142 results about "Object contour" patented technology

The invention provides an AI-based factory abnormal behavior identification monitoring method and system. The method comprises the steps of obtaining continuous video stream data collected by a plurality of monitoring devices in a target factory in real time; performing time dimension segmentation and space dimension preprocessing on the continuous video stream data to obtain a multi-frame behavior feature set corresponding to the dynamic behavior sequence, and inputting the multi-frame behavior feature set into a pre-trained abnormal behavior recognition model, performing cross-modal association fusion on the skeleton key point features, the environment associated object contour features and the action trajectory features to generate a target behavior feature vector, performing similarity matching with a preset abnormal behavior feature library, determining an abnormal behavior category, and triggering a corresponding early warning signal according to the behavior type when it is determined that the target behavior feature vector belongs to the abnormal behavior category. And real-time alarm display is carried out. According to the invention, the problem that the existing monitoring system is insufficient in accuracy and real-time performance can be solved.

The invention discloses a light rail vehicle safety monitoring system based on multi-sensor fusion, and relates to the technical field of light rail vehicle safety monitoring, the system is composed of a plurality of functional modules, and the system comprises a multi-mode sensing module for capturing images of the surrounding environment of a light rail in real time and delimiting a visual mark area; performing three-dimensional scanning on the visual marking area, generating environment point cloud data in real time, and acquiring a dynamic object distance by using a Doppler compensation algorithm; the data processing module is used for calculating object deformation between adjacent frames by adopting an optical flow method and a CNN deformation detection network; carrying out object relative motion calculation according to object shape analysis; carrying out instance segmentation based on MaskR-CNN, and outputting an object contour; meanwhile, a dynamic target clustering algorithm is used for conducting multi-target tracking on the object contour, and three-dimensional coordinates and motion vectors of the object are output in combination with environment point cloud data; and the fusion and analysis module performs time synchronization based on the PTP protocol.

The invention is suitable for the technical field of mechanical arm control, and provides a self-adaptive mechanical arm clamping control method and system based on image recognition, the system comprises a mechanical arm, and the mechanical arm comprises a mechanical arm body and a mechanical clamping jaw; the image acquisition module is used for acquiring a 2D image and depth information of an object; the image processing module is used for carrying out preprocessing, contour extraction and pose coordinate determination on the acquired image; the data transmission module is used for transmitting the contour, size and pose coordinate information of the object processed by the image processing module to the controller; the controller is used for receiving information sent by the data transmission module and controlling movement and clamping force of the mechanical arm; according to the mechanical arm, the path can be automatically planned and proper force can be applied according to the characteristics of different objects, the flexibility, the accuracy and the safety of the clamping operation of the mechanical arm are improved, and the mechanical arm is suitable for various automatic scenes.

The invention discloses an image recognition security and protection processing method and system for a security and protection screen, relates to the technical field of security and protection screens, solves the problem that original object contour feature processing is rough, and can effectively lock a dynamic object by analyzing pixel point differences between adjacent frame images acquired by a high-definition camera. When the dynamic features are confirmed, difference regions and contour regions are accurately divided through point location analysis and gradient calculation of the system, and the feature region with the largest area is selected from a plurality of feature regions to serve as the dynamic features, so that the recognition accuracy is ensured, and the misjudgment rate is reduced; in each contour partition, feature difference values of different internal connecting lines are compared, and the connecting line with the minimum difference value is selected as an average refining connecting line, so that refining processing of the dynamic feature edge contour is realized, the extracted feature contour is more accurate, and the real form of an object can be better reflected.

The invention discloses a grape fruit grain segmentation method based on multi-source feature fusion, relates to the technical field of image processing, and establishes an instance segmentation network MFFNet based on multi-source feature fusion, and the MFFNet cooperatively utilizes strong semantic features extracted by Swin Transform and explicit boundary information generated by parallel edge branches. According to the method, fruit grain segmentation is carried out on multi-scale information of ASFF adaptive fusion, targeted feature fusion is carried out in different stages by adopting a staged feature fusion strategy, and the precision and robustness of grape fruit grain segmentation in a complex scene are improved. Through the MFFNet, the segmentation problems of dense grape fruits, shielding, fuzzy boundary and the like in a complex field environment are effectively solved. The staged feature fusion strategy improves the perception and description ability of the model to the object contour, and well solves the problems of instance adhesion and unclear boundary.

The invention belongs to the technical field of smart cities, and relates to a smart city public service opinion feedback multi-dimensional evaluation system, which comprises a data acquisition and processing module for analyzing object contour features in public service original data to generate a structured data pool; the space-time atlas modeling module is used for generating a demand migration feature vector with a weight coefficient; the dynamic scheduling engine module is used for calculating a municipal resource real-time scheduling path according to the demand migration feature vector, and generating an emergency execution scheme when a resource gap value exceeds a preset threshold value; the task execution monitoring module is used for splitting an emergency execution scheme into interactive task nodes, acquiring running state data of an execution terminal in real time, and triggering an automatic replacement path supplement mechanism when detecting that the overheating duration of equipment exceeds a safety critical value; and the feedback optimization evaluation module is used for generating an urban governance health index report containing a compensation scheme and a priority adjustment factor. According to the invention, the service can be recovered only through manual intervention when abnormal events are solved.

The invention relates to the technical field of size detection, and provides a size detection method, system and device and a medium, and the method comprises the steps: obtaining a shadow image of a to-be-detected workpiece, carrying out the gray conversion or binarization processing, and smoothing the image through a Gaussian kernel filter, so as to reduce the noise and enhance the edge information; performing edge detection by using an improved Canny operator to extract object contour pixel points, calculating by using an interpolation algorithm to obtain a higher-precision object contour image, gradually traversing the contour image according to a preset track and a reference template by means of a sliding window algorithm, and calculating similarity to select a target local area; and finally, finding the highest point and the lowest point of the object contour image in the target local area, and integrating the data of each target local area according to the height difference, thereby accurately obtaining the overall size of the workpiece to be measured, and improving the accuracy and convenience of size measurement.

An AI auxiliary image annotation method based on semi-supervised learning comprises the following steps: acquiring a to-be-annotated image, extracting multilayer features of the image by using a deep learning network, and generating a feature map; the global context information and the local feature information are fused, and enhanced feature representation is generated and used for contour prediction; on the basis of enhanced feature representation, modeling a relationship between vertexes by adopting a graph convolutional network, and constructing a dynamic graph structure; according to the dynamic graph structure, calculating self-adaptive matching weights among the vertexes to realize a dynamic vertex pairing strategy; predicting an initial contour position by combining global and local features through a dynamic vertex pairing strategy; based on the predicted initial contour, a deformable convolutional network is adopted to realize progressive optimization of the contour; in the progressive optimization process, vertex position information is updated in real time, and the graph structure relation is adjusted; and finally, outputting an accurate object contour labeling result, and completing an image labeling task.

The invention relates to the field of image processing, in particular to a visual image processing method and device, equipment and a program product. The method comprises the following steps: acquiring a binocular image, and generating a first disparity map through a stereo matching network; obtaining a second disparity map through extreme value filtering; generating an edge mask through edge detection; performing edge filtering on the second disparity map according to the edge mask to obtain a third disparity map; and horizontal displacement is obtained according to a parallax value in the third parallax image, parameters are calibrated through a camera, the horizontal displacement is converted into depth information, and a three-dimensional point cloud is generated according to the positions of the pixel points and the depth information. According to the method, a real boundary is extracted through edge filtering, parallax of a boundary area is forcibly corrected by using an edge mask, a transition zone of network prediction is suppressed, burrs or outliers of point clouds after conversion can be effectively reduced, scattered point cloud noise is eliminated, the object contour boundary is clearer, and the perception precision of a visual system 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 discloses a multi-instance object image controllable generation method based on decoupling contour representation, and belongs to the field of image generation, and the method comprises the following steps: data collection and arrangement; data enhancement; contour representation processing; pre-training a contour encoder; calculating by a contour fusion module; the global information injection module fuses global and image features; contour position coding; object space perception calculation; constructing a model; and generating an image. According to the invention, through a coding mode based on contour boundary point coordinates, accurate representation and decoupling of contour information of different objects in the same coordinate space are realized, so that the model can clearly distinguish each object, and common problems of object confusion, dislocation and the like in a traditional method are avoided. When a scene image containing a plurality of objects in different shapes is generated, the contour of each object can be accurately presented, the position and the size of each object can be accurately controlled, and the generated image is rich in details and accords with expectation.

The invention discloses a visual prosthesis control method and system and a storage medium, and relates to the field of medical instruments, and the method comprises the steps: extracting at least one initial target image based on a target object selection result fed back by a user, carrying out the region-of-interest extraction of the initial target image, and obtaining a target object contour image, grouping each contour pixel point based on the brightness values of the contour pixel points in the target object contour map to obtain a plurality of data groups and a stimulation sequence thereof, mapping a stimulation current matrix of each data group based on the stimulation current value and the space coordinate of each contour pixel point, and obtaining a stimulation current matrix of each data group based on the stimulation sequence; and controlling the release of the stimulation current in the electrode array based on each stimulation current matrix in sequence. Redundant information is reduced based on user feedback and region-of-interest extraction, the stimulation sequence based on the average brightness value of each data set is configured, the stimulation current is controlled to be released in the electrode array based on each stimulation current matrix in sequence, and the perception quality of the brain to the target object is improved.

The invention discloses a virtual defect image generation method and device, equipment and a storage medium, and relates to the field of image processing. Obtaining a contour image and a semantic image of the material plate, marking a target contour object in the contour image, and obtaining use case data for the target contour object; updating a target contour object in the semantic image and / or the contour image based on the use case data to generate a virtual material number graph; setting a defect contour and generating a material label graph in the virtual material number graph, and obtaining virtual defect data used for synthesizing a virtual defect; and synthesizing a virtual defect graph based on the contour labels in the material label graph and the virtual defect data. According to the scheme, the contour map and the semantic map can be made based on the preferential material plate, the virtual material number graph is expanded and generated through object contour mapping, targeted training of all defect types is considered, even a defect graph of a specific neighborhood is generated in a targeted mode for high-incidence defects of a production line, the detection capability of the model in characteristic type deficiency is emphatically optimized, and the detection efficiency is improved. The production and detection efficiency of a production line is improved.

The invention discloses a method and equipment for extracting object contour vector data based on an orthoimage, and the method comprises the steps: S1, inputting the orthoimage into a Mask2Former model based on a VisionTransform architecture, and carrying out the segmentation of ground feature elements in the image; s2, based on a Douglas-Peucker algorithm, adaptively extracting geometric constraints including key endpoints, later angles and side lengths in an original boundary after surface feature element segmentation, rejecting abnormal sides and unreasonable angles, finally extracting a main direction from relatively long sides, and sequentially executing collinear line segment fusion and line segment intersection operation to ensure polygon closure and direction consistency; and S3, performing contour reconstruction operation of the ground feature elements based on the main direction and the key points. The method can reduce the dependence on manpower, achieves the efficient and automatic extraction of the object contour vector data based on the orthoimage of the unmanned aerial vehicle, and improves the working efficiency.

The invention discloses a contour extraction method based on edge detection, and belongs to the technical field of image processing, and the method comprises the following steps: obtaining a plurality of images to be subjected to contour extraction, and real labeled contours corresponding to the images to be subjected to contour extraction; performing noise level quantization and edge ambiguity quantization analysis on each image to be subjected to contour extraction, and calculating to obtain an image complexity level of each image to be subjected to contour extraction; performing complexity level classification on each image to be subjected to contour extraction, and constructing a complexity classification image data set; constructing a contour extraction model based on edge detection; a trained contour extraction model is obtained; and obtaining a new image to be subjected to contour extraction and an image complexity level, inputting the image to be subjected to contour extraction into the trained contour extraction model for contour prediction, and obtaining a predicted image contour of the image as a contour extraction result. The problem that it is difficult to accurately extract the object contour in a complex image scene is solved.

The invention discloses a neighbor soft package two-dimensional segmentation and three-dimensional positioning combined method, which comprises the following steps of: segmenting a two-dimensional picture by using a Segment Anything algorithm and carrying out preliminary screening according to the contour area of a segmented object; performing coordinate range screening, determining and calculating an external rectangle and an area thereof through processing of a two-dimensional minimum compact bounding box, and outputting a minimum area by rotating a line to be greater than 90 degrees; obtaining an included angle between an object center coordinate and x and y through a bounding box, calculating a result through a formula, and performing range screening to segment a recognition object; processing the soft package by using a two-dimensional minimum bounding box algorithm to obtain a center coordinate and an included angle; acquiring internal and external parameter matrixes, two-dimensional coordinates and depth of the camera, and converting two-dimensional image coordinates into three-dimensional coordinates in a camera coordinate system and a world coordinate system; and calculating a neighbor point cloud normal vector to obtain a three-dimensional attitude of the soft package center point. According to the positioning method disclosed by the invention, the problem of low positioning precision of the soft package is solved by a minimum compact bounding box algorithm, and a more accurate grabbing pose is conveniently provided for the robot.

Embodiments of the disclosure provide a line effect processing method and apparatus, an electronic device, a storage medium, a computer program product, and a computer program. The method includes: collecting a plurality of contour points formed by an object contour of a target object in an image to be processed; performing point expansion processing on the plurality of contour points to obtain a plurality of vertices; generating, based on the plurality of vertices, texture curves corresponding to adjacent pairs of vertices, and obtaining at least one texture curve; utilizing the at least one texture curve to generate an effect line frame; and mapping the effect line frame onto the image to be processed to obtain a target image corresponding to the image to be processed. Problems of missed gaps in line effect displaying are solved, and smooth display accuracy of a line effect is improved.

The invention discloses a semantic segmentation method and device based on a lightweight coding-decoding network, equipment and a medium. The method comprises the following steps: acquiring image data; preprocessing the acquired image data; inputting the preprocessed image into an encoder for low-resolution feature extraction; the extracted low-resolution features are input into a decoder for feature detail recovery, and a feature map with the same resolution as the original image is obtained; and inputting the feature map with the same resolution as the original image into a semantic segmentation classifier to obtain pixel-by-pixel classification prediction. Therefore, the method can achieve higher class precision and more detailed object contour prediction, thereby achieving better balance between speed and precision.

The invention provides an object contour minimum bounding box generation method and device, a storage medium and equipment, and is applied to the technical field of computers. The method comprises the following steps: screening out a maximum distance point and an angular point of a target point cloud cluster, and determining a pre-selected long edge of the target point cloud cluster; performing linear fitting on at least one long side point cloud in the target point cloud cluster, wherein the linear distance between the long side point cloud and the pre-selected long side point cloud is within a preset distance threshold value to obtain a course angle; and according to a rotation matrix constructed by the course angle, carrying out translation and rotation transformation on the target point cloud cluster, then calculating the contour size and the center point of the point cloud cluster, and generating an object contour minimum bounding box of the target point cloud cluster. The method determines the pre-selected long edge by screening the maximum distance point and the angular point, improves the course angle precision by combining the threshold filtering point cloud and straight line fitting, and performs coordinate transformation on the point cloud based on the rotation matrix to generate the minimum bounding box closely fitting the object contour, thereby improving the course detection and bounding box precision, and improving the course detection precision. And more reliable environment information is provided for intelligent driving perception.

The invention relates to the field of computer vision and image processing, in particular to a three-dimensional matching method based on wavelet transform and superpixel segmentation, which can improve the problems of poor depth estimation precision, blurred depth map contour and failure in multi-texture area depth estimation of the existing three-dimensional matching method. According to the method, different frequency domain sub-band information of the image is obtained by introducing a wavelet transform image processing method, the guiding effect of a low-frequency structure is enhanced, and the cost body aggregation process is participated; and meanwhile, fine modeling is carried out on the boundary of an image structure by combining a superpixel segmentation technology so as to enhance the integrity of an object contour in a depth image and the accuracy of parallax estimation. The method shows higher matching precision and boundary definition in a complex scene, has good calculation efficiency and deployment feasibility, and can provide reliable support for an embedded stereoscopic vision system with high precision and low resource consumption.

The invention relates to the technical field of image processing, in particular to a computer image processing system based on deep learning, which determines a first contour boundary of a target object through a preprocessing module according to a deep learning model, and draws brightness expression curves of pixel points of each sub-region in different direction dimensions through a feature extraction module. Marking a non-confidence sub-region in the aggregation analysis region through a feature aggregation module, obtaining brightness expression curves of pixel points subjected to exposure adjustment in different direction dimensions through a local processing module so as to determine contour boundary fitting points, and fitting contour boundary sections of the non-confidence sub-region, and updating the first contour boundary through the output end and outputting a second contour boundary of the target object. According to the method, local region optimization is carried out aiming at insufficient local features in a target object contour boundary extraction process under a complex illumination condition, and high-precision extraction and dynamic updating of a target object contour in a complex environment are realized.

The invention relates to the technical field of sewing automation, and discloses an automatic sewing identification method and system based on image identification. The method comprises the steps of obtaining surface image data of a target sewing object, and performing feature extraction on the surface image data to generate a geometric contour data set; according to sewing parameter configuration information input by a user, executing path generation operation to obtain an initial sewing path coordinate sequence; based on spatial distribution characteristics of the geometric contour data set, performing path density adaptive adjustment on the initial path, and generating a corrected sewing path coordinate sequence; performing smoothness optimization processing on the corrected path to obtain a final sewing path coordinate sequence; and converting the final path coordinate sequence into visual graphic data, and rendering and outputting the visual graphic data in real time. According to the method, the contour of the sewing object can be accurately obtained, different sewing requirements can be flexibly met, the path density and smoothness are optimized, the accuracy and efficiency of sewing path generation are improved, the defects of a traditional method are overcome, and the method is suitable for various sewing processing scenes.

The invention belongs to the technical field of rain and fog image processing, and particularly relates to a rain and fog image restoration method and system based on space-frequency collaborative awareness and electronic equipment, and the method comprises the steps: obtaining a rain and fog image, and respectively extracting the spatial domain features and frequency domain features of the rain and fog image; fusing the spatial domain features and the frequency domain features based on an adaptive channel attention mechanism to obtain dual-domain fusion features; the dual-domain fusion features with recovered resolution and the spatial domain features with the same size are added through residual connection to obtain a feature map, convolution processing is performed on the feature map to obtain a clear image, jump connection of traditional U-net is replaced with a residual layer structure in residual connection design, the gradient disappearance problem is relieved through residual connection, and the model training efficiency is improved. According to the design, cross-level fusion of shallow-layer edge information and deep-layer semantic features is achieved, and compared with original jump connection, recovery of local details (such as object contours) and removal of rain and fog textures are better facilitated.

An image augmentation device is provided, which includes a memory and a processor. The processor is configured for performing following operations: extracting a first object contour from a first image mask, wherein the first object contour corresponds to a first label; superimposing the first object contour to a superimposed region in a second image mask according to an augmentation parameter to generate a third image mask, wherein the augmentation parameter includes a contour scaling parameter, a contour moving distance, a contour rotation angle and a range which can be superimposed, and the third image mask includes the first object contour and a second object contour in the second image mask; and generating a sample image which corresponds to according to the first object contour and the second object contour in the third image mask by a generative adversarial network model for performing machine learning.

The invention discloses a TSTFP-Net-based semantic segmentation method for a low-illumination superposition rain image, and the method aims at an image segmentation scene shot in a weak light and rainy environment, and solves a problem that the semantic segmentation precision is reduced because of insufficient illumination, raindrop interference and a significant distribution difference between a source domain and a target domain. The method comprises the following steps: 1) processing an existing data set; 2) designing a semantic segmentation network based on low illumination and rain superposition of TSTFP-Net; 3) training the RIPLM network; 4) training a trunk segmentation network; 5) training a space domain and frequency domain double-teacher network; the method comprises the following steps of (1) training an FSFM module, (2) training a comparison learning module, (3) training a domain discriminator network, and (4) testing a semantic segmentation network with low illumination and rain overlay. The method can still accurately identify the contour and category of an object in complex weather, improve the domain generalization performance, reduce the domain gap, and further improve the precision of semantic segmentation in the low illumination and rain overlay environment.

The invention relates to a parameter capturing model training method and device, electronic equipment and a storage medium. The method comprises the following steps: acquiring multiple groups of training parameters and training scenes, wherein the training scenes comprise multiple target objects; for each group of training parameters, determining an object grabbing score of the target object grabbed by the end effector based on the training parameters; based on the object capture scores corresponding to the multiple groups of training parameters, determining a total point capture score corresponding to the sampling points of the target object; based on the scene point cloud corresponding to the training scene, the object contour mask and the total point capture score corresponding to each sampling point, determining a training sample corresponding to the training scene; training the initial grabbing parameter model by using a training sample to obtain a target grabbing parameter model; the target grabbing parameter model is used for determining target grabbing parameters of the end effector. By adopting the method, the accuracy of capturing the parameter model can be improved.

The invention provides a scatter diagram generation method based on an object contour diagram, and belongs to the technical field of graphic processing, and the method comprises the steps: contour preprocessing: carrying out the smooth denoising processing of a contour diagram, carrying out the Gaussian blurring of the contour diagram, converting the contour diagram into a binary image, extracting a contour set of a connected region in the binary image, and carrying out the Gaussian blurring of the contour image; calculating the surrounding area of each contour and carrying out noise filtering; uniform sampling and initial scatter point generation: constructing a pixel point set, and calculating a maximum distance and a minimum distance of sampling; and scatter-contour matching optimization: traversing the contour line and the scatter set, and eliminating interference scatters to obtain a final scatter diagram of the object contour. Through the methods of uniform sampling, contour point matching, repeated point de-duplication and the like, a specified number of uniformly distributed scatter points can be quickly generated for a contour map. According to the method, the defect of scatter point generation in the current method is overcome, model training is not needed, generalization is high, and the method can be widely used in profile diagrams of various graphs.

The application relates to an object pose recognition method and device, computer equipment, a storage medium and a computer program product. The method is applied to an automatic guided vehicle and comprises the following steps: dividing point cloud data into multiple point group data; determining a point group difference value between a pose recognition parameter of each point group data and a historical point cloud cumulative value; selecting object point group data, in which a point group difference value meets an object contour condition, from each point group data; determining object pose reference lines in different directions according to the object point group data, and determining the pose of the object point group data relative to the automatic guided vehicle through the object pose reference lines. The object pose can be more accurately recognized by using the method.

The invention discloses an image importance detection method for cultural heritage and historical ancient city protection. The method comprises the following steps: constructing an image importance detection model based on Transform and a stable diffusion model; wherein the image importance detection model comprises a coding network and a decoding network; the coding network comprises an input module, a coding noise adding module, a condition module and a coding noise removing module; the decoding network comprises a decoding noise adding module, a decoding noise removing module and an output module; and then, obtaining a to-be-identified historical relic image, inputting the to-be-identified historical relic image into the image importance detection model, and outputting a contour graph and an importance graph to realize image importance detection. According to the method, the global long-range dependency relationship can be effectively utilized, more effective region importance and object contour line detection and recognition can be carried out, the robustness and generalization ability are high, and meaningful objects can be stably detected under various conditions.

The present invention relates to the technical field of peeping detection, and discloses a peeping detection method for a monitoring screen area based on real-time visual recognition, including: determining the screen area, judging whether there is a suspected peeping behavior according to the overlapping result of the current image and the previous image of the acquisition device, identifying the gesture actions of the screen image, determining the object contour of the screen image according to the edge detection algorithm, and looking up the object contour value in the preset object contour table, obtaining the lens brightness value according to the effective lens pixel points, based on the gesture actions, using the peeping behavior model to output the contour weight factor of the object contour value and the brightness weight factor of the lens brightness value, comparing the peeping behavior value with the peeping behavior threshold, and judging whether to send a peeping warning message to the signal jammer according to the comparison result. The present invention comprehensively measures the peeping behavior through the object contour value, the lens brightness value and the gesture actions, and improves the reliability of peeping detection.