47 results about "Probability hypothesis density filter" patented technology

The invention relates to a multi-target tracking system in a dynamic video sequence, belonging to the technical field of image processing. The multi-target tracking system comprises an input module, a moving target detection module, a PHD (Probability Hypothesis Density) filter module and an output module, wherein the moving target detection module comprises a background area initialization submodule, a background area updating submodule, a foreground image extraction submodule, a morphology processing submodule and a connected domain analysis submodule; and the PHD filter module comprises a Gauss element parameter prediction submodule, a Gauss element updating submodule, a Gauss element trimming submodule and a state extraction submodule. The invention provides more reliable measurement information for the video tracking system through the improved moving target detection module, solves the problems of possible complex computation of particle filters and probability hypothesis density filters and unreliable state extraction under the condition that targets are crossed, ensures the effectiveness and the reliability of the video tracking system, avoids the data association computation and is widely used in various fields of multi-target video monitoring.

The invention discloses an extended target probability hypothesis density filtering method based on cubature Kalman filtering. The method comprises the steps of (1) pre-setting the Gaussian mixture form of posterior strength of the moment k-1 and obtaining the mean value and covariance of the ith Gaussian item, (2) conducting one-step prediction on the weight, mean value and covariance of the ith Gaussian item obtained from the first step, and (3) conducting measurement updating on the prediction result obtained from the second step to obtain the estimated value of each Gaussian component (as stated in the specification) of the moment k. According to the extended target probability hypothesis density filtering method based on cubature Kalman filtering, extended target tracking is achieved under a nonlinear system, extended target tracking is achieved when the Jacobian matrix of a nonlinear function does not exist or is hard to solve, and a new realization approach is provided for extended target tracking under nonlinear conditions.

The invention discloses a multi-target tracking method in a radar system. According to the method, on the premise of processing radar measurement values according to sectors, cross-sector multi-target tracking is achieved by conducting corresponding time prediction, measurement correction, fused trimming and target state extraction on gauss components of different sectors of radar, radar full-plane multi-target tracking under the asynchronous measurement condition is finally achieved, and therefore the problem that the traditional probability hypothesis density filter can not be directly applied to the radar system is solved. The measurement values are not required to be simultaneously measured at the same time in the whole process, the number-variable multi-target tracking problems can be solved, and compared with a traditional method such as joint probabilistic data association and multiple hypothesis tracking, the calculation burdens are small; in addition, track starting, maintaining and ending are naturally completed without being independently listed.

The invention relates to the technical field of target tracking, in particular to a method for tracking multiple targets under unknown measurement noise distribution. Through the advantages and disadvantages of a risk value measurement particle, the particle is evaluated through a risk evaluation function and weight updating is performed. The entire process is independent of the measurement noise distribution and multiple targets can be tracked stably under unknown complex measurement noise distribution. The method has higher robustness and stability compared with those of the ordinary particle probability hypothesis density filtering technology.

The invention discloses an infrared weak and small target detection and tracking method. According to an improved four-order partial differential equation method, an original infrared image is processed to obtain an infrared image with background suppression and target enhancement achieved, the position information and number information of candidate targets in the obtained infrared image are extracted according to a block adaptive threshold segmentation method, finally multi-target state and number estimation is carried out on the extracted position information and number information of the candidate targets according to Gaussian Mixture Cardinalized Probability Hypothesis Density (GM-CPHD) filter, and the states and number of multiple infrared weak and small targets are accurately and stably estimated through the GM-CPHD filter. The invention further discloses an infrared weak and small target detection and tracking device. Through the method and device, implementation is easy, the effect is obviously superior to that of a traditional background suppression method, the traditional multi-target tracking data association problem is avoided, and the multi-target states and number changing along with time can be estimated in time more stably.

The invention discloses an improved Gaussian mixed potential probability hypothesis density filtering method. The method comprises the following steps: 1) forming a target state set and a target strength function; 2) initializing probability hypothesis density and potential distribution of an initial target; 3) carrying out predication on the probability hypothesis density and potential distribution of the target state set at the time of k+1 to obtain probability hypothesis density and potential distribution at the time of k+1; 4) updating the probability hypothesis density and potential distribution of the target state set at the time of k+1 to obtain probability hypothesis density and potential distribution at the time of k+1, carrying out unbiased conversion on a true covariance matrix and true deviation, and setting an ellipsoid threshold value to simplify a measurement set and reduce observation number of a current observation set; 5) carrying out trimming and combining on Gaussian items of the target strength function, and extracting target state estimation and carrying out performance evaluation; and 6) repeating the steps 3)-5), and tracking the target until the target disappears. The method facilitates direct application of radar data information, and reduces calculation amount of a filter.

The invention is suitable for the field of multi-sensor information fusion and provides a probability hypothesis density filter target information maintaining method. The probability hypothesis density filter target information maintaining method includes: step 1, forecasting posterior moments and Gaussian items at the current moment according to posterior moments and Gaussian items of the last moment; step 2, updating the posterior moments and the Gaussian items according to the posterior moments and the Gaussian items of the current moment and a measurement set of the current moment; 3, cutting down or combining the updated Gaussian items; step 4, extracting a weight Gaussian item as output of a filter according to the cut down and combined Gaussian items, wherein means and variances in the corresponding Gaussian items are state estimation and error estimation of a survival target. By means of the hypothesis density filter target information maintaining method, information of a target with detection leaked is retained in a posteriori updating moment by amending an updating function of a probability hypothesis density filter, so that information of the target with detection leaked cannot be missing, effectiveness in target number estimation and target state extraction is improved, and further a multi-target tracking capability of the Gaussian probability hypothesis density filter is improved.

The invention discloses a flight track extraction method based on a probability hypothesis density filter associated with global time and space. The flight track extraction method comprises the following steps: S1, extracting a target state by using the probability hypothesis density filter; S2, measuring consistency and calculating consistency confidence; and S3, obtaining a global flight track extraction strategy. By adopting the scheme, the target state is firstly obtained by using the probability hypothesis density filter; then the consistency between a predicted peak value and an estimated peak value is measured by using global time-space information and the consistency confidence is calculated; simultaneously four decision rules based on expert knowledge of the flight track extraction are shown, namely a rule for judging inseparable targets, a rule for judging homology of the targets, a rule for judging disappearance of the targets and a rule for judging novel targets; a global flight track extraction strategy is shown based on the consistency confidence and four decision rules, thereby extracting flight tracks of a plurality of targets, improving the flight track extraction effect, improving the flight track extraction accuracy and playing an important role in multi-target tracking engineering application.

The invention provides a track identifying method of a probability hypothesis density filter suitable for the technical field of multi-sensors. The track identifying method includes the first step of determining a predicted gauss item and an identity identification of the gauss item according to a gauss item and an identity identification of the gauss item at the previous moment and adding a non-exclusive identity identification to each newly produced gauss item, the second step of determining updated gauss items and identity identifications of the updated gauss items according to the predicted gauss item, the newly produced gauss items and the corresponding identity identifications, the third step of cutting and combining the updated gauss items and the identity identifications, the fourth step of regulating the identity identifications whose weight is larger than a preset weight threshold value according to the cut and combined gauss items and the corresponding identity identifications, and the fifth step of extracting the gauss items whose weight is larger than the preset weight threshold value to serve as output of the filter and outputting the corresponding identity identifications. According to the method, the identity identifications are added to the gauss items, target states of different moments are correlated, and accordingly motion trails of targets are obtained.

The invention provides a target tracking-before-detecting method based on Gaussian cardinalized probability hypothesis density filter, comprising the following steps: initializing: letting k represent a k time wherein k starts from one and belongs to a set of 1, 2,...,D; determining Nk targets corresponding to the k time; recording the moving state of the pth target at the k time as what is described in the figure; calculating the likelihood function corresponding to the intensity of the moving state Xk of the Nk targets at the (i,j) position contained in a radar observation area in the k-time Cartesian coordinate system under the moving state of the pth target at the k time; then calculating the average state estimations of the Nk targets at k time and the covariance estimations of the Nk targets at k time; and calculating in sequence the probability pk(Nk) of the target number (Nk) at the k time and the estimated value of the targets Nk contained in the radar observation area Nx X Ny in the k-time Cartesian coordinate system; letting one added to k to obtain the estimated value of target N1 included in the radar observation area Nx X Ny in the Cartesian coordinate system at 1 time to the estimated value of the target number ND included in the radar observation area Nx X Ny in the Cartesian coordinate system at D time.

The invention discloses a multi-target tracking method based on measure-driven target birth intensity PHD (MDTBI-PHD), mainly solves the problems in estimating the multi-target motion state and the number of targets when the target birth intensity is unknown. The method comprises the steps of estimating the target state and the clutter state respectively by using an augmented state-space method, thereby avoiding the interference of unknown clutter to the target intensity estimation; constructing a target birth measure set, and estimating the target birth intensity by using a measure-driven method, thereby avoiding dependence on the priori knowledge of the target birth intensity; and implementing the above method using Gaussian mixture probability hypothesis density filter. The method has the advantage of being sensitive to the change in the number of targets, and meanwhile can reduce the computational complexity and significantly improve the tracking accuracy.

The invention belongs to the field of radar target tracking, and particularly relates to a multi-maneuvering-target Doppler radar tracking method based on Gaussian mixture probability density hypothesis. The method comprises the steps: firstly, introducing pseudo measurement to replace target radial velocity measurement obtained by a Doppler radar, then, introducing a measurement conversion methodbased on predicted value information to process position measurement and pseudo measurement, and meanwhile, carrying out decorrelation on the pseudo measurement and the position measurement; by adopting a Gaussian mixture probability hypothesis density filtering method and by means of a multi-model framework, aiming at the correlation between Gaussian components and models, carrying out differentprocessing on the Gaussian components of surviving, newly-born and derivative targets; for Gaussian components irrelevant to the models, namely, the newly-born and derivative Gaussian components, directly estimating the states thereof; for the Gaussian components related to the models, that is, the surviving Gaussian components, obtaining the model probability of each model filter and the model condition distribution of the updated components, and then, fusing the models and the condition distribution of the updated components to obtain state estimation, wherein introducing sequential filtering during filtering of the weight, the mean value, the covariance and the like of the Gaussian components, and obtaining position estimation according to position measurement; and performing sequential processing on the position estimation by using pseudo measurement to obtain final state estimation.

The invention discloses a multi-target track-before-detect method based on probability hypothesis density filtering. The method deeply analyzes the problem of a traditional method and points out the essential reason of the problem of the traditional method that the traditional method considers that one target influences the observation of a whole area and also considers that the number of each frame of false alarm can be approximated into one constant value, i.e. the traditional method does not comply with two pieces of basic hypothesis for realizing PHD (Probability Hypothesis Density) filtering: firstly, one target only can generate one observation, and secondly, the number of each frame of false alarm must submit to Poisson distribution on an aspect of time. In order to solve the problems, the invention puts forward the multi-target track-before-detect method based on the probability hypothesis density filtering, can improve the accuracy of target number estimation, enhances detection and tracking performance, and also achieves an effect on lowering a calculated amount.

The invention discloses a dual radar modified sequential Gaussian mixture probability hypothesis density (GM-PHD) filtering method. The traditional dual radar GM-PHD is only suitable for the case thata measurement target is located in a common measurement region of two radars, when the target is not in the common measurement region, a target loss problem is prone to occur. The dual radar modifiedsequential GM-PHD filtering method is based on a finite statistical theory, performs prediction, updating, trimming fusion, maintaining fusion and target state extraction on Gaussian components corresponding to each radar measured values, realizes multi-target tracking, cannot loss a target in a non-common measurement region, and expands the application range of sequential GM- PHD. Compared to traditional methods, the computational complexity of the dual radar modified sequential GM-PHD filtering method does not change much.

The invention discloses a multi-moving target situation awareness method, a device and a system. The multi-moving target situation awareness method comprises steps: a target measurement collection is acquired; the target measurement collection is sent to at least two parallel sequential Monte Carlo-probability hypothesis density filters for processing, and a processing result is outputted; and the processing result is comprehensively analyzed, and a moving situation of each target in the target measurement collection is acquired. The parallel sequential Monte Carlo-probability hypothesis density filters are adopted for analyzing and processing multiple moving targets, true paths of the multiple targets are accurately and clearly reflected by multi-moving target situation awareness, and the number of targets at each moment can be accurately estimated.

The invention discloses a probability hypothesis density filter radar system error fusion estimation method based on ADS-B. The method includes the steps that a radar system error observation equation based on ADS-B is set up; a radar system error state equation is set up; a radar system error state and observation finite set is constructed; error fusion estimation is conducted on a probability hypothesis density filter radar system. According to the estimation method, firstly, the ADS-B of a target and the observation of a radar are converted to be in a rectangular coordinate system with the radar as the center, due to the fact that the monitoring precision of the ADS-B is far superior that the positioning precision of the radar, on the basis of not considering the positioning error of the ADS-B, the radar system error observation equation and the radar system error state equation based on the ADS-B are set up, then a probability hypothesis density filter is used for conducting fusion estimation on the measuring difference of the ADS-B and the radar, and therefore an estimation result of the radar system error is obtained. The probability hypothesis density filter radar system error fusion estimation method has the advantages of being high in estimation precision, good in estimation performance and the like.

The invention discloses a weak target detection and tracking method based on box particle probability hypothesis density filter; the method comprises the following steps: using a box particle PHD filter method to process a weak target under low signal to noise ratio conditions, using mean value filtering to weaken single point edge noises so as to highlight a target located area, and taking the target located area as section measurement according to the maximum uncrossed principle; using dozens of box particles to replace several hundreds of point particles, thus effectively reducing the calculating complexity, improving the calculating efficiency, and obtaining a more accurate target state in the same time.

The invention discloses a probability hypothesis density filtering and smoothing method based on segmentation RTS (Rauch-Tung-Striebel). Through combination of a probability hypothesis density (PHD) filter and an RTS smoother, a probability hypothesis density filtering and smoothing algorithm based on the RTS is provided. The fact that a relatively high output delay problem exists in a smoothing process is taken into consideration, so a segmentation thought is employed, and the probability hypothesis density filtering and smoothing algorithm based on the segmentation RTS is provided. Evaluation values needing to be smoothed is segmented; track-evaluation association is carried out through adoption of a Hungary algorithm; and RTS smoothing is carried out on associated evaluation values segment by segment. Compared with a PHD filtering result, the probability hypothesis density filtering and smoothing method based on the segmentation RTS provided by the invention has the advantages thata target state can be evaluated precisely, and the problem that the timeliness is poor resulting from directly applying the RTS smoothing can be effectively avoided.

The invention discloses a multi-target tracking method through multi-search particle probability hypothesis density filter, and belongs to the field of radar data processing. The multi-target tracking method based on particle probability hypothesis density filter has an obvious defect of rapid degeneration of particle diversity caused by resampling when leak detection of targets occurs and then the phenomenon of target loss occurs, and thus the algorithm is difficult to meet multi-target tracking of low target detection probability. The multi-search particle probability hypothesis density filter aims at solving the problem. The multi-target tracking method is simple in structure, rapid in computation and simple in hardware realization; meanwhile, the limitation of the application based on the general particle probability hypothesis density filter method is overcome, and adaptability to a nonlinear and non-Gaussian system is high so that the multi-target tracking method has relatively high engineering application value and popularization prospect.

The invention discloses a GMPHD video multi-target tracking method fusing correlation filtering, and belongs to the technical field of computer vision, mode recognition and information processing. According to the method, a target is tracked by adopting a related filtering thought, and intersection ratio judgment of image information is added to carry out target template updating-free and parameter processing on the shielded target, so that pollution of a target template is reduced, and mistaken following frames are reduced. According to the method, a shielded target is put into Gaussian mixture probability hypothesis density filtering to perform position prediction updating operation, and if the target reappears in the later period, the target label is associated again, so that fragmentedtracks are reduced, and the defect of missing detection of a detector is overcome. Finally, a result on an MOT17 data set proves that compared with the best tracking algorithm GMPHDOGM17 related to GM-PHD at present, the multi-target tracking accuracy MOTA index is improved to 50.3 from the original 49.9.

The present invention discloses a probability hypothesis density filtering radar spatial error registering method under an ECEF (earth-centered Earth-fixed coordinate) coordinate system. According to the method, the observation data of a target in two observation radars are transformed to rectangular coordinate systems with the radars adopted as centers respectively; the data are transformed to the ECEF coordinate system, so that the common-observation relationship of the target between the two radars is built; on the basis of the common-observation relationship, a radar system error observation equation and a state equation under the ECEF coordinate system are established; a probability hypothesis density filter is used to perform fusion estimation on radar system errors under the ECEF coordinate system; and therefore, spatial error registration of the two radars is realized. The method of the invention has the advantages of no need for data association processing, high estimation accuracy, good estimation performance and the like.

The invention provides an unknown clutter passive cooperative positioning method based on sparsity order optimization. According to the method, a clutter density estimator with optimized sparsity order is embedded into a Gaussian mixture probability hypothesis density filter to estimate the multi-target state and number. Firstly, target measurement is removed through a threshold technology and target state feedback, so that clutter measurement is obtained, secondly, a sample is selected from the clutter measurement, and GA-SVR is used for fitting a sample, an extreme point of a fitting curve is obtained through a gradient method, and the x-coordinate of the extreme point is downwards rounded to obtain the optimal sparsity order n. According to the method, the passive cooperative positioning multi-target tracking performance under unknown clutters can be effectively improved, and the problem of multi-target tracking is solved.

The invention discloses a static fusion method based on a Gaussian mixture probability hypothesis density filter, and the method comprises the steps of migrating a multi-target tracking algorithm in atime dimension to a sensor dimension, and enabling a multi-source target static fusion problem to be regarded as a multi-target tracking problem with a time interval of 0; modeling a state model andan observation model of a target on the basis, performing static fusion after a high-speed effective tracking algorithm, namely Gaussian mixture probability hypothesis density (GMPHD) filtering, is modified according to the models, obtaining a fusion result with higher precision, and therefore, the operation efficiency is greatly improved. The method is suitable for input measurement under variousconditions, and can effectively improve the positioning precision and the operation efficiency of a fusion result.

The invention provides a multi-group target tracking system and method. The system comprises a data obtaining unit, which is used for obtaining state data of multiple groups of targets at one moment;a data modeling unit, which is electrically coupled to the data acquisition unit and is used for establishing a state model based on the acquired state data; a data prediction unit, which is electrically coupled to the data modeling unit and is used for estimating the state model by means of a gamma Gaussian inverse Wishart probability hypothesis density filtering algorithm and generating prediction data; and a data updating unit, which is electrically coupled to the data prediction unit and is used for introducing the prediction data to update the state model. The system and method are to solve the group maneuvering problem and the model probability prior solidification problem in the prior art.

The invention discloses a multi-target tracking method based on an adaptive extended Kalman probability hypothesis density filter, and belongs to the technical field of multi-target tracking. Firstly,a two-point difference algorithm is utilized to initialize new target strength, and then a target maximum speed constraint algorithm is utilized to eliminate wrong new target strength. Besides, in order to eliminate the interference of clutter measurement values, an improved measurement value classification algorithm is utilized to respectively extract a survival target measurement value and a new target measurement value from the measurement value set, and then the survival target measurement value and the new target measurement value are respectively utilized to update the survival target and the new target, so that the precision of the algorithm is improved. According to the method, the problem that the EK-PHD filter cannot track the target under the condition that the strength of thenew target is unknown is solved.

The invention discloses a parallel computing particle probability hypothesis density filtering multi-target tracking method. In view of the problems that the traditional particle probability hypothesis density filtering can not achieve complete parallel computing and the real-time performance is poor, a multi-core processor is used to realize parallel computing of all processes from importance sampling to a state estimation stage, so as to improve the real-time performance of multi-target tracking. Each parallel processing unit obtains initialized particles and then independently performs importance sampling, weight calculation normalization, resampling and local estimation, the local weight of the resampled particle before resampling is kept and recorded, a central unit calculates the particle weight ratio of each parallel unit participating in the state estimation according to the kept local weight, and then, the local estimation value is weighted and fused according to the weight ratio to obtain final multi-target state global estimation. The real-time performance of multi-target tracking is improved through parallel computing, and the application prospect is good.

The invention discloses a multi-sensor multi-target tracking method based on posterior track estimation, and the method comprises the steps: obtaining the posterior track probability hypothesis density of each sensor and a corresponding posterior track estimation set through a track probability hypothesis density filtering method, and constructing an association cost matrix between different sensors; dividing posterior tracks in the posterior track estimation sets of different sensors into associated tracks and non-associated tracks; fusing the probability hypothesis densities of the associated tracks among different sensors to obtain a fused associated track probability hypothesis density; error tracks in the non-associated tracks among different sensors are removed, the probability hypothesis density of the non-associated tracks after the error tracks are removed is fused with the probability hypothesis density of the fused associated tracks, and the probability hypothesis density of the fused tracks among the different sensors is obtained; and according to the fusion track probability hypothesis density among different sensors, extracting to obtain a multi-sensor estimated track state. According to the invention, the communication load is obviously reduced, and the problem of missing detection in GCI fusion is solved.

The invention discloses a PHD filtering method for tracking non-rigid multiple extended targets, which comprises the following steps of: firstly, modeling key points of target extension by using a space-time Gaussian process, and constructing states of the extended targets by using internal reference points of kinematics; employing a probability hypothesis density filter to propagate a first moment of the non-rigid multi-extension target random finite set; establishing a measurement tracking gate to select targets and corresponding measurement, then grouping the targets and the measurement to construct the suitable prediction likelihood of optimal division of the non-rigid multi-extension targets, and providing filter recursion and necessary approximation and hypothesis; and finally, converting the posterior density is converted into a Gaussian mixture form, so that the closed-form solution of the non-rigid target and the closed-form solution of the corresponding smoothing filter are obtained, and the number, the motion state and the shape of the non-rigid target can be accurately estimated in the presence of clutter, leak detection and noise.