43results about How to "Realize multi-target tracking" patented technology

The invention discloses an UAV onboard multi-target detection tracking and indication system and method, and belongs to the technical field of target detection tracking and indication. The invention discloses an UAV onboard multi-target detection tracking and indication system comprising a multi-target detection and tracking system and a multi-target laser indication system. The multi-target detection and tracking system comprises an infrared camera, a visible light image sensor and a high-speed parallel image processing and tracking feedback control circuit. The multi-target indication systemincludes an integrated laser, a fast reflector, a fast reflector control module, and a laser control module. In order to improve laser indication accuracy, the UAV onboard multi-target detection tracking and indication system also includes a laser pointing control system. The invention also discloses an UAV onboard multi-target detection tracking and indication method based on the UAV onboard multi-target detection tracking and indication system. The UAV onboard multi-target detection tracking and indication system and method realize multi-target all-weather detection tracking and high-precision stable laser indication under the condition of a UAV onboard platform.

The invention discloses a video multi-target tracking method based on multi-Bernoulli characteristic covariance which belongs to the technical field of artificial intelligent and intelligent information processing and aims mainly to solve the problems that in complex environment, the targets in video multi-target tracking are very close to each other and that the size change and tracking to them are not accurate. This method, according to the multi-Bernoulli filtering framework, is introduced by an integration graph and employs a particle filter method to track video multi-targets whose number are in constant change in combination with the characteristic covariance technology. Based on that, a self-adaptive mechanism and a size self-adaptive mechanism for targets in close contact are proposed to realize the self-adaptive processing of targets in close contact and the tracking windows respectively. Finally, a particle filter method is utilized to realize the self-adaptive recognition and tracking of the video multi-targets through their movement trajectories. The method of the invention is highly robust and has a good anti-interference ability. The method meets the design requirement of a practical engineering system and has good engineering application value.

The invention discloses a vehicle retrograde motion intelligent detection method based on tracking track analysis. The method comprises the following steps: detecting a target vehicle in a video imageby utilizing a primary classifier and a secondary classifier; extracting a target vehicle area, distributing a nuclear correlation filtering tracker for the target vehicle; wherein each target vehicle is matched with one nuclear correlation filtering tracker, the nuclear correlation filtering tracker is used for obtaining a tracking area of the target vehicle, then the movement track growth direction of the target vehicle is obtained according to the tracking area and compared with the initially marked retrograde movement direction, and if the movement track growth direction is the same as the initially marked retrograde movement direction, the target vehicle retrograde movement occurs; And if not, the target vehicle does not run reversely. By means of the mode, real-time detection of vehicle retrograde running is achieved, and the problems existing in manual detection are solved; And meanwhile, the reliability of vehicle detection and recognition is improved by using the cascade classifier, and simultaneous tracking of multiple vehicle targets is realized.

The invention discloses a convex optimization method for a three-dimensional (3D)-video-based time-space domain motion segmentation and estimation model. The method is implemented by the following steps of: 1) establishing the 3D-video-based time-space domain motion segmentation and estimation model according to an active contour theory and a mapping relationship between a background three-dimensional motion parameter and a two-dimensional light stream; 2) converting the model into a corresponding horizontal set description equation, calculating a corresponding gradient descent equation, calculating an equivalent equation of the gradient descent equation, calculating an energy function corresponding to the equivalent equation, and performing convex relaxation on the energy function to obtain a convexly-optimized time-space domain motion segmentation and estimation model; and 3) introducing a cost variable into the further relaxation of the convexly-optimized time-space domain motion segmentation and estimation model, minimizing the convexly-optimized time-space domain motion segmentation and estimation model by adopting a multi-variable alternate iteration algorithm, and performing iterative convergence to obtain a final split surface according to a selected threshold function. The method has the advantages of high adaptability to changes in a target number, independence of a segmentation result on an initialized contour, and high operation efficiency.

The invention discloses a multi-target detection and tracking method under conditions of low observability and high clutter, and belongs to the technical field of radar and sonar. The idea of the method is that a plurality of measurements from different propagation paths to a receiver are considered as the possible target measurements during the processing of the target-measurement correlation, and are enabled to be correctly correlated with each known multi-path measurement function, thereby obtaining the accumulation of target information, and improving the target detection capability; and then the target tracking is carried out in a mode of a sliding window. The method employs the target information which is transmitted to a sensor through different paths, enables the measurement information to be correctly correlated with each known multi-path measurement function, thereby obtaining the accumulation of target information, and improving the detection capability of a target under the conditions of the low observability and high clutter. The method can effectively reduce the impact between the adjacent objects.

The invention belongs to the technical field of intelligent information processing, and relates to a multi-target tracking method based on a variational Bayesian label multi-Bernoulli superposition model. The noise covariance of the superposition model is estimated. On the basis of an original superposition model, the covariance of measurement noise is unknown, unknown parameters are estimated based on variational Bayes, the prediction and updating process of the superposition model marked with the multi-Bernoulli filter is achieved, state extraction is conducted, and therefore the tracking problem of the superposition model under unknown measurement noise is solved. The method has the characteristics of wide application range, strong robustness, high estimation precision and the like, caneffectively solve the problem of non-cooperation in an actual superposition model scene, realizes multi-target tracking and parameter estimation in a complex scene, can meet design requirements, andhas a good engineering application value.

The invention discloses a kernel correlation filtering multi-target tracking method fusing motion information, and belongs to the field of computer vision and intelligent information processing. On the basis of detection and tracking, the KCF is introduced to track multiple targets, excessive dependence on a detector is reduced, and accurate tracking of the multiple targets is achieved; in the tracking process, speed information and an SCCM mechanism are combined into a tracking framework, so that the problems of tracking of a shielded target and drifting of a tracking frame are solved; and finally, the false target is judged by adopting the IOU and the historical trajectory information, so that trajectory fragments are reduced. Experiments show that the method has good tracking effect androbustness, and can widely meet the actual design requirements of intelligent video monitoring, man-machine interaction, intelligent traffic control and other systems.

The invention provides a multi-target tracking method and device based on video. The method comprises the following steps of: extracting a target module, initiating target parameters, detecting Adaboost, carrying out dynamic prediction on a particle set according to a motion model, updating a weight value of each mixing component, updating a motion state of each target, and updating the module. A multi-mode problem in multi-target tracking can be effectively solved, the effectiveness of calculation is maintained, the multi-target tracking can be realized, and simultaneously, the requirement of instantaneity can be ensured.

The invention relates to a mask wearing condition monitoring system and method based on a probabilistic neural network, and the system comprises: an input management module which is responsible for the reading and preprocessing of an input video; the target detection and segmentation module that is responsible for detecting and identifying each target in the video and segmenting the pedestrian mask part for detection; the multi-target tracking association module that performs many-to-many association on the detection result in each video in the video; the mask monitoring module that is used for carrying out low confidence coefficient and track discretization after completing preliminary track association; the output module that is arranged at a unified data center node, supports multi-pathcross-camera data output, performs multi-target tracking on each scene, and uniformly converges and outputs a calculation result; the system setting module that is used for configuring training of anetwork model used in the plurality of modules. Technical support is provided for the fields of video monitoring, behavior analysis and the like.

The invention relates to a dynamic quantity sound source tracking method based on a microprocessor array. The method comprises the following steps: using a beam-forming algorithm to process a microphone array receiving signal and calculate a spatial spectrum; calculating the matching probability of spatial spectrum peaks and tracking sound sources so as to realize data association between the spatial spectrum peaks and the tracking sound sources; updating the particle weight of each tracking sound source and updating sound source positions according to the matching result; detecting the probability that the spectrum peaks are judged to be new sound sources, activating particle filters of the new sound sources and evaluating the possibility that the particle filters exist; monitoring the active states of the tracking sound sources and deleting non-active sound sources. By using the method, multi-target real-time tracking with a dynamic sound source quantity can be realized, and the method is widely applied in complex scenes such as human-machine interaction, teleconferences and virtual reality.

The invention discloses a target tracking method in a multi-camera scene based on SIFT, and the method comprises the steps: obtaining a target detection total data set which is formed by photographing pictures containing different types of detection targets; using the target detection total data set to train a target detector YOLO-V5s model; obtaining a target tracking re-identification data set, wherein the target tracking re-identification data set is formed by taking photos containing different types of tracking targets and extracting the part containing the tracking target in each photo; training a target appearance feature extraction network in a DeepSort algorithm by using the target tracking re-identification data set; and acquiring a video shot by splicing multiple cameras by using an SIFT algorithm, and tracking a tracking target in the video by using the trained YOLO-V5s model in combination with the trained DeepSort algorithm. According to the invention, a larger target detection range can be obtained, and the target tracking precision is improved.

The invention discloses a target tracking method in a multi-camera scene, which realizes splicing of pictures of different cameras by combining YOLO-V4 with an improved DeepSort algorithm and an image splicing algorithm, and finally realizes multi-target tracking in a spliced video. In the aspect of data sets, a self-made intelligent trolley data set and a self-made vehicle re-identification data set containing an intelligent trolley are adopted. According to the vehicle re-identification method, multi-target tracking in a multi-camera scene is realized by constructing a rich data set, improving a model and splicing and fusing pictures, and the vehicle re-identification accuracy is well improved.

The invention discloses a multi-target data association method based on a photoelectric sensor. The multi-target data association method comprises the following steps: judging angle measurement information according to the angle measurement information acquired by different sensors and judging whether the angle measurement information comes from the same target or not; obtaining three-dimensionalspace coordinate information of the target after judging that the three-dimensional space coordinate information comes from the same target; and updating the track of the target based on a nearest neighbor plot-track association algorithm of the weighted sum distance. According to the invention, multi-target tracking can be realized under the condition that only the measurement information of theazimuth angle and the pitch angle of the target can be obtained. The multi-target data association method based on the photoelectric sensor can be widely applied to the field of multi-target trackingand multi-sensor information processing.

The invention discloses a power distribution network field operation video intelligent safety control method driven by a knowledge graph, and the method comprises the steps: firstly constructing a power operation safety control knowledge graph, and aims to decompose objects, flows, matters needing attention and other semantic information in an operation task, thereby enabling the operation to be more visual and visualized; establishing a basis for subsequent personnel information security matching, skill point action matching and dynamic evaluation; dividing a power grid operation site area and a video intelligent management and control category; finally, using a deep learning algorithm, monitoring videos of all areas of an operation site are combined, safety control is conducted on operators in the whole process in real time, original passive on-site operation manual supervision is changed into active intelligent automatic monitoring, support is provided for early warning of on-site operation safety risks, and therefore the risks of the on-site operators are reduced, and the safety of the operators is improved. Therefore, the method is of great significance to the improvement of the safety management and control level of the power distribution network on-site operation.