126 results about "Image boundary" patented technology

The invention relates to the technical field of topographic change analysis, in particular to an unmanned aerial vehicle-based topographic change detection system, which comprises a slope direction sensing track control module, a texture structure extraction module, a crack evolution track construction module, a direction trend comparison module and a patrol recheck positioning module. According to the method, a continuous elevation point column of an unmanned aerial vehicle scanning area is extracted, laser reflection point coordinates are fused, a space relation of transition point distribution is constructed, dynamic adjustment of a ground-imitated flight path is achieved, and a texture structure area with continuous directivity is recognized in combination with a high-angle image boundary communication relation; texture boundary evolution is compared at different time nodes to form a crack path, the stability of the path and the slope direction is judged through an included angle sequence, recognition and sorting of areas with the consistent direction are completed, a space comparison result is registered in a three-dimensional coordinate system, terrain change areas are accurately marked, and rapid positioning and continuous tracking of high-risk areas are achieved.

The invention provides an image region-of-interest extraction method and system based on a Mama architecture, and relates to the technical field of image processing, and the method comprises the steps: obtaining a to-be-extracted image; performing multi-scale feature extraction on the to-be-extracted image through a local enhancement module; multi-scale semantic enhancement features are generated through a cross-scale self-attention module, and dimension reduction processing is performed through a feature conversion module; local detail features are generated through an adaptive detail enhancement module; performing global context enhancement processing on the local detail features through a pyramid pooling module to generate context enhancement features; performing up-sampling processing on the context enhancement features; the context enhancement features after up-sampling processing are fused through a self-adaptive global-local fusion gating module; and carrying out image extraction based on the decoded fusion features. The image segmentation precision is improved, the model is light in weight, reasoning is fast, and the problems of image boundary blurring and scale variability are effectively solved.

Systems and methods for image processing (e.g., image extension or image uncropping) using neural networks are described. One or more aspects include obtaining an image (e.g., a source image, a user provided image, etc.) having an initial aspect ratio, and identifying a target aspect ratio (e.g., via user input) that is different from the initial aspect ratio. The image may be positioned in an image frame having the target aspect ratio, where the image frame includes an image region containing the image and one or more extended regions outside the boundaries of the image. An extended image may be generated (e.g., using a generative neural network), where the extended image includes the image in the image region as well as generated image portions in the extended regions and the one or more generated image portions comprise an extension of a scene element depicted in the image.

The invention relates to the technical field of wearable monitoring, in particular to farm intelligent wearable inspection equipment, which comprises a track thermal difference recognition module, an abnormal target labeling module, a high fever behavior screening module, an image information extraction module, a structure label generation module and an individual inventory recognition module. According to the invention, by extracting the infrared temperature measurement data and calculating the thermal fluctuation difference value, the positioning of a temperature abnormal section is realized, the identification capability of tiny thermal environment variation is enhanced, the high-heat target area is positioned, the image filing is completed, and the target screening accuracy and the image utilization efficiency are improved. Static high-heat individuals are identified based on the calorific value state and the spatial position stability, the discovery probability of potential abnormal objects is improved, through image boundary contour extraction and structural feature comparison, the individual identification efficiency is improved, and image information and dynamic path data are synchronously analyzed, so that the monitoring result is more continuous and complete, and the detection accuracy is improved. And the intelligent response capability and the information processing depth are improved.

The invention provides an urban digital twinning modeling method and system based on multi-source heterogeneous data fusion, and relates to the technical field of digital twinning, and the method comprises the steps: obtaining a registration image and a registration point cloud of a target region, respectively extracting an image boundary candidate and a point cloud boundary candidate, generating a boundary uncertain band for each boundary candidate, and carrying out the boundary uncertain band extraction; determining an uncertain band width and an offset direction identifier; boundary matching pairs are formed according to the spatial intersection relation, and intersection sections with opposite bias direction identifications are determined as a confrontation boundary section set; for each confrontation boundary section, determining an image confidence weight and a point cloud confidence weight, and generating a boundary judgment result according to the image confidence weight and the point cloud confidence weight; and performing topological consistency verification on a boundary judgment result, and determining a unique fusion boundary. According to the method, under the condition that local physical characteristic difference and systematic geometric deviation exist in multi-source data, the geometric edge fusion precision and topology reliability can be improved, and the stability and availability of the urban digital twin model are enhanced.

The invention discloses an ultrasonic image boundary perception segmentation method, system and device, and relates to the field of image processing, and the method comprises the steps: obtaining an ultrasonic image, and sequentially processing the ultrasonic image through a block embedding module and an image coding module to obtain spatial domain features; performing frequency domain feature extraction processing on the spatial domain feature by using a preset multi-scale frequency extraction strategy to obtain a fine-grained high-frequency feature, a coarse-grained high-frequency feature and a low-frequency feature, and obtaining a frequency fusion feature in combination with a preset frequency alignment strategy; a preset frequency guide boundary refining strategy is combined to determine refining features; and according to the refined features and a preset boundary guiding decoding strategy, determining a segmentation mask representing a result of segmenting the target region under the ultrasonic image from the background region. A low-frequency structure and multi-scale high-frequency boundary details are explicitly separated through frequency domain decomposition, an ultrasonic image boundary sensing segmentation scheme based on frequency guidance is provided, the boundary sensing ability is high, and the cross-domain generalization ability is high.

The invention discloses a steel structure drilling deviation real-time correction method based on visual identification, and relates to the technical field of drilling deviation analysis, and the method comprises the steps: collecting a steel structure image before drilling, and calculating an image center coordinate value and image boundary included angle information of a current drilling target position based on a hole position contour and an edge line segment in the image; constructing a deviation reference set containing the historical drilling task number, the image center coordinate value and the image boundary included angle information, and based on the current image center coordinate value and the image boundary included angle information, retrieving a plurality of spatially adjacent reference records from the deviation reference set to generate a first local deviation estimation value; an image sequence before feeding of the drill bit in the drilling process is collected, and a second local deviation estimated value is output based on a local image block of a contact area between the edge of a drill bit cone and the surface of the component in the image sequence; real-time deviation correction and path continuity control in the steel structure drilling process are achieved.

This disclosure provides a video processing method and apparatus. An example method includes: determining whether a coded block includes samples outside an image boundary; and, in response to the coded block being determined to include samples outside an image boundary, performing quadtree segmentation of the coded block regardless of the value of a first parameter, wherein the first parameter indicates whether the quadtree is permitted for segmenting the coded block.

A high-robustness mixed data enhancement synthesis method, comprising the steps of: 1. preprocessing the input original image to generate an accurate binary mask; 2. performing transformation enhancement on the synthesized image, including but not limited to rotation, shearing, erosion and dilation operations; 3. through a probability-weighted random allocation mechanism, randomly generate high-quality irregular detection targets in stress-concentrated positions in terms of the number of detection objects and the physical layer, simulate real fault scenarios; 4. using a matching algorithm to paste the samples of the detection targets onto the background image, and introducing a boundary overflow processing mechanism to ensure that the detection targets do not exceed the ROI or image boundary, and to ensure that the position of the detection target corresponds to the image, and to ensure that the label does not change. This method is applied to application scenarios with scarce data sets in industrial scenarios, significantly improving the diversity and complexity of the data set, thereby greatly improving the generalization ability and detection robustness of the subsequent target detection model in complex environments.

The application discloses a kind of high-efficiency boundary detection methods based on angle entropy weighted spring network, it is related to image boundary detection technical field, the method includes: for each data point in data set, search its multiple near neighbors;Calculate the elasticity of each near neighbor to current data point, elasticity is based on Hooke's law simulates spring connection;The resultant force of all elasticities is calculated as basic boundary score;Calculate the angle entropy of current data point, angle entropy is based on the angle information of near neighbor distribution Quantification direction distribution;With angle entropy weighted basic boundary score to obtain final boundary score;Select boundary point by pre-set threshold, complete boundary detection.The application quantifies local stress imbalance by spring network, and utilizes angle entropy to amplify boundary feature, while ensuring linear time complexity, significantly improves the discrimination of boundary point detection.

This application provides a method for adding watermarks to an image. Belonging to the field of information security, this method involves dividing an image into blocks to obtain a target number of non-overlapping blocks. The texture complexity is determined based on the number of pixels at the block boundaries and the size of the blocks. The watermark embedding amount is determined based on the texture complexity of each block, the average watermark embedding capacity, and the initial texture complexity. The average watermark embedding capacity is determined based on the total watermark information capacity and the target number. Singular value decomposition is performed on each block to obtain the maximum singular value. The singular value after watermark embedding is determined based on the maximum singular value of each block and the watermark information corresponding to the watermark embedding amount. Inverse singular value decomposition is performed on the singular values ​​after watermark embedding in each block, and each block after watermark embedding is reconstructed to obtain a watermarked image. This method can improve the amount of watermark information and the quality of the watermarked image.

The application discloses a kind of based on pattern recognition's storage tank corrosion defect C scan analysis method, including the following steps: obtaining storage tank C scan image, decay characteristic time chart, amplitude chart and wall thickness estimation chart;With the preprocessed data set input image boundary passage of cross-modal corrosion boundary perception network;With the preprocessed data set input physical decay passage;Based on the cross-modal fusion module of cross-modal corrosion boundary perception network, cross-modal attention fusion is executed;With the cross-modal fusion feature set input boundary refinement module, generate corrosion region initial segmentation chart and corrosion boundary confidence chart;Based on corrosion boundary confidence chart, corrosion region initial segmentation chart is executed iterative boundary refinement processing;According to target corrosion region segmentation chart and preprocessed data set, establish spatial correspondence.This application cross-modal corrosion boundary perception network realizes the fine identification and quantitative analysis of storage tank corrosion region.

A block partitioner includes a quad splitter structured to partition a target block obtained by recursive partitioning in half in both a horizontal direction and a vertical direction to generate four blocks, and a binary / ternary splitter structured to partition the target block obtained by recursive partitioning into two or three in the horizontal direction or the vertical direction to generate two or three blocks, and the binary / ternary splitter disallows partitioning of the target block in the horizontal direction when partitioning of the target block in the horizontal direction causes the target block obtained by partitioning to be located beyond a right side of a picture boundary, and disallows partitioning of the target block in the vertical direction when partitioning of the target block in the vertical direction causes the target block obtained by partitioning to be located beyond a lower side of the picture boundary.

The invention provides an engineering specification document processing method and system, and relates to the field of data set production, and the method comprises the steps: generating a chapter index sequence based on the hierarchical number characteristics of an engineering specification document; the method comprises the following steps: converting an engineering specification document into a high-resolution image, identifying text content based on an OCR technology, extracting a numbering mode from the OCR identification content through a regular expression, and accurately matching the numbering mode with a chapter index sequence so as to obtain page numbers and coordinate information of identified chapters; the missing position is calculated according to the coordinate information of the adjacent chapters, and interpolation completion is carried out on the missing coordinate information; calculating the image boundary of each chapter according to the complemented coordinate set, cutting the picture of the engineering specification document according to the image boundary, setting a boundary protection mechanism in the cutting process to maintain the integrity of the text content of each chapter, and finally forming a structured image library which can be directly used for knowledge management and intelligent retrieval. And the engineering specification retrieval efficiency and accuracy are improved.

The invention provides a bacterial colony image segmentation method, system and device based on AI, and a medium, and belongs to the technical field of bacterial colony selection, and the method comprises the steps: building a bacterial colony image collection environment; the method comprises the following steps: placing a culture dish with a bacterial colony to be detected in a bacterial colony image acquisition environment, acquiring an RGB image of the culture dish, extracting an effective area of the culture dish, removing background interference, segmenting the effective area into sub-graphs with the same size, and setting a sub-graph overlapping area; reasoning each sub-graph by using a pre-trained bacterial colony detection model, and outputting a mask and a bounding box of a bacterial colony; combining reasoning results of the sub-graphs, and deleting target bacterial colonies of which the distance from a central point to a sub-graph boundary is smaller than a set pixel threshold value; and calculating the diameter, roundness and proximity of the bacterial colonies based on the masks of the bacterial colonies, and screening out the bacterial colonies meeting conditions based on a preset threshold value. The accuracy and integrity of bacterial colony selection are improved, and the selection efficiency is improved.

This application discloses a fisheye image correction method applied to an in-vehicle stitching system. The method includes acquiring a fisheye image F using an in-vehicle fisheye lens; determining the edges of black borders within the fisheye image F; performing curve fitting on the pixels at the edges of the black borders; mapping the fitted curve to the image boundary of the fisheye image F to obtain a correction mapping relationship; and acquiring a fisheye image T to be corrected using the in-vehicle fisheye lens, and performing mapping correction on the corrected fisheye image T using the correction mapping relationship. This application also provides a computer-readable storage medium.

The invention relates to the technical field of intelligent information processing, in particular to an enterprise financial report authenticity verification method fusing OCR and financial logic verification, and the method comprises the steps: carrying out the preprocessing and structure analysis of a financial report image file, generating a regionalized image set with metadata, carrying out the feature extraction of the regionalized image set, and carrying out the verification of the authenticity of the financial report. The method comprises the following steps: obtaining high-dimensional visual features, carrying out serialized decoding on the high-dimensional visual features, outputting a character probability sequence, decoding the character probability sequence to obtain a corresponding identification value, binding the identification value with bounding box coordinates, identification confidence and accounting subject identifiers of the identification value in an original financial report image, and constructing a structured financial data unit; and a multi-dimensional fusion verification mechanism is synchronously triggered, and finally, various verification results are comprehensively analyzed, and financial report authenticity risk rating is generated, so that deep fusion of financial report image identification results and financial logic rules is realized, and the automation degree and risk identification capability of enterprise financial report authenticity verification are effectively improved.

The invention discloses a unified pinhole-fisheye depth estimation method based on synthesis distortion and frequency alignment, which is characterized in that a novel pinhole-fisheye reverse mapping method is provided to solve mapping conflicts, the problem of texture discontinuity is solved based on a super-resolution network, and synthesis of a high-quality fisheye image and a depth truth value thereof is realized. The diversity of synthetic data is improved by supporting various fisheye projection models and focal length disturbance. In addition, the invention provides an omnidirectional frequency alignment loss function based on wavelet transform, the distribution of the prediction result and the real depth on the multi-directional high-frequency component is effectively aligned, and the depth estimation precision of the fisheye image boundary is improved. Furthermore, the pinhole depth basic model is finely adjusted through a balanced sampling and playback strategy, a unified depth estimation framework of the pinhole and the fisheye can be realized without internal reference of a camera, leading performance is obtained on a plurality of unseen data sets, and the method has wide application value.

The application discloses a dual-energy ray bone density evaluation method based on a flat panel detector and relates to the technical field of ray evaluation.The direct conversion mode of the flat panel detector is used to realize high-precision collection of low-energy and high-energy ray images, and noise and resolution loss caused by traditional indirect imaging are significantly reduced;one-time and two-time corrections are completed in an edge computing node, the problem of gray scale discontinuity and energy ratio distortion in the joint area of bones and soft tissues is effectively solved, and the authenticity of the image boundary is improved;further dynamic smoothing processing of the energy ratio is combined to obtain a stable ratio image, the stability and repeatability of bone density measurement are significantly enhanced, and more accurate bone density evaluation is realized.

The application discloses a security and warning robot instruction execution system for quick tracking of a holder, which acquires continuous image frames, a current angle of the holder and a circumferential obstacle distance, extracts center coordinates, scale variation, a motion direction and confidence of a warning target, constructs a warning intention propagation graph containing a target, a warning sector, a holder margin, a chassis pass, an obstacle suppression and a loss risk node to obtain a tracking intention value, generates a boundary pressure value according to a distance from the target to a boundary of the image, an interframe displacement towards the boundary, a remaining rotation angle of the holder and a circumferential passable distance of the chassis, determines a current main tracking target, and generates a pre-rotation angle of the holder, a segmented rotation speed curve and a chassis compensation rotation angle, and generates a target residual sector and recaptures the target when the confidence is reduced, so that the tracking stability in a quick moving and short-time shielding scene is improved.

The present invention concerns an apparatus configured to partition a picture into leaf blocks using recursive multi-tree partitioning, block-based encode the picture into a data stream using the partitioning of the picture into the leaf blocks, wherein the apparatus is configured to, in partitioning the picture into the leaf blocks, for a predetermined block which extends beyond a boundary of the picture, reduce an available set of split modes depending on a position at which the boundary of the picture crosses the predetermined block in order to obtain a reduced set of one or more split modes, wherein the apparatus is configured to signal a selected split mode in the data stream.

This invention belongs to the field of visual 3D reconstruction technology and discloses a method, system, terminal, and storage medium for 3D Gaussian reconstruction of fisheye images. The method includes: acquiring a fisheye image; generating a sparse 3D point cloud corresponding to the fisheye image based on a motion recovery structure algorithm; optimizing the 3D Gaussian reconstruction algorithm based on a reversible residual network and an octahedral projection model; and reconstructing a 3D model of the fisheye image based on the optimized 3D Gaussian reconstruction algorithm and the sparse 3D point cloud. This invention reduces the 3D reconstruction error of fisheye images by simulating camera distortion parameters of a lens distortion model using a reversible residual network and incorporating these parameters into the 3D Gaussian reconstruction algorithm. Replacing the planar perspective model of the 3D Gaussian reconstruction algorithm with an octahedral projection model reduces stretching and image boundary distortion. The optimized 3D Gaussian reconstruction algorithm is used to generate the 3D model of the fisheye image, improving the realism of the 3D model reconstructed from the fisheye image.

The application discloses an RGB-D saliency detection method and device, electronic equipment and medium, wherein the method comprises: preprocessing an input image; generating a superpixel segmentation map and extracting a boundary; inputting the preprocessed RGB image and depth image into an encoder to obtain multimodal features; fusing the RGB features and depth features; inputting the fused features into a decoder for decoding to output a predicted saliency map. The RGB features are input into a boundary perception module to output a predicted RGB image boundary map; the fused features are input into the boundary perception module to output a predicted saliency map boundary; the saliency target prediction result is supervised by using a fine-labeled segmentation map, the RGB image boundary is predicted by using the boundary map, and the learning of the model is guided by using a loss function. The application provides boundary guidance for the network through a superpixel generation algorithm, greatly improves the perception of the network to the edge, effectively improves the segmentation quality, and can be widely applied to the field of image processing.

A computer-implemented system and method for detecting a boundary in an image are provided. The system includes at least one processor and memory in communication with said at least one processor, wherein the memory stores instructions, when executed at said at least one processor, cause said system to: receive or access a first image comprising a first polygon structure; generate, using a data model representing a neural network, a second image based on the first image by splitting the first polygon structure in the first image, wherein the second image comprises a first portion and a second portion partitioned by a line across the first polygon structure; and generate, based on the second image, a geo-image comprising corresponding spatial-reference information for one or more pixels in the geo-image, the geo-image comprising one of the first portion and the second portion in the second image.

The invention discloses a post-pond transplanting ridge searching navigation line identification method based on circularity and edge pixel point jump, and belongs to the technical field of ridge culture agriculture visual identification. The method comprises the steps of image acquisition, image boundary smoothing processing, color conversion, threshold segmentation, denoising processing, edge detection, midpoint extraction and navigation line fitting. A depth camera is used for obtaining a depth image, then a traditional RGB image is converted into an HSV image to reduce the influence of an illumination condition on furrow recognition, an H component is used for carrying out next-step processing, threshold segmentation is carried out to obtain a binarized image, a morphological denoising method and an area threshold method are used for removing most noise interference in the image, and the image recognition accuracy is improved. The method comprises the following steps: firstly, performing depth denoising on an image based on circularity calculation, extracting a ridge pit boundary by utilizing edge pixel point jump, then solving a midpoint coordinate point of an image pixel by utilizing abscissas of two boundary points of a ridge pit, and finally realizing navigation line fitting of the ridge pit by utilizing the point coordinates and utilizing a least square method.

The invention relates to the technical field of deep learning, in particular to a PET-CT multi-modal image fusion method and system based on deep learning, and the method comprises the following steps: obtaining PET and CT images, carrying out the normalization and registration, fusing a tumor signal distribution map and an edge response map, generating a focus displacement map, adjusting the image boundary, calculating the similarity, and generating a boundary guide fusion map. Image pyramid decomposition and consistency correction are carried out, pixel-by-pixel difference is carried out, an over-threshold residual error region is reconstructed, and a reconstructed image is output. According to the method, the image alignment precision is improved through normalization and registration, a signal distribution diagram and an edge response diagram are fused to enhance tumor features, a focus evolution displacement diagram is generated to refine tumor tracking, a fusion diagram structure is adjusted based on similarity, a boundary is corrected, noise is eliminated, pixel-by-pixel difference and regression reconstruction errors are achieved, and high-quality reconstruction is ensured. And the image structure consistency, the boundary definition and the feature extraction precision are optimized.

The present application relates to a kind of image classification methods based on separable convolution block and spatial reduction attention mechanism, belong to image classification field.Crossover depth separable convolution and improved spatial reduction attention mechanism are added to PVT model, reduce model training time, and while reducing attention calculation, the original information of feature map is not lost basically at the same time amount of calculation.Crossover depth separable convolution block embedding and spatial reduction attention mechanism based network model is built, including block embedding module, linear projection module, position information embedding module and spatial reduction attention mechanism module;Image classification is carried out to improve the calculation rate when image classification and preserve original boundary information, so as to achieve the overall improvement effect.The present application effectively improves the problems such as huge model calculation and image boundary information loss, reduces the amount of calculation of model and improves the model classification performance.

The invention discloses a vehicle panoramic perception result determination method and device, equipment and a storage medium. According to the technical scheme, real-time detection is carried out on the tiny target and the large target in the vehicle environment image through the tiny-scale detection model and the multi-scale detection model, the target detection accuracy is improved, the high-precision panorama perception requirement of the intelligent vehicle is met, and the vehicle experience is improved. The problem that the target detection accuracy is low when a tiny target and an ultra-large target with the width and the height close to the image boundary exist at the same time is solved.

The application provides a city digital twin modeling method and system based on multi-source heterogeneous data fusion, relates to the technical field of digital twin, and the method acquires registered images and registered point clouds of a target area, respectively extracts image boundary candidates and point cloud boundary candidates, generates boundary uncertainty bands for each boundary candidate, determines uncertainty band width and bias direction identifier; form boundary matching pairs according to spatial intersection relationship, and determine the intersection sections with opposite bias direction identifiers as the set of opposite boundary sections; for each opposite boundary section, determine the image confidence weight and the point cloud confidence weight, and generate a boundary decision result accordingly; perform topological consistency verification on the boundary decision result to determine the unique fused boundary. The method can improve the accuracy and topological reliability of geometric edge fusion and enhance the stability and usability of the city digital twin model in the case of local physical characteristic differences and systematic geometric deviations in multi-source data.