499 results about "Binocular stereo" patented technology

The invention discloses a visual detection system for an automobile door panel assembly, and relates to the technical field of automatic detection, and the visual detection system comprises the steps: installing a high-precision industrial camera, and carrying out the synchronous collection of the multi-view image data of an automobile door panel; preprocessing the acquired multi-view image data of the automobile door panel; carrying out detail enhancement on the preprocessed multi-view image data of the automobile door panel by adopting a super-resolution algorithm; integrating the enhanced multi-view image data of the automobile door panel into an automobile door panel panorama through binocular stereoscopic vision matching, and constructing an automobile door panel three-dimensional model; the automobile door panel three-dimensional model is compared with a standard automobile door panel three-dimensional model, and the buckle defect in the automobile door panel is recognized; and carrying out position marking and defect type analysis on the identified defect points, and generating an automobile door panel detection report according to the defect type analysis. According to the invention, details are enhanced by using a super-resolution algorithm, defects are accurately positioned by using a three-dimensional model, and the detection efficiency and accuracy are improved.

The invention discloses a double-station robot sorting optimization method and system based on digital twinning and a terminal, double improvement of sorting efficiency and safety is achieved by constructing a digital twinning driven collaborative operation system, and the method comprises the steps that firstly, a laser radar and a polarization camera are used for forming a composite sensing unit; geometric morphology, material reflection characteristics and spatial pose data of a target object are synchronously obtained, and are input into a digital twin engine after time synchronization processing; an engine constructs a virtual sorting scene containing a material attribute database based on a physical rendering technology, sub-millimeter-level space registration is achieved through feature fusion of a binocular stereoscopic vision depth map and geometric parameters, high-precision digital twin stations are generated, a system plans a space-time constraint trajectory of a double-station robot in a virtual environment, and a target object is obtained. And an integrated discrete event simulation engine performs operation time sequence conflict prediction, and when a space overlapping risk is detected, an alternative path containing a dynamic obstacle avoidance strategy is generated through a trajectory re-planning algorithm.

The invention discloses a multi-modal fusion real-time environment monitoring visual robot system, and relates to the technical field of real-time vision. The system comprises a multi-mode sensing module, and is equipped with various sensors such as a binocular stereo camera and a laser radar to collect environment data. The heterogeneous data preprocessing unit cleans and downsamples data such as images and point clouds; the space-time alignment fusion module realizes multi-source data space-time registration and synchronization; the environment semantic understanding engine constructs an environment semantic graph through a multi-branch network in combination with an attention mechanism; the abnormal event detection unit identifies abnormity based on a historical data model; the path planning and decision-making module integrates multiple targets to generate an optimal path; the autonomous movement execution module controls the robot to move and operate; and the cloud cooperative control center supports model updating and remote intervention. According to the invention, through cooperative work of all the modules, full-process intelligentization of environment monitoring data acquisition, processing, analysis and decision making is realized.

A slope monitoring method based on optical flow estimation and binocular vision belongs to the technical field of deformation monitoring and measurement, and adopts the technical scheme that a binocular stereoscopic vision system is built, and camera acquisition parameters are calibrated; transmitting the image to a server in real time through the Internet of Things; distortion correction and stereo matching are carried out to obtain a depth map; monitoring points are manually or automatically arranged on the corrected image, and dense optical flow estimation tracking pixel point movement is carried out; world coordinates of observation points and movement conditions of each frame are obtained, and displacement changes of slope monitoring points are obtained by combining coordinates before and after displacement; and obtaining a displacement rate by combining monitoring time and displacement, judging slope safety, obtaining a displacement track, and performing linear interpolation to obtain a slope deformation cloud picture, thereby realizing the monitoring purpose. The slope displacement monitoring method has the advantages that slope monitoring can be carried out more flexibly and efficiently, the safety of workers is improved, the slope monitoring cost is reduced, real-time slope displacement monitoring is achieved, and the intensive degree of slope monitoring points is greatly improved.

The invention belongs to the technical field of plate shape monitoring and visual analysis, and discloses a stainless steel medium-thickness plate double-face contour measurement method and system, and the method comprises the steps: S1, completing the calibration of a double-line-scan digital camera, and obtaining the upper and lower surface images of a medium-thickness plate; s2, carrying out binocular stereoscopic vision three-dimensional reconstruction to generate a point cloud picture, and converting the point cloud picture into a point cloud picture under the same coordinate system for display; s3, eliminating transverse rotation and deviation according to point cloud center line fitting and plane fitting of the upper surface and the lower surface of the medium-thickness plate; s4, comparing the geometrical relationship between the point clouds of the upper and lower surfaces to quantify local vibration and thickness change, correcting vibration errors, and obtaining the thickness of the medium-thickness plate; and S5, extracting an edge point set from the point cloud image to obtain the overall contour of the medium-thickness plate, and obtaining the position of a shear line based on the width and edge linearity detection of the medium-thickness plate. By the adoption of the technical scheme, the overall contour, overall thickness distribution and accurate shearing position of the medium-thickness plate can be obtained.

The invention provides a structural deformation monitoring method and system based on binocular vision and coordinate transformation, and the method comprises the steps: carrying out the calibration of a binocular stereoscopic vision left camera and a binocular stereoscopic vision right camera based on a calibration plate, and obtaining a plurality of calibration plate images of the calibration plate under different postures; collecting a target structure image based on a binocular stereoscopic vision camera; calibration parameter calculation is carried out based on the plurality of calibration plate images, and a rigidity conversion matrix when the calibration plates coincide with the target structural surface is obtained; coordinate transformation is carried out on the measuring points of the target structure image based on the stiffness transformation matrix; and calculating in-plane displacement and out-of-plane displacement of the target structure based on the converted three-dimensional coordinates. According to the method, the 2D checkerboard target is utilized, the three-dimensional world coordinate system located on the left camera system is converted into the three-dimensional world coordinate system on the structural surface through coordinate conversion, the morphological characteristics and displacement information of the structural surface are more visually shown, the calculated amount is small, the precision is high, the distance measuring range is large, and the cost is low.

The invention relates to the technical field of agricultural machinery control, in particular to a multi-variety vegetable harvester dynamic identification and feeding control system based on image processing, which comprises an image acquisition module, an environment sensing module, an image processing module, a data fusion module, a dynamic identification and tracking module and a feeding control module. Through binocular stereo vision and millimeter wave radar fusion, three-dimensional space positioning and dynamic tracking of the brassica oleracea corm are realized. Track prediction is carried out in combination with extended Kalman filtering and a long-short-term memory network, and the clamping path of the mechanical arm is optimized; the clamping force is adaptively adjusted through fuzzy PID control, and accurate feeding of the cabbages is achieved in combination with speed control of the conveying belt. The automatic cabbage harvesting machine improves the automation level of cabbage harvesting, effectively reduces the damage rate, improves the operation stability and harvesting efficiency, and is suitable for intelligent harvesting operation of a large-scale cabbage planting base.

The invention discloses an ice surface defect detection and quantification system and method based on deep learning. The system comprises a data acquisition module, a data enhancement module, a neural network training module, a binocular stereoscopic vision-based detection module and an information fusion and analysis module. The method comprises the following steps: firstly, acquiring image data of ice surface defects by using various camera devices through a data acquisition module, and covering different illumination conditions and defect types; and then, the data enhancement module performs rotation, overturning, contrast adjustment and noise addition on the acquired data so as to improve data diversity and generalization ability of the neural network. And then, a multi-scale feature extraction module (MSTG-Net) and a detail enhancement convolution module (MSDSC) are introduced into a neural network training module, and the segmentation feature extraction of the ice surface defect is realized by training the improved YOLOv8 neural network. The detection module based on binocular stereo vision acquires an ice surface image through a binocular camera, and obtains defect mark information and a depth map in combination with the trained neural network model. And finally, further determining the three-dimensional position, the surface area size and the spatial morphological characteristics of the defect by combining the information fusion and analysis module with a binocular stereoscopic vision distance measurement result. According to the method, high-precision identification and positioning of ice surface defects are realized, reliable data support is provided for subsequent ice surface repair planning, and the method has the advantages of high precision, strong robustness, wide applicability and the like.

The invention discloses a transcranial magnetic stimulation target region recommendation method based on craniocerebral position estimation. The method comprises the following steps: constructing a face key point cloud based on binocular stereo vision; reconstructing the face key point cloud into a cranial surface model by using a pre-trained generative model; the method comprises the following steps: collecting magnetic resonance imaging data of patients with various diseases, establishing a transcranial magnetic stimulation target region standardized template with disease specificity, and determining a stimulation target region of each disease in a standard space; and based on the transcranial magnetic stimulation target region standardization template, mapping the corresponding stimulation target region from the standard space brain region to the scalp of the patient according to the disease type of the patient so as to realize recommendation. According to the method, the graph neural network is innovatively combined with the SHAP algorithm, the scientificity and interpretability of target spot selection are improved, important brain region target spots are obtained by calculating the contribution value of classification model features to classification decision of each sample, and then intervention target spots are sorted and selected to obtain an optimal treatment target spot recommendation scheme.

A novel knowledge distillation method from binocular parallax to monocular depth comprises the following steps: S1, training a parallax depth conversion module: inputting first training data into a pre-trained binocular stereo parallax teacher model to obtain a parallax map; the first training data comprises a left eye color image and a right eye color image; inputting the disparity map and the original disparity reference map into a disparity depth conversion module to obtain a pseudo label depth map; calculating a loss function according to the pseudo-label depth map, and updating parameters of a parallax depth conversion module; s2, mixing the pseudo-label depth map output by the parallax depth conversion module with the original true label depth map to obtain a mixed label depth map; s3, training a monocular depth estimation student model: training the monocular depth estimation student model through the second training data; the second training data comprises a mixed label depth map and a corresponding original reference map, and the method can achieve the training on a data set with a lower data size, and improves the global precision.

The invention discloses a pavement disease size accurate quantification method based on binocular vision, and the method specifically comprises the steps: collecting pavement image data through a vehicle-mounted binocular camera, achieving the real-time detection of a disease target through an improved RT-DETR model, and outputting the disease type and detection frame information; after a pavement disease target is detected, pixel-level segmentation is carried out on a disease area in the detection frame based on a semantic segmentation model, and a disease contour mask is extracted; an improved IGEV-Stereo stereo matching algorithm is used for calculating a disparity map, depth information is output based on camera calibration parameters and a disparity estimation result, and conversion from pixel coordinates to three-dimensional coordinates is achieved; and the binocular depth information and a disease detection segmentation result are combined to realize accurate size quantification of typical road surface diseases with different characters. Through the binocular stereoscopic vision technology, automatic detection and precise quantification of pavement diseases can be realized, the cost can be effectively reduced while the disease quantitative evaluation precision is improved, and decision support is provided for road management and maintenance.

The invention belongs to the field of helicopter rotor blade motion parameter measurement, and relates to a blade deformation measurement method based on rotation-following vision. The method comprises the steps that LED mark points are installed on the upper surfaces of the paddles; a follow-up rotating binocular stereoscopic vision system used for shooting the upper surfaces of the blades obliquely and downwards is mounted above the propeller hub; images of the LED mark points are collected through a rotation-following binocular stereoscopic vision system according to the test state; identifying pixel coordinates of the LED mark points in the image; according to an internal reference matrix, a rotation matrix and a translation matrix, obtained through pre-calibration, of the following-rotation binocular stereoscopic vision system, the pixel coordinates of the LED mark points are converted into three-dimensional coordinates under a coordinate system of the following-rotation binocular stereoscopic vision system; and converting the three-dimensional coordinates of the LED mark points in the coordinate system of the follow-up rotation binocular stereoscopic vision system into the three-dimensional coordinates of the LED mark points in the coordinate system of the propeller hub, thereby calculating the deformation of the propeller blade.

The invention provides an OPM-MEG and MRI registration method and system based on a binocular camera, and the method comprises the steps: firstly carrying out the calibration of the binocular camera, and obtaining a calibration parameter through a checkerboard calibration method; then, binocular images of the testee wearing the brain magnetic measurement device are collected; performing three-dimensional reconstruction, obtaining an initial disparity map through epipolar correction and an SGBM binocular stereo matching algorithm, optimizing the disparity map to generate a three-dimensional point cloud, and performing noise reduction and smoothing processing by using MLS and Poisson surface reconstruction technologies; in the face point cloud registration stage, point cloud is cut by taking a nose tip point as a sphere center, ICP registration is carried out, and a transformation matrix from a camera coordinate system to an MRI coordinate system is obtained; and finally, in a helmet positioning stage, detecting a three-dimensional coordinate of a ball target on the helmet, calculating a transformation matrix from a helmet coordinate system to a camera coordinate system, and unifying the helmet coordinate system to an MRI coordinate system. According to the invention, a one-button registration process can be realized, the registration time is shortened, the registration precision is improved, and a basis is provided for improving the source positioning precision.

The invention discloses a water level measurement method based on binocular stereoscopic vision. The method comprises the following steps that model selection is conducted on a binocular camera according to an actual measurement scene; horizontally erecting a binocular camera on a river bank, adjusting a pitch angle to enable left and right views to cover a water surface area under a water level full range, and synchronously acquiring water surface images of the left and right views by using the binocular camera; acquiring a corresponding disparity map pixel by pixel by adopting an iterative optimization semi-global stereo matching method; based on a calibration result of the binocular camera, obtaining a three-dimensional point cloud coordinate value of the water surface image by using a relationship between parallax and depth; performing plane fitting on the point cloud by using an RANSAC algorithm to generate a horizontal plane equation; calculating the distance from the optical center of the left camera to the horizontal plane, and converting the elevation of the camera to obtain the water level. According to the invention, the shore-based arrangement of the instrument can be realized, the water level can be measured remotely in an inclined manner, the measurement does not depend on any reference object in contact with a water body such as a water gauge, the frame-while-frame measurement can be realized in severe environments such as a high flood period, and the device is particularly suitable for water level emergency monitoring.

The invention provides a macrobrachium rosenbergii quality grading system based on machine vision. The macrobrachium rosenbergii quality grading system based on machine vision comprises a prawn body placing platform used for placing macrobrachium rosenbergii to be tested; the three-dimensional acquisition module comprises a structured light projection device, a binocular stereo camera and a flight time depth sensor; and the three-dimensional reconstruction module is used for constructing the data of the three-dimensional acquisition module into a complete three-dimensional point cloud model through a data fusion algorithm and a point cloud registration algorithm. According to the macrobrachium rosenbergii quality grading system based on machine vision, the overall scheme has robustness and expandability, the system is suitable for various large-scale farms and grading scenes, and the intelligent level of macrobrachium rosenbergii quality control is greatly improved.

The invention discloses a power transmission line reconstruction method combined with multi-parallel-line laser, and belongs to the field of power transmission line modeling, and the method comprises the steps: determining a target power transmission line region, and carrying out the three-dimensional scanning of a multi-parallel-line laser scanner, and obtaining the point cloud data of the region; shooting a plurality of groups of binocular stereo images by using a binocular camera to obtain regional image data; establishing a registration coordinate system, and registering the point cloud and the image data to obtain an initial three-dimensional model of the power transmission line; constructing a key component set, and carrying out geometric recognition and segmentation on the models to obtain a plurality of key component models; and optimizing the plurality of key component models to obtain a plurality of optimized component models, and performing model optimization on the initial three-dimensional model of the power transmission line to obtain an accurate three-dimensional model of the power transmission line. The technical problems of low precision and poor efficiency of three-dimensional model reconstruction of the power transmission line in the prior art are solved, and the technical effects of improving the three-dimensional model reconstruction precision of the power transmission line and realizing efficient and accurate three-dimensional reconstruction of the power transmission line are achieved.

The invention relates to a binocular stereoscopic vision three-dimensional reconstruction method and system based on a polarization state. The method comprises the following steps: acquiring polarization images of a polarization camera under different polarization degrees; according to the obtained polarization image, the angle of a normal vector is calculated based on the Fresnel theory, and the angle of the normal vector comprises the zenith angle and the azimuth angle of the surface of the object; the ambiguity of the azimuth angle is eliminated through the projection included angle of the normal vector on the YOZ plane, and mismatching is eliminated by using the angle constraint of the normal vector; and performing gradient integration on the angle of the normal vector after mismatching elimination to complete three-dimensional reconstruction. According to the method, the projection included angle is utilized to assist in judging the quadrant of the normal, so that mismatching of the corresponding points is eliminated, and the measurement efficiency is improved.

The invention relates to a vehicle-mounted road defect detection system and method based on binocular vision and deep fusion, and the system comprises a binocular camera module, a main control platform, a positioning module, a communication module and a power supply control module. The system realizes automatic detection and severity quantitative evaluation of various pavement defects such as cracks, pits and ruts, obtains defect positions in combination with a positioning module, and uploads the defect positions to a cloud platform in real time through a communication module, thereby forming a closed-loop urban road defect information perception and management system. According to the method, the quantitative evaluation capability of the identification accuracy and severity of the road defects is remarkably improved, and automatic identification, accurate positioning and real-time cloud synchronous uploading of the road defects are realized; the method has the advantages of flexible deployment, efficient operation, high adaptability and the like, is particularly suitable for urban road intelligent inspection and maintenance management scenes, and has good engineering application prospects and popularization values.

The invention relates to the technical field of computer vision, and particularly provides a tunnel video stream three-dimensional modeling system based on binocular stereo matching and SLAM (Simultaneous Localization and Mapping), which comprises a multi-modal image acquisition unit, a binocular infrared camera and an RGB (Red, Green and Blue) camera are configured to synchronously acquire an infrared image pair, an RGB image pair and inertial measurement unit data of a tunnel environment, the multi-source data is uploaded to the cloud processing platform in real time through the wireless transmission module; based on a dynamic switching mechanism of environmental perception, performing adaptive fusion pose estimation of a feature point method and a direct method on input multi-modal image data; the stereo matching and dense reconstruction unit is used for calculating a disparity map by adopting an improved self-adaptive window stereo matching algorithm, generating a three-dimensional point cloud in combination with the pose information output by the pose estimation unit, and constructing a global consistency point cloud model of the tunnel scene through a time sequence point cloud registration and splicing algorithm; and a multi-algorithm target detection and semantic fusion unit. The method can meet the requirements of precision and robustness of tunnel reconstruction.

The invention provides an intelligent and accurate fruit picking method based on visual perception, relates to the field of fruit picking, and adopts a picking robot comprising a movable chassis (1), a walking mechanism (2), an industrial computer (3), a collecting frame (4), a picking mechanical arm (5), an end effector (6), a monocular vision camera (7), a binocular stereo camera (8) and a blowing assembly (9). According to the picking robot, through recognition of the binocular stereo camera (8), by means of blowing of the blowing assembly (9) and through a timestamp, the relative motion state between a picking point and sheltered fruits after blowing is obtained, and therefore sensing and precise clamping and shearing of the picking point in the motion state are achieved under the condition that the fruits are sheltered and subjected to complex motion after sheltering is removed; the picking efficiency is improved, the equipment damage probability is reduced, and the harvesting cost is controlled.

The invention provides a binocular stereo matching method based on matching cost aggregation. The binocular stereo matching method comprises the following steps: step 1, selecting an image; 2, constructing a parallax cube; 3, selecting a plurality of paths, and updating the parallax cube; 4, determining an optimal disparity value to obtain a final disparity map; according to the binocular stereo matching method based on matching cost aggregation, the real-time problem of three-dimensional reconstruction can be solved, and the binocular stereo matching method is mainly used for application scenes with high requirements for three-dimensional reconstruction efficiency, such as automatic driving. According to the method, the platform portability is considered, the resolving efficiency is emphasized, meanwhile, the excellent reconstruction quality is guaranteed, and the practicability is high.

The invention provides a multi-modal image fusion endoscope operation target positioning device and method, and the method comprises the steps: calibrating internal and external parameters of a binocular camera and a fluorescence camera, and obtaining a pose conversion matrix between the fluorescence camera and the binocular camera; using a U-Net network model to identify and track the position of the fluorescent mark point in the fluorescent image in real time, and obtaining the two-dimensional image coordinate of the fluorescent mark point; processing a left image and a right image of the binocular camera by using a region-based local stereo matching method in binocular stereo imaging to obtain a point cloud image of a target space; and obtaining three-dimensional pose information of the fluorescent mark point by combining the two-dimensional image coordinate of the fluorescent mark point and the point cloud image of the target space. The binocular stereoscopic vision and the fluorescence imaging technology are combined, the three-dimensional pose and depth information of the fluorescence mark point is obtained by registering the information between the fluorescence camera and the binocular camera, more accurate and real-time vision assistance is provided for an operation, and the safety and efficiency of the operation are improved.

The invention relates to the technical field of intelligent management, in particular to a garden nursery stock intelligent management method based on stereoscopic perception, and the method comprises the steps: obtaining the three-dimensional point cloud data of a garden based on a binocular stereoscopic vision technology, analyzing the distribution of oriental cherry trees in the garden through combining with an image segmentation and clustering algorithm, and obtaining each oriental cherry management region; the method comprises the following steps: acquiring environment data and image data of each region, and calculating a regional oriental cherry moisture demand coefficient according to first oriental cherry data; secondly, constructing an oriental cherry anomaly recognition model, processing regional image data by using a deep convolutional neural network, recognizing regional pest anomaly features and calculating regional oriental cherry anomaly coefficients; and finally, according to the moisture demand coefficient and the regional oriental cherry abnormal coefficient, comprehensively calculating a regional oriental cherry scheduling urgent degree coefficient, and further realizing intelligent resource allocation and scheduling according to the regional oriental cherry scheduling urgent degree coefficient. According to the invention, intelligent scheduling of garden resources can be realized, so that the garden management efficiency is improved.

The invention provides a power distribution construction safety distance monitoring method, device and equipment and a storage medium, and the method comprises the steps: inputting an obtained power distribution operation scene image into a stereo matching neural network model for binocular stereo matching calculation, outputting a three-dimensional information disparity map, and converting the three-dimensional information disparity map into a power distribution operation depth map; based on the power distribution operation scene image and the target detection neural network model, key operation elements and two-dimensional coordinate information are determined; screening the depth values of the pixel points in the key job element target frame to obtain image coordinates of the screened pixel points; performing coordinate system conversion on the image coordinates of the screened pixel points and the depth values to obtain initial three-dimensional point cloud data of the key job elements; clustering and screening the initial three-dimensional point cloud data to obtain effective three-dimensional point cloud data; and determining a first nearest point coordinate, a second nearest point coordinate and a third nearest point coordinate from the effective three-dimensional point cloud data, and determining a safety distance among the operator, the construction machinery and the electrified equipment.

The present invention relates to a vehicle control method, a binocular stereoscopic vision-based road unevenness feature detection method, system and device, and a computer readable storage medium. The binocular stereoscopic vision-based road unevenness feature detection method comprises the steps: S1, road surface unevenness feature detection; S2, generating a binocular disparity point cloud; S3, feature region point cloud projection; and S4, region feature calculation: calculating unevenness information of a corresponding region point cloud. The vehicle control method comprises the steps: T1, performing ego-vehicle trajectory prediction on the basis of the motion state of the vehicle; T2, on the basis of the predicted ego-vehicle trajectory, calculating a degree of correlation to a corresponding region point cloud; and T3, sending to an ADAS controller the concavity and convexity and height position of the corresponding region point cloud and the acquired degree of correlation as summarized information of each road unevenness feature. The present invention can effectively identify road unevenness information, reduce calculation resource consumption, and thus improve the comfort of intelligent driving.

The invention provides a cross-domain binocular stereo matching method based on multi-scale information dynamic fusion and feature deviation correction, and belongs to the field of computer vision. Specifically, the model adaptively fuses a convolutional neural network (CNN) and a Transform structure according to dynamic change of a scene so as to capture local detail information and global context dependence at the same time. Secondly, through a multi-stage progressive cost body fusion and aggregation mechanism, full fusion of multi-scale matching cost bodies is promoted, redundant information is effectively inhibited, and matching precision is improved; and finally, guiding the model to learn more stable domain invariant representation by gradually reducing the deviation between different domain features, thereby remarkably enhancing the cross-domain robustness and adaptability. Compared with the most advanced method for performing cross-domain experiments in different scenes, the method disclosed by the invention shows better performance compared with an existing advanced model: 1) the cross-domain stereo matching precision is remarkably improved in a complex scene, and stronger domain migration capability and generalization stability are shown; and 2) while the cross-domain robustness is maintained, relatively high accuracy and calculation efficiency are maintained in a non-cross-domain environment, and double consideration of cross-domain and non-cross-domain performance is realized.

The invention relates to the technical field of computer vision, in particular to a method and a device for constructing a binocular stereo matching recursive network based on deep learning, in a feature extraction stage, a self-adaptive spatial pyramid pooling layer based on a kernel normalization convolution layer is provided, multi-scale feature extraction is performed on an image, and in a matching cost calculation stage, a self-adaptive spatial pyramid pooling layer based on a kernel normalization convolution layer is provided. A new attention approach is utilized in combination with a regularized cost volume to produce a more accurate similarity measure. And in the main network framework, a GRU (Gated Recirculation Unit) based on kernel convolution is adopted as an iterative update operator, features are retrieved from the stereo matching cost body, and parallax is updated through regression iteration. The GRU can consider the wider context around each pixel, the fusion of local and global information is promoted, the more comprehensive understanding of the feature space is provided, and the reasoning ability of the network for the scene structure of a larger area is enhanced.

The invention relates to the technical field of computer vision, in particular to a three-dimensional reconstruction method and device based on optical polarization and stereoscopic vision principles. The three-dimensional reconstruction method sequentially comprises the steps of image preprocessing, polarization parameter extraction, binocular stereo matching, disparity map generation and optimization, normal vector correction, depth information fusion and three-dimensional point cloud reconstruction and modeling generation. A polarization normal vector field is constrained and corrected as a'skeleton ', and the azimuth angle pi ambiguity problem which puzzles polarization three-dimensional reconstruction for a long time is solved; high-frequency surface normal details contained in corrected polarization information are used as textures to enhance and fill up depth information loss of binocular vision in weak texture and repeated texture areas, the inherent limitation of a single sensing technology in a three-dimensional reconstruction task is overcome, and the three-dimensional reconstruction of the surface of an object, especially a diffuse reflection object, is realized. And high-precision and high-integrity three-dimensional shape recovery is realized.

The invention discloses a ring surface outer pipeline welding seam surface grinding device and a grinding quality detection method thereof, and relates to the technical field of pipeline grinding, the ring surface outer pipeline welding seam surface grinding device comprises a grinding unit, a holding arm unit and a motion control system, the grinding unit is composed of a binocular stereoscopic vision system and a flexible grinding head device, the arm holding unit is composed of a pressure sensor set, an electromagnetic adsorption device and a mechanical arm set, and the motion control system is composed of a driving box, a wheel set and a frame set. The electromagnetic adsorption device is adopted, during work, the binocular stereoscopic vision detection system determines the range of a to-be-polished area, the controller drives the polishing device to move, the induction type pressure sensor feeds back collected pressure signals of an electromagnet and the outer wall of a pipeline to the control module, and the magnitude of electromagnetic force is controlled by changing the current of a coil; in addition, due to the fact that the polishing device is of a semi-open structure through the hinge device, multi-dimensional safe and efficient polishing operation is achieved by means of the electromagnetic force adjustable adsorption device.

The invention belongs to the technical field of computer vision, and discloses an underwater binocular positioning method and system based on optical target prior constraint. The method comprises the following steps: establishing a binocular stereo imaging model considering camera distortion; the method comprises the following steps: constructing an image data set of an underwater prior optical target, extracting potential feature points through feature transformation identification, performing feature point matching on a denoised image in the image data set of the underwater prior optical target, and performing verification by using prior position information of an optical target point as a depth constraint condition. Eliminating wrong matching with overlarge position deviation in the matching feature points; and combining the projection position of the target to be measured with the prior projection position of the target to establish a re-projection error function, and iteratively optimizing the parameters of the binocular stereoscopic vision model by taking the collinear relationship generated by the optical target as a multi-target geometric constraint condition until the re-projection error meets the precision requirement. According to the method, the feature point identification and positioning precision of the underwater robot in a complex water body environment is improved.