405 results about "Video recognition" patented technology

The invention provides an image text recognition method and system based on a mask diffusion model, a storage medium and equipment, and belongs to the technical field of image or video recognition or understanding. According to the method, multi-scale visual features of an image are extracted through a visual encoder, a mask diffusion decoder is combined, a diversified mask strategy and random character replacement disturbance are adopted in a training stage, denoising loss and auto-reflection loss are calculated respectively, and a model is optimized in a combined mode; in the reasoning stage, starting from a full mask state, a complete text sequence is recovered through multi-round iterative denoising. According to the method, the one-way modeling limitation of a traditional autoregression model is broken through, all-around context-dependent modeling is achieved, an autoreversion error correction mechanism and a block low-confidence mask strategy are introduced, and the recognition accuracy and reasoning efficiency in complex scenes such as shielding and fuzzy scenes are remarkably improved. The method provided by the invention reaches a leading level on a plurality of public data sets, and has the advantages of high precision and high speed.

The invention relates to the technical field of machine learning, and particularly discloses an intelligent parking guidance and reverse vehicle searching system based on multi-source heterogeneous data fusion. The system comprises a data perception fusion layer, a dynamic prediction decision-making layer and a user service interaction layer, realizes multi-step advanced probability prediction and dynamic optimal path planning of a parking space state through fusion of geomagnetic detection, video identification, payment flow, traffic situation and activity information, and realizes the optimal path planning of the parking space state based on multi-mode induction and live-action AR reverse vehicle searching guidance. And the parking efficiency and the user experience are improved.

The invention relates to the technical field of intelligent operation and maintenance of photovoltaic power stations, in particular to an AI video recognition system for photovoltaic power station equipment inspection, which comprises a data acquisition unit, a data preprocessing unit and a defect recognition unit, and is characterized in that characteristics from one mode are used as query, evidence information serving as keys and values is searched from corresponding area characteristics of other modes, and the data acquisition unit is used for acquiring data; the method is used for synergistically diagnosing compound and early defects with weak or invisible characteristics in a single mode. According to the method, cross-modal collaborative reasoning can be realized: when a suspicious feature is found in one modal, whether evidence features capable of mutually verifying exist in the same position in other modals or not can be inquired, the diagnostic logic of field experts is simulated, different physical phenomena can be associated, and the probability of mutual verification is reduced. Therefore, early-stage or composite defects which are extremely difficult to find in any single mode can be accurately identified. Therefore, the problem that the recognition performance is reduced due to environmental interference such as illumination and shadow can be fundamentally solved.

The invention relates to the technical field of video recognition and analysis, and discloses a video recognition and analysis method based on physical characteristics. Video frame pixels are mapped to a two-dimensional coordinate system with the upper left corner as an original point, and mirror image expansion and median filtering are carried out on a gray level image; constructing a binary image based on a gray threshold value, and analyzing and extracting a target region by using a four-neighborhood connected domain; using neighborhood search and polar angle sorting to close the tracking contour, and generating equidistant re-sampling points based on Euclidean distance and an interpolation method; the curvature of the re-sampling points is estimated through a three-point difference algorithm, and zero denominator is avoided through numerical protection; performing discrete Fourier transform on the curvature sequence to extract a frequency spectrum, and normalizing an amplitude to form a standardized feature vector; and finally, inter-frame similarity is calculated based on the feature vector, and the most similar frame is automatically retrieved. By processing unified data standards in stages, edge noise is suppressed, sampling uniformity is ensured, and feature stability and cross-frame comparability are improved.

A system and method for training, assessing, and providing feedback on cardiopulmonary resuscitation (CPR) performance based on at least one video of a CPR training session performed by a trainee on a non-mannequin training object. A preprocessing module is configured to process the at least one video to generate a standardized video. A marking module is configured to use pose estimation to mark points for body movements during the CPR training session based on the standardized video, and a computing module configured to compute body movement parameters for CPR based on the marked points. A classification module implements a machine learning model that classifies CPR compressions on the non-mannequin training object based on the computed body movement parameters, thereby generating compression classifications, wherein the machine learning model is trained to extract CPR-specific features. An editor module maps metrics over the standardized video based on the compression classifications and generates a feedback video based on the mapped metrics. An analysis module identifies deviations from CPR guidelines based on the feedback video and generates analysis results based on the deviations. A feedback module provides performance feedback to the trainee based on the analysis results.

The invention relates to the technical field of image processing, and discloses a sorting video recognition processing method and system based on YOLO. The method comprises the steps of processing mask data by adopting a YOLO-Inpaint algorithm, embedding image restoration loss calculation in a YOLO network, and inferring a mapping relation of complete article features from incomplete features through adversarial training learning to obtain a restoration feature graph. And calculating an article integrity score based on the repair feature map, and integrating the ratio of the visible pixel area to the predicted total area and the similarity of the repair area and the real area. And the integrity score and the detection confidence coefficient are fused according to the adaptive weight of the shielding degree, and a compensated recognition result is generated. And constructing a time-sequence-labeled identification data stream which comprises an article category, a position coordinate, a shielding state and a confidence coefficient, and outputting a processing report. According to the method, the problems of accuracy and robustness of object shielding recognition in the sorting video are solved, and the recognition success rate and the processing efficiency under the shielding condition are improved.

The invention relates to the technical field of image or video recognition or understanding, and discloses a bus driver driving state evaluation method, which comprises the following steps: acquiring a face image, an electrocardiosignal and voice audio data of a bus driver, performing timestamp alignment and preprocessing, and constructing a multi-mode driving state data set; extracting multi-modal features of the collected data, and mapping the multi-modal features to a unified feature space; fusing multi-modal features through a cross-modal attention mechanism, dynamically adjusting attention weight based on an emotional change difficulty index, and extracting emotional change key features; and predicting a two-dimensional continuous emotion value of the driver by using the fusion features, calculating a long-time-sequence emotion driving risk score based on an emotion stimulation dynamic model, and performing evaluation and early warning of a driving state. The problems of single-mode analysis, lack of long-period early warning and driving state static recognition in the prior art are solved, and the purposes of accurate evaluation, high safety, multi-mode fusion and long-time-sequence prediction are achieved.

The invention discloses a berth positioning system and method based on video recognition, and relates to the technical field of intelligent transportation, and the system comprises a calibration module which generates berth calibration data based on a berth type, and stores the data to a cloud server; the edge calculation service module is configured to judge the spatial relationship between the vehicle and the berth through a berth type adaptive algorithm based on the real-time latitude and longitude coordinates of the inspection vehicle and the berth calibration data; and the vehicle identification module records license plate information when determining that the vehicle is in the berth, and actively snapshots an empty berth image when no vehicle is in the berth and uploads the empty berth image to the cloud. According to the scheme of the invention, the accuracy, environmental adaptability and long-term operation stability of parking state recognition can be remarkably improved, and a comprehensive solution with high robustness and low misjudgment rate is provided for urban intelligent parking management.

The invention provides a spinach cleaning quality detection method based on image recognition, and relates to the technical field of image or video recognition. A closed-loop control framework integrating active detection, cross-modal sensing, physical modeling decoupling and intelligent decision optimization is constructed; cleaning efficiency and quality bottlenecks caused by single information dimension and passive control mode in the prior art are fundamentally solved; the self-adaptive cleaning device has the final beneficial effects that the energy consumption and the water consumption of a unit product are greatly reduced by precisely putting water flow energy according to requirements while the cleaning cleanliness stability under complex working conditions is remarkably improved, and self-adaption and self-optimization of the whole cleaning process are realized.

The invention discloses a large video model training method and a related device, and relates to the technical field of video recognition, and the method comprises the steps: collecting a training video data frame to obtain an image frame, inputting a preset prompt word, a user question and the image frame into a large image model, and obtaining a thinking chain and a question answer. Performing cold start on the video large model based on the thinking chain and the question answer to enable the video large model to have thinking chain output capability; and combining training video data and questions to generate space and time disordered data and thinking chain data. And inputting the three types of data into the model to obtain corresponding outputs, calculating the accuracy of each output, obtaining space and time accuracy reward values, and training the model through a group relative strategy optimization algorithm in combination with a thinking chain consistency reward value to obtain an inference video large model. According to the method, the trained video large model can have thinking reasoning capability based on thinking chain implementation.

The invention discloses an intelligent recognition system for environmental protection violation behaviors based on multi-modal AI, belongs to the technical field of environmental protection monitoring and video recognition, and aims to solve the problems of data fragmentation, incomplete violation recognition, low precision and insufficient supervision efficiency in traditional environmental protection monitoring. The system comprises a data analysis module, a dynamic sampling module, a data preprocessing module and an intelligent identification module. Analyzing the production log data, and generating a sewage discharge sequence of a sewage discharge port; generating sampling parameters in combination with the emission sequence and historical violation characteristics, and extracting video frames to form a monitoring frame sequence; positioning a sewage draining exit to obtain a dynamic monitoring area, and identifying emission frames to construct an emission frame data set; processing key evaluation indexes of emission frames, identifying excessive emission and abnormal frames, marking violation behavior types, and generating a violation identification report; according to the invention, full-dimension and high-precision recognition of environmental protection violation behaviors is realized, a traceable and efficient decision support is provided for environmental protection supervision, and the supervision efficiency is remarkably improved.

The invention relates to the technical field of safety management, and discloses a venue safety emergency large model construction method based on artificial intelligence, comprising the following steps: step 1, deploying a composite sensing device in a preset area of a venue, the composite sensing device comprising an infrared thermal sensing detection module and a video identification module, the infrared thermal sensing detection module is used for collecting human body thermal sensing track data, and the video recognition module is used for collecting human body image data and marking behavior characteristics in the thermal sensing track data and the image data in real time. According to the method, the technical scheme of constructing the dynamic event evolution diagram is adopted, the technical effect of automatically constructing the event causal chain is achieved by mapping the personnel behavior state and the regional position into the diagram nodes and establishing the space-time correlation edge, and compared with the technical scheme of isolated analysis of surface features in the prior art, the method has the advantages that the efficiency is high; the defect that a child is misjudged due to the fact that a binding relation between a guardian moving track and ticket business cannot be associated is overcome.

The invention discloses a multi-modal video recognition and low-code complex time sequence intelligent algorithm arrangement method, which comprises the following steps of: firstly, jointly constructing a safety production basic algorithm library by using a multi-modal video recognition technology and experience accumulated in previous projects, then analyzing requirements, determining a requirement function and a mixed algorithm requirement, and constructing a safety production basic algorithm library; a reasonable large model architecture is designed; a perfect time sequence algorithm model is established by deeply analyzing the time sequence dependency relationship under various scenes; machine learning and artificial intelligence technologies are utilized to realize dynamic optimization of an algorithm execution sequence; and finally, a friendly interface of the main body is developed and provided by using a low-code technology, so that the main body can complete the arrangement of a complex time sequence algorithm through simple operation. The invention aims to improve the accuracy of video recognition through multi-modal data fusion, simplify the arrangement process of a complex time sequence intelligent algorithm by using a low-code technology, and accelerate the development and deployment of the algorithm.

An apparatus of a vehicle comprises a memory storing at least one instruction and a processor configured to execute the at least one instruction. The at least one instruction may be configured to cause, when executed by the processor, the apparatus to: via a tuning parameter learning model for image correction, generate, based on received video data, a tuning parameter for adjusting image signal processing (ISP) for correcting the received video data; correct, based on the tuning parameter, the received video data; identify, via a video recognition model, at least one object in at least one image corresponding to the corrected video data; and control, based on the identified at least one object, autonomous driving of the vehicle.

The invention discloses a video recognition-based man-machine cooperative processing method for ditches and rivers in an urban area, and relates to the field of video recognition or understanding, and the method comprises the steps: constructing a data collection and analysis closed loop covering the whole periods before, during and after heavy rainfall: collecting images before and after treatment through an unmanned aerial vehicle and a camera before rainfall, generating a treatment operation influence map for quantitatively evaluating the treatment effect; in rainfall, the fixed camera is used for continuously monitoring the water level and the flow velocity; after rainfall, images are collected again, and the abnormal river reach is recognized. According to the method, camera data is taken as a boundary condition, a hydrological model is driven to perform inversion in combination with a governance influence map, missing heavy rainfall process data is deduced, and finally, accurate attribution of abnormal reasons is realized, and a differentiated governance scheme is generated. According to the method, the problem of monitoring data missing in severe weather is effectively solved, and quantitative evaluation of the treatment effect and scientific diagnosis of the abnormal cause of the river channel are realized.

The invention discloses an analysis method and system based on AI video recognition behavior monitoring, and relates to the technical field of video recognition, and the method comprises the steps: collecting video data, calculating the overall pixel gradient change of adjacent frames, marking preliminary candidate frames, calculating the edge saliency intensity of each frame, and analyzing a salient region point set; clustering and calculating the geometric center point of each cluster as an anchor point value, and forming an anchor point set through the center points of each cluster; according to the method, concentrated extraction of salient points is realized through a density-based clustering algorithm DBSCAN, so that an extraction result has logic uniqueness and local integrity of a salient target, a local feature map is extracted by using a shallow convolutional network, target texture and edge information is effectively captured, bounding box optimization is realized by rotating an IoU loss function, and the robustness of the target is improved. And the fitting precision of the boundary of the salient region is improved.

Example systems, computer readable medium and methods for automatically generating portrait videos from landscape videos include receiving a landscaped video, identifying one or more objects to track, and tracking the one or more objects by moving a cropping window. A user interface is presented that provides options for a user to automatically convert a landscape video into a portrait video. The options include selecting one or more objects in the landscape video to track and tracking based on a flight plan used to capture the video. The automatic tracking uses a hierarchy of a person for tracking where a face is used if the person is close, an upper body is used if the person is farther away, and a whole body is used if the person is even farther away. A hierarchy of a person includes indications of which parts of the hierarchy to exclude first.

The invention belongs to the technical field of intelligent traffic and passenger flow monitoring, and discloses a passenger flow statistics fusion system and method based on multi-modal big data. The system comprises an infrared induction data acquisition module, a video identification data acquisition module, a WiFi probe data acquisition module, an entrance and exit environmental factor sensing module, a confidence coefficient calculation module and a data fusion module. The system calculates the confidence coefficient of various data sources by collecting various types of passenger flow data and the shielding intensity, crowd density and bandwidth state in the current environment, and performs weighted fusion according to the confidence coefficient to obtain a passenger flow statistical result with higher credibility. The system improves the accuracy and stability of passenger flow statistics in a complex environment, and is suitable for dynamic monitoring and management of a high-density passenger flow scene.

The invention relates to the technical field of computer vision, in particular to a video recognition method and system for zoo animals, and the method comprises the steps: data collection, animal detection and tracking, animal feature extraction, individual recognition and decision making, system optimization and continuous learning, and data management and application. According to the video identification method and system for the zoo animals, non-intrusive and high-precision animal individual identification can be realized based on deep learning and computer vision technologies and in combination with unique biological characteristics of the animals, so that efficient, accurate and non-intrusive animal management is realized, and the video identification efficiency is improved. Therefore, powerful technical support is provided for zoo operation and scientific research.

The invention provides a video processing method and device, electronic equipment and a storage medium, and belongs to the technical field of multimedia. The method comprises the steps that in a video editing interface, a first video and a video script obtained by recognizing the first video through the artificial intelligence technology are displayed, and the video script is used for describing behaviors executed by a main object in the first video in the first video; displaying the modified video script in response to a modification operation on the behavior of the main body object in the video script; and in response to the video processing operation, based on the modified video script, displaying that at least two items in the audio, the image and the subtitle in the first video are modified by adopting an artificial intelligence technology to obtain a second video, and the behavior of the main object in the second video is matched with the modified video script. The method not only conforms to the video processing intention of the user object and ensures the accuracy of video processing, but also does not need to re-record the video by the user object, is simple to operate and can improve the video processing efficiency.

The invention discloses a video action recognition method based on an auxiliary time modeling module, and belongs to the field of mode recognition. According to the method, a series of time modeling modules (ECTLs) are designed, and the advantages of Transform and LSTM in the aspect of processing sequence information are efficiently combined, so that the potential of CLIP in the aspect of improving the efficiency and universality of an action recognition technology is explored. The ECTL module can extract local and global features from the sequential sequence, thereby improving the efficiency of sequential modeling. Experiments carried out on three data sets (HMDB-51, UCF-101 and Someting-Someting V2) show that the ECTL series time modeling module provided by the invention has effectiveness, and has universality in various video recognition environments. According to the method, the extremely competitive performance is kept under complete supervision, and the accuracy which is equivalent to or even higher than that of the most advanced method is realized on three public reference data sets.

The invention discloses a video event association reasoning method based on a knowledge graph. The method comprises the following steps: S1, collecting video data and preprocessing the video data; s2, constructing an event graph according to a preprocessing result, and generating a state representation sequence through a graph convolutional network; s3, constructing a DreamerV3 world model, inputting a state representation sequence, and generating an event trajectory set; s4, performing trajectory value evaluation operation on the event trajectory set, and constructing a target path based on an evaluation result; s5, executing a structure alignment operation in the event graph according to the target path, and generating a reasoning knowledge graph; and S6, based on the inference knowledge graph, outputting a video event association result according to a time sequence. According to the method, video identification, map modeling and trajectory prediction methods are fully fused, and an efficient closed loop of key event extraction and map-level causal reasoning is realized.

The invention provides a video semantic token compression method, a video recognition method and electronic equipment, and the method can comprise the steps: processing an original video, and obtaining a to-be-coded video; wherein the frame number of the to-be-coded video is less than that of the original video; performing semantic coding on the video to be subjected to semantic coding to obtain an original token data set corresponding to the video to be coded; and compressing the original token data set to obtain a target token data set. Through the method, the video and the semantic token of the video can be compressed, and the calculation amount required by semantic understanding of the large model can be reduced if the target token data set is subsequently used as the input of the large model, so that the large model can efficiently and accurately understand the long video, and the storage pressure of equipment is reduced.

The invention provides a natural shoreline and mangrove forest supervision method and system based on video recognition, and the method comprises the steps: S1, installing and deploying information collection equipment, and setting a preset position and a cruise plan according to different monitoring targets; s2, accessing original sample data acquired by information acquisition equipment, reinforcing an original sample data set through a three-level data enhancement framework, and constructing a shoreline target sample data set; s3, optimizing by taking the YOLOv8n model as a basic model, and constructing a shoreline target recognition model; s4, constructing a target entity-spatial constraint-ownership state triple verification rule fusion judgment model as a behavior recognition model; and S5, inputting a recognition result of the shoreline target recognition model into a behavior recognition model, judging an illegal behavior through a triple verification rule, and generating early warning information. According to the invention, the working efficiency of natural shoreline supervision can be improved, and a replicable solution is provided for fine management of the coastal zone and the mangrove forest.

The invention discloses an intelligent industrial safety production supervision platform based on intention recognition and a large field model. The intelligent industrial safety production supervision platform comprises a multi-source multi-modal data fusion module, a multi-modal video recognition module, an intelligent algorithm arrangement module, an intention recognition module, a large field model integration module and an emergency response module. The intention recognition technology is combined with an industrial field large model, comprehensive multi-modal data fusion and intelligent decision support are achieved in industrial safety production for the first time, and the method is not only suitable for high-risk industries such as electric power and petrochemical engineering, but also can be applied to safety production management in the fields such as machine manufacturing and food processing; besides, the system platform adopts a modular design and a low-code development environment, a user can flexibly configure a monitoring algorithm and an emergency scheme according to own requirements, and the system is suitable for various industrial environments and production scenes.

The invention belongs to the technical field of electric power safety, and discloses an electric power operation risk supervision video identification method and system based on an AI technology, and the method comprises the steps: carrying out the parallel analysis through a double-flow pyramid network architecture, extracting dynamic features and static features in a multi-source video stream, carrying out the feature fusion, and generating different feature axes; identifying dynamic and static risk factors in the feature axis by constructing a hybrid encoder, predicting a cascade evolution path of the risk in combination with a preset risk diffusion simulator, and generating a risk diffusion probability topological graph; and evaluating the generated risk diffusion probability topological graph by using an entropy decision algorithm, calculating the risk level of the power operation scene, triggering a preset hierarchical response mechanism based on the risk level, and performing visual display of the risk level in the video. According to the method, the double-flow pyramid network architecture is adopted to perform parallel extraction and fusion analysis on the dynamic and static features, and the real-time performance and accuracy of risk monitoring can be improved.

The invention discloses a freight yard gantry crane operation path optimization method based on linkage node recognition, and particularly relates to the technical field of freight yard gantry crane operation path optimization, which comprises the following steps: acquiring image data of a target gantry crane operation node based on the video recognition capability of a skyeye platform, extracting continuous actions in the image data through a time sequence, and determining the operation path of the target gantry crane; an action chain is formed, the action chain comprises lifting appliance descending, container moving and vehicle entering and exiting, and whether the target gantry crane operation node meets the completion condition or not is preliminarily judged based on the integrity of the action chain; on the basis of preliminarily identifying the completion condition of the operation node, the verification intervention signal is generated, the natural response behavior of the operation personnel or equipment is collected, the response feature vector is constructed and compared with the preset standard vector, so that whether the node really completes the operation or not is judged, and false path scheduling is prevented from being triggered due to false action.

The invention relates to the technical field of image or video recognition or understanding, and discloses an areca yellows early warning method based on regular economic forest multi-dimensional symptom analysis, in the method, the unique morphology of areca and various morphological changes which are invisible under trees but visible in remote sensing during the period that areca suffers from yellows are fully utilized, and the areca yellows early warning effect is achieved. And the multiple symptoms are combined to form a multi-symptom matching degree, so that the misrecognition problem can be eliminated in an auxiliary manner when specific symptoms exist. When the specific symptoms do not exist, whether the disease of the current areca-nut forest belongs to the infectious disease or not is indirectly judged according to the distribution characteristics of the multi-symptom matching degree in time and space by utilizing the distribution characteristics and the disease course development rule when the areca-nut is cultivated as an economic arbor, and if yes, the disease of the current areca-nut forest belongs to the infectious disease; and considering that the areca yellows occupy a large proportion in the areca infectious diseases, the pathogen detection cannot be infeasible due to entrainment of a large number of irrelevant diseased plant samples when the pathogen detection is carried out at the moment. The above points are combined to realize early warning of areca yellows.

The invention relates to the technical field of nondestructive testing, and discloses an aero-engine endoscopic eddy current testing system based on voice interaction and video recognition, which comprises an endoscopic eddy current integrated probe, a voice input module, a processor, an AI image recognition module, a parameter storage module, a display module and a video playing module. The integrated probe integrates an endoscopic optical unit and an eddy current detection unit, and synchronous acquisition of images and eddy current signals is realized. The AI image recognition module analyzes the real-time endoscopic image and automatically recognizes the type and the position of the workpiece; the processor intelligently calls matched detection parameters from the parameter storage module according to the recognition result, automatically configures the working mode of the probe and can control the video playing module to call and display corresponding operation guidance videos according to voice instructions. According to the invention, a traditional detection mode is changed, and automatic setting of detection parameters and intelligent guidance of a detection process are realized through deep fusion of voice interaction and AI visual identification.

An electronic device includes a display and at least one processor operatively connected to the display. The at least one processor may play a workout video through the display, identify the speed of a workout move included in the workout video, identify an audio signal that is distinct from the workout video, identify a target play speed of the workout video at which the speed of the workout move corresponds to the tempo of the audio signal, and control a play speed of the workout video such that the play speed corresponds to the identified target play speed of the workout video.