569 results about "Track algorithm" patented technology

The present invention relates to an adaptive sensing-based lightweight monitoring method for a fine crack in a complex background region. The method comprises the following steps: step S1, on the basis of region division, performing automatic acquisition of crack information, wherein PTZ camera sensors are used to automatically perform block-wise acquisition on crack regions; step S2, performing an adaptive complex scale calibration process, using a multi-scale template matching algorithm to adaptively correct distortion information of all regions, and performing real-scale conversion from pixel precision; step S3, constructing a lightweight crack segmentation network to process data processed in step S2; and step S4, by means of a quantitative crack-tracking algorithm based on Euclidean distance similarity classification, performing real-time monitoring on each piece of crack dynamic information. Compared with the prior art, the present invention has advantages such as achieving efficient, accurate, and online monitoring and analysis of cracks.

The invention relates to the technical field of marine monitoring, in particular to a marine target visual servo tracking method, which comprises the steps of multi-modal sensor fusion, a self-adaptive visual tracking algorithm, a servo control and visual collaboration mechanism and a shielding processing and target re-identification strategy. The problem that tracking is unstable under the conditions of illumination change, ship body shaking, target shielding and the like in a traditional method is solved. The IMU, the GNSS and the visual data are fused through extended Kalman filtering, ship body shaking is compensated, and the target state estimation precision is improved; the improved D-Fi ne target detection model is combined with an online feature updating mechanism to dynamically adapt to the appearance change of the target; the prediction and correction control strategy and the double-closed-loop PI D controller cooperate to adjust the camera holder, and the tracking delay is reduced; the multi-clue shielding detection and space-time joint feature matching technology ensures accurate re-identification of the target after shielding is removed. The real-time performance and robustness of the tracking system on an embedded platform are improved, and an efficient and stable target tracking solution is provided.

The invention discloses a millimeter wave radar human body tumble detection method based on multi-feature fusion, and the method comprises the steps: obtaining an original point cloud frame from each radar, and carrying out the timestamp calibration and space coordinate system conversion; performing noise filtering, ground segmentation and human body point cloud extraction on each frame of point cloud; dividing the preprocessed point cloud into a static cluster and a dynamic cluster; filtering false dynamic clusters; extracting a residual dynamic cluster set, and identifying and tracking the human body dynamic clusters in continuous frames by adopting a tracking algorithm; extracting a corrected time sequence feature from the tracked human body cluster; inputting the time sequence characteristics into a pre-trained deep time sequence network, learning a falling time sequence dependency relationship, and outputting an abnormal index; and performing multi-source fusion with the abnormal index to obtain a comprehensive index to judge whether to trigger an alarm. The method can adapt to a complex home environment, improves the detection accuracy and real-time performance, and reduces the false alarm and missing alarm.

The invention discloses a river and lake water level and flow prediction method and system in the field of remote sensing hydrology and hydraulic engineering, and the method comprises the steps: obtaining satellite radar altimeter data, river and lake water system data, rainfall inversion data and surface temperature inversion data based on a target river, a target lake and a watershed range file; taking the cross interval as a virtual hydrometric station observation point, extracting and processing waveform data, calculating water surface elevation information through a wavelet tracking algorithm, constructing an elevation profile group to establish an initial water level time sequence, and generating a river and lake water level inversion data set through fitting and elevation reselection. Meanwhile, NDWI data are obtained from satellite radar altimeter data, water surface width data are extracted, an inversion data set is formed by combining actually-measured river channel section data and a Manning formula inversion river channel flow rate, and finally various kinds of data are input into a pre-constructed intelligent prediction model to obtain future river and lake water level and flow rate change conditions. According to the invention, the precision of river and lake water level and flow prediction in areas lacking data is effectively improved.

The invention relates to the technical field of target tracking, in particular to a photoelectric tracking algorithm and system for self-adaptive target tracking, and the system comprises an image collection module, a cross-spectrum self-adaptive sensing module, an image processing and target recognition module, a self-adaptive tracking algorithm module and a servo control module, and constructs a closed-loop feedback link to achieve the synchronous optimization of parameters in a whole link. The algorithm comprises the steps of image acquisition and adjustment, target detection, parameter optimization, servo driving and feedback. The system can accurately estimate a high-speed turning target, is adaptive to complex environments such as low illumination / haze and the like, completes sheltered target recapture within 0.5 s, has servo compensation precision of + / -0.1 degree, is suitable for airport bird monitoring, port unmanned aerial vehicle tracking and highway vehicle monitoring scenes, and significantly improves tracking stability and practical value.

The invention discloses a high-precision continuous tracking method for a strong-maneuvering infrared weak and small target under a space-based detection visual angle, and the method comprises the steps: obtaining the uncertainty measurement of a target at a non-maximum suppression stage of a detection result, and enabling the uncertainty measurement to act on a target state updating and data association stage; therefore, the experience distribution of the detector is fully transmitted to the tracking process, and the tracking accuracy is improved. In order to cope with a complex target maneuvering state, an interactive multi-model technical route is adopted to design a tracking algorithm; in order to reduce the dependence on a prior motion model, a dynamic Markov transfer matrix construction method is designed, and the model transfer probability is updated in a mode of comprehensively modulating historical dynamic information and current static information; in a data association stage, targets with different scales are associated by integrating advantages of IoU and NWD, uncertainty is transmitted to a cost calculation process, and tracks and targets which are indefinitely matched are processed based on scale invariant and energy invariant hypotheses, so that high-precision continuous tracking of the targets is realized.

The invention relates to the technical field of fault equipment detection, and discloses a power equipment fault detection method and system based on high-precision temperature measurement, and the method comprises the steps: scanning the surface temperature of power equipment through a high-precision temperature measurement device, obtaining a temperature distribution image, processing the temperature distribution image, and obtaining a temperature distribution matrix; carrying out gradient analysis and temperature difference identification on the temperature distribution matrix to obtain a hot spot candidate area; performing clustering analysis on the hot spot candidate region to obtain an initial hot spot region, and generating a static boundary by adopting a boundary tracking algorithm; correcting the static boundary according to the temperature change of each boundary point to obtain a hot spot area; and performing fault risk assessment according to the load data of the power equipment and the temperature data of the hot spot area. According to the invention, through temperature data analysis and region boundary correction, the hot spot region can be accurately identified, and in combination with a machine learning algorithm, the accuracy of power equipment fault risk assessment is improved, and safe and stable operation of a power system is ensured.

The invention discloses a pedestrian multi-target detection and tracking algorithm based on cross-layer fusion and dynamic adjustment, and aims at the problem that a deep convolutional neural network in a backbone network is frequently excessively parameterized, a composite scaling mechanism is introduced to reduce the parameter quantity and improve the feature extraction capability of the network; meanwhile, considering that targets with different scales exist in a data set, a lightweight cross-scale feature fusion module is fused at the neck of the network, so that the adaptability of the model to scale change is enhanced; besides, aiming at the problems of small targets and shielding targets, a new loss function InnerWiseWIoU is designed, and in combination with internal region optimization and context information, the target positioning precision is improved. According to the method, the new model is applied to pedestrian tracking, the shielding degree between pedestrians is calculated and the matching threshold is dynamically adjusted aiming at the condition that a fixed confidence threshold in a tracking task is not suitable for centralized and rapid change of a target, so that the tracking performance is effectively improved, and the adaptability in a complex scene is improved.

The invention relates to a tracking navigation method for positioning a crop center point based on laser radar and image fusion, and belongs to the technical field of navigation, and the method comprises the steps: carrying out the marking of a farmland image after distortion correction, generating a training sample, inputting the training sample into a U-net model, carrying out the training, obtaining a crop segmentation model, and carrying out the classification of crops and weeds; performing feature fusion on the filtered laser radar point cloud data and the farmland image after distortion correction to obtain crop point cloud with semantic information; calculating a crop center point according to the crop point cloud with the semantic information; and providing navigation for the agricultural robot by taking the crop center point as a preview point. The spatial structure information of the target crop is obtained through the laser radar and the camera, the semantic point cloud is generated through feature fusion, the crop center point is dynamically extracted from the point cloud to serve as a real-time preview target, a fixed preview distance mechanism in a tracking algorithm is replaced, centimeter-level precise navigation is achieved, and the precision of the target crop is improved. And the operation precision and the operation safety of the agricultural robot are greatly improved.

A video file may be presented via a user application that displays one or more video frames of the video file. A user request to perform an object detection for objects of a specific object type in a video frame of the video file may be received from the user application. A machine-learning model of a plurality of machine-learning models that is configured to detect objects of the specific object type may be applied to the video frame to detect an object of the specific object type in the video frame. Each of the plurality of machine-learning models may be trained to detect objects of a corresponding object type. Subsequently, an object tracking algorithm may be applied to one or more additional video frames of the video file to track the object of the specific object type across the one or more additional video frames.

The invention relates to a target tracking system based on multi-modal deep feature proxy cross attention guidance, and the method combines and uses a visible light image and a thermal infrared image, and improves the performance of a tracking algorithm under a complex illumination condition through an innovative proxy cross attention mechanism. The system comprises a multi-modal detector and a data associator, wherein the detector consists of a double-branch feature extraction network, a proxy cross attention feature enhancement module and a feature pyramid sharing convolution module; the data correlator comprises a neural Kalman filter and a low-confidence-coefficient-based detection box reuse strategy, the challenge of a complex motion scene is solved by dynamically adjusting process noise and observation noise parameters, and the long-term tracking performance is improved. According to the method, multi-modal data fusion is carried out on a deep feature level, so that target detection and tracking are optimized, and the tracking precision and stability in a complex environment are improved.

The invention provides a three-dimensional tracking method applied to a single-photon laser radar to detect a far-field dynamic unmanned aerial vehicle point target, and belongs to the technical field of laser radar detection and target tracking. The problems that in the prior art, under the long-distance condition, the target imaging size is smaller than one pixel, and a detection and tracking algorithm based on shape, texture or edge features fails are solved. The method comprises the following steps: constructing a six-dimensional state vector containing a three-dimensional position and a three-dimensional speed of a target; predicting a state vector according to a state transfer function containing air resistance and turning acceleration; and an extended Kalman filtering framework is utilized to update the predicted state and obtain state estimation of the target, and the extended Kalman filtering framework comprises a self-adaptive process noise adjustment mechanism based on innovation feedback and is used for adjusting a process noise covariance matrix on line. The method is mainly used in the field of detection and multi-dimensional tracking of airspace moving targets.

The invention discloses an optical imaging zoom point selection correction method and system, which can correct the deviation between a theoretical model and actual lens characteristics in real time through a peak point tracking algorithm and offset vector calculation in a second stage, effectively solve the problem of individualized errors caused by lens group assembly tolerance, material refractive index difference and the like, and improve the correction accuracy. The stratified sampling strategy of the first stage is combined with the slope interpolation of the third stage, untested points in a theoretical curve are filled, a high-precision focusing curve is generated, the full range of zooming is covered, and local focusing failure caused by traditional uniform sampling is avoided; according to the method, curve jitter caused by test noise or mechanical vibration is eliminated, a smooth zoom focusing curve is output, the anti-interference capability in actual control is improved, frequent jitter of a focusing motor is avoided, a search strategy is dynamically adjusted in combination with definition data, an actual peak point is quickly locked, and the problem of overmodulation or slow convergence caused by a traditional fixed step length is solved.

The invention discloses a cigarette hard carton damage defect detection system and method, and the system comprises an image collection unit which comprises a shooting angle adjustable camera shooting assembly and an array type LED light source assembly; the image processing unit is used for carrying out gray processing, edge detection, dynamic binarization, denoising, sharpening enhancement and the like on the acquired image, and the image detection unit is used for generating a dynamic mask area to carry out defect detection on the image based on a detected straight line and dynamic rectangular tracking algorithm. Performing spatial positioning on spots in the image based on a rectangular matching algorithm of coordinate transformation, and when detecting that a certain spot is not in any rectangular range, judging whether the spot is a defect or not by combining a grading algorithm based on a color block area threshold value; and the result execution unit is used for outputting the defect detection result to the elimination mechanism, the early warning mechanism and the display assembly. According to the invention, the embedded visual detection module is used for carrying out machine visual online detection on the damage and scratch defects of the carton, so that the sound-light alarm and elimination of the defective cigarette carton are realized, the defective cigarette carton is prevented from entering the next process, the long-time tracing caused by difficulty in finding problems is also avoided, and the production efficiency is improved. And the control capability of a workshop aiming at the carton defect quality problem is greatly improved.

The invention provides a power grid frequency detection method, system and device and a medium, and the method comprises the steps: carrying out the narrowband signal decomposition of an obtained input voltage signal of a power grid through a null-space tracking algorithm, and obtaining a fundamental main mode component of the input voltage signal; performing low-pass processing on the fundamental wave main mode component to obtain a fundamental wave signal of the fundamental wave main mode component; performing discrete Fourier transform on the fundamental wave signal to obtain phase information of the fundamental wave signal; according to the phase information of the fundamental wave signal, calculating phase deviation information of the fundamental wave signal in pre-selected adjacent cyclic waves to obtain frequency estimation information of the power grid; according to the invention, narrow-band decomposition is carried out on an input voltage signal through a null-space tracking algorithm, so that a fundamental main mode can be accurately extracted; and phase deviation information can be determined based on discrete Fourier transform, so that when the interference frequency of the power grid is close to the fundamental frequency, high separation precision can be kept, and the accuracy of power grid frequency estimation can be improved.

The invention discloses an agricultural machinery high-precision automatic driving intelligent control system, which belongs to the technical field of agricultural machinery vehicle automatic driving and comprises a data acquisition module, a path planning module, a control module and an instruction execution module. The data acquisition module is used for acquiring a dynamic data set in real time and transmitting the dynamic data set to the path planning module, and the path planning module is used for receiving and analyzing the dynamic data set and generating an optimal path; the control module is used for synchronously receiving the optimal path, the real-time vehicle body posture dynamic data and the environment change data and generating a control instruction set; the instruction execution module receives the control instruction set, synchronously adjusts the steering angle and the running speed of the agricultural machine, generates actual execution parameters and feeds back the actual execution parameters to the control module. Through arrangement of the control module, the data acquisition module, the path planning module and the instruction execution module, the collision hidden danger caused by insufficient environmental adaptability of a traditional fixed track algorithm is solved.

The invention discloses an anti-interference target tracking algorithm based on Transform. The algorithm comprises the following steps: collecting a video image; performing feature extraction and training on the sample data; continuously tracking the selected target; a coping strategy mechanism for appearance change, similar object interference, shielding and other phenomena; an attention module is used for increasing the extraction capability of the network on target features; the built target tracking network is trained; and if the interference phenomenon does not occur, the algorithm directly returns and displays the tracking result on the computer. If the algorithm judges that interference occurs, the algorithm adopts an interference coping strategy, and after interference influence is eliminated, a tracking result is returned and displayed on the computer. The method has high application value in the fields of unmanned aerial vehicle tracking, video monitoring, intelligent driving and the like, and interested objects appearing in tasks in the related fields can be tracked more continuously and accurately.

The invention discloses a self-adaptive environment image processing method for industrial gas detection, which relates to the field of industrial gas detection and comprises the following steps: extracting feature points of a current frame by utilizing a feature point detection algorithm according to image frames of an industrial monitoring video stream, and acquiring coordinates of feature points between frames by combining an optical stream tracking technology; generating affine transformation parameters, performing transformation decomposition on the coordinates of the feature points by using the affine transformation parameters, and extracting pose change parameters to obtain an image motion track; predicting an image motion track by using an image track processing algorithm, and comparing the predicted image motion track with a real-time image motion track generated by the current frame; and optimizing the affine transformation parameters, and carrying out frame-by-frame processing on the video stream by using the optimized affine transformation parameters. Through feature point detection and an optical flow tracking algorithm, image motion information in an industrial monitoring video can be extracted, and industrial application requirements of high real-time performance and high stability are met.

The invention provides an unmanned aerial vehicle laser obstacle clearance system and method for improving laser obstacle clearance efficiency, the obstacle clearance system comprises an unmanned aerial vehicle, an airborne computer carried on the unmanned aerial vehicle, a stability augmentation holder, a holder camera and a laser transmitter, and the unmanned aerial vehicle, the holder camera and the laser transmitter are all fixed on the stability augmentation holder; the pan-tilt camera is used for collecting a trunk swing video stream in real time and transmitting the video stream to the visual module of the airborne computer; the visual module extracts tree trunk skeleton tracking points in the video stream; a control module of the airborne computer generates a holder direction control instruction and a laser trigger signal according to the tracking point coordinate position; the stability augmentation holder carries out real-time attitude adjustment according to the holder direction control instruction; and the laser emitter performs laser cutting on the swing tree trunk after receiving the laser trigger signal. According to the invention, real-time tracking of the target local area of the swing tree trunk is realized by using the unmanned aerial vehicle and the tracking algorithm, and the laser obstacle removing precision and the temperature rising efficiency are improved.

The invention provides a financial statement automatic generation method and system based on machine learning, and the method comprises the steps: carrying out the matching and conversion of semantics of different fields through employing an intelligent recognition technology according to a preliminary data format difference classification result, generating a standardized data structure set, and obtaining a unified field mapping relation; performing dynamic recombination on the cross-system business process by adopting a process tracking algorithm through a complete business data link, obtaining a time sequence relation and a data flow path of each business link, and obtaining a complete process tracking graph; key financial indexes and associated data of all links are obtained through the repaired process data set, and a data integration module is adopted for multi-dimensional summarization to obtain a preliminary financial data report framework; and through the final financial data content, obtaining a preset report template and an output rule, and generating a financial report output result conforming to a specification.

The invention relates to the technical field of computer vision, in particular to a target tracking method based on adaptive template updating and lightweight Transform. The invention provides an efficient and robust target tracking algorithm for solving the problems that a traditional twin network is insufficient in robustness in a complex scene and a Transform architecture is highly dependent on computing resources. Firstly, a supervision feedback module is designed, supervision information related to a task is introduced in a feature extraction stage, and a network is guided to be more focused on a target area, so that the feature discrimination capability is improved, and background interference is effectively suppressed; secondly, a lightweight Transform structure is constructed, the calculation complexity and the parameter scale are remarkably reduced while the global modeling capability is maintained, and the balance between the model performance and the calculation efficiency is achieved; and finally, designing a self-adaptive template updating mechanism, and dynamically updating the template content in combination with the state information of the current frame and the environment change, thereby enhancing the adaptability of the model to the target appearance change and reducing the tracking drift risk.

A video file may be presented via a user application that displays one or more video frames of the video file. A user request to perform an object detection for objects of a specific object type in a video frame of the video file may be received from the user application. A machine-learning model of a plurality of machine-learning models that is configured to detect objects of the specific object type may be applied to the video frame to detect an object of the specific object type in the video frame. Each of the plurality of machine-learning models may be trained to detect objects of a corresponding object type. Subsequently, an object tracking algorithm may be applied to one or more additional video frames of the video file to track the object of the specific object type across the one or more additional video frames.

The invention discloses a security unmanned aerial vehicle system based on STM32, and relates to the technical field of security unmanned aerial vehicles. The strategic value of the unmanned aerial vehicle in the public safety field is verified, the technical iteration and application deepening of the unmanned aerial vehicle can continuously promote the modernization of social governance, and the novel 7 * 24-hour all-weather security unmanned aerial vehicle system is provided for solving the problems that an existing unmanned aerial vehicle is poor in endurance, single in application scene, low in working efficiency and the like. The problem of endurance is solved by adopting a new technology and a new thought, and the system has the advantages of low cost, low power consumption, high efficiency, energy conservation, environmental protection and the like; the YOLOv5 algorithm and the ArcFace algorithm can form complementation in function, complete more complex tasks together, and realize more efficient and accurate face recognition and pedestrian analysis; the Hungary tracking algorithm and the Kalman filter are combined and used, so that multi-target tracking can be effectively realized, and the tracking accuracy and reliability are improved.

The invention belongs to the technical field of crossing of artificial intelligence and computational mechanics, and particularly relates to an artificial intelligence-based MPM simulation method and system for a mold bag sand cofferdam, a three-dimensional simulation system of the mold bag sand cofferdam is constructed in combination with an MPM method, and the penetration state of sand and geotechnical cloth is detected in an accelerated manner by utilizing a penetration tracking algorithm and spatial Hash. An artificial intelligence algorithm is introduced, a reward function is designed with minimum penetration depth and penetration times as targets, penetration prediction of MPM particles is driven, the speed and stress of the MPM particles are further dynamically regulated and controlled on the basis of a predicted contact model and the artificial intelligence algorithm, penetration is prevented, numerical stability is ensured, and the prediction accuracy of the MPM particles is improved. The calculation precision is improved by adopting implicit integration and gradient descent methods; and bidirectional coupling of the sand and the geotechnical cloth is realized through a visualization module. The problems that in a traditional method, coupling simulation of sandy soil and a flexible structure is low in efficiency, easy to penetrate, unstable in numerical value and the like are solved, and the accuracy and reliability of simulation analysis of the mold bag sand cofferdam structure under the complex working condition are remarkably improved.

The invention relates to the technical field of photovoltaic power generation, and discloses a photovoltaic power generation efficiency management intelligent optimization method, which comprises environmental data acquisition and illumination prediction, photovoltaic panel angle intelligent optimization, temperature adjustment and heat dissipation optimization and maximum power point tracking. Firstly, environment data are collected, and the illumination intensity at the next moment is predicted through an illumination prediction function; then, based on the predicted illumination intensity, the angle of the photovoltaic panel is optimized through an intelligent algorithm to minimize the illumination error. Thirdly, designing a temperature change and cooling power regulation function, optimizing a photovoltaic panel heat dissipation system, and ensuring that the temperature is within the optimal working range; and finally, the input voltage, current and power information of the inverter are utilized, and the input voltage of the inverter is adjusted in real time through a maximum power point tracking algorithm, so that optimal power output is realized. According to the method, the efficiency and the stability of the photovoltaic power generation system are comprehensively improved through environment data analysis and an intelligent optimization technology.

The invention discloses an unmanned aerial vehicle autonomous information perception path optimization method fusing sparse Gaussian estimation and RLSAC, and particularly relates to the technical field of unmanned system multi-target tracking, and the method comprises the steps: employing Kalman filtering to carry out the optimization processing of target initial data obtained by a ground laser radar, achieving the filtering of the noise of a target, obtaining the dynamic trajectory of the target, and obtaining the optimal path of the target. Therefore, the output insufficiency of the RLSAC to the dynamic trajectory is made up. The data after Kalman filtering processing is used as original input data of a multi-unmanned aerial vehicle target tracking I PP algorithm based on information path planning, the own unmanned aerial vehicle adopts the I PP algorithm to track a target in real time and transmit target related data to provide input data for RLSAC, the RLSAC is used to process the input data, model fitting is carried out, and the RLSAC is used to carry out target tracking. And the RLSAC outputs rewards of different geometric morphology models adopted by the own unmanned aerial vehicle during mode recognition, and the model which is more consistent with the output of the RLSAC is determined by calculating the confidence coefficient.

The invention discloses a dynamic visual guidance method and system for cooperative assembly of multiple mechanical arms, and belongs to the technical field of industrial robot visual guidance, and the method comprises the steps: synchronously collecting an image sequence through a binocular visual system, synchronously recording the pose sequence of each mechanical arm through a robot controller, and obtaining an observation data set; obtaining a two-dimensional trajectory flow of vehicle body assembly area feature points and instrument body feature points through an optical flow tracking algorithm, and obtaining a hand-eye matrix correction through a nonlinear optimization algorithm; obtaining an updated hand-eye matrix and a local pixel compensation field through a spatial interpolation algorithm based on the re-projection error of each feature point; according to the method, self-adaptive compensation of flexible deformation of a vehicle body and dynamic reflection of an instrument is realized by adopting unsupervised clustering and spatial interpolation according to a re-projection error, and the precision of cooperative assembly of multiple mechanical arms and the self-adaptive capability of a system in a complex industrial environment are improved.

The invention discloses a traffic asset identification and tracking method based on YOLO and ByteTrack algorithms, and aims to improve the intelligence and working efficiency of road inspection, when an inspection vehicle enters a designated inspection area, a visible light camera is started to collect video images in real time; a YOLOv8 deep learning algorithm is used to carry out target detection on a video frame, a ByteTrack tracking algorithm fused with a motion compensation mechanism is used to carry out stable tracking on a detected target, and at the same time, a unique ID is allocated to each target. And cutting corresponding video frames based on a target detection frame generated by the tracker, naming the video frames according to the target ID and the event category, and uploading the video frames to an inspection system. Target tracking stability is improved through motion compensation, ID change caused by camera jitter or detection errors is effectively reduced, repeated reporting of inspection events is reduced through tracking information, and accuracy and efficiency of an inspection system are remarkably improved.

The invention discloses a multi-unmanned aerial vehicle target tracking method based on cross-view collaboration, which is suitable for robust target monitoring in a dynamic scene. The method comprises the following steps: firstly, establishing a spatial mapping relation among visual angles of multiple unmanned aerial vehicles through feature point matching and projection transformation; after each unmanned aerial vehicle node independently executes deep learning-based target detection, converting a detection frame center point coordinate of an auxiliary view angle into a main view angle coordinate system, and calculating an Euclidean distance between the detection frame center point coordinate and a main view angle detection result; a main visual angle potential leak detection area is identified by setting a dynamic distance threshold value, and a local high-sensitivity re-detection module is triggered based on leak detection confidence; and finally, realizing space-time tracking of observation data by adopting a single-view target tracking algorithm. According to the invention, the problem of leak detection caused by single-view-angle shielding is effectively solved through a cooperative verification mechanism of a cross-view-angle detection frame center point space relationship, and the continuity and reliability of target tracking in an unmanned aerial vehicle view angle complex environment are remarkably improved while the real-time performance is ensured.

The invention provides a video point trajectory constraint-based video description method and system. The method comprises the following steps of: sampling a key frame image and acquiring a space-time trajectory of continuous inter-frame pixel points by using a point tracking algorithm; performing average pooling operation on the visual features of the corresponding frames of the same track fragment; performing semantic alignment on the text features, the visual features and the track features and then performing multi-head attention feature fusion; semantic correlation score calculation is carried out on the visual areas corresponding to the track fragments, correlation scores are arranged in a descending order, the correlation scores are accumulated, and a threshold value is set; jointly optimizing a video point tracking model by using language generation loss and focusing loss; and decoding the multi-source features after focusing optimization to obtain a final video description result. The method introduces a video point trajectory aggregation strategy, explicitly models the dynamic characteristics of a target in a space-time dimension, retains the spatial appearance and time coherence of an object, and effectively solves the problems of semantic fracture and description fragmentation in a complex scene.