71 results about "Filling-in" patented technology

The invention requests to protect a brain tumor segmentation network method in a missing mode based on mode perception, aims to solve the problem of missing modes in multi-mode MRI data, and belongs to the technical field of computer vision and medical image processing. The method comprises the following steps: step 1, providing a dynamic sampling encoder module which dynamically extracts available modal features and explicitly fills missing modal information through cross-modal feature mapping to enhance the robustness of a model; 2, a modal perception attention fusion module is designed, the module combines local and global attention mechanism adaptive fusion features, and the long-tail tumor region segmentation capability is further improved. And step 3, a two-stage training strategy is designed and comprises self-supervised modal reconstruction pre-training and supervised segmentation fine tuning, and the self-supervised modal reconstruction pre-training and the supervised segmentation fine tuning are supervised by a double-task loss function so as to optimize the network to adapt to various missing modal scenes.

The invention belongs to the technical field of remote sensing image segmentation, and particularly relates to a double-branch water body small target image segmentation method based on feature efficient interaction, which comprises the following steps: preparing a data set, constructing a network model, training the network model, selecting a proper loss function and evaluation index, and determining a segmentation model. According to the multi-scale detail feature interactive aggregation encoder, efficient fusion of multi-scale features and effective supplement of implicit relative position encoding information are achieved; the long-distance feature efficient capture encoder increases the edge segmentation effect on the small target image of the water body by processing features along a specific space direction; the cavity space convolution pyramid module based on the large convolution kernel is used for improving the ability of the model to learn a large-scale effective receptive field; the non-significant feature extraction module is used for filling up the deficiency of non-significant features. The whole network adopts a learning strategy of parallel connection of global and local features and series connection from a large scale to a small scale, and effectively extracts different scale information of the water body image.

The invention relates to the field of computer graphics and real-time rendering, and discloses a naked eye VR visual angle continuous roaming real-time tracking system, which comprises a multi-source viewpoint acquisition unit, a rendering feedback prediction unit connected with a depth buffer area and a template buffer area, and a viewport consistency arbitration unit. According to the method, motion vectors are purified according to object attributes of a template buffer area, a viewpoint is reversely deduced by means of data hierarchical solution of a depth buffer area, and a viewport consistency arbitration unit compulsively uses the reversely deduced viewpoint to take over rendering when the deviation of external sensing data exceeds a limit. The image visual residual inertia is used for filling the blank of external sensor data, and rendering collapse caused by sensor signal interruption or asynchronization is avoided.

The invention discloses a stable diffusion model lawn supplementing method based on hidden space feature editing. The method comprises the following steps: processing an input image by using a semantic segmentation network, and generating a binary mask of a lawn area; inputting the input image, the binary mask and the text prompt into a pre-trained neural network model; performing low-resolution lawn restoration, and outputting a low-resolution restoration image; the neural network model is a repair model which takes a Stable Diffusion XL network as a generation trunk and an Inpaintg-ControlNet network as a condition control network, and the like. The invention provides a lawn restoration method of a stable diffusion model based on hidden space feature editing, which can realize pixel-level accurate positioning and editing of a lawn area and effectively avoid any interference to a non-lawn area in an image.

The invention provides an unmanned aerial vehicle remote sensing image detection method and system, computer equipment and a medium, and belongs to the field of remote sensing image target detection.The method comprises the steps that a remote sensing target detection model SOE-YOLOv8 is provided, a sampling module is asymmetrically filled, filling in four different directions is carried out, and a remote sensing target detection model SOE-YOLOv8 is obtained; the forced convolution kernel pays attention to context information of different edge regions, adaptively adjusts weights through an adaptive feature splicing module so as to improve feature expression ability, increases weights in corresponding directions of a target when detecting an inclined target, inhibits background noise and enhances related features of the target; through a dimension perception selective integration module, the detection performance in a complex scene is remarkably improved; through the standardized Gaussian Wasserstein distance, loss evaluation is carried out, distribution similarity measurement is calculated, a label distribution and feature learning mechanism is optimized, and the robustness and precision of small target detection are improved.

The invention discloses a photo-induced quantum dot in-situ growth method for coupling and filling element defects and application, relates to the technical field of optical element defect detection, and solves the problems of poor selectivity, low resolution and low sensitivity of a thermally induced in-situ growth material in defect detection of an optical element in the prior art. According to the invention, the CsPbBr3 quantum dot precursor is selectively filled in the defect area on the surface of the element, and the CsPbBr3 quantum dot precursor is induced to grow in situ at the defect position under the irradiation of a specific light source, so that a remarkable local photothermal effect is generated at the defect position. According to the method, the problem that a traditional thermal induction method is insufficient in positioning selectivity in a defect area is effectively solved by utilizing a selective growth mechanism, high-sensitivity, high-selectivity and high-spatial-resolution positioning detection on microdefects of the optical element can be realized, and the method can be applied to defect positioning of the optical element; and a new technical approach is provided for visual characterization and high-precision positioning of defects.

The application discloses a high-precision three-dimensional reconstruction method based on sparse depth, an RGB-D image sequence of sparse depth is input, the missing depth information in each frame of sparse depth map is effectively filled in by a depth completion network trained on a public data set, and a pre-trained visual large model Dinov2 is used to infer the semantic information of each frame; for each frame of RGB image, a joint sampling method based on feature point sampling and uniform sampling is used to extract the sampling points of light, and spatial three-dimensional points are generated through importance sampling and random sampling strategy; the position coding is carried out on the three-dimensional space sampling points to code the low-frequency features, the high-frequency features are processed by using hash coding, and the two are fused and input into a neural network to predict the color value, SDF value and semantic value; according to the current scene representation information, the pose of the current frame is rendered and optimized, the predicted value is compared with the real color, depth and semantic information of the current frame through the global key frame as a sliding window, the implicit scene representation is optimized, and a high-fidelity three-dimensional model is reconstructed from the sparse depth RGB-D image sequence. The application effectively solves the problem of insufficient reconstruction precision of the traditional three-dimensional reconstruction technology under the condition of sparse depth data, and has a wide application prospect, and is especially suitable for the field of three-dimensional reconstruction based on RGB-D data.

The present invention discloses a method and system for reconstructing optical images with high spatiotemporal resolution based on the fusion of multi-source remote sensing data, aiming to solve the problem of missing spatiotemporal information of optical remote sensing images due to cloud cover, sensor failure and the like in the prior art. The method is as follows: 1. Acquisition, preprocessing and data pair construction of multi-source remote sensing data. 2. Constructing and training an optical image reconstruction model, the model comprising a generator that receives multi-source data as input and a discriminator. 3. Reconstructing optical images of missing phases using the trained generator to generate a complete continuous optical image time series with high spatiotemporal resolution. The present invention can effectively fill in the gaps in optical image data, and the generated images have high spatial and spectral accuracy. It can effectively process data of different resolutions and reduce the influence of geometric registration errors through a multi-input network structure, providing reliable data guarantee for remote sensing applications that rely on continuous optical observations.

The application relates to the technical field of defect detection, and discloses an intelligent identification and distribution visualization method and system for weld defects, which comprises the following steps: collecting a weld radiographic testing image, obtaining a smooth image after denoising and performing multi-scale feature extraction, and obtaining a weld segmentation image through semantic segmentation; adopting a convolutional neural network to classify different defect types, performing pixel labeling, and obtaining a defect labeling image; performing uniform grid division on the weld segmentation image, obtaining an optimized sub-region set after filtering boundary pixels; calculating the number proportion of defect pixels and total pixels in the sub-region, comparing the number proportion with a preset proportion threshold value, and completing classification after comparison, mapping to a two-dimensional matrix to obtain a density distribution matrix; aggregating proportion values along the length and width directions of the weld, and filling in missing data points to obtain a continuous proportion value sequence; and assigning corresponding colors to the proportion values, and obtaining a defect distribution heat map after rendering. The method can realize high-precision detection and visual presentation of weld defects.

1. Name of the product in this design: Image clearing dynamic graphical user interface image area of ​​electronic device. 2. Purpose of this design: An electronic device. 3. The key design feature of this product is its graphical user interface. 4. The image or photo that best illustrates the design points: Design 1 Interface Change State Diagram 3. 5. Design 1 is designated as the basic design. 6. Purpose of the graphical user interface: The overall purpose of the interface is to remove specific areas from the image based on user operations and dynamically fill in the removed areas; the local purpose of the interface is to display the different stages of the image cutout process through particle and light effects in the image area. 7. Human-computer interaction method of graphical user interface: In Design 1, the main view is the main interface displayed by the electronic device. When the user long-presses the area of ​​the image area in the main view that needs to be cleared, the interface will automatically jump to the interface change state diagram 1, interface change state diagram 2, interface change state diagram 3 and interface change state diagram 4 in sequence; among them, the gray area in the interface is used to display the image. In Design 2 and Design 3, the main view is the main interface displayed by the electronic device. When the user long-presses the area in the image area of ​​the interface that needs to be cleared in the main view, the interface will automatically jump to the interface change state diagram 1, interface change state diagram 2 and interface change state diagram 3 in sequence; the gray area in the interface is used to display the image. In Design 4, the main view is the main interface displayed by the electronic device. When the user long-presses the area in the image area of ​​the interface that needs to be cleared in the main view, the interface will automatically jump to the interface change state diagram 1 and interface change state diagram 2 in sequence; the gray area in the interface is used to display the image. 8. The areas in the diagram that are covered with a semi-transparent blue coating represent parts that do not require protection.

The present application relates to the technical field of marine engineering digital twin and three-dimensional visualization technology field, provide a kind of based on right triangle grid's dredging topography twin's generation method and system, comprising: obtaining regular grid elevation or sampling height data after pre-processing that have been unified to engineering plane coordinate system, determine the space range covered read space range and density parameter, based on space range and total row number and total column number of calculated grid;In the space covered, generate regularly arranged vertex, construct the topological structure of isosceles right triangle grid;With vertex as unit, several Chunk blocks are divided into global grid with fixed size, establish the triangular index mapping relationship of global row and Chunk block, the three vertices of isosceles right triangle and elevation or height data are filled in each Chunk block in parallel, generate local VBO, IBO and AABB;Batch rendering and coloring according to height color table are executed to all Chunk blocks, and the visualization of topography is realized.

The application provides a periodic image hole filling method and system, and relates to the technical field of digital image processing.A periodic image hole filling method comprises the following steps: preprocessing a periodic image based on a preset image processing model to obtain a hole region, a gray reference region and a gray statistical curve of the periodic image; performing smoothing processing on the gray statistical curve to obtain a period segmentation node; obtaining filling data based on the hole region and the period segmentation node; and filling holes of the periodic image according to the filling data.The periodic image hole filling method solves the technical problem that a sawtooth is generated in an edge excessive region after a periodic image hole is filled in the related art, which affects filling accuracy.

The invention provides a track detection camera real-time calibration method and device and a medium, and the method comprises the steps: obtaining track image data collected by a camera, and carrying out the preprocessing of an image, including Bayer conversion, filtering and vignetting correction; configuring a track edge candidate region in the image, and extracting track edge features; carrying out error detection filtering and leakage detection filling on the extracted edge features, and fitting to obtain a track edge line; constructing an energy equation of camera parameters based on the prior size information of the track edge line; solving the energy equation by adopting a nonlinear optimization method and taking a historical frame optimization parameter as an initial value to obtain camera internal and external parameters of a current frame; and converting the image point cloud into a three-dimensional point cloud of a real scale space by using internal and external parameters of the camera. According to the technical scheme, the camera can be calibrated frame by frame in real time, the real-time calibration parameters of the camera are obtained, the camera can adapt to environment changes in real time, and the measurement precision is improved.

The present application discloses a method for removing image content, which relates to the field of computer vision. The method includes: starting a camera application and displaying a photo preview interface of the camera application. Obtaining a first preview screen and a first reference frame screen captured by the camera, and determining that the first object in the first preview screen is the object to be removed. Based on the first reference frame screen, determining the content to be filled in the first preview screen, wherein the content to be filled is the image content of the second object in the first preview screen that is blocked by the first object. The terminal generates a first repair screen based on the content to be filled and the first preview screen. In this way, image content that the user does not want in the picture or video taken by the user can be removed.

The application provides a non-spherical surface shape detection distortion correction method, device, equipment and medium, and relates to the optical technology field. The method determines the first CCD coordinates corresponding to each data point without distortion by operating the preset distortion model of the distortion parameter, determines the second CCD coordinates in the case of distortion, and then fills in each data point without distortion. The mapping relationship between the actual mirror and the CGH is used to establish the relationship between the actual mirror and the CCD coordinates by taking the CGH as a medium, so as to complete the correction of the imaging distortion and output more accurate non-spherical surface shape data. The above process does not need to set up physical markers on the actual mirror, avoids the influence of the physical markers on the high-precision optical element, simplifies the operation process of the non-spherical surface shape detection, and improves the distortion correction efficiency and accuracy.

The invention relates to an image multi-modal layered labeling method and a semi-automatic image data labeling method.The image multi-modal layered labeling method comprises the following steps that image-level labeling is established, at least one image-level selection box is established for each effective image so as to cover the whole effective image, the overall description of the effective image is filled in, and the image-level labels are obtained; if the current effective picture contains the key elements, marking that the current effective picture contains the key elements; independently labeling each key element in the effective picture by using a labeling box; performing visual information decoupling and text cleaning on the overall description of the effective picture, removing non-visual description, extracting and filling visual description, and extracting and filling subject nouns in the visual description; and filling in a logic process from the visual description derivation to the diagnosis conclusion. Non-visual description is eliminated, pure visual description is reserved, model illusion can be eliminated, and the robustness of the model is improved. The interpretability of the model is enhanced by filling in a logic process from visual description derivation to diagnosis conclusion.

The invention provides a periodic image hole filling method and system, and relates to the technical field of digital image processing, and the method comprises the steps: carrying out the preprocessing of a periodic image based on a preset image processing model, obtaining a hole area, a gray scale reference area and a gray scale statistical curve of the periodic image; smoothing the gray statistical curve to obtain a periodic segmentation node; based on a hole region and the periodic segmentation node, obtaining filling data; and filling the holes of the periodic image according to the filling data. According to the periodic image hole filling method, the technical problem that the filling precision is affected due to the fact that sawteeth are generated in the edge transition area after periodic image hole filling in the prior art is solved.

The present application relates to a kind of high light removal method and system of face image based on regional hierarchical processing, belong to digital image processing technical field.It includes the following steps: obtaining face image and generating face target area mask, image is converted from RGB color space to LAB color space in limited area, and overexposure and near overexposure area are divided based on brightness threshold value;Fill in repair is carried out to overexposure area using the statistical characteristics of near overexposure area on brightness and color difference channel;Further generate highlight area mask according to brightness distribution, calculate dynamic adjustment coefficient, and adaptively update brightness and color difference channel in highlight area, finally the repair result is converted back to RGB color space and output, effectively remove the high light of face and keep natural skin color.The present application can effectively retain face details and color consistency while suppressing overexposure highlight, improve the naturalness and robustness of the result of removing highlight.

The present invention provides a method and system for filling holes in virtual viewpoint synthesis. The method includes: a foreground extraction step, extracting a foreground mask that causes large holes in the virtual viewpoint from the original viewpoint image; a background completion step, first using an improved spatiotemporal filling method for filling holes, specifically registering several previous and subsequent frames to the current frame. During the registration process, optical flow tracking is performed on the feature points of the matching pairs to eliminate incorrect matching pairs. The homography matrix calculated after the optical flow tracking is then used for image splicing to achieve partial background filling; then, an improved spatial filling method is used for filling holes, specifically assigning a priori values ​​to the initial bias of pixels when using the PatchMatch algorithm to obtain a completed occluded background; and a background re-transformation step, transforming the completed occluded background into the virtual viewpoint. The present invention can improve the quality of virtual viewpoint synthesis.

The application discloses a kind of sea surface flow field inversion and reconstruction method based on stationary satellite high frequency image, belong to marine remote sensing technical field.It includes the following steps: satellite image is preprocessed and locally enhanced;Initial inversion is carried out using dense optical flow algorithm for single time phase flow field;The flow field data cube of space-time grid uniformization is constructed;Time series reconstruction is carried out using two-way Kalman filtering, through forward recursion, backward recursion and two-way fusion, using time series information to correct abnormal vector and fill in missing data;Finally, based on the physical consistency principle of seawater approximate incompressibility, the reconstructed flow field is subjected to iterative divergence constraint, and the non-physical disturbance is weakened.The application effectively solves the problem that the traditional optical remote sensing flow measurement method is easy to produce abnormal vector and poor spatiotemporal consistency under complex sea conditions such as cloud cover and weak texture by introducing time series correlation and physical constraint, significantly improves the accuracy, stability and physical reasonableness of sea surface flow field inversion.

This application relates to the field of robot vision inspection technology, and provides a vision-based intelligent robot control method and system. The method includes: acquiring original image data of multiple workpieces to be grasped within a material frame; determining information-deficient regions in the original image data based on the original image data; the image information includes brightness values ​​and texture features; filling in the information-deficient regions with pixels in the surrounding area of ​​the information-deficient regions in the original image data to obtain repaired image data; the width of the surrounding area is greater than a preset width threshold; generating and issuing a workpiece grasping command to the robot based on the repaired image data. This improves the accuracy of robot control when the robot grasps workpieces.

This invention relates to the field of water conservancy engineering technology, specifically to a method and system for detecting cracks in water conservancy projects based on intelligent visual recognition. The method includes the following steps: acquiring crack images and enhancing brightness and contrast in different regions; extracting and optimizing crack edges; performing morphological processing for noise reduction and smoothing; segmenting crack regions and determining connectivity; analyzing geometric features to detect connectivity; and calibrating crack positions to obtain localization results. This invention, through dynamic adjustment of image brightness and contrast, makes cracks more prominent in complex backgrounds, especially under uneven lighting, reducing the difficulty of crack identification caused by inconsistent lighting. Optimizing crack edges through morphological operations fills in missing parts of the crack, smooths the crack outline, and eliminates noise, thereby ensuring the clarity and integrity of crack features. In the crack segmentation stage, adaptive thresholding and geometric feature analysis accurately determine crack fracture and connectivity, achieving the localization of crack endpoints and their continuous regions.

The invention provides a security inspection equipment imaging quality evaluation method and system, and the method comprises the steps: carrying out the three-stage preprocessing of Canny edge detection, Hough posture correction and polynomial distortion compensation on the X-ray images of a first test body and a second test body, and guaranteeing the regularity of the images; and then a plurality of function test areas are accurately segmented by using light YOLOv8n, and manual area finding is replaced. Then, a five-dimensional algorithm is started in parallel according to the national standard, and the pixel missing rate of a linear resolution area and a penetrating resolution area is measured by combining self-adaptive threshold value and curve fitting; the spatial resolution area uses RANSAC to supplement broken lines and count; a CNN residual high-frequency model is used for evaluating the definition of the penetrating power area, and contour features are superposed to calculate the integrity; organic matter equal four-resolution areas are converted into HSV, and the number of color blocks and the average saturation are calculated through K-means clustering; and converting the two resolution regions of the material into gray scale contrast difference. And in combination with preset parameters, the five-dimensional result is mapped into a quality judgment result, experts are not needed, human eye related errors are eliminated, and full-automatic efficient acceptance of the large hub security inspection equipment is realized.

This invention discloses an image processing method, system, device, and medium, relating to the field of image processing technology. The method includes the following steps: acquiring an image to be processed; extracting channel features from the odd-field and even-field data of the image to be processed, and weighting and fusing the odd-field and even-field data based on an adaptive channel weight allocation strategy to generate multi-scale features; extracting multi-level contextual information from the multi-scale features, analyzing the multi-level contextual information in conjunction with interpretability rules to obtain target object detection results, and dynamically adjusting image restoration parameters according to ambient light intensity, filling in missing areas using background color; semantically aligning the target object detection results with the filled image, and adjusting the brightness threshold of dark areas to generate the final image. This invention achieves accurate matching of cross-modal information in complex scenes and robust processing of complex lighting / occlusion scenes.

The present application relates to image multimodal layered labeling method and semi-automatic image data labeling method, wherein the image multimodal layered labeling method comprises the following steps: establishing picture level labeling, establishing at least one picture level bounding box for each effective picture to cover the whole effective picture, and filling in the overall description of the effective picture, if the current effective picture contains key elements, indicating that the key elements are contained; using the labeling box to independently label each key element in the effective picture; decoupling the visual information of the overall description of the effective picture and cleaning the text, removing the non-visual description, extracting the visual description and filling in, and extracting the subject noun in the visual description and filling in; and filling in the logical process from the visual description to the diagnostic conclusion. Eliminate non-visual description, keep pure visual description, help to eliminate model illusion, improve model robustness. By filling in the logical process from the visual description to the diagnostic conclusion, the model interpretability is enhanced.

This invention provides a projection-based method and system for compressing and transmitting LiDAR point clouds, belonging to the field of LiDAR data processing technology. The method includes: receiving and initializing raw 3D point cloud data from a LiDAR system; mapping the point cloud data in a spherical coordinate system to 2D image pixel coordinates; generating a dual-output depth image and a pseudo-color visualization image; filling in void regions using a distance-weighted interpolation algorithm; and compressing the pseudo-color image using JPEG and pushing it via the RTSP protocol. This invention, through spherical projection dimensionality reduction, a dual-output mechanism, and distance-weighted interpolation with horizontal surround characteristics, reduces the transmission bandwidth from 5.22MB / s to 39KB / s while ensuring the integrity of spatial information, achieving low-latency, high-compression-ratio, and cross-platform compatible real-time transmission of LiDAR point clouds.

The present application relates to computer graphics and real-time rendering field, disclose a kind of naked-eye VR view angle continuous roaming real-time tracking system, comprising: multi-source viewpoint acquisition unit, rendering feedback prediction unit and viewport consistency arbitration unit connected with depth buffer and template buffer, the present system is through rendering feedback prediction unit, according to the object attribute purification motion vector of template buffer data hierarchical solution to deduce viewpoint, viewport consistency arbitration unit is when external sensing data deviation overrun, forcedly use deduced viewpoint to take over rendering, the present application constructs independent hot backup and abnormal takeover mechanism based on rendering pipeline data, uses picture visual residual inertia to fill in external sensor data blank, avoid rendering collapse caused by sensor signal interruption or asynchronization.

The invention discloses a camera point supplementing method for global tracking. The method comprises the following steps: constructing a three-dimensional model of a region, and carrying out projection to generate a grid coordinate system of a two-dimensional base map; solving a homography matrix of the existing camera according to a mapping relation between the acquired data of the existing camera and the three-dimensional model; calculating an accurate sensing area of the existing camera, and combining the grid areas with the continuous relation to form a closed sensing area; based on the obtained closed sensing area, carrying out reverse deduction to obtain a plurality of blind areas to be compensated which need to be monitored; and calculating the average sensing range of the existing cameras, calculating the minimum number of camera compensation points in each to-be-compensated blind area, and performing deployment and control. The method has the beneficial effects that the relation between the visual angle of the camera and the spatial range can be effectively recovered in the absence of detailed external parameters, so that the blind area range is judged; therefore, judgment and complementation of visual dead angles can be completed without replacing hardware of the original monitoring system.

The application discloses a kind of based on point mark image target object auxiliary marking method, comprising: a kind of image pre-segmentation method, semi-automatic generation target object mask method.Based on connected region segmentation algorithm, image is pre-segmented, and superpixel macroblock is obtained;Then the representative color of each superpixel macroblock and id are filled in the pixel points covered by it to generate pre-segmentation image and pre-segmentation mask respectively;Manual distinguishes target object on the pre-segmentation image of front interface and marks point, and the mapping relationship of superpixel macroblock id and target object id is automatically generated according to the marking point;Based on this mapping relationship, target object mask output is obtained, and the meta information of mark point is also output.The method for generating mask and marking point meta information that can be used for supervised image segmentation training in the application only needs simple manual marking, and provides sufficient data support for supervised learning model in image segmentation field.

The application provides an image reconstruction method, system and two-photon microscopic imaging system, comprising: based on the data obtained by MEMS Lissajous scanning, and using a Gaussian weighted interpolation algorithm, a two-dimensional image is reconstructed and generated and subjected to frequency domain transformation, based on multiple global statistical characteristic indexes of power spectrum, a comprehensive scoring function for evaluating image quality is constructed; through iterative search, a target phase compensation value is determined to make the comprehensive scoring function optimal; using the target phase compensation value, the phase of the scanning track is corrected and the finally reconstructed image is output. The scheme of the application establishes a MEMS second-order frequency-phase response model to provide an accurate initial range for the search algorithm; a Gaussian weighted interpolation strategy is introduced to fill in the sampling holes and suppress high-frequency sampling noise; the global statistical characteristics are used to significantly improve the robustness in a low signal-to-noise ratio environment; a "global sparse scanning + robust golden section search" algorithm is proposed to greatly reduce the calculation amount.