Particle filter tracking method for adaptive adjustment of tracking window size

Abstract

The invention discloses a particle filter tracking method for adaptive adjustment of tracking window size and is mainly used for solving the problem of real-time adaptive adjustment of a particle filter tracking window. The window adjustment method comprises the following steps of: firstly establishing a state transfer equation of a moving target under a particle filter tracking framework, and initializing a target state; then predicting a set of particles at the next moment by applying the state transfer equation, calculating likelihood function values of the particles and updating weights of the particles according to the likelihood function values of the particles, and calculating the average distance d between the particles with great weights and the center of the set of particles; and finally linearly changing the window size along with the value d to realize the adaptive adjustment of the tracking window. The particle filter tracking method realizes the adjustment of the particle filter tracking window by using a small amount of calculation. Compared with similar methods, the particle filter tracking method has the advantages of simple calculation and high tracking speed.

Description

technical field [0001] The invention relates to the technical field of video target tracking, in particular to a particle filter tracking method with adaptive adjustment of tracking window size. Background technique [0002] Target tracking is an important branch of machine vision research and the basis of advanced machine vision research. It has a wide range of applications, such as military guidance, road traffic monitoring, industrial automation production monitoring, passenger flow statistics, etc. Among the many tracking algorithms proposed by people, particle filter-based and Mean-Shift algorithms are research hotspots. [0003] In the particle filter target tracking algorithm, the target tracking window is determined by the initial size of the tracking target, and the size of the tracking window remains constant throughout the tracking process. However, when the size of the moving target becomes smaller and smaller, if the tracking window is fixed, some background no...

AI Technical Summary

Problems solved by technology

[0010] The Chinese invention patent with the application publication number CN102184554A discloses a method for adjusting the particle filter tracking window. This method uses the Otsu threshold method to segment the target area, and calculates the segmentation marker map to obtain the target scale observation value, and then uses the Kalman The filtering method adjusts the tracking window adaptively. The problem is that the size of the segmented target area is greatly affected by the noise, so the error of using it as the observed value is relatively large, and because the Kalman filtering method is used to adjust the window, the increase of Calculations;

[0011] The Chinese invention patent with application publication number CN102005055A discloses a method for adjusting the particle filter tracking window. The method calculates the covariance matrix of the particles based on the EM algorithm, and then calculates the global kernel window width of the particle set according to the covariance matrix. The calculation of the particle filter algorithm increases the amount of calculation, which affects the tracking speed of the algorithm.

[0012] The existing tracking window adjustment method requires more complex calculations, thus affecting the tracking speed

Images