79results about How to "Improve object tracking performance" patented technology

The invention relates to a distributed MIMO radar target tracking accuracy joint resource optimization method. The method includes the following steps that: an objective function is derived, a resource allocation optimization model containing three variables, namely, array element, power and bandwidth is constructed; the model is simplified, an array element selection problem is converted into a convex optimization problem from a 0-1 problem through convex relaxation; and the optimization model is decomposed into iterative optimization sub-problems of array element selection and power allocation through using cyclic minimization, the problems are solved through using an SPCA (sequential parametric convex approximation) method until the objective function is not optimized any more, final array element selection and power allocation results are obtained, and a final bandwidth allocation result is calculated, and a target position at a next time point is estimated. According to the distributed MIMO radar target tracking accuracy joint resource optimization method of the invention, the accuracy of the target tracking of an MIMO radar can be effectively improved through resource allocation, and the utilization rate of resources can be also improved; and an appropriate array element subset is selected, and power and bandwidth resources are allocated optimally, and therefore, the accuracy of the target tracking can be further improved under a situation that radar system resources are limited.

The invention provides an MCAUPF based target tracking method. An MCAUPF is used to complete state estimation in the target tracking process. In the target tracking process, target tracking equation and measurement equations are reconstructed into a nonlinear recursion model, in the framework of unscented particle filtering, a significance probability density function needed by filtering particlesgenerated by maximum correntropy unscented Kalman filtering is used, a Kullback-Leibler distance resampling method is used to resample the generated particles, the state of the tracking target is estimated according to a UPF algorithm flow, and the target is tracked in real time. The MCAUPF method is applied to target tracking in which outliers occur in noise measurement, the precision is highercompared with that of present PF, improved PF and robust filtering, and the computing complexity is lower than that of a present improved particle filter algorithm.

The invention provides a target tracking method, device and system. The method comprises the following steps of detecting a target object in a current frame image by a detection model, and judging whether the target object is a designated target or not; if yes, acquiring the tracking target matched with the specified target in the specified frame image before the current frame image; if the matching tracking targets are a plurality of, determining a designated target from the plurality of tracking targets according to the relative positions of the detection data of the plurality of tracking targets; estimating a position of a tracking target other than a specified target in a current frame image based on a position of the tracking target other than the specified target in the specified frame image among a plurality of tracking targets, determining an occluded target from the tracking target other than the specified target according to an estimation result. The invention can realize thetracking of an unoccluded target object and an occluded target, reduces the occurrence probability of tracking loss, and improves the accuracy rate of target tracking.

An image monitoring system may include an object area detection processing section and an object area tracking processing section. The object area detection processing section may be configured to detect object area information from an image obtained by an input image signal on a basis of the input image signal. The object area tracking processing section may be configured to perform object area tracking processing by associating a current object area with a past object area on a basis of the object area information detected by the object area detection processing section. The object area tracking processing section may associate the current object area with the past object area using unique identifying information. The object area tracking processing section may continue to retain identifying information given to a predetermined object area that has disappeared, and may give the retained identifying information to the predetermined object area when the predetermined object area reappears.

The invention discloses a target tracking method based on local feature learning. A target object and background are resolved into a large number of local units having dimension and shape invariance which serve as training samples of a target and background classification model, and a mode of deep learning is adopted to learn local expression of the target object and background from the training samples. The degree of confidence that each specific area in an image belongs to the target object is judged, thereby realizing accurate positioning of the target object. Since local expression obtained by learning of a large number of samples has a high target identification capability, the tracking method has relatively high adaptability to situations such as target deformation and target occlusion. When an object appearance model is updated, local areas with a relatively high target confidence degree are extracted as training samples of the model, and the model is updated. In the process of target tracking, the method continuously updates the appearance model, learns key characteristics of the target object, and can obtain a relatively good tracking effect under the circumstance that an appearance change is relatively big.

The invention discloses a target predicting and tracking method based on a probability graph model, which builds a target tracking model by a probability graph model modeling method, estimates the target position by the particle filtering prediction algorithm, and infers the possible positions of the target at any time by a distributed particle filtering tracking method, thus realizing the target tracking. The invention is not restricted by linearization error and Gaussian noise assumption, does not need to assume the distribution form of sample points, can deal with any density distribution and flexibly adapt to non-linear dynamic models and multimodal observation models of target tracking. The method can obtain higher overall predicting degree of accuracy and show better target tracking performance.

The invention belongs to the field of multi-sensor information fusion and provides a target tracking method and system transmitting edge distribution and existence probability. The method includes the following steps that firstly, according to edge distribution and existence probability of all targets at the previous moment, edge distribution and existence probability, at the current moment, of the targets already existing at the previous moment are predicated, edge distribution and existence probability are appointed to newborn targets at the current moment, updated edge distribution and existence probability are measured and determined in combination with the positions at the current moment, then edge distribution and existence probability of all the targets are determined at the current moment, finally, the targets with existence probability smaller than a first threshold value are cut out according to edge distribution and existence probability of all the targets at the current moment, edge distribution and existence probability of all the targets after cutting serve as recursion input at the next moment, and meanwhile the targets with existence probability larger than a second threshold value are extracted as output at the current moment.

The invention discloses a fast correlation neighborhood feature point-based sliding window target tracking method. The method includes the following steps of: S1, target window template generation; S2, fast correlation neighborhood feature point extraction; S3, optimal point of interest screening; S4, point of interest sliding window search; S5, feature point template matching update; and 6, decision voting output. The invention also discloses a fast correlation neighborhood feature point-based sliding window target tracking system. With the method and system of the invention adopted, the problem of poor real-time performance and low stability of target tracking in a complex condition can be solved. According to the method and system, fast correlation neighborhood feature points, adopted as point of interests, are detected, and therefore, the robustness of target feature description in a complex condition can be enhanced; point of interest screening is carried out through using a window cross-correlation relation, and therefore, the accuracy of the target description can be improved; and sliding window search and adaptive multi-scale template matching online update are adopted in template construction, and target window output is realized through adopting decision voting, and therefore, the accuracy and stability of target tracking in the complex condition can be improved.

The invention discloses a target tracking method and system. The method comprises the steps: carrying out the position coordinate tracking of a target object through a preset tracking algorithm, and obtaining the tracked position coordinates; determining the scale of the current target object, and carrying out the tracking of the target object according to the scale and the tracked position coordinates. Therefore, after the tracked position coordinates of the target object are obtained through the tracking algorithm, the scale of the current target object will be further determined, and then the tracking of the target object is carried according to the determined scale of the current target object and the tracked position coordinates, thereby enabling the final tracking effect to truly reflect the scale of the current target object, and greatly improving the target tracking effect.

The invention discloses a passive multi-source multi-target tracking method based on dynamic multidimensional allocation, relates to the field of passive multi-source multi-target tracking and aims at solving the problems of low correlation accuracy of a target track and high time complexity of an algorithm in an existing passive multi-source multi-target tracking algorithm. The passive multi-source multi-target tracking method disclosed by the invention comprises the following steps: firstly, corresponding to a preselection wave gate of a track p of a target; secondly, constructing a cost function and a two-valued variable; thirdly, obtaining a (S+1)-D allocation formula and giving out constraint conditions; fourthly, carrying out dimension reducing processing on the (S+1)-D allocation formula to obtain a two-dimensional allocation formula; fifthly, calculating a dual solution of the two-dimensional allocation formula; sixthly, updating a lagrangian multiplier by using a subgradient vector; seventhly, obtaining an allocation combination of the track p and a corresponding observation value; eighthly, carrying out maximum likelihood estimation by using a likelihood function; ninthly, estimating a target state according to a kalman filtering method and updating the track by using a state estimating value, thus realizing multi-target tracking. The passive multi-source multi-target tracking method disclosed by the invention is applied to the fields of aviation and airborne radar.

The invention provides a deep clustering based polar coordinate dynamic programming passive coherent location method. The method comprises the steps of firstly, providing a process of environment variable learning so as to be used for determining the expansion amount of space transfer, and solving an effective transfer range in a nonlinear measurement space by combining an error propagation theory; secondly, adding an anti-interference factor into a backtracking threshold determined according to an extreme value theory for the first time, and finally determining a backtracking threshold of a deep clustering based dynamic programming algorithm; thirdly, providing a deep clustering based dynamic programming algorithm, and performing target detecting and tracking in the measurement space; and finally, performing dimension augmentation on a target state, thus comprising the state of a radiation source, providing an interaction prediction covariance based extended Kalman filtering algorithm, and performing joint estimation on the augmented state.

The invention relates to a wireless sensor node optimization selection method orientated at visual tracking. The wireless sensor node optimization selection method orientated at visual tracking comprises the following steps that whether triggering conditions meeting node selection exist at the current time point is checked; each video sensing node reconstructs image frames, a motion target is extracted by adopting a self-adaptation Gaussian mixture background model, the sizes of target blocks are counted and a hue histogram model of the target blocks is built; the nodes achieve target motion prediction by utilizing unscented kalman filtering at the same time; relation among targets in different image planes is achieved through homography conversion among node visual angles, and the corresponding relationship among multiple adjacent targets is accurately determined by combing the matching of the hue histogram; a confidence coefficient model of the nodes is built according to the sizes of the target blocks and indetermination characteristic of tracking, and the node with the largest determination degree is selected as the best node and takes over a subsequent tracking task. The wireless sensor node optimization selection method orientated at visual tracking effectively overcomes partial shielding in the motion process of the targets, has good target tracking performance, and meanwhile has low communication expenses and calculation complexity.

The invention relates to a fulfilling method for enhancing the ability of a video sensor network to cover the motion tracks of mobile targets, which comprises two operation steps, namely, a preparing stage and an adjusting stage. The invention is characterized in that: proceeding from the directivity perceiving feature of video sensor node and based on the hypothesis that the initial positions of the video sensor nodes are invariably fixed, the invention revolutionarily puts forth a simple handling approach, by which the motion tracks of the continuously movable targets are discretized into a plurality of motion track points; and then a concept of center of mass is introduced, and the issue of strengthening the coverage is transformed into another issue, which centers on the calculation of the virtual resultant forces between the centers of mass and the motion track points as well as among centers of mass. Moreover, the sensing directions of the video sensor nodes, which are likely to detect the movable targets across the entire network, are adjusted to enable the nodes to participate into the monitoring over the mobile targets. The invention has the advantages of improving the probability of coverage when the mobile targets cross the video sensor network along any tracks, enhancing the tracking ability of the video sensor network over the mobile targets, and ensuring the integrity and reliability of the monitoring data of the mobile targets.

The invention provides a Camshift-based target tracking method. The method includes the following steps that: a search window is initialized; the color probability distribution map, namely, reverse projection map, of a search area is calculated; the meanshift method is adopted to obtain the new size and position of a target, namely a new window; the previously obtained window is adopted as a new window for a next frame of video image; and with the window obtained through using the meanshift method adopted as a center, the size of the search window is adaptively adjusted, and an actual target window under an ideal condition is estimated by using a color histogram, similarities between models in the result of the meanshift method, the adjusted search window and the estimated target window and the actual target model are compared, and the window corresponding to the model which is the most similar to the actual target model is adopted as a result. With the method of the invention adopted, the stability of target tracking can be improved.

The invention provides an unmanned aerial vehicle target tracking method and device and an unmanned aerial vehicle. The method comprises the following steps: acquiring an image acquired by a sensor for acquiring the image of an unmanned aerial vehicle; determining first spatial position information of the target object according to an image area, corresponding to the recognized target object, in the image; predicting the next spatial position information of the target object according to the first spatial position information; and controlling the unmanned aerial vehicle to fly towards the target object according to the next spatial position information. According to the method of the embodiment, the target tracking effect can be improved.

The invention discloses a moving target track tracking method based on track correlation. The pulse compression is performed on an obtained FMCW radar intermediate frequency echo signal to obtain thedistance information of a target; and then the signal-to-noise ratio is judged to obtain a long-distance low signal-to-noise ratio signal and a close-distance high signal-to-noise ratio signal; CFAR detection and linear difference is performed on the high signal-to-noise ratio signal, and multi-frame accumulation is performed on the high signal-to-noise ratio signal; and finally, a track correlation processing method is used for target correlation, and sudden changes in altitude are removed to obtain a continuous moving target track. The track tracking method greatly improves the detection accuracy, the altitude value of the target can be calculated in real time, and the moving track of the target is marked; and meanwhile, the problems of discontinuous tracking of the target track and disorder of the target track are solved.

The invention provides an electric power business hall passenger flow statistics method and system based on target dynamic tracking. The method includes detecting a pedestrian target in the video frame image after image preprocessing; performing de-duplication processing on the detected targets existing in the tracking target list; tracking targets in the tracking target list by using a KCF algorithm and updating target parameters; obtaining the average moving speed of the pedestrian target in the video by analyzing the target parameters; predicting a video frame number and a specific positionwhen the pedestrian target arrives at the counting line according to the average moving speed; using a target detection model of a KCF algorithm to verify the position of the video frame and the pedestrian target existence condition in the neighborhood range of the video frame, and performing counting statistics on the business hall passenger flow according to the verification result and the initial position of the target; according to the invention, the target tracking process is reduced, the calculated amount of the system is reduced, the target tracking performance of the system is improved, and the high execution rate of the system is ensured.

The embodiment of the invention relates to the technical field of unmanned aerial vehicles, and specifically discloses a target tracking method and device, and an unmanned aerial vehicle. The method is applied to an unmanned aerial vehicle comprising a visible light camera and an infrared camera. The method comprises the steps of: controlling the visible light camera to perform visual tracking fora target object, recording first tracking information of the target object in real time, controlling the infrared camera to perform infrared tracking for the target object, and controlling second tracking information of the target object in real time; determining if the visible light camera loses the target object, controlling the visible light camera to relock the target object according to thesecond tracking information and continuously perform visual tracking; or if the infrared camera loses the target object, controlling the infrared camera to relock the target object according to the first tracking information and continuously perform visual tracking. According to the technical scheme, the target tracking method and device, and the unmanned aerial vehicle provided by the embodimentcan improve the target tracking performance of the unmanned aerial vehicle and can expand the application range.

The invention discloses a video monitoring method and a system. According to the method and the system, the operation speed and the response speed of the video monitoring system can be increased, the target tracking performance is optimized, and the timeliness of target tracking is improved. The method includes: a second video monitoring device receives coordinates of the current physical positions of a to-be-tracked object sent by a connected first video monitoring device at regular intervals of the preset duration, the visual angle of the first video monitoring device is fixed, the visual angle of the second video monitoring device is adjustable, and the first video monitoring device and the second video monitoring device perform coordinate calibration of the physical positions by employing the same coordinate calibration method; and the second video monitoring device determines the to-be-tracked object from two or more moving objects according to the received coordinates of the more than one current physical positions, and locks and tracks the to-be-tracked object.

The embodiment of the invention discloses a particle filtering method and device based on a TSK fuzzy model and a storage medium. The method includes: constructing an importance density function of particle filtering based on the TSK fuzzy model; extracting N particles from the importance density function to form a particle state set of the target; calculating weights of particles in the particlestate set, and normalizing the weights; and calculating the state and covariance of the target at the moment k based on the normalized weight and the particle state set. Through the implementation ofthe invention, the TSK fuzzy model is introduced to model the dynamic system of the target, an effective importance density function is constructed, the robustness and diversity of particle sampling are effectively improved, and the target tracking performance in a non-linear non-Gaussian environment is enhanced.

The invention provides a monitoring device linkage method and device. The method comprises the steps of calculating a relative location relationship between a target object in the field of view of a first monitoring device, and the first monitoring device according to the state parameter of the first monitoring device; calculating a first motion parameter of a second monitoring device according to the relative location relationship between the target object and the first monitoring device, a predicted relative location relationship between the first monitoring device and the second monitoring device, and the state parameter of the second monitoring device; controlling the second monitoring device to adjust the state parameter according to the first motion parameter, wherein the first motion parameter is used for enabling the second monitoring device to adjust the state parameter according to the first motion parameter and then capture the image of the target object. According to the method, self-adaptive high-precision device linkage and the tracking of the moving target object can be realized at the same time, further when the method is carried out on multiple monitoring devices, a multi-device self-adaptive linkage scheme can be realized.

The invention discloses a target tracking method and device, and belongs to the technical field of electronics. The method comprises the following steps: determining the position of a detection framein each frame of video image of the video and the category and image features of a target in each detection frame through a target identification model, wherein the target recognition model is obtained by training a neural network model by using a target detection data set and a retrieval data set, each piece of target detection data is marked with a detection frame position and a corresponding target category, and each piece of retrieval data is marked with a target identifier; and tracking the target contained in the video according to the position of the detection frame in each frame of video image of the video and the category and image features of the target in each detection frame. The target recognition model can locate the target and extract the image features of the target at thesame time, the image features of the target can be directly provided for evaluation reference of target tracking under the condition that extra time consumption and resource consumption are not increased, and the real-time performance and accuracy of target tracking can be improved.

The invention provides a target locating and tracking method based on a high-effect radio frequency network. The radio frequency network comprises at least two nodes. The nodes are distributed around a monitoring region and form more than two directed links. Each node sends a first message to a base station, wherein the first message includes a base station identification number, a sending node identification number and a receiving node identification number, and an attenuation link is formed by the sending node identification number and the receiving node identification number. By the high-effect radio frequency network, energy loss can be reduced and service life of the nodes can be prolonged.

The invention discloses a peripheral target vehicle state estimation method based on vehicle-to-vehicle communication and GMPHD filtering. The method comprises the following steps of initializing multi-target Gaussian mixture posterior intensity, acquiring point cloud data, preprocessing the point cloud data, converting the point cloud data into a vehicle body coordinate system, predicting multi-target prior intensity, obtaining multi-target posterior intensity, performing information communication and coordinate time synchronization on peripheral target vehicles, converting a synchronized measurement set into the vehicle body coordinate system, updating the multi-target posterior intensity by applying the measurement set, and obtaining final multi-target posterior intensity; setting an abandon threshold value, a merge threshold value and the maximum Gaussian mixture number, and pruning and merging the multi-target posterior intensity; and extracting a target state to obtain a multi-target state collection. Compared with the prior art, the peripheral target vehicle state estimation method can integrate vehicle perception information and state information sent by a peripheral targetvehicle in the Internet of vehicles environment, and improve the accuracy and robustness of target positioning and tracking.

The invention discloses a perception fusion method for an automatic driving system, and the method comprises the steps: calibrating a millimeter-wave radar and a camera based on a laser radar; based on a laser radar, calibrating information collected by the millimeter wave radar and the camera, respectively identifying a target in each piece of information, and obtaining target parameter information of each target; on the basis of the target parameter information, adopting an unscented Kalman filtering algorithm and a global matching method to carry out sensing fusion and matching, and obtaining the state update quantity of the matched and tracked target at the current moment; and updating the matched and tracked target based on the state update quantity of the matched and tracked target, comparing the tracked target at the current moment with the tracked target at the previous moment, and identifying a new tracked target. According to the invention, based on the data received by the plurality of sensors, the targets of the plurality of sensors are fused by using the unscented Kalman filtering algorithm, the target detection and target tracking effects are improved, and the stable operation of the automatic driving perception function is ensured.

The invention discloses an optical object tracking method and an optical object tracking system. The optical object tracking method includes: acquiring a continuous image group containing a reference light source in a time period, and recognizing image feature changes of the continuous image group, wherein the reference light source is maintained in a constant-brightness mode and generates known brightness changes within the time period. The optical object tracking method further includes: comparing the image feature changes with the known brightness changes to determine whether starting of a preset function is needed or not, wherein the preset function refers to calculation of position information of the reference light source in image of the continuous image group. By means of the reference light source flickering randomly, object tracking efficacy of the optical object tracking system is effectively improved, and software and hardware of the optical object tracking system are simplified to reduce product cost.

The invention provides a multi-target tracking method and device in video monitoring. The method comprises the steps of: obtaining initial information of at least one interest object in a tracking video, wherein the initial information includes initial positions, speeds, accelerations and characteristic information; according to a sequence from big to small in characteristic saliency, selecting N0(t) interest objects from the interest objects as an initial tracking layer, wherein the N0(t) is determined by speeds and accelerations of interest objects, and the characteristic saliency is determined by the characteristic information; carrying out tracking on the N0(t) interest objects in the initial tracking layer one by one according to the initial positions of the N0(t) interest objects; and according to the correlation degree between untracked interest objects and the integral tracked objects, the untracked interest objects are selected into the next new tracking layer in sequence for tracking layer by layer. By adopting the multi-target tracking method and device in the video monitoring, the effectiveness and the accuracy of multi-target tracking are improved.

The present invention relates to a tracking method for a variable number of maneuvering targets. In a particle state prediction and update step, the sampling of a predicted particle state set is carried out according to a particle being variable, the association degree of a current observed value and maneuvering target state particles is considered, an association problem between an observed set and a maneuvering target state sampling particle set is solved by utilizing a fuzzy auction algorithm and a particle swarm optimization theory, a judging criterion for appearance and disappearance of the maneuvering targets is given, and the update of particle weight is realized; a hybrid sampling particle set is re-sampled by utilizing a sequential importance sampling theory, and a particle set containing model information and state information as well as approaching state posteriori probability distribution of each maneuvering target is obtained; the influence of the particle being variable is considered, and a maneuvering target local state posteriori estimated value and a mean-square error are obtained through particle state fusion according to a target model probability; and finally, local tracking information of sensors is subjected to weighting fusion, so as to obtain a global state estimated value of each maneuvering target, and realize accurate estimation of number and state variation of the maneuvering targets.

The invention provides a tracking method and device and intelligent equipment, and relates to the technical field of machine vision. The method comprises the following steps: determining a to-be-tracked target object based on an acquired image of intelligent equipment; acquiring a tracking template of the target object according to the proportion of the target object in the acquired image; whereinthe resolution of the tracking template is related to the proportion; and tracking the target object based on the tracking template. The target tracking effect of the intelligent equipment can be effectively improved, and the tracking reliability is improved.