent rescue robot which comprises a data acquisition module used for acquiring a field image of a task environment to be executed; the data analysis module is used for determining the qualification of the contour edge of the field image based on the edge contour continuity evaluation value of the field image, and determining whether the feature is a human body feature according to the human body contour evaluation value of the feature of the field image; the data adjustment module is used for determining an adjustment mode according to the ratio of the edge contour continuity evaluation value to a preset edge contour continuity evaluation value on the basis of the condition that the contour edge of the field image is unqualified; the data processing module is used for determining whether a fuzzy region exists in the field image or not based on the color characterization parameter of the field image and determining whether the fuzzy region is a shielding region or not according to the ambiguity evaluation value; and the data optimization module determines an optimization mode according to a difference value between the ambiguity evaluation value and a preset ambiguity evaluation value. According to the invention, the identification efficiency of the intelligent rescue robot is improved.

The invention discloses a transparent element assembling method and system, and the method comprises the steps: firstly building an ideal three-dimensional model of a transparent element and a to-be-assembled part, then measuring the posture of the transparent element (the posture comprises a pitch angle, a roll angle and a depth distance), and adjusting the posture of the transparent element to a to-be-assembled posture based on the ideal three-dimensional model, it is ensured that the initial posture of the element in the three-dimensional space meets the assembly requirement; secondly, the two-dimensional relative poses (including the rotation angle and the translation amount) of the transparent element and the contour edge of the part to be assembled are detected, and the position deviation of the transparent element and the contour edge of the part to be assembled in the plane is accurately obtained; finally, according to the detected two-dimensional relative poses, the relative positions of the transparent element and the to-be-assembled part are adjusted so that the transparent element and the to-be-assembled part can reach the proper to-be-assembled position, according to the method, through staged posture adjustment and posture detection, the positioning problem in the transparent element assembling process is effectively solved, and the method can be widely applied to automatic assembling scenes of various transparent elements.

The present application relates to the technical field of mask lithography, and discloses a hot spot layout correction method, device and equipment in a mask layout and a storage medium, the hot spot layout correction method comprises determining a hot spot layout in an initial mask layout; determining the environmental graphic parameters of each contour edge segment in the hot spot layout; constructing a basic optimization objective function according to the edge placement error varying with the displacement amount of the contour edge segment; superimposing an optimization compensation term on the basic optimization objective function according to each environmental graphic parameter to obtain an optimization objective function; optimizing the displacement amount of each contour edge segment according to the optimization objective function to obtain the corrected displacement amount corresponding to each contour edge segment, so as to determine the corrected layout of the hot spot layout. The present application realizes the automatic correction of the hot spot layout without manual operation, improves the correction efficiency of the hot spot layout, and ensures the accuracy and reliability of the correction result.