239 results about "Image alignment" patented technology

The invention discloses a multi-frame photoacoustic image reconstruction method based on optical flow alignment and depth feature fusion. The method comprises the following steps: S1, obtaining photoacoustic signal data; s2, reconstructing a photoacoustic cross-sectional image; s3, performing optical flow calculation and image alignment; and S4, training the deep feature fusion network. According to the method, through optical flow motion correction and a potential space learning mechanism, space-time information and complementary features in the aligned multiple frames of images are dynamically integrated, complementary information in the multiple frames of images is adaptively fused, noise is suppressed, and finally reconstruction of high signal-to-noise ratio and high spatial resolution images of biological tissues is achieved. Experimental results show that the method can significantly improve image quality, recover image distortion and detail loss caused by motion and noise, and provide a new effective scheme for promoting robust clinical application of a photoacoustic imaging technology.

The invention relates to the technical field of image processing, and discloses a fluorescent lifetime microscopic image large-view-field splicing method and system and electronic equipment, and the method comprises the steps: constructing a tissue region mask for a to-be-spliced image, and constructing a vignetting model in a tissue region to complete vignetting correction; counting brightness indexes in an organization area to determine a global brightness reference, and realizing image group brightness unification through global zooming and single image brightness adaptive correction; further realizing alignment of adjacent images through feature point matching, and finishing image block splicing in combination with a minimum color difference suture fusion method to obtain an image band; an overlapping area is extracted from an image belt to generate an effective content mask, relative displacement is estimated by using a phase correlation method after low-frequency suppression, image belt splicing is completed by multiplexing a minimum color difference suture fusion method, a large-view-field spliced image is output, the automation degree and robustness of splicing are improved, the spliced image is geometrically consistent and visually seamless, and the splicing efficiency is improved. And the requirements of medical research on high-resolution and large-field-of-view fluorescence lifetime microscopic images are met.

The invention discloses a multi-modal fine-grained semantic alignment method and device based on a graph neural network, and relates to the field of multi-modal deep learning. Firstly, deep feature extraction is performed on input multi-modal original data, and then word-level text features and local image features are constructed into a cross-modal graph structure. And performing weighted aggregation on node neighborhood information of the cross-modal graph structure through the graph attention network. And finally, carrying out weighted fusion on the text alignment features and the image alignment features. According to the cross-modal feature fusion method, the word-level text features and the local image features are uniformly abstracted into the graph structure nodes for refined alignment, a more accurate cross-modal semantic corresponding relation can be captured, the heterogeneity problem in expression modes and semantic structures is effectively relieved, and the accuracy and reliability of cross-modal feature fusion are improved. The graph attention network can adaptively adjust the weight distribution of information propagation, highlights the effect of key features in the alignment process, and ensures that the model makes full use of important semantic relationships.

The invention discloses a hierarchical three-dimensional scene generation method and system based on spatial super-division, and belongs to the technical field of computer graphics, and the method comprises the steps: carrying out the preprocessing of a scene image, and obtaining a high-resolution object image; generating initial rough scene voxels for the scene image, and screening rough voxels and structural latent variables aligned with the high-resolution object image from the initial rough scene voxels to construct a hierarchical scene tree; inputting the high-resolution object image and the rough voxel into a voxel super-resolution model, and generating a fine voxel which keeps geometric consistency with the rough voxel; performing scale alignment and attitude registration based on the rough voxels and the fine voxels; and generating fine voxels of the sub-components recursively by taking the rough voxels of the current node as conditions based on the hierarchical scene tree, and finally assembling to generate a high-resolution three-dimensional scene. According to the method, a high-quality three-dimensional scene with high visual fidelity, fine geometric details and global structure consistency can be efficiently and automatically reconstructed from a single RGB image.

The invention provides a camera pose optimization method and a point cloud and image alignment method, and relates to the technical field of computer vision and three-dimensional reconstruction. The camera pose optimization method comprises the steps of obtaining point cloud data collected by a laser radar in a target scene and a pose sequence of the laser radar when the point cloud data are collected; based on the pose sequence and external parameter data between the laser radar and the camera, camera poses corresponding to multiple frames of target images collected by the camera in the target scene are determined, and space coordinates of the same target feature points in the multiple frames of target images are determined; performing joint optimization on the camera poses and the space coordinates of the target feature points corresponding to the multiple frames of target images; and on the basis of the camera poses corresponding to the optimized multiple frames of target images, incrementally registering new images collected by the camera in the target scene, and iteratively executing the steps of determining the space coordinates of the target feature points and optimizing the camera poses for a registered image set including the new images.

The invention discloses a rigid registration method and system for visible light and infrared images of an unmanned aerial vehicle, and the method comprises the steps: gradually achieving the high-precision cross-modal image alignment through rough registration and fine registration based on a bimodal feature interaction and causal attention enhancement mechanism; therefore, a stable spatial transformation result can still be obtained under the complex conditions of low signal-to-noise ratio, local shielding and the like. According to the method, high-precision rigid registration of the infrared and visible light images of the unmanned aerial vehicle under the condition of significant modal difference and scale difference can be realized, a registration task is modeled as a long sequence dependence problem by referring to a vision mechanism of step-by-step partition comparison when people observe a large scene, and a CALS module is introduced to gradually focus a local area, so that the registration accuracy is improved. Confusion caused by information overload in traditional global modeling is avoided, and therefore balance is achieved between global scale consistency and local geometric accuracy.

The invention relates to the field of medical image processing, in particular to a cross-modal medical image alignment method based on anatomical feature one-dimensional distribution, and the method comprises the steps: obtaining a computed tomography (CT) image sequence and a magnetic resonance imaging (MRI) image sequence of a hip joint region of a patient; extracting a CT one-dimensional anatomical feature sequence representing the morphological change of the first side femoral head from the CT image sequence; extracting an MRI one-dimensional anatomical feature sequence representing the morphological change of the first side femoral head from the MRI image sequence; based on the CT one-dimensional anatomical feature sequence and the MRI one-dimensional anatomical feature sequence, a similarity alignment process is executed; wherein the similarity alignment process comprises the following steps: determining an optimal similarity alignment offset between a CT image sequence and an MRI image sequence, and aligning the MRI image sequence and the CT image sequence according to the optimal similarity alignment offset. According to the method, the cross-modal image alignment problem is subjected to dimensionality reduction into one-dimensional anatomical feature sequence matching, so that the precision, efficiency and robustness of registration are remarkably improved.

The invention discloses an anomaly detection method and system based on image alignment and feature difference analysis, and relates to the technical field of computer vision and industrial automation detection, and the method comprises the steps: obtaining a to-be-detected image and a reference image, and carrying out the image alignment processing of the to-be-detected image and the reference image; obtaining a region of interest corresponding to the reference image, wherein the region of interest is a preset region; cutting in the aligned to-be-detected image according to the region of interest to obtain a local image corresponding to the region of interest; obtaining a reference feature corresponding to the region of interest, wherein the reference feature comes from a pre-established reference feature library; extracting features of the local image as to-be-detected features; and comparing the to-be-detected feature with the reference feature to obtain a comparison result, and judging whether the target part existing in the local image is abnormal or not. The problem of dependence of offset interference and abnormal samples is solved, and the method is adaptive to multiple industrial scenes.

The invention discloses a template matching defect detection method and system for BC battery piece silk-screen printing, belongs to the technical field of defect detection, and aims at solving the problems that an existing detection method is weak in pertinence and low in positioning precision, and the general trend of defects is difficult to recognize. The method comprises the following steps: acquiring a template image, marking a calibration point and constructing a calibration data set; generating a functional area mask for the functional area of the template image, and obtaining a functional area template sub-image; an observation image is collected and cut into observation sub-images, calibration points are matched, an affine mapping matrix is generated and corrected, and a calibration observation image aligned with the template image is obtained; comparing the calibration observation image with the template image to generate an effective difference image, and determining an abnormal region; and combining a defect feature library to identify defect types, dividing grades, and retrieving abnormal functional area observation sub-graphs to judge defect general trends. According to the invention, accurate positioning, classification and trend identification of BC battery piece printing defects are realized, the detection efficiency and reliability are improved, and data support is provided for production optimization.

The invention belongs to the field of food detection, and mainly relates to a quality detection system for baked food production, which is characterized in that free format requirements such as characters and pictures submitted by customers are converted into structured production instructions through an order analysis module; a cake image is collected through a rotatable camera, a deep learning model is used for image alignment, edge detection, feature point extraction and OCR technologies are comprehensively used, and decoration elements and character areas are accurately positioned; the key points of the real object are geometrically compared with the ideal layout to automatically judge whether the pattern and the text content are correct or not and whether the placement position is accurate or not; finally, all detection results are fed back in real time in a visual mode, and the whole process is associated with order information for data storage, so that full-process automatic quality inspection, early problem interception and production quality tracing from orders to finished products are realized, and high-cost rework caused by appearance errors is effectively reduced.

Embodiments of the present disclosure disclose image alignment methods, devices, electronic equipment and computer readable media. A specific implementation of the method comprises: respectively performing distortion correction on a first image and a second image to obtain a third image and a fourth image; performing feature point matching on the third image and the fourth image to obtain a feature point information group sequence; determining whether a homography matrix corresponding to a current rotation angle exists; in response to the homography matrix not existing, determining the homography matrix according to the feature point information group sequence; and performing perspective transformation on the fourth image according to the homography matrix, so that the fourth image after the perspective transformation is aligned with the third image. This implementation does not need to set a complex hardware structure, thereby achieving efficient and accurate image alignment.

The invention relates to the technical field of image data processing, in particular to a production speed regulation and control method and system for packaging bag printing. The method comprises the following steps: acquiring and segmenting a printed packaging bag image, and correcting rotation offset by using a histogram of oriented gradients to ensure image alignment. Then, the corrected image is converted into an LAB color space, the LAB color space is compared with a preset standard template, the ink color unevenness is calculated, and the printing quality is quantified. And finally, a relation model of the printing quality and the production speed is constructed based on historical data, the target production speed is determined, regulated and controlled, and balance of the printing efficiency and the quality is achieved. According to the invention, the influence of artificial experience dependence and image dislocation on quality evaluation is overcome, and the accuracy and stability of printing speed regulation and control are improved.

The invention discloses a target detection method and device and electronic equipment, and relates to the technical field of image target detection, and the method comprises the steps: carrying out the image alignment of a first image and a second image, and carrying out the frame difference analysis and optical flow analysis of an image alignment result, and obtaining an initial frame difference graph and an initial optical flow graph; filtering the initial frame difference graph and the initial optical flow graph based on a mask matrix to obtain a target frame difference graph and a target optical flow graph; carrying out image interception on the first image, the target frame difference graph and the target optical flow graph based on a historical prediction result and a frame difference analysis result; and dynamically splicing the first image, the target frame difference graph, the target optical flow graph and sub-images obtained by image interception to obtain a target image, and inputting the target image into a detection model for target detection. The technical problems of low detection accuracy and high detection cost of the moving target in the low-altitude scene in the prior art are solved.

The invention discloses a task-aware robot availability prediction method based on retrieval enhancement and cross-image alignment. The method comprises the following steps: constructing an accessibility memory bank containing object appearance characterization and interaction prior; performing semantic filtering on the memory bank based on the task tag, and retrieving a reference sample according to appearance similarity; a prediction process is divided into two parts, namely static contact point positioning and dynamic action direction prediction, contact points are correspondingly positioned by querying dense features of an image and a reference sample, and an action direction is regressed by a retrieval-enhanced cross-image alignment model; the model performs action vector modulation on the reference features, and performs gating cross-attention fusion with the query image features, and the fusion weight is jointly determined by the similarity and the learnable weight; and finally, the two-dimensional contact point and the action direction are lifted into a three-dimensional pose and a three-dimensional trajectory, and executable grabbing and post-contact actions are generated. According to the method, cross-category robust generalization under the condition of few samples can be realized, and the operation success rate of the robot is improved.

The application discloses an expert knowledge guided ovarian mass ultrasound standard section acquisition method and system, and belongs to the technical field of image acquisition, which comprises the following steps: pouring field expert knowledge and text-image alignment knowledge into a knowledge enhanced large model, then generating an initial maximum diameter standard section of an ovarian ultrasound mass by using a constructed Bi-SwinGRU model, taking the initial maximum diameter standard section as a reference, recognizing uncollected image features based on text-image alignment semantics of real-time imported ultrasound section images, and outputting feature collection prompts in a text form by combining expert knowledge and thinking based on the uncollected image features, so as to guide the collection of the maximum diameter standard section, and realize accurate collection of the maximum diameter standard section of the ovarian mass.

An image inspection apparatus (100) that compares a read image generated by reading a recording medium on which an image is formed with a reference image to detect misregistration of an image formation position on the recording medium, the apparatus including a controller (81) that: aligns an image included in the read image and an image included in the reference image; and detects the misregistration of the image formation position based on a deviation between a reference position in the read image and a reference position in the reference image in an aligned state. The controller selects a reference position used for detecting the misregistration from reference positions included in the read image and the reference image based on a position of an area in which an image is formed on the recording medium.

The invention belongs to the technical field of medical image registration, and discloses an unsupervised registration method and related device for multi-task medical image alignment, and the method comprises the steps: collecting a floating image and a reference image to be registered, and inputting the floating image and the reference image into a pre-trained medical image registration model to output a registration deformation field; the model adopts a multi-level pyramid encoder-decoder framework, and integrates a multi-scale fusion module, a local-global attention module and a feature interaction and fusion module. By adopting the method, the problems of high calculation cost, cooperative imbalance of global and local features, weak feature interaction, low multi-scale fusion efficiency and limited cross-domain generalization ability in the prior art are effectively solved, the registration precision and robustness are remarkably improved, the adaptability to complex deformation and cross-modal scenes is enhanced, and the method is suitable for being popularized and applied. The method optimizes the feature interaction and multi-scale fusion efficiency, improves the cross-equipment and cross-center generalization ability, meets the requirements of clinical high precision, high robustness and strong generalization, and is suitable for various registration application scenarios.

The invention discloses a cotton field disease and pest control pesticide spraying amount prediction system based on an unmanned aerial vehicle multispectral remote sensing image, and belongs to the field of cotton field disease and pest control pesticide spraying amount prediction, and the system comprises a data acquisition module, an image alignment module, a panoramic spectral image generation module, and a disease and pest recognition and pesticide spraying amount prediction module. The data acquisition module controls the unmanned aerial vehicle to acquire a cotton field multispectral remote sensing image, records an image space-time position through a GPS and an IMU, and collects image processing modes in different historical weathers; the image alignment module aligns the image with the shooting time, divides small areas of the cotton field and associates real-time weather; the panoramic spectral image generation module screens high-quality images and splices the high-quality images into a cotton field panoramic image; the disease and insect pest recognition module recognizes disease and insect pests through deep learning, quantifies the stress degree and generates a variable pesticide application prescription map; the system can improve image quality and processing efficiency, realizes accurate pesticide spraying amount prediction, reduces pesticide waste, and effectively solves the problem of low pesticide application accuracy in traditional cotton field pest and disease control.

Methods and systems for image alignment are provided. In one embodiment, a method includes individually aligning candidate alignment target images in an image generated for a sample to corresponding set images of set alignment targets selected to have mutually linearly independent vectors between a position of the set alignment targets and a reference position. The method also includes, for any of the candidate images successfully aligned to their corresponding set images, individually determining coordinates of the reference position from coordinates of the aligned candidate images and their corresponding mutually linearly independent vectors. In addition, the method includes determining final coordinates of the reference position in the image generated for the sample from the individually determined coordinates of the reference position.

According to the microscope time sequence scanning image splicing method based on global position optimization, relative displacement is calculated through a phase correlation method, global position optimization is carried out in a spatial dimension, and a global position parameter with the minimum error is obtained. And then aligning the image at each moment to the global optimal position of the first frame by using a feature point-based time sequence alignment method to realize the minimization of time sequence jitter. The method can eliminate accumulative errors, improve splicing precision and reduce artifacts in dynamic analysis, and is suitable for the field of microscope image processing and splicing.

This application relates to an image alignment control method, apparatus, computer device, and storage medium. The method is applied to a smart device equipped with a first camera and a second camera, and includes the following steps: after entering a powered-on state, controlling the second camera to extend via a connecting rod; rotating the first and second cameras by an angle, and controlling the first and second cameras to capture images at each angle to obtain corresponding images; comparing the images captured by the second camera with the images captured by the first camera at the corresponding angles; determining the first deviation data for each second camera at each angle based on the image comparison results, and correcting and aligning the images captured by the second camera based on the first deviation data. This solution can achieve reliable image alignment, thereby ensuring the imaging effect of the device.

The invention relates to the technical field of image data processing, in particular to a vehicle-mounted thermal imaging target AI identification method supporting aerial imaging display. The method comprises the following steps: acquiring a thermal imaging video stream and vehicle speed data shot in a vehicle driving process; identifying each suspected pedestrian area of adjacent image frames in the thermal imaging video stream, and calculating the area similarity of each corresponding suspected pedestrian area of the adjacent image frames; performing picture consistency matching on the adjacent image frames to enable the adjacent image frames to complete image alignment processing; reflecting the change trend of the whole picture by using the vehicle speed data, and obtaining a pedestrian body region and a pedestrian limb region in the aligned picture by combining the region similarity of each suspected pedestrian region in the aligned picture; and determining whether pedestrian features exist in the aligned picture or not, and marking and displaying the pedestrian features when the pedestrian features exist in the corresponding region. Based on the technical scheme of the invention, accurate pedestrian recognition can be realized.

A method for aligning multiple depth cameras in an environment based on image data can include accessing, by one or more processors, a first plurality of point cloud data points corresponding to a first pose for a subject and a second plurality of point cloud data points corresponding to a second pose for the subject. The method can include determining, by the one or more processors, a frame of reference for the image data based on at least one of the first pose or the second pose. The method can include transforming and aligning, by the one or more processors, at least one of the first plurality of point cloud data points or the second plurality of point cloud data points to the frame of reference.

The application discloses a kind of measurement method and system of tire belt steel wire angle, including characteristic mark step: in the A surface and B surface of tire section, respectively mark tread pattern design width point, and confirm corresponding endpoint by on-off detection, mark as steel wire corresponding point;Image acquisition step: respectively to the image acquisition of tire section A surface and B surface of completion mark, obtain image;Image registration step: based on tread pattern design width point, A surface image and B surface image are registered in space, coordinate extraction step: from the A surface image and B surface image after registration, based on the mark of steel wire corresponding point, extract position coordinate;Angle calculation step: according to position coordinate, the angle of this steel wire is calculated.The application extracts steel wire coordinate by collecting section image and carrying out image alignment based on tread pattern design width point to calculate angle, realizes non-contact high-precision measurement, significantly improves measurement efficiency and data traceability.

A system and technique may be used to detect whether a skimmer is present in a card reader of an automated teller machine (ATM). An example technique may include capturing an image of a portion of the card reader, generating a set of key points of the image, and aligning key points of the set of key points of the image to corresponding key points in a baseline image to create an aligned image. The example technique may include comparing the aligned image to the baseline image to determine whether a skimmer is present, and outputting a result of the determination.

To provide guidance and feedback which makes early detection of relative movement between a patient's position and / or anatomical features associated with the patient and an imaging device, and simultaneously corrects difference.SOLUTION: Disclosed embodiments determine, at an early stage, suitability of an intraoperative image for further intraoperative surgical analysis. The determination of suitability may be made by using a first angle (such as a first obturator angle) based on at least three pelvic feature points in a preoperative image and a corresponding second angle (such as a corresponding second obturator angle) based on at least three corresponding pelvic feature points in an intraoperative image, and by comparing the first angle and the corresponding second angle to determine intraoperative image suitability. The first intra-operative image is indicated as suitable for further intraoperative analysis when an absolute value of a difference between the first angle and the corresponding second angle does not exceed a threshold. When the intraoperative image is determined as unsuitable for further intraoperative analysis, an indication of a movement direction for a fluoroscopy camera used to capture the intraoperative image is provided.SELECTED DRAWING: Figure 25

The invention relates to an intrusion judgment method and device based on a radar track and a visual track, computer equipment and a medium. The intrusion judgment method comprises the following steps: encoding radar point cloud data into a radar feature map aligned with a camera image through voxel reconstruction; performing fusion visual analysis based on the radar feature map and the camera image to obtain a visual track of the target; performing clustering analysis on the radar point cloud data to obtain a radar track of the target; cross-modal matching is carried out on the radar track and the visual track to obtain a matching point pair, and the matching point pair comprises the radar track and the visual track of the to-be-detected target; based on the radar trajectory, obtaining a first intrusion result of the to-be-detected target in a three-dimensional space of the preset area, and based on the visual trajectory, obtaining a second intrusion result of the to-be-detected target in a two-dimensional space of the preset area; and when both the first intrusion result and the second intrusion result represent that the to-be-detected target has the intrusion behavior, determining that the to-be-detected target has the intrusion behavior. The method can accurately judge whether the target to be detected has the intrusion behavior or not.

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 discloses a multi-modal feature alignment method and device for heterogeneous data and a medium, and relates to the field of intelligent retrieve.The method comprises the steps that an input multi-modal document is analyzed, and text elements and non-text elements are recognized; extracting original text features, original image features and original structured text features; constructing a shared semantic space, mapping the original image features to the shared semantic space through an image mapping network, and mapping the original text features and the original structured text features to the shared semantic space through a text mapping network; based on the mapping features in the shared semantic space, through a predefined multi-objective optimization function, executing multi-objective joint optimization to drive feature alignment; and generating a query vector based on the aligned image alignment feature and text alignment feature. By constructing a unified shared semantic space and utilizing multi-target joint optimization to drive feature alignment, it is ensured that original information of non-text elements is completely reserved in the feature mapping process.

The invention discloses an alignment device for tinned PCB laser imaging and a use method. The alignment device comprises a workbench, and first alignment plates which are symmetrically arranged are arranged at the positions, corresponding to the long edges, above the workbench; the second alignment plates which are symmetrically arranged are arranged at the positions, corresponding to the short edges, above the workbench. A moving plate is arranged below the second alignment plate, first sliding plates which are symmetrically arranged are connected to the position, corresponding to the moving plate, below the second alignment plate, and a supporting plate is connected to the position, away from the second alignment plate, of one side of the moving plate; the lead screw drives the adjusting plate to move on the outer side of the lead screw in the rotating process, the adjusting plate can drive the first sliding plate and the second alignment plate to move to be close to or away from the middle of the workbench in the moving process, and therefore the PCBs of different proportions can be aligned; consecutive operation can be achieved, and therefore efficiency can be improved.