80 results about "Visual matching" patented technology

The invention provides a welding path planning method based on multi-modal fusion and deep learning, and relates to the technical field of welding automation, the method comprises the following steps: synchronously acquiring three-dimensional coordinate information of a welding workpiece through a laser ranging module and a CCD visual sensor, and fusing laser height parameters and visual plane feature points; an initial path is generated based on the dynamic welding adaptation coefficient and the welding environment complexity, and laser errors, visual matching deviation and environment interference intensity are quantified; laser and visual data are dynamically weighted by using multi-modal feature fusion weights, spatial-temporal features are extracted through a deep learning model, and a welding track is optimized; and the six-axis mechanical arm is controlled to execute welding operation, weld quality is monitored in real time, and path parameters are adjusted in a closed-loop mode. According to the method, the problem of insufficient precision of a single sensor is solved through a multi-modal data complementation mechanism, sudden obstacles and temperature fluctuation are dealt with in combination with dynamic parameter calculation and real-time path correction, and the local path optimization capability is improved by using the convolutional neural network and a self-attention mechanism.

The invention discloses an underwater LED fish gathering lamp control method based on a light field scattering model, and the method comprises the following steps: obtaining water optical characteristic parameters and environmental parameters of a target water area in real time, and building an underwater light field scattering model through employing a radiation transfer equation in combination with a Monte Carlo simulation method; with improvement of target fish school gathering efficiency and reduction of energy consumption as optimization objectives, environment parameters are taken as constraint conditions, illumination effects under different LED parameter combinations are simulated through a light field scattering model, and LED control parameters with optimal comprehensive performance are determined; a control instruction is generated according to the LED control parameters, and the working state of each LED unit is adjusted; meanwhile, fish school gathering data and current change values of water temperature, salinity and turbidity are monitored in real time and fed back to the light field scattering model for parameter correction, and closed-loop control is formed. The underwater LED fish gathering lamp control method solves the technical problems of low light field distribution prediction precision, insufficient fish school visual matching performance, poor environmental adaptability and unbalanced energy consumption optimization in the existing underwater LED fish gathering lamp control method.

The invention provides a part free edge visual matching method and system based on model matching, and the method comprises the steps: analyzing an assembly three-dimensional model or a cutting two-dimensional model of a ship part, and building a part model database; determining a visual matching template database according to the two-dimensional contour of the part and the part features; collecting an image of a to-be-detected part; determining part features of the to-be-detected part; matching the part features of the to-be-tested part with the visual matching template database, and determining a matching result; if the matching result is successful matching, determining pixel coordinates of the to-be-detected part in the image of the to-be-detected part according to a visual matching template corresponding to the image of the to-be-detected part; and converting the pixel coordinates of the to-be-detected part in the image of the to-be-detected part into actual coordinates of a physical space, and determining the free edge of the to-be-detected part on the machine tool of the ship. According to the method and the device, the free edges of various parts of the ship are positioned, and the efficiency and the quality of ship manufacturing are improved.

The invention discloses a multi-scale machine vision matching method, and belongs to the technical field of vision matching, and the method comprises the steps: determining an original scale template image I1 based on a search image I0 inputted by a user, and obtaining a template feature point set under each layer of pyramid of the original scale; calculating a template scale step s; calculating upper and lower limits of a template scale; calculating a scale factor set S; traversing the scale factor set S to obtain each layer of pyramid feature point set corresponding to different scale factors; performing feature correction on each layer of pyramid feature point set of which the scale factor is less than 1; calculating a pyramid image sequence set of a search image I0 input by a user; performing top layer template similarity calculation to obtain a top layer matching result; performing neighborhood suppression on all the top-layer matching results to obtain a top-layer matching result set; and traversing all pyramid layers by adopting a layer-by-layer approximation mode to obtain a final matching result. According to the invention, targets of multiple scales can be effectively positioned.

The invention discloses an AI-based creative thinking auxiliary generation method and system, relates to the technical field of artificial intelligence, and solves the problem of inaccurate sound and picture matching in a traditional method by performing timestamp alignment and feature extraction on audio data and a visual image frame and calculating the correlation between the audio data and the visual image frame by using a cross-modal attention mechanism. According to the method, user eye movement track data is introduced, real attention points of a user are mapped into a visual sequence, and an optimized weight matrix is generated by constructing attention masks and fusing model attention weights, so that a generation result is more in line with perception key points of the user. And meanwhile, a feedback mechanism is established based on the synchronization error score, and when the sound and the picture are detected to be asynchronous, the visual frame timestamp can be dynamically adjusted, so that the self-adaptive correction of the content is realized. On the whole, the method has remarkable advantages in the aspects of improving modal alignment precision, enhancing user perception consistency and optimizing generation result naturalness.

The invention discloses a brain-computer interface system based on audio-visual matching error monitoring, which is characterized in that: according to EEG data obtained by three feedback paradigms of audio-visual matching errors, important ERP components related to the audio-visual matching errors are extracted by applying ERP analysis, so that neural responses in a specific time window are interpreted; then, on the basis of time-frequency analysis, dynamic changes of the electroencephalogram signals in different wave bands are captured, and the reliability and stability of the electroencephalogram activity in different feedback modes are revealed by combining ITPC of the electroencephalogram activity; and finally, through an improved EEGNet model, automatic feature extraction and efficient classification of the signals are realized, and the robustness and generalization ability of the system are improved. The electroencephalogram signals have high nonlinearity and specificity, by combining ERP analysis, time-frequency analysis and a deep learning model, potential neural mechanisms of audio-visual matching errors can be more effectively understood, the performance of a brain-computer interface system is improved, and the method is applied to clinical research of error processing related obstacles.

The invention discloses an aircraft visual navigation method based on double matching and dynamic satellite map loading. The method comprises the following steps: firstly, preparing a satellite slice database offline for dynamic loading during subsequent online matching; when visual matching positioning is carried out, feature extraction and matching are carried out on a current frame and a satellite map which is dynamically spliced and loaded according to course angle and height information, and latitude and longitude positioning information P1 is obtained through calculation; meanwhile, the current frame and the previous frame are subjected to same feature extraction and matching processing, and latitude and longitude positioning information P2 is obtained through calculation; and finally, inputting the double-matching positioning result into a Bayesian fusion module for fusion optimization to obtain a final positioning result of the current frame. The method provided by the invention effectively solves the problem that the traditional single matching method is easy to cause unstable positioning and failure, and greatly improves the precision and robustness of matching positioning.

The invention discloses an unmanned aerial vehicle rear-end loopback detection method and device based on multi-sensor fusion extended Kalman filtering. The method comprises the following steps: constructing a linearization state equation and a linearization observation equation of a target unmanned aerial vehicle; collecting visual feature points and laser radar feature points; performing target unmanned aerial vehicle rear-end loopback detection based on the visual feature points and the laser radar feature points to obtain a visual matching degree and a laser radar matching degree; calculating a visual proportion coefficient and a laser radar proportion coefficient in the loopback detection fusion process; taking the linearized state equation and the linearized observation equation as initial pose states of the target unmanned aerial vehicle, and iteratively updating a Kalman filtering gain by using a visual proportion coefficient and a laser radar proportion coefficient; the pose state of the target unmanned aerial vehicle is corrected and updated according to the Kalman filtering gain, and back-end loopback detection is carried out on the target unmanned aerial vehicle again based on the pose state of the target unmanned aerial vehicle; according to the invention, the error of back-end loopback detection in the mapping process of the unmanned aerial vehicle is reduced.

The utility model discloses visual mouth matching equipment for vacuum cup processing, which comprises a rack, and a turntable mechanism, a cup mold ejecting mechanism, a mouth matching front detection support frame and a pressing support frame which are arranged on the rack, a mouth matching front detection mechanism is arranged on the mouth matching front detection support frame, and a pressing mechanism is arranged on the pressing support frame. The cup mold ejection mechanisms comprise cup molds used for fixing vacuum cups, the cup mold ejection mechanisms are correspondingly arranged below the opening matching front detection mechanism and the pressing mechanism, and the opening matching front detection mechanism comprises a front detection camera with a lens facing the corresponding cup mold ejection mechanism. The downward pressing mechanism comprises a downward pressing driving source arranged on the downward pressing supporting frame, a downward pressing telescopic rod driven by the downward pressing driving source to move up and down, a downward pressing guiding assembly arranged on the downward pressing telescopic rod and moving up and down along with the downward pressing telescopic rod, and a downward pressing buffering assembly arranged at the end of the downward pressing guiding assembly. The utility model solves the problems of low production efficiency, unstable rate of finished products and the like in the production process of the conventional opening matching procedure of the vacuum cup.

The invention discloses a scene fusion supply and demand panoramic visualization matching decision-making system based on big data, and relates to the technical field of information sharing. The scene fusion supply and demand panoramic visual matching decision-making system comprises a data acquisition module, a portrait construction module, a matching module, a scene fusion module, an information sharing platform, an advanced decision-making module and a visual display module. The portrait construction module is used for constructing each supply end portrait by using the collected supply data of the supply ends and constructing each demand end portrait by using the demand data of the demand ends; the supply end and the demand end are combined to generate a supply-demand group; the scene fusion module is used for analyzing the influence of different environmental data on all supply end portraits and demand end portraits in the market, and fusing the influence of different environmental data to construct a comprehensive influence model; and the advanced decision module is used for making adjustment decisions of all supply and demand groups in advance through the information in the information sharing platform.

The invention relates to the technical field of stereoscopic vision, and discloses a synchronous speed visual matching method and system, electronic equipment and a storage medium, and the method comprises the steps: synchronously collecting a stereoscopic video sequence with a predefined frame rate; executing multi-target hybrid tracking and motion induction detection, and outputting target motion information including position and velocity vectors; extracting hierarchical motion features of the target from continuous multiple frames of the stereoscopic video sequence, and performing unified space-time coding; under geometric constraints of stereoscopic vision, scale cosine similarity, direction similarity and trajectory consistency measurement are calculated and serve as observation evidences to be input into the probabilistic reasoning model for fusion, and a posterior probability representing matching reliability is output; the weight distribution of the speed similarity and the direction similarity is adjusted according to the motion characteristics of the targets in the scene, and the stable corresponding matching relation between the left view target and the right view target is established. According to the method, high-time-resolution information can be utilized, motion features and geometric constraints can be effectively fused, and the method has self-adaptive capacity.

The invention discloses a method for calculating the position of a cutting channel through wafer visual automatic alignment. A low-precision camera and a high-precision camera are adopted to make a low-precision visual matching template and a high-precision visual matching template for the alpha face and the beta face respectively, and then the difference value from the corresponding visual matching template to the cutting channel intersection position is calibrated; determining the center position O of the wafer; taking a position A at the edge of the wafer in the alpha surface of the wafer, calculating a position B and a position C, calculating angle deviation values of the cutting channels AB and BC, and rotating the wafer to enable the cutting channels AB and BC to be matched with the alpha surface high-precision visual matching template; calculating horizontal positions Xi, starting positions Yi and ending positions Yi of all cutting channels on the alpha surface of the wafer; and calculating the horizontal position Xj, the starting position Yj and the ending position Yj of all cutting channels on the beta surface of the wafer by adopting the same method. The low-precision camera is matched with the high-precision camera, so that the accuracy of the position of the cutting channel is ensured.

The invention relates to the technical field of target detection, in particular to a low-slow-small flight target detection method based on visual matching, which comprises the following steps: running a real-time target detection thread and a periodic salient target detection thread in parallel, and performing low-slow-small target detection on each frame of visual image by the real-time target detection thread; the periodic salient target detection thread is executed once every a preset period, sky segmentation and saliency detection are carried out on the current frame of visual image, salient candidate targets in a sky area are extracted, the salient candidate targets are added into a candidate target queue, and priority ranking is carried out on the salient candidate targets; and if the significant candidate target with the highest priority enters a preset countering distance, detecting the confidence coefficient of the significant candidate target through a real-time target detection thread, and performing countering decision. According to the method, the problems of difficult identification, easy tracking loss, insufficient tail end precision and the like of the low-slow small flight target in a complex background are effectively solved.

The invention discloses a generative machine vision matching method, and belongs to the technical field of vision matching, and the method comprises the steps: calculating the matching degree of a top-layer search image based on the node direction of a pixel node of a top-layer template image and the gradient direction of the pixel node of the top-layer search image, and reserving a result of which the matching degree is greater than a user input threshold value; on the basis of the matching result of the top-layer search image, calculating the matching result of the next-layer search image layer by layer downwards until the calculation of the matching results of all sampling layers of the search image is completed; and based on matching results of all sampling layers of the search image, drawing rectangles in combination with the width and the height of the template, and determining a final matching result of visual matching according to the overlapping rate of the rectangles. According to the method and the device, a user can be supported to draw a template through a mouse or generate templates in other shapes in a self-defined manner for visual matching under the condition that a real template picture does not need to be provided.

The invention provides an unmanned aerial vehicle navigation deception detection method based on cross-view visual matching. The method comprises the following steps: acquiring a first aerial image collected by an unmanned aerial vehicle at the current moment and unmanned aerial vehicle positioning information; acquiring a first satellite image of the corresponding area based on the unmanned aerial vehicle positioning information; respectively inputting the first aerial image and the first satellite image into a pre-trained first twin network branch and a pre-trained second twin network branch, generating a first aerial image cross-view matching feature and a first satellite image cross-view matching feature, and sharing parameters of the first twin network branch and the second twin network branch, each of the ViT and the ViT comprises a ViT encoder and a cross-view feature mapping module; and determining whether a navigation spoofing attack exists based on the first aerial image cross-view matching feature and the first satellite image cross-view matching feature. By implementing the method, the accuracy and generalization ability of navigation deception detection can be effectively improved.

The invention relates to an inertial vision fusion positioning method and system based on graph optimization in an indoor cross-floor environment, and the method comprises the steps: obtaining the short-distance relative pose estimation data of a pedestrian based on the multi-source motion parameters of the pedestrian; based on the priori three-dimensional feature map and the image of the current scene of the pedestrian, acquiring visual absolute pose data; and based on a pre-constructed factor graph model, fusing the relative pose estimation data and the visual absolute pose data, and obtaining the global optimal three-dimensional position and pose of the pedestrian through nonlinear optimization solution. According to the method, continuity of deep coupling inertial navigation and absolute precision of visual positioning are optimized through the factor graph, inertial navigation accumulated drift is effectively inhibited, the problem of visual matching failure is solved, decimeter-level positioning precision (the average error is 0.24 m) and 100% continuous positioning are realized in an indoor cross-floor complex scene, and positioning robustness is remarkably improved.

The application discloses a method for cluttered scene target object grasping based on visual-language-action joint modeling. The application uses object-centered representation to realize a method for cluttered scene target object grasping based on visual-language-action joint modeling, processes the object-centered representation through a pre-trained visual-language model and a grasping model, obtains visual-language features and grasping features of each bounding box, and uses a transformer to implement cross-attention mechanisms among visual-language-action multimodality, generates visual-language-action cross-attention features, and then generates decisions and executes, so that higher sample utilization is realized, and additional data collection and training in the simulation-physical migration process are avoided; compared with a two-stage strategy, visual attributes and planner screening rules for language-visual matching need not be artificially designed, so that more flexible language instructions can be adapted, and better task generalization is achieved.

The present invention provides a visual positioning method based on a point cloud map, including a point cloud map generation module, a visual inertial odometer construction module, and a visual matching and positioning module based on the point cloud map. Among them, the point cloud map module establishes a high-precision point cloud map by fusing laser, IMU, and GPS information; the visual inertial odometer construction module first extracts visual feature points, uses the optical flow method to track the feature points of the front and rear frames, and performs pre-integration on the IMU, and fuses with the visual feature points to construct a visual inertial odometer, outputs the initial pose of each frame, and restores the depth of the feature points; the visual matching and positioning module based on the existing map extracts a sub-map from the point cloud map according to the initial position, projects the feature points restored by vision into the 3D space map coordinate system, and queries the nearest points in the current sub-map. For the nearest points matched by vision and the map, the RANSAC algorithm based on dual quaternion is used to optimize the pose of the current frame.

The invention discloses a vehicle assembling method, device and equipment and a storage medium, and relates to the technical field of vehicles, and the vehicle assembling method comprises the steps that when a to-be-assembled vehicle is located at a target station, corresponding to-be-installed part information is determined according to vehicle identification chip information; acquiring a corresponding to-be-installed part image according to the to-be-installed part information; performing visual matching according to the to-be-installed part image and the to-be-installed part information, and determining a target part matching result; and determining a corresponding target vehicle assembly result according to the target part matching result. The chip information of the to-be-assembled vehicle is identified, the corresponding to-be-installed part information is determined, then the to-be-installed part image is collected for visual matching, and finally the vehicle part assembling result is determined according to the visual matching result, so that vehicle part mismatching, misassembling and neglected assembling caused by human factors are avoided, and the assembling efficiency is improved. And correct parts are ensured to be assembled on a correct vehicle.

The invention provides a same-view-field cross-lens real-time target tracking system and method based on visual matching, and relates to the technical field of multi-target tracking, and the system comprises an acquisition module, a target detection module, a feature extraction module, a trajectory tracking module, a database management module and a cross-lens matching module. Wherein the trajectory tracking module can track a detected target in combination with target feature information, generate an identity label and output trajectory information, so as to ensure identity continuity and trajectory integrity of the target under the same lens; the database management module is used for uniformly storing the identity label, the feature information and the track information of the target and providing reliable data support for cross-lens tracking; and the cross-lens matching module is used for matching targets under different lenses based on the information in the database management module so as to realize cross-lens association of multiple lenses in the same view field. Through cooperation of multiple modules, multi-lens cross-lens real-time tracking under the same field of view can be effectively realized, and the accuracy and real-time performance of target matching are improved.

The present invention provides a control method and system for collecting soft tissue secretions based on a robotic arm. The method includes: obtaining a region to be extracted by using a stereo vision matching algorithm, and obtaining the specific position of the target detection region by using a threshold segmentation algorithm; using flexible tactile sensors distributed in different orientations at the end of the robotic arm to collect contact force vectors; combining a fuzzy control algorithm with a PID force-displacement control to achieve stable contact control of the soft tissue during the process of collecting secretions by moving the end of the robotic arm, and maintaining the stability of the contact pressure during this period. Implementing the embodiments of the present invention combines stereo vision with multi-dimensional touch to control the robotic arm to extract oral soft tissue secretions, improving the safety, reliability, automation, and intelligence of oral soft tissue secretion extraction.

This invention discloses a low-altitude visual matching navigation method, device, system, and storage medium, comprising: in an offline phase, optimizing an aerial image sequence into a high-fidelity 3DGS map model; in an online phase, rapidly fusing coarse poses for rendering using inertial pre-integration and global descriptor retrieval; based on the coarse pose, using the 3DGS differentiable rasterization pipeline to synthesize a high-fidelity new perspective reference image in real time by jointly optimizing the pose increment and reference view fusion weights; obtaining a 2D-2D correspondence between the real-time image and the new perspective reference image through deep learning matching, and converting the depth map synchronously generated by the 3DGS model into a 2D-3D association; and calculating the high-precision visual pose of the aircraft through the PnP algorithm and nonlinear optimization. This invention can improve the problems of low quality of new perspective reference images, insufficient multi-sensor information fusion, and limited positioning accuracy in complex low-altitude scenarios.

The invention discloses a multi-scene-oriented security robot intelligent inspection method and system, and relates to the field of intelligent security, and the method comprises the steps: collecting machine body vibration data generated by interaction between a robot and the ground in an environment with good visibility, and building a standard fingerprint band and a historical image; in a low-visibility environment, the robot collects a fuzzy visual image and vibration data in real time, generates a current inspection fingerprint and matches the current inspection fingerprint with a standard fingerprint band; and when deviation is detected, calling a visual memory library and determining the closest reference landmark by utilizing a deep learning visual matching network, so that the steering angle of the robot is adjusted, and the robot is enabled to be close to the central path of the fingerprint zone again. The method can be applied to industrial control software, is deployed on security robots of multiple models to realize process control of inspection, solves the problem that the security robots are unstable in inspection positioning in a low-visibility environment, and realizes path self-correction and intelligent inspection control based on vibration fingerprints and depth vision matching.

The invention relates to an automatic driving vehicle-mounted sensor fusion positioning method, device and equipment and a medium, and the method comprises the steps: fusing a coarse positioning result corresponding to I MU data and a coarse positioning result corresponding to wheel speed meter data based on a preset Kalman filtering algorithm, recalculating the position deviation, and obtaining the position deviation of the I MU data; carrying out constraint calculation on the relative displacement so as to determine fusion positioning information of the autonomous vehicle; and determining a vehicle moving track of the autonomous vehicle according to the fused positioning information, and performing map matching or visual matching based on the vehicle moving track to complete map matching positioning of the autonomous vehicle. According to the invention, high-precision positioning of data fusion of the automatic driving vehicle-mounted sensor can be realized without the help of high-cost and high-precision inertial navigation equipment under the condition that the GNSS signal is unavailable.

The invention belongs to the technical field of intelligent traffic, and particularly discloses an unlisted non-motor vehicle driver identification method and system and a program product, and the method comprises the steps: carrying out the time-space correlation matching of a front image and a back image of a driving non-motor vehicle and a driver, and an electronic license plate number of the non-motor vehicle; the method comprises the following steps: acquiring a front image and a back image of a non-motor vehicle, performing visual matching on the corresponding front image and back image, performing space-time and visual double-base judgment based on results of the two matching modes, determining a target front image corresponding to a target back image, finally performing face recognition on the target front image, and determining identity information of a corresponding non-motor vehicle driver without listing a tag. According to the method, the identity of the non-motor vehicle driver can be accurately and efficiently traced and identified, so that a complete evidence chain is formed, and an effective basis is provided for traffic management personnel to treat and punish non-motor vehicle non-listed behaviors.

The invention discloses a ship robot double-wire welding process and welding equipment based on model driving and visual matching fusion. The ship robot double-wire welding process comprises the following steps that an adaptive robot double-wire welding process is selected; visual scanning, model introduction and fusion are completed by the structured light shooting camera; clicking the welding plan to generate a welding operation file; a welding seam information json file is imported, and a robot welding operation list is generated; gun cleaning operation is executed, and welding is conducted after gun cleaning is completed; a welding task is automatically executed according to the welding operation list, gun cleaning is automatically completed according to the length of the welding seam, and welding of the next welding seam is continued; through the model driving and visual matching fusion algorithm, collaborative operation of model importing and visual scanning is achieved, the success rate of one-time workpiece recognition is high, the recognition precision is high, the welding seam positioning time is shortened, the arcing rate of the robot is increased, and the overall welding capacity and the unit area output efficiency are improved.

Method (200) for recognizing visually matching tones, wherein the method comprises: Receiving visual training data (105) at a visual coder (110) that has an initial machine learning (ML) model; Identifying data corresponding to a visual object in the visual training data (105) using the first ML model; Receiving audio training data (107) synchronized with the visual training data at an audio forwarding regulator (115) which has a second ML model, wherein the audio training data (107) has a visually matching tone and a visually mismatched tone, both of which are synchronized with one and the same frame in the visual training data (105) which contains the visual object, wherein the visually matching tone corresponds to the visual object, whereas the visually mismatched tone is generated by a sound source which is not visible in the same frame; Filtering data matching the visually appropriate tone from an output of the second ML model using an information bottleneck (120); and Training a third ML model following the first and second ML models (235) using the data corresponding to the visual object and data corresponding to the visually inappropriate tone.

The invention provides a spatial positioning method and device, equipment and a storage medium. The method comprises the following steps: constructing a neural radiation field based on three-dimensional point cloud data and a multi-view discrete image of a target scene; obtaining user behavior data which comprises a basic pose image shot by a user and / or editing data when a virtual object is edited, extracting an effective pose from the user behavior data, and generating a virtual shooting pose set; generating a rendered image set bound with a pose through the neural radiation field; and finally, receiving a positioning request, and determining a target pose in combination with the to-be-positioned image, the rendering image set and the three-dimensional point cloud data. The target pose is determined in combination with the multi-modal data, the precision limitation of a single-modal sensor in a complex scene is broken through, visual matching accumulative errors and dynamic environment interference are eliminated, the computing resource consumption is reduced while the positioning precision is improved, meanwhile, the effective pose is extracted in combination with the user behavior data for rendering, and the user experience is improved. And the problem of complex or inaccurate calculation caused by redundant rendering is avoided.

The system receives an indication of a sensitive operation. The system obtains a unique ID of a user's UE. Based on the unique ID of the UE, the system retrieves a visual authentication method including a visual ID. The system records the visual ID, and retrieves a corresponding stored visual ID. The system performs a liveness check associated with the visual ID, to determine whether the visual ID is a recording or a live version of the visual ID. Upon determining that the visual ID is the recording, the system refuses to authenticate the user. Upon determining that the visual ID is the live version of the visual ID, the system compares the visual ID and the corresponding stored visual ID to determine whether the visual ID and the corresponding stored visual ID match. Upon determining that the visual ID and the corresponding stored visual ID match, the system authenticates the user.

The invention discloses an audiovisual matching method based on a relation perception attention correction network. Internal association between semantic features of different modes is explored by utilizing correction attention between relation perception modes and correction attention in a self-adaptive mode. The correction attention mechanism between the relation perception modes guides features of different modes to perform target-specific interaction through correlation information obtained by a relation perception network so as to identify potential semantic relations. In view of significant information difference possibly existing between different modes, correction of attention in the adaptive mode serves as an auxiliary mechanism, the features can be independently and effectively processed, and then feature representation of each mode is deeply mined. In addition, the two attention modules are integrated with self-adaptive attention correction units, so that attention distribution is optimized, feature representation is enhanced, and meanwhile, the influence of interference information on the network is reduced. By enhancing the contrast between positive and negative samples, the relative distance stretch metric loss helps to improve the embedded characterization of the audiovisual features. According to the method, testing is carried out on a plurality of public reference data sets, various scenes are covered, and good performance is shown.