A Moving Target Tracking Method Based on Gaussian Improved Particle Swarm Filter

Abstract

The invention belongs to the field of signal processing, and in particular relates to a moving target tracking method based on a Gaussian improved particle swarm particle filter for quickly tracking a moving target whose appearance changes. The present invention includes: first selecting the target in the first frame of the video, and initializing the feature base; collecting the next frame of image during the tracking process, and obtaining the feature base of the tracking target; filtering through the Gaussian improved particle swarm particle filter, and storing the target corresponding to the highest probability The image window of the particles; when the number of collected images reaches the required number, update the characteristic base of the observation value. The invention mainly aims at the fast tracking of the moving target whose appearance changes by the mobile robot, and proposes a method capable of improving the fast tracking of the moving target in a complicated environment. The invention uses the low-dimensional feature base space to represent and describe the object to be tracked, and adopts the Gaussian improved particle swarm particle filter to track the position of the tracked target, thereby increasing the speed of tracking the moving target.

Description

technical field [0001] The invention belongs to the field of signal processing, and in particular relates to a moving target tracking method based on a Gaussian improved particle swarm particle filter for quickly tracking a moving target whose appearance changes. Background technique [0002] Tracking of moving targets has a wide range of applications in many areas, such as public safety, government surveillance, and military. [0003] In recent years, particle filter-based target tracking technology has been developed in the long run. Particle filter realizes state estimation by tracking system state probability distribution, and realizes conditional probability transfer through Bayesian theorem. A large number of particles are generated by means of Monte Carlo simulation, and the probability distribution of the state is approximated by their dispersion, so it can be used with any nonlinear system that can be represented by a state-space model, as well as those that are dif...

AI Technical Summary

Problems solved by technology

The resampling proposed in this invention is an algorithm that has already been proposed, and the invention does not use the current observation value in the importance sampling, which will increase the sensitivity of the method to noise, and resampling will make the calculation amount greatly increase, making the real-time performance decrease

Images