Particle filtering tracking method and tracking device

Abstract

The invention discloses a particle filtering tracing method and a tracing device thereof. The method comprises the steps as follows: Step S1, samples are collected from the original image frame to obtain an initial particle set; Step S2, foreground and background classifications are carried out to the pixel of the original image frame to obtain a probability graph; Step S3, particles are spread according to a system dynamic model and sampling is carried out, thus obtaining the second particle set; Step S4, the weight of the second particle set is obtained from a system observation model builton the probability graph and the third particle set is obtained by normalizing the weight of the particles; Step S5, importance sampling is carried out to the third particle set, thus obtaining an importance sampling particle set; and Step S6, the system state is calculated by the importance sampling particle set and output. The method and the device improve the calculation efficiency of particlefiltering, improve the robustness and are not easy to lose objects.

Description

technical field [0001] The invention relates to an image processing method, in particular to a tracking method and a tracking device of a moving object. Background technique [0002] Tracking of moving objects is an important part of many computer vision application systems, such as video surveillance, human-computer interaction, driver assistance systems, robot navigation, etc. The main purpose of moving object tracking is to obtain the trajectory of the moving object in the image sequence or video, and further, sometimes it is required to obtain the scale, shape or three-dimensional pose information of the moving object. After decades of development, a large number of tracking methods have emerged in the field of moving object tracking, among which a class of probabilistic tracking methods represented by particle filtering has attracted extensive attention from researchers and developers. [0003] Particle filter is a Sequential Monte Carlo algorithm, which approximates t...

AI Technical Summary

Problems solved by technology

[0011] The deficiency of the particle filter tracking method of the prior art is: first, the particle filter algorithm is carried out on the original image, and the defect of doing like this is: the original image is not specially designed for the tracking of the moving object, in the original image The moving object on the surface may have very low contrast compared with the surrounding background, which is not conducive to tracking the moving object; second, the observation model uses information such as color histogram; however, it is very consuming to calculate the color histogram for each particle , it will affect the real-time performance of the algorithm, and the use of the color histogram only considers the color of the moving target, but does not consider the part of the background information that is the same or similar to the color of the moving target, so it is easy to lose the tracking target, that is, because the background is not considered information makes such existing methods less robust

Images