36 results about "Bernoulli filter" patented technology

The invention discloses a multi-Bernoulli filter distributed fusion method, which includes using a multi-Bernoulli filter for each sensor node of the sensor network to realize local filtering, and receiving the multi-Bernoulli filter of its adjacent nodes through a communication link The posterior distribution, and then based on the generalized covariance intersection information entropy, realize the joint grouping of target tracks between sensors, and finally calculate the parameters of the multi-Bernoulli filter distributed fusion under each target track grouping, and obtain the fusion posterior Multi-Bernoulli distribution parameters. The invention can efficiently realize the multi-Bernoulli filter distributed fusion under the condition of a large number of targets, and has the advantages of small approximation error, low implementation complexity and the like.

The invention provides a multi-cell automatic tracking method based on a multi-Bernoullie filter with a label, which comprises a step of cell set initialization by the multi-Bernoullie filter with the label, a step of prediction by the multi-Bernoullie filter with the label and a step of updating by the multi-Bernoullie filter with the label. Cell set initialization adopts the k shortest path, a new cell set is generated according to the newly-appearing cell detection probability, weights and cell probability densities of the former-frame cell set and the newly-detected cell set are used, a predicted cell set in the current frame, the cell predicted probability density in the cell set and the cell set weight are acquired, the posterior probability density of the cells in the cell set and the weight of the cell set are calculated and updated through similarities, and cell tracking can be completed with no need of relevance.

Aiming at the problem of spectrum sensing in cognitive radio, the present invention designs a joint spectrum sensing and link state estimation method device. This method for the first time incorporates a link state information between the primary user and the secondary user as another hidden state in spectrum sensing, and thus establishes a general dynamic state space model; based on Bayesian Based on the sequential estimation framework and Bernoulli random finite set theory, a Bernoulli filtering mechanism is designed, which can obtain the joint blind estimation of the primary-secondary link state information while sensing and detecting the working status of the licensed frequency band (algorithm process has been Attached). The method makes full use of the primary-secondary link state information, thereby significantly improving spectrum sensing performance; at the same time, the obtained primary-secondary link state information can also be used for subsequent resource optimization; in addition, the joint user perception and The method and device for estimating link state information is versatile and can be extended to other primary-secondary link state information.

The invention discloses a distributed field-of-view complementary multi-Bernoulli correlation arithmetic mean fusion tracking method. First, the sensor's sensing range is expanded by complementing each sensor's field of view, so that the measurement of each sensor covers the entire tracking scene; Secondly, each sensor runs local multi-Bernoulli filtering separately, and associates Bernoulli components corresponding to the same target to the same subset by flooding communication correlation with the filtering posterior results, and performs a The arithmetic mean fusion completes the fusion state estimation. The main feature of this method is that the measurement information between different sensors is complemented by the complementary field of view, and at the same time, it will not be shared repeatedly, which can effectively reduce the amount of calculation and improve the calculation efficiency; The Bernoulli components are linked together and then fused by arithmetic mean, which can effectively improve the tracking performance.

The invention discloses an extended target tracking method based on GLMB filtering and Gibbs sampling, target number estimation, and extended target shape estimation problems, and proposes a multi-extended target tracking method based on a label random finite set framework. The method mainly includes Two aspects: dynamic modeling of multiple extended targets and tracking estimation of multiple extended targets. Firstly, combined with the generalized label multi-Bernoulli filter, the measurement finite mixture model of the extended target is established, and the parameters of the finite mixture model are deduced by using Gibbs sampling and Bayesian information criterion to learn and track the state of the multi-extended target, and then adopt The equivalent measurement method is used to replace the measurement produced by the extended target, and the shape of the extended target is modeled by ellipse approximation to realize the estimation of the extended target shape. The simulation experiment shows that the method provided by the present invention can effectively track multiple extended targets, accurately estimate the state and shape of the extended targets, and can obtain the track trajectory of the target.

The invention provides a fault tracking and positioning method based on the Internet of Things, comprising the following steps: constructing a multi-dynamic sensor network; coupling an anti-swarming distributed cooperative search algorithm and a multi-Bernoulli filtering algorithm to realize fault tracking and obtain real-time sensor data; Establish a weight clustering analysis model for the sensor data of historical normal samples, and calculate the control limit of the control quantity; perform fault detection on the acquired real-time sensor data according to the established weight cluster analysis model and control quantity, filter out the fault data segment and Obtain the corresponding control quantity; build a random forest classification regression algorithm based on the process variable and control quantity of the fault data segment; obtain the variable importance measure of the process variable through the random forest classification regression algorithm, and determine the fault variable according to the variable importance measure. Positioning, the method provided by the present invention has high tracking and positioning accuracy, and has obvious advantages.

The invention discloses a multi-sensor self-adaptive angle control method based on GLMB filtering, comprising the following steps: S1, using a generalized label multi-Bernoulli filter to perform local filtering; S2, performing distributed fusion based on a generalized covariance cross criterion; S3, multi-target sampling processing; S4, use generalized label multi-Bernoulli filter for pseudo-prediction; S5, calculate the ideal measurement of each sensor; S6, group the sensors in pairs, generate a decision set, Perform distributed fusion processing on the sensor pairs; S7, calculate the Cauchy-Schwartz divergence under the control decision combination of paired pairs; S8, select the local optimal sensor control decision for each sensor pair; S9, combine all the local optimal sensor control decisions; The optimal sensor control decisions are combined into one set. The invention can ensure small error loss, and solve the multi-sensor self-adaptive angle control based on GLMB filtering, which can be realized in both precision and calculation cost.