53 results about "Image deformation" patented technology

The invention, which belongs to the technical field of remote sensing image processing and geological disaster monitoring, discloses a high-risk slope multi-temporal remote sensing image deformation monitoring and early warning system comprising a multi-platform remote sensing data acquisition module, a deep neural network registration module, a sub-pixel-level deformation detection module, a meteorological data fusion module and a dynamic early warning module. The centimeter-level registration precision is realized by introducing an adaptive feature registration mechanism based on a graph attention network, the deformation estimation precision of a texture sparse region is improved by adopting a time-space domain combined sub-pixel-level deformation field estimation algorithm, and the early warning timeliness is improved by establishing a dynamic early warning threshold self-adjustment mechanism fused with meteorological data. The system supports multi-scale multi-platform cooperative monitoring, the early warning advance period is prolonged by 7-15 days, and the system is successfully applied to high-risk slope monitoring projects along the expressway.

The invention belongs to the technical field of data processing, and provides a scale detection and scale conversion method and system based on an ultrasonic image, and the method comprises the steps: constructing a discrimination function fusing multi-modal features according to the spectral basis functions of different scales, and determining the scale contour of the scales in the ultrasonic image; the design can dynamically adapt to the spectral characteristics and structural characteristics of the scale, can effectively cope with the interference of scale diversity, ultrasonic speckle noise and the like of equipment of different manufacturers, and ensures the stability of scale positioning and subsequent conversion; the method comprises the following steps of: establishing a biomechanical-image deformation coupling model by taking a scale contour of a scale in an ultrasonic image as a geometric reference and combining biomechanical characteristics of an ultrasonic probe and tissues to obtain a corrected pixel distance between adjacent scale points of the scale; and curvature compensation for eliminating refraction errors of ultrasonic waves in curved tissues is introduced to correct a scale pixel-millimeter conversion coefficient, so that the pixel-millimeter conversion error rate is reduced to 1.2%, which is far lower than the level of a traditional method.

The invention relates to the technical field of automatic visual detection, and discloses a bottle body label quality detection-based model automatic generation method, which comprises the following steps of: acquiring a multi-angle image of a bottle body, carrying out denoising, enhancement and brightness equalization processing, extracting a curved surface label through target detection and a perspective transformation algorithm, and carrying out image processing on the curved surface label; further generating a distortionless complete label image through an image registration and fusion technology, and taking the distortionless complete label image as a training sample set of a label defect detection model constructed based on a double-branch deep learning network; and performing hot updating on the label defect detection model through a closed-loop feedback and continuous learning mechanism. According to the method, the technical problems of imaging deformation of the curved surface label, diverse defects, difficulty in detection and performance degradation after model deployment are effectively solved, and high-precision, high-robustness and sustainable-evolution automatic label quality detection is realized.

The invention provides an endoscope image stabilization control method, which comprises the following steps of: acquiring an original dithering image sequence and inertial data of an endoscope, fusing the original dithering image sequence and the inertial data, acquiring an initial mixed motion field, performing operation semantic segmentation on the original dithering image sequence, and acquiring a collaborative attention map in combination with a physical motion mechanism. Obtaining a space-time adaptive filtering kernel through the collaborative attention map, filtering the initial mixed motion field according to the space-time adaptive filtering kernel, obtaining a stable motion field, carrying out image deformation and interpolation synthesis processing on the stable motion field, and obtaining a stable image sequence; through the integrated technical scheme of operation semantic segmentation, dynamic image stabilization optimization and accurate jitter suppression, the problem of interference of endoscope jitter on the neurosurgery operation visual field is solved.

The invention provides a defect detection method and device, electronic equipment and a storage medium. The method comprises the following steps: acquiring a template image and a product image of a to-be-detected product; performing feature matching on the template image and the product image, and performing non-rigid alignment on the product image based on feature point pairs obtained by feature matching to obtain an alignment image of which the geometric view angle is consistent with that of the template image; and defect detection is carried out based on the aligned image and the template image to obtain a detection result, so that the problem of image deformation caused by the fact that the surface of the to-be-detected product is a curved surface and the shooting angle is not correct in a traditional scheme is effectively solved, and local and global deformation of the product image can be corrected, thereby greatly improving the alignment precision of defect detection, and improving the detection accuracy of the to-be-detected product. Therefore, the real difference of the product surface can be more focused during defect detection, non-defect geometric deformation interference is effectively shielded, and the accuracy and robustness of defect detection of complex surfaces of household electrical appliances and the like are greatly improved.

An apparatus includes a memory to store first volumetric image data representing a first volumetric image of an anatomical region having a volume of interest (VOI). The first volumetric image includes several first voxels at first locations in the first volumetric image. The first volumetric image data includes first uncertainty values corresponding to the several first voxels. The first uncertainty values represent probabilities that the corresponding plurality of first voxels are part of the VOI. The memory is to store second volumetric image data representing a second volumetric image of the anatomical region. The apparatus includes a processing device operatively coupled to the memory. The processing device is to determine a deformation field to map the several first voxels to second locations in the second volumetric image. The deformation field is based in part on the first uncertainty values.

The invention provides a chromosome straightening method and device, electronic equipment and a storage medium. The chromosome straightening method comprises the steps that a chromosome image forms a binary image; s2, acquiring a central axis of each single chromosome in the binary image; s3, segmenting the chromosome image into a plurality of first-level triangles; s4, mapping the ordinates of the first-level central axis sampling points corresponding to the first-level triangles to be the same to form second-level central axis sampling points; s5, according to transformation of each first-level central axis sampling point and the corresponding second-level central axis sampling point, obtaining a corresponding second-level triangle coordinate through the first-level triangle through a rigidity-keeping image deformation algorithm; calculating an affine transformation matrix corresponding to each first-level triangle according to transformation of the first-level triangles and the corresponding second-level triangles, and performing affine transformation on the chromosome image region corresponding to each first-level triangle to obtain a straightened chromosome image; the chromosome straightening method provided by the invention can effectively correct the chromosome image.

The embodiment of the invention provides a method and device for image processing, equipment, a storage medium and a system for augmented reality. In the method, grid information corresponding to the pose of the augmented reality device is acquired. The grid information is used for image deformation processing of the augmented reality equipment. A first set of pixels of a first image acquired by the augmented reality device in the first memory is caused to be stored to the second memory based on the grid information. The read speed of the second memory is higher than that of the first memory. Further, positional information of a first pixel in the first image corresponding to a second pixel in the image-deformed second image is determined, and a value of the second pixel is acquired by reading a value of the first pixel from the first memory or the second memory based on the positional information. The priority of reading the value of the first pixel from the second memory is higher than that of the first memory. In this way, the delay of video perspective of the augmented reality device can be effectively reduced.

The invention relates to the field of medical image processing, and discloses a medical image data processing method and system, and the method comprises the following steps: S1, obtaining a multi-time-point medical image sequence; s2, constructing a finite element mesh model and endowing the finite element mesh model with biophysical attributes; s3, solving a physical optimization problem to calculate a displacement field; s4, calculating a stress field, a strain field and a physical residual field; s5, outputting a basic analysis result; the system comprises an image acquisition module, a model construction module, a physical solving module, a multi-scale coupling module, a risk assessment module, a treatment planning module and an output module. According to the method, physical constraints based on a biomechanical model are constructed, and a physical residual field is calculated in combination with a reverse image deformation technology, so that the transformation from traditional geometric morphology measurement to deep information mining based on a biophysical model is realized, and the information dimension and accuracy of an image processing result are improved.

This invention discloses a method, apparatus, device, and storage medium for medical image deformation registration. The method includes: acquiring a first modal image and a second modal image, and generating a mask image based on the noise region features of the second modal image; constructing a hybrid similarity measure function based on the first and second modal images, the hybrid similarity measure function including a similarity measure term based on image context information and a similarity measure term based on local gradient features; iteratively optimizing the registration of the first and second modal images based on the hybrid similarity measure function and the mask image to determine the spatial transformation parameters for aligning the second modal image with the first modal image; and transforming the first modal image according to the spatial transformation parameters to obtain the registered first modal image. Compared with the prior art, this invention improves the accuracy and robustness of multimodal medical image deformation registration.

The invention belongs to the technical field of PET (positron emission tomography), and relates to a PET probe extensive field image multi-model adaptive deformation and FPGA (field programmable gate array) curing method, which comprises the following steps of: acquiring single-lift event data of a PET probe; generating an original extensive field image based on the single-lift event data; performing coordinate traversal transformation on the original extensive field images by adopting an image deformation model in a preset multi-model transformation library; performing off-line analysis on each model and the parameters thereof in the multi-model transformation library by utilizing a preset evaluation index, and screening out an optimal model and an optimal parameter; and transforming the original pan-field image by using the optimal model and the optimal parameters to generate an enhanced image and a crystal position lookup table, and solidifying the enhanced image and the crystal position lookup table to the FPGA. The method has the beneficial effects that the local adhesion area can be stretched in a targeted manner through the nonlinear transformation model, the distinction degree of crystal event clusters is remarkably improved, and the LUT calibration precision is fundamentally improved.

The invention relates to the technical field of image processing, in particular to a target picture and template picture specific area association matching method and system.The method comprises the steps that a template picture, a specific area mask corresponding to the template picture and a target picture are received as input, and multi-scale features are jointly extracted through a neural network model; and in combination with area prior information provided by the mask, predicting a central point offset proportion of a corresponding area in the target image. And calculating an accurate position coordinate in the target picture by combining the predicted offset based on the center coordinate and the width and height parameters of the specific region in the template, and further positioning the matched region and generating an association result. According to the method, the mask is fully utilized to guide the network to pay attention to the key area, and the matching robustness and the positioning precision under the condition that the image has deformation, illumination change or noise interference are improved. Compared with a traditional method, error accumulation caused by dependence on dense feature matching or key point detection is avoided, and end-to-end rapid and accurate area positioning is achieved.

The invention relates to the technical field of oral cavity imaging, in particular to an intelligent electronic tongue depressor and a high-precision three-dimensional oral cavity imaging method. According to the structural features of different frames of optical slice images, the local deformation possibility of each matching feature point in each frame of optical slice image is obtained, and matching error feature points forming a deformation target area are screened out; for any frame of optical slice image, according to the local deformation possibility of different matching feature points in the deformation target area, the positions of the different matching feature points in the three-dimensional point cloud model and the change trend of a corresponding moving path when the tongue depressor intersects with the deformation target area, the matching error consistency of the deformation target area on each frame of optical slice image is obtained; and in combination with the point cloud coordinates of the matching error feature points in the deformation target area and the position distribution of the deformation gravity center, obtaining point cloud correction coordinates of the matching error feature points. According to the method, the point cloud correction coordinates of each image deformation area are accurately obtained, and the accuracy of oral cavity three-dimensional model construction is improved.

A projected image deformation correction result evaluation method and apparatus, a device, and a storage medium. The method comprises: obtaining corner point coordinates of a projected image, wherein the projected image is a quadrilateral, and the corner point coordinates are coordinates corresponding to four corners of the quadrilateral in a rectangular coordinate system; obtaining four included angles between four sides of the projected image and the corresponding sides of an absolute rectangle, wherein the absolute rectangle refers to that one group of opposite sides in the four sides of the rectangle is parallel to the X-axis in the rectangular coordinate system, the other group of opposite sides is parallel to the Y-axis in the rectangular coordinate system, the X-axis is the horizontal direction, and the Y-axis is the vertical direction; determining values of the four included angles on the basis of the corner point coordinates; and evaluating a correction result of the projected image on the basis of the values of the four included angles.

The invention belongs to the technical field of infrared-visible light real-time registration, and particularly relates to a real-time registration method for an infrared-visible light dual-spectrum camera, and the method comprises the following steps: collecting an infrared-visible light image to be registered; random affine transformation and random elastic deformation are applied to original infrared images in the strictly registered RoadScene and TNO data sets to obtain deformed infrared images, and the deformed infrared images and the deformed infrared images form an unregistered infrared-visible light image data set; the method comprises the following steps: establishing an improved Dense Matcher network; performing improved Dense Matcher network training by using the unregistered infrared-visible light image data set to obtain an optimal infrared-visible light image deformation field prediction model; and predicting a deformation field of the infrared-visible light image to be registered by using the optimal infrared-visible light image deformation field prediction model, and performing spatial transformation on the infrared image according to the deformation field to obtain a registered infrared-visible light image.

The invention discloses a longitudinal submerged arc welding pipe steel plate compression molding deformation detection image distortion repairing method. An image detection device, a PC terminal, an application server and a background database of the steel plate pressing deformation image online detection system are sequentially connected, application software is installed in the application server, and the PC terminal is in communication connection with the image detection device; the longitudinal submerged arc welding pipe steel plate compression molding deformation detection image distortion repairing method comprises the following steps: establishing a longitudinal submerged arc welding pipe steel plate compression molding model database; forming parameters are monitored, collected and analyzed in real time; calculating steel plate pressing deformation characteristic data; and predicting and guiding the rolling reduction of the forming die. The method has the beneficial effects that the steel plate pressing deformation detection image distortion repairing system is established, the welded pipe forming process is corrected in real time according to image deformation distortion monitored in the longitudinal submerged arc welded pipe forming steel plate pressing process, the oil and gas pipeline manufacturing process is optimized, the management and control capacity is improved, and the product quality is guaranteed.

The invention belongs to the technical field of image processing, and discloses a 4D-CBCT dynamic reconstruction method which comprises the following steps: constructing a priori motion model based on principal component analysis in an off-line manner for describing a main mode of organ respiratory motion; designing and training a deep learning network fusing a convolutional long short-term memory network and a double attention mechanism, wherein the network can learn nonlinear mapping from a two-dimensional projection image sequence to a PCA coefficient representing three-dimensional motion; in the online reconstruction stage, real-time projection is input into a network to predict a PCA coefficient, and a high-quality 4D-CBCT image sequence is quickly reconstructed through PCA inverse transformation and image deformation. According to the invention, the CBAM module enhances local features and inhibits noise in an encoder, and the SPA module carries out global context compensation after ConvLSTM. According to the method, the precision and robustness of 4D-CBCT dynamic reconstruction are improved, and complex respiratory movement can be accurately captured.

The application discloses a filtering stabilization algorithm based on a head-mounted display, and relates to the technical field of head-mounted displays; the algorithm is realized based on a head-mounted display system, and the system comprises the following modules: a data acquisition module, which acquires original image data from a high-resolution image sensor built in a head-mounted display device, and synchronously acquires data of a head posture sensor to acquire posture information of the head-mounted display device in a three-dimensional space; and a preprocessing module, which adopts an adaptive median filtering algorithm to perform denoising processing on the original image data.The application constructs accurate rotation and translation matrices, accurately transforms from a world coordinate system to a head-mounted display device coordinate system through reasonable matrix operation, guarantees correct mapping of an image in a spatial position, performs image distortion correction by using an efficient bilinear interpolation algorithm, ensures that pixel values in new coordinates can be quickly and accurately calculated when image deformation caused by head shaking is processed, effectively reduces image blur and distortion, and improves the definition and accuracy of the image.

An alignment system (1) comprises an input image generation device (10) and a learning device (20). The input image generation device (10) comprises a simulated image generation unit that generates a first simulated image indicating a target range and corresponding to a first domain on the basis of first ground surface information indicating a ground surface in the target range, and generates a second simulated image indicating the target range and corresponding to a second domain on the basis of second ground surface information indicating the ground surface in the target range. The learning device (20) comprises an image deformation unit that deforms at least one of the first simulated image and the second simulated image on the basis of the amount of positional deviation set between the first simulated image and the second simulated image.

A projected image deformation correction result evaluation method includes: obtaining corner point coordinates of a projected image, wherein the projected image is a quadrilateral, and the corner point coordinates are coordinates corresponding to four corners of the quadrilateral in a rectangular coordinate system; obtaining four included angles between four sides of the projected image and the corresponding sides of an absolute rectangle, wherein the X-axis is the horizontal direction, and the Y-axis is the vertical direction; determining values of the four included angles on the basis of the corner point coordinates; and evaluating a correction result of the projected image on the basis of the values of the four included angles.

According to the robust image watermarking, embedding and extracting method and system, the storage medium and the computer equipment provided by the invention, after the original image and the to-be-embedded watermark information are acquired, the original image is converted into the grayscale image for watermark embedding, so that the relatively high resistance of the grayscale image to geometric transformation and noise interference is effectively utilized, and the robustness of the watermark is improved. A chain reaction caused by damage of a certain channel in a traditional RGB three-channel balanced embedding method is avoided, and the robustness of the watermark is improved. Meanwhile, the multi-scale feature fusion mode enables the watermark information to be more uniformly distributed in different scale features of the image, and the stability of the watermark in a complex environment is further enhanced. In the extraction stage, after the to-be-detected image is converted into the YUV color space, geometric correction is carried out on the Y channel image in the converted to-be-detected image, image deformation caused by attacks can be accurately recovered, and complete extraction of watermark information is ensured.

The goal is to enable even someone unfamiliar with camera specifications to appropriately photograph structural defects such as narrow cracks. [Solution] An inspection device for inspecting the condition of a structure comprises: an imaging unit that photographs at least a part of the structure and generates an image; a distance measuring unit that measures the distance of at least a part of the structure and generates depth data; a storage unit that stores a reference distance, which is the distance from the imaging unit, that can image deformation of a predetermined width in the structure with a predetermined number of pixels or more; and a control unit that identifies the measured distance to the object to be photographed on the structure based on the depth data, and if the measured distance is greater than the reference distance, it makes a predetermined first notification.

The application provides a wafer image deformation correction method, which comprises the following steps: providing a wafer with a wiring pattern on a surface; obtaining a design layout and a scanning image of the wafer, wherein the design layout comprises a design pattern corresponding to the wiring pattern, and the scanning image comprises a scanning pattern corresponding to the wiring pattern; aligning the center of the scanning image with the center of the design layout, and superimposing the scanning image and the design layout; extracting the design pattern and the scanning pattern respectively, wherein the design pattern and the scanning pattern do not completely coincide; and correcting the scanning pattern according to the design pattern, so that the corrected scanning pattern coincides with the design pattern, thereby obtaining a corrected image of the test wafer, wherein the corrected image comprises the corrected scanning pattern. The application corrects the scanning image of the wafer with pattern deformation, and obtains a corrected image which is more in line with the actual situation.

The invention discloses a camshaft surface defect detection method and device, and the method comprises the steps: S1, collecting a camshaft surface image through an industrial camera, completing the preprocessing through cylindrical surface distortion correction, oil film reflection suppression and noise removal, and solving the problems of curved surface image deformation and illumination interference; s2, performing defect labeling on the enhanced image data; s3, constructing a camshaft defect detection model based on an improved YOLOv8n algorithm; s4, training a camshaft defect model by using the image data subjected to defect marking in the step S2 to obtain a trained camshaft defect detection model; and S5, shooting a camshaft defect image by using a camera, and sending the image into the trained camshaft defect detection model to complete defect detection. According to the invention, high-precision and high-efficiency non-contact detection of defects such as cracks, wear, pits and the like of the camshaft is realized, and effective technical support is provided for quality control of core parts of an automobile engine.

The invention belongs to the field of biological and ecological toxicity monitoring and aquatic organism behavior analysis, and particularly discloses an improved fish behavior analysis tank based on a camera shooting technology and a behavior parameter evaluation method. The problems of monitoring dead angles, false positive and the like caused by image deformation when a common camera is used for monitoring and identifying a fish body are solved. The device comprises a cylinder body and a partition plate assembly, the inner wall of the partition plate assembly is parallel to a camera shooting light path, the partition plate assembly is composed of a plurality of partition plates with gradually-changed thicknesses, and the thicknesses of the partition plates are gradually reduced in the camera shooting light path direction. The partition plate assembly comprises a left partition plate and a right partition plate, excrement grooves are formed in the bottoms of the left partition plate and the right partition plate, and a plurality of evenly-distributed water inlet and outlet through holes are formed in the upper middle portions of the left partition plate and the right partition plate. When the camera shooting light path is in the vertical direction, the partition plate assembly further comprises a front side partition plate and a rear side partition plate. When the camera shooting light path is in the horizontal direction, the partition plate assembly further comprises a bottom partition plate and a top partition plate.

The application discloses a kind of full-automatic left atrial myocardial strain analysis methods based on deep learning, belong to medical image processing field, comprising: constructing joint group registration and segmentation model, the joint group registration and segmentation model includes group registration encoder, segmentation encoder, inter-task attention module, group registration decoder, segmentation decoder, image deformation module and mask deformation module;Define loss function, the loss function includes registration loss, smoothing loss, segmentation loss and inter-task loss;Constructing cardiac cine image dataset, and based on the cardiac cine image dataset training and testing the joint group registration and segmentation model;For newly collected cardiac cine image sequence, the trained model is used to predict motion field sequence, inverse motion field sequence and mask sequence, and left atrial myocardial strain map, strain curve and strain rate curve are calculated.The application improves the accuracy, efficiency and repeatability of strain analysis.

The invention provides a printing extraction method and device applied to the surface of a commodity, equipment and a medium, and the method comprises the steps: carrying out image matting to obtain a printing pattern, and converting the printing pattern from an RGB color space to an HSV color space to obtain a conversion image; clustering pixels in the conversion graph by using a K-means color clustering algorithm to obtain a classification graph; determining a first control point of the classification graph; a second control point of the target image, the coordinates of the first control point of the classification image and the coordinates of the second control point of the target image are input into a TPS image deformation algorithm, and a tiled printed image is obtained; lear feature extraction is carried out on the tiled printed image by using ipadator, the feature vector and the printed image are input into the sdxl model, the required pattern is generated, the printed pattern can be rapidly and accurately extracted from any commodity image, and the efficiency and quality of related work of commodities are improved.

The application discloses a wedge mirror frosting and dewing judgment method and device based on image processing and belongs to the technical field of atmospheric optical measurement equipment. The method comprises the following steps: collecting images of stars after being split by a double-wedge mirror in real time and converting the images into gray-scale images; identifying star targets by adopting a YOLO series light model and extracting starlight spots of regions of interest; calculating three characteristic indexes of spot gray-scale uniformity, spot edge sharpness and star target contrast for each spot region of interest; obtaining an ambient temperature and a dew point temperature, combining preset judgment thresholds of the three characteristic indexes in S3, adopting a continuous frame judgment rule and judging whether frosting and dewing phenomena exist on the surface of the double-wedge mirror; and according to a frosting and dewing judgment result, controlling the start-stop and heating power of a transparent heating film on the surface of the double-wedge mirror, simultaneously monitoring image deformation states in a heating process and dynamically adjusting heating parameters. The application can realize automatic and accurate judgment of frosting and dewing of the double-wedge mirror.

The present application relates to the field of image processing and UV printing technology, and particularly relates to a UV printing control method and system based on visual positioning, the method comprising: acquiring a fixed image of a printing substrate, comparing the fixed image of the printing substrate with a standard template, generating a local deviation map of the printing substrate, and determining a deviation source area according to the local deviation map of the printing substrate; acquiring a to-be-printed image, performing image structure analysis based on image structure features of the to-be-printed image, and generating an image deformation cost map and a key structure protection mask. The present application can adapt to the actual form of the printing substrate while reducing the risk of stretching, compressing or distorting key image structures, and improving the registration stability and image structure preservation effect of the UV printed pattern near the local boundary, the periphery of the hole site and the irregular contour.

This application relates to the field of virtual try-on, and more particularly to a method for training a virtual try-on model, a method for generating try-on images, and related apparatus. The method includes constructing a virtual try-on model using a clothing distortion network and a try-on generation network. Human body keypoint images and clothing images are input into the clothing distortion network to obtain distorted clothing images. Real try-on images, distorted clothing images, and model identity images are input into the try-on generation network. By fusing semantic features from the distorted clothing feature image, the model identity feature image, and clothing attribute information, a reconstructed feature image is generated. The reconstructed feature image is decoded by a decoder to generate a virtual try-on image. The clothing distortion network and the try-on generation network are iteratively trained using a loss function until the virtual try-on model converges, resulting in the virtual try-on model. This method effectively improves the realism and quality of virtual try-on, providing better service quality and user experience for online try-on.