538 results about "Three dimensional vision" patented technology

The invention relates to a non-standard special-shaped part high-precision measurement and verification method based on three-dimensional vision, and belongs to the field of three-dimensional reconstruction and computer vision. Point cloud data of a non-standard special-shaped part are obtained through multi-angle scanning by means of multi-camera-projector collaborative layout in combination with a structured light coding technology and a phase shift code technology; the method comprises the following steps of: acquiring point cloud data, ensuring comprehensiveness and accuracy of the data, realizing high-precision alignment of the point cloud data through matching an underivative optimization algorithm with image feature points, and performing fusion, denoising and registration processing on the acquired point cloud data to ensure that point clouds acquired from a plurality of angles can be effectively fused so as to reconstruct a high-precision three-dimensional model. In the measurement and verification stage, the deviation between the point cloud data and the CAD model is measured through a standard coordinate system, multi-stage registration and a high-precision algorithm, rapid detection and size verification of the geometric deviation of the industrial workpiece are achieved, and reliable technical support is provided for quality control.

The invention discloses a body-in-white welding seam recognition and automatic welding method based on a machine vision technology, particularly relates to the technical field of computer vision and image processing, and is used for solving the problem of welding seam track recognition accuracy caused by insufficient processing capability of an existing three-dimensional vision recognition method on incomplete and uncertain point cloud data. Through the steps of multi-view point cloud acquisition and registration, probabilistic confidence evaluation, region growth of track continuity constraint, multi-track fusion optimization and the like, accurate identification of a body-in-white welding seam track under a complex working condition is realized; firstly, multi-view point cloud data are obtained, probabilistic registration is carried out to generate a confidence evaluation result, then candidate tracks are generated based on confidence weighting and semantic constraint, finally, an optimal track is generated through intelligent optimization and converted into a welding instruction which can be executed by a robot, and the accuracy and robustness of weld joint recognition are effectively improved.

The invention provides a pipe thread positioning and measuring method and system based on three-dimensional vision, and belongs to the technical field of industrial automation and three-dimensional vision. According to the method, a measuring system composed of a robot measuring system and a three-dimensional visual locating system is built, and hand-eye calibration is completed; the method comprises the following steps: acquiring a three-dimensional point cloud model of the end part of a pipe thread workpiece, and extracting an end surface normal vector, a tangent vector and a circle center coordinate through preprocessing, cylinder fitting, near-end surface point cloud extraction and end surface circle fitting so as to establish a measurement reference coordinate system; and converting the reference coordinate system to a robot base coordinate system based on a hand-eye calibration result and the tail end pose of the robot, generating a robot moving trajectory, guiding the robot to be positioned to a measurement position, and completing preparation before measurement. According to the invention, automatic high-precision extraction of the pipe thread measurement reference is realized, the problems of non-uniform reference, large error and low efficiency in traditional measurement are solved, the method is suitable for various pipe thread workpieces, and a reliable basis is provided for subsequent high-precision measurement.

The invention discloses a three-dimensional vision calibration and positioning method and system based on an event camera, and belongs to the field of positioning. The method comprises the steps of constructing an array containing a plurality of unique frequency coding light sources, constructing a binocular event camera system, collecting light source array light signal event streams, extracting pixel positions, frequencies and timestamp information, determining pixel coordinate centroids of the light sources through frequency and pixel clustering analysis, and establishing frequency and centroid mapping. Auxiliary light sources with the same coding rule are arranged on the surface of a target object, and after a binocular system collects light signals of the auxiliary light sources, the center-of-mass coordinates of the light sources are matched according to the mapping relation. And based on event camera calibration parameters and binocular external parameters, correcting and stereoscopically matching centroid coordinates, and calculating three-dimensional coordinates of the auxiliary light source through triangulation. And in combination with the known geometric topological relation of the auxiliary light source on the surface of the object, the spatial pose of the target object is solved by using a rigid transformation algorithm, and high-precision positioning is completed. According to the embodiment, the calibration precision can be improved in a complex application environment.

The invention discloses a three-dimensional visual detection method, system and equipment for industrial application, and relates to the related field of three-dimensional computer vision, and the method comprises the steps: obtaining a three-dimensional point cloud data set of a target product, carrying out the multi-dimensional feature analysis based on the three-dimensional point cloud data set, and constructing a multi-modal feature vector set; performing three-dimensional visual defect detection according to the multi-modal feature vector group to generate a defect detection quality report; and making a control decision instruction according to the defect detection quality report, and automatically controlling the target product through the control decision instruction. The technical problem that existing industrial application-oriented three-dimensional visual detection is insufficient in detection accuracy and comprehensiveness is solved, and the technical effect of improving the detection accuracy and comprehensiveness is achieved.

The invention provides a railway turnout identification method and device, equipment and a medium, and the method comprises the steps: S1, constructing a classification network, obtaining training data, carrying out the customization training of the classification network, and obtaining a trained railway turnout classification model; s2, obtaining turnout rail shape three-dimensional visual point cloud data in real time, and processing the turnout rail shape three-dimensional visual point cloud data through the railway turnout classification model to obtain turnout rail shape classification data; s3, adding the turnout rail shape classification data to a preset sliding window queue, and enabling the preset sliding window queue to keep a preset length; and S4, performing statistical analysis on the sliding window queue to obtain a railway turnout classification result. According to the technical scheme provided by the invention, the accuracy of visual turnout identification in the railway industry can be improved by comprehensively considering the turnout identification algorithm of the left and right rail states in the time window.

The invention discloses an electronic cabin assembly quality detection method based on three-dimensional vision. The method comprises the following steps: firstly, generating data for comparison during detection for a standard electronic cabin; shooting all to-be-detected objects in the to-be-detected electronic cabin to obtain a depth image and a color image of each to-be-detected object; according to the depth image and the color image of the to-be-detected object, combining with the data of the standard electronic cabin for comparison during detection; and whether screws of the electronic cabin to be detected are neglected and not installed in place, whether cable plugs are neglected and not installed in place, whether connectors and cable plugs in the wiring unit are wrongly matched, and whether the bending radius of cables is too small are detected in sequence. Through fusion of depth information, three-dimensional quantitative detection of the assembly quality is realized, and the detection precision, reliability and automation level are significantly improved.

The invention relates to the technical field of garbage recognition, and particularly discloses an intelligent garbage recognition system and method based on multi-modal fusion, a hardware sensing unit is composed of a three-dimensional vision module, a touch sensing module and a near infrared spectrum module, and point cloud data, a pressure distribution matrix and spectrum data of garbage are obtained respectively; and an edge computing unit is equipped to carry an NPU acceleration chip for data processing. And the software module realizes multi-modal feature fusion by using ResNet-50, LSTM (Long Short Term Memory), 1D-CNN (Convolutional Neural Network) and Transform Encoder, and is also provided with a multi-sensor timestamp synchronous controller. And the dynamic feature database starts online learning to update the material feature library when the garbage recognition confidence coefficient is lower than 85%. During feature fusion, a cross-modal feature mapping relation is established, the weight is adjusted through an attention mechanism, and dynamic learning adopts a model lightweight method based on knowledge distillation. And starting an arbitration mechanism to correct the classification result when different modal identification results conflict. According to the method, the advantages of different modal data are fully utilized, and the recognition capability in a complex environment is improved.

The invention discloses a non-ferrous metal ingot detection method and equipment based on three-dimensional vision. Belongs to the field of nonferrous metal detection. The method comprises the steps of construction of a three-dimensional pixel model with laser projection structured light projection and line laser scanning camera shooting, adaptive filtering preprocessing, establishment of a three-dimensional point cloud model through deep learning, feature information extraction, deep learning algorithm identification and classification, production process problem analysis and optimization and the like in sequence. Therefore, high-precision, multi-angle and all-directional detection of the non-ferrous metal ingot is realized. The equipment comprises a multi-axis collaborative three-dimensional visual inspection subsystem consisting of a mounting bracket, an X-Y-Z servo motion platform, a line laser scanning camera and the like. According to the method, the quality detection precision and defect detection rate of the non-ferrous metal ingot can be effectively improved, detailed product quality analysis is provided, and the method plays an important role in improving the product quality, optimizing the production process and reducing the production cost.

The invention relates to an end-to-end automatic driving planning method based on three-dimensional visual language pre-training, and the method comprises the following steps: obtaining a multi-view bird's-eye view feature X, and inputting the multi-view bird's-eye view feature X into a BEV encoder in a pre-trained BEV former to calculate a global visual feature; integrating the real description of the environment into a BERT structure-based text attention block to obtain a text representation; calculating grouping alignment loss; inputting the global visual features and the prompt features into a language model, outputting word vectors by the language model, and training a generative end-to-end automatic driving planning model pre-trained based on a three-dimensional visual language; and inputting the actual multi-view image into the automatic driving planning model, and outputting an actual automatic driving planning path. Compared with the prior art, the method has the advantages of improving the real-time performance and generalization ability of path planning and the like.

The invention discloses a digital twin manufacturing and testing system for key parts of aerospace equipment, which belongs to the field of intelligent manufacturing and comprises an intelligent manufacturing module, a surface detection module, a nondestructive inspection module and a mechanical detection module. The intelligent manufacturing module digitalizes the manufacturing process by means of a digital twinning technology, so that the processing process state and the subsequent detection result can be mapped conveniently; the surface detection module is mainly used for visual detection, two-dimensional or three-dimensional visual information is used, an artificial intelligence algorithm is combined, and surface machining quality detection is achieved; the nondestructive inspection module is used for detecting internal defects of the parts by virtue of internal defect inspection equipment; the mechanical detection module is mainly used for simulating detection of mechanical characteristics and life cycle of parts under dynamic working conditions of stress, temperature and the like. According to the method, digital mapping of the intelligent manufacturing process, the surface characteristics, the internal characteristics and the mechanical properties is realized by constructing a neural network architecture, so that the aim of gradually replacing physical testing with digital testing is fulfilled.

The invention discloses a method for calculating the volume of a muck pile in a strip mine area based on three-dimensional visual point cloud. The method comprises the following steps: step 1, extracting frames from a video of a whole muck pile shot by an unmanned aerial vehicle to obtain a muck pile picture, and processing the collected picture and video by using colmap to obtain internal reference and pose information of a camera; then, taking the picture and camera internal reference and pose information as input to be transmitted into a neural radiation field model, completing three-dimensional reconstruction of the muck pile, and obtaining point cloud data of the muck pile; step 2, processing the point cloud data of the muck pile to obtain morphological parameters of a longitudinal profile, a volume and a throwing distance of the muck pile; according to the method, three-dimensional reconstruction is performed on the muck pile through the neural radiation field, the muck pile point cloud is generated, and then the muck pile point cloud is processed, so that the muck pile volume and other morphological parameters are obtained quickly and accurately.

The invention provides a three-dimensional vision and infrared thermal imaging combined calibration method and system and a medium, and the method comprises the steps: carrying out the processing of a calibration plate based on a special infrared response material or an electric heating insert, and obtaining a visible light angular point pattern and a calibration plate of a thermal response structure region; shooting an image of the calibration plate based on a three-dimensional vision camera, and extracting image coordinates; acquiring an infrared image of the calibration plate to obtain a thermal image, and analyzing distribution information of thermal characteristic elements on the calibration plate; establishing a nonlinear mapping relationship between the image coordinates and the distribution information of the thermal characteristic elements; analyzing a projection error based on the nonlinear mapping relation, and performing parameter calibration on the three-dimensional vision camera and the infrared thermal imager based on the projection error; the thermal imager and the three-dimensional vision camera are responded at a time through the bimodal calibration plate, a nonlinear injection relation is established, parameter calibration is accurately carried out on the three-dimensional vision camera and the infrared thermal imager according to projection errors, the calibration precision is improved, and the method is suitable for complex environments such as industry and medical treatment.

The invention discloses a method for measuring the navigation pose of a battery replacement robot based on three-dimensional visual features. The method comprises the following steps: 1, acquiring a color image and a depth image of a scene by using a 3D visual sensor; 2, positioning a locking and unlocking area in the color image through a convolutional neural network, and projecting depth data of the area into a point cloud by combining internal parameters of a camera; 3, performing statistical filtering and voxel filtering preprocessing on the initial locking and unlocking point cloud; 4, collecting five key points from the target point cloud by using an FPS sampling method, solving the coordinate of each key point in the target point cloud object coordinate system, and taking the coordinate as a directional vector; 5, points in the scene point cloud are converted into an object coordinate system, candidate points of the key points are screened according to the direction and length of the steering vector, and the key points of the scene point cloud are determined from the candidate points; and 6, fitting the pose of the scene and unlocking the point cloud by adopting a least square method according to the corresponding relation between the two groups of key points of the target point cloud and the scene point cloud, and optimizing the pose estimation result by using an ICP algorithm.

The invention discloses a complex stacking scene robot disordered grabbing method based on three-dimensional vision, and the method comprises the steps: firstly obtaining an original point cloud through a depth camera, and obtaining a high-quality point cloud after statistics and radius filtering, moving least square smoothing, background plane removal and grid simplification; then, hand-eye calibration of participation eyes inside the camera outside the hand is completed, and alignment of vision and a robot coordinate system is achieved; single workpieces are separated through Euclidean clustering; in the key point extraction stage, geometric significance and super voxel density screening are integrated, weighted three-dimensional SIFT and multi-scale SHOT descriptors are adopted for feature expression, and rough registration and symmetric ICP fine registration are combined to obtain the six-degree-of-freedom pose of a workpiece; and based on the self-adaptive curvature and the force closing score, a grabbing posture is optimized, a collision-free path is planned by using an RRT algorithm, and grabbing is executed. The method does not need pre-modeling, can adapt to shielding, symmetry and density change scenes, is low in deployment cost, is simple to maintain, and is suitable for industrial disordered feeding and discharging and logistics sorting.

The invention discloses an unmanned aerial vehicle non-cooperative environment autonomous landing method and system based on three-dimensional visual semantic information, and relates to the technical field of unmanned aerial vehicle landing recognized.The method comprises the steps that during emergency landing, a semantic segmentation model is used for generating a two-dimensional semantic segmentation map and obtaining semantic tags, and a landing area is automatically screened and tracked; fusing vision and inertia information through an airborne binocular camera and an IMU (Inertial Measurement Unit), outputting an accurate pose and depth map, and inputting a map construction algorithm to reconstruct a landform map of a landing area; projecting the semantic tag to a three-dimensional space and fusing the semantic tag with a terrain map to generate an evaluation map; and finally, an optimal landing point is determined by a landing point selection algorithm, and autonomous landing is completed through a flight control system. The method has the characteristics of high precision and fast reasoning, gets rid of dependence on external signals, significantly improves the landing capability under complex terrains, and guarantees the safety and reliability of autonomous landing.

The invention discloses an interaction method and device, a storage medium and equipment. The method comprises the steps of displaying a three-dimensional environment generated by augmented reality equipment; displaying a three-dimensional virtual keyboard in the three-dimensional environment, wherein a plurality of virtual keys are displayed on the virtual keyboard; identifying a stamping gesture of the current object; determining an interaction event between the current object and the virtual keyboard according to a stamp gesture, a stamp gesture collision body corresponding to the stamp gesture and a hovering collision body corresponding to a virtual key in the virtual keyboard; in response to the interaction event, interaction feedback information is output, and the interaction feedback information at least comprises the visual indication. According to the virtual keyboard with the three-dimensional visual effect, interaction feedback information can be output, interaction convenience and input efficiency of the virtual keyboard are improved, and interaction intuition and naturalness are improved.

The invention belongs to the crossing field of three-dimensional scene understanding and multi-modal artificial intelligence, and discloses a self-adaptive three-dimensional large language model system based on query guidance. A query-guided adaptive cutting module and a multi-modal object-level feature enhancement module are provided, the problems of redundant information processing and semantic information insufficiency in a three-dimensional scene are solved respectively, and the method comprises four steps of three-dimensional visual language alignment, query-guided adaptive cutting, multi-modal feature enhancement and language model training optimization. Through joint processing of multi-modal data, the method can realize efficient and accurate three-dimensional scene understanding under the condition of no or less annotation of a data set. Experimental results show that the performance of the method on a plurality of standard data sets is superior to that of the prior art, particularly in three-dimensional questions and answers and scene description tasks, remarkable precision improvement and robustness are shown, and particularly, higher accuracy and interpretability are achieved when complex three-dimensional structures and fine grit problems are processed.

The invention relates to the field of three-dimensional vision measurement, in particular to an aluminum alloy CNC product defect detection method and system based on structured light scanning. The aluminum alloy CNC product defect detection system based on structured light scanning comprises an image acquisition and analysis module, a self-adaptive illumination control module, a point cloud reconstruction module, a defect detection module and a report generation module. According to the method, through texture direction perception and matrix type light source dynamic angle modulation, and in combination with a light intensity-wavelength self-adaptive scheme of saturated pixel grading, mirror surface glare of the aluminum alloy high-reflection surface is greatly reduced; in cooperation with multi-exposure gray scale weighted fusion, a stripe image which is uniform in contrast and free of overexposure can be obtained in single scanning; in the same scanning time, compared with a fixed illumination scheme, the point cloud integrity rate can be improved, and a more reliable three-dimensional data basis is provided for subsequent defect detection.

The invention provides a welding seam identification and trajectory optimization method and system based on three-dimensional vision, and the method specifically comprises the following steps: firstly, collecting the three-dimensional point cloud data of a welding seam, extracting the local geometric features of the welding seam, and carrying out the feature enhancement of the local geometric features, so as to screen out representative effective edge points; and then, collecting multi-source point cloud data of the welding seam area, constructing a multi-scale welding seam feature recognition network, inputting effective edge points obtained by screening into the network, realizing accurate recognition of the welding seam area, and further extracting a linear welding seam area. According to the recognition result, a fitting algorithm is adopted for conducting curve fitting on the weld joint area, and a preliminary welding track is generated; and for the condition that multiple sections of welding seams are intersected, an optimal fitting point searching mechanism is further introduced, the preliminary track is optimized, and the continuity and integrity of a welding path are ensured. And finally, based on the center line of the welding seam, the posture of the welding gun is precisely planned in combination with the dihedral structure model, and therefore high-precision track control and posture adjustment are achieved.

The invention provides a complex heterogeneous molded surface three-dimensional vision measurement method and system, and the method comprises the following steps: (1), extracting part design model features through a semantic segmentation and intelligent recognition method, and building a to-be-measured feature set; (2) a three-dimensional measurement viewpoint planning method based on reinforcement learning is adopted to obtain a viewpoint sequence and a motion path; (3) when the local measurement system reaches a preset viewpoint, a self-adaptive single-pixel imaging feature matching method is adopted for mutual reflection and semitransparent molded surfaces; for other molded surfaces, a parameter adaptive fringe projection three-dimensional measurement method is adopted to obtain a single-viewpoint three-dimensional point cloud; (4) converting the single-viewpoint three-dimensional point cloud into a global coordinate system to realize three-dimensional point cloud splicing; (5) traversing each viewpoint of the viewpoint sequence, and executing the steps (3) and (4) to obtain a multi-viewpoint spliced three-dimensional point cloud; and (6) based on the point cloud error estimation, fusing the multi-view three-dimensional point cloud. The problem of accurate acquisition of the three-dimensional point cloud of the complex heterogeneous molded surface is solved.

The invention discloses a lithium battery laser welding detection method, and belongs to the technical field of nondestructive testing. The method comprises the following steps: acquiring a three-dimensional path point sequence of a to-be-detected welding seam; controlling the transient thermal excitation unit to apply transient thermal excitation along the path; a high-speed multispectral imaging unit is used for synchronously collecting the thermal radiation attenuation process of the weld joint under at least two different spectral bands, and a space-time spectrum data cube is formed; and finally, inputting the data cube into an artificial intelligence model for defect identification. The system correspondingly comprises a three-dimensional visual sensor, a transient thermal excitation unit, a high-speed multispectral imaging unit and a synchronous control and data processing unit. According to the method, different types of welding defects such as air holes and incomplete fusion can be effectively distinguished by analyzing the heat diffusion characteristic difference under different spectral bands, the interference of uneven surface emissivity is inhibited, and the online detection accuracy and reliability of the three-dimensional complex welding seam are improved.

The invention relates to a method (100) for providing an automobile data recorder playback function in an on-board display unit (50) by generating a visual resimulation from individual video recordings of a multi-camera system (60) of a vehicle (30), the method comprising the following steps: providing a plurality of video recorders (60); the step is performed by the vehicle data processing device (10): (101) providing video recordings of different cameras (61) of the multi-camera system (60), the video recordings being obtained from camera recordings of the environment of the cameras (61); providing (102) vehicle signal data, the vehicle signal data being specific for a state of the vehicle (30); (103) providing environmental model data, which is specifically for the modeling of the surroundings of the vehicle (30); (104) performing synchronous playback based on the provided environmental model data, the provided vehicle signal data and the provided video record; (105) generating a three-dimensional visual reconstruction model of the recorded environment based on the synchronized playback; a control for interactive transformation of the perspective of the generated three-dimensional visual reconstruction model is provided (106), the control being temporally coupled with the playback control.

The invention relates to the technical field of automobile part optimization, in particular to an intelligent self-adaptive screw locking control method for a motor, which can acquire an image of an assembly area in real time and establish a three-dimensional coordinate model of a screw hole pose through a camera integrated at the front end of the motor and a three-dimensional visual recognition algorithm. According to the method, the position and the posture of the screw hole can be determined with high precision, so that the vertical alignment relation between the screwdriver blade and the target screw hole is ensured, and the consistency and the accuracy of locking operation are improved.

The invention discloses a badminton tracking method and system based on three-dimensional vision, and relates to computer vision. Badminton motion video data of a left view and a right view are synchronously collected through a binocular camera, a left view frame sequence and a right view frame sequence which are aligned in time are obtained, and the collected video data are preprocessed; obtaining standardized left and right view input frame data; according to input frame data, constructing a 2D key frame detection model based on an improved YOLOv8 network, and extracting the sum of 2D key frame coordinates of left and right views; according to the 2D key frame coordinate sum of the left view and the right view, 3D space positioning is carried out on the badminton through stereo matching and track prediction, 3D track data are obtained, according to the 3D track data, a broken track is repaired through a frame leakage compensation mechanism, and a final 3D track is obtained. In order to solve the problem that in the prior art, key frames are prone to being lost and track breakage is caused during high-speed movement such as smash, the breakage track of the shuttlecock is repaired.

The invention relates to a mutual inductor self-adaptive mechanical arm appearance detection device fused with three-dimensional visual compensation, which structurally comprises a base, a mounting cavity is formed in the base, a first belt conveyor and a second belt conveyor are respectively arranged on two sides of the base, a detection table is rotatably arranged at the top of the base, and a mounting plate is fixedly connected to the top of the base. And the camera is fixedly mounted on one side of the mounting plate. Through rotation of the rotating arm, the mutual inductor on the first belt conveyor can be transferred to the detection table, the detection table can be driven to rotate by a circle, all-directional image acquisition of the mutual inductor is completed in cooperation with a camera, and the detection precision is improved; meanwhile, in the clamping process of the mutual inductor, the mutual inductor on the detection table can be automatically pushed to the second belt conveyor through the push block to be conveyed, the detection efficiency and accuracy are improved, and manual intervention and cost are reduced.

The invention provides a multi-robot 3D vision cooperation space attitude sensing calibration method and system, and relates to the technical field of multi-robot cooperation work. And collecting three-dimensional visual data in real time in the cooperative working area by using at least three robots. Setting a calibration precision factor to quantify an attitude estimation error, and preprocessing the collected data, including noise reduction and normalization; the weight of each data source is calculated to reflect the importance of the data source in the fusion process, a fused attitude estimation model is generated through dynamic Kalman filtering, and historical information is introduced. A motion estimation error is set and monitored in real time, when the error exceeds a threshold value, a calibration re-correction mechanism is triggered, the robot is reconfigured for data collection and weight updating, and it is ensured that the calibration precision continuously meets the requirement. According to the method, the attitude sensing precision of the multiple robots in a complex environment is improved, the data processing complexity is reduced, the real-time performance of the system is enhanced, and a foundation is laid for wide application of cooperative work of the multiple robots.

The invention relates to the technical field of visual recognition, and particularly provides a bottle sorting-oriented three-dimensional visual posture recognition method and related equipment, and the method comprises the steps: obtaining three-dimensional point cloud data and RGB image data of bottles on a conveyor belt, carrying out the preprocessing, feature extraction and geometric analysis of the three-dimensional point cloud data, and obtaining six-degree-of-freedom posture parameters; and the six-degree-of-freedom attitude parameters are projected to RGB image data, the overlap ratio of the projection contour and the actual bottle contour is calculated, and when the overlap ratio is larger than a preset threshold value, it is confirmed that the attitude is effective, and the corresponding attitude parameters are output to a mechanical arm sorting control system. Six-degree-of-freedom posture parameters of the bottle body can be obtained through posture alignment and template point cloud registration, reliability confirmation of a posture recognition result is achieved through overlap ratio verification in an RGB image data domain, and the defect that the posture recognition result is not reliable under the conditions that the posture change of the bottle body is large, the appearance of the bottle body is diversified and the bottle body is inclined or shielded and interfered is overcome. The problems of insufficient recognition precision and incomplete attitude parameters exist.

The invention discloses a spatial non-cooperative dynamic target approach visual detection method, and relates to the technical field of optical visual sensor detection, and the method comprises the steps: building a spatial non-cooperative dynamic target approach visual detection system software and hardware frame structure, and employing a deep residual shrinkage network to achieve target detection in a strong noise interference environment; establishing a coaxial three-dimensional vision measurement model to realize rapid measurement of target three-dimensional point cloud and motion parameters; the motion state of the target is subjected to reinforcement learning training, the motion condition of the target is sensed and predicted, and support is provided for a flight carrier to intelligently sense the target and make a maneuver decision. The method is suitable for high-precision and intelligent detection in the approaching process of a space dynamic target, and has important significance in guaranteeing detection and perception in the fields of spacecraft short-distance capture, near-earth asteroid early warning, air refueling docking and the like.

The invention discloses a monocular endoscope real-time depth estimation method and system based on self-supervised learning, and the method comprises the steps: firstly inputting a reference image and a target image, carrying out the preprocessing of the target image, and generating an image block sequence; and inputting the image block sequence into the depth map generation network, and outputting a depth map. And extracting feature graphs of the reference graph and the target graph by using an encoder, and respectively inputting the feature graphs into the appearance flow graph generation network and the optical flow graph generation network to obtain an appearance flow graph and an optical flow graph. Inputting the feature map into a body motion estimation network to obtain relative motion transformation of the endoscope, and inputting the relative motion transformation into a camera internal reference decoder to obtain a camera internal reference matrix; and reconstructing a target map based on the depth map, the relative motion transformation and the internal reference matrix, carrying out brightness calibration in combination with the reference map and the appearance flow map, and supervising and optimizing the depth map generation network through the optical flow map and the visibility mask to realize the real-time depth estimation of the monocular endoscope. According to the invention, depth information can be rapidly output in a medical scene with a high real-time requirement, and accurate three-dimensional visual assistance is provided.