Intelligent video tracking method based on immune genetic particle filtering

Abstract

The invention relates to an intelligent video tracking method based on immune genetic particle filtering, which comprises the following steps: initializing and sampling particles, calculating the weighted value of the particles, estimating and outputting the state, calculating the number of effective particles, and if the number of effective particles is smaller than the preset threshold, optimizing the set of immune genetic particles, otherwise, directly sampling particles again. The invention adopts the principle of the immune genetic algorithm, and realizes the optimization of the particle set through fitness calculation, memory unit updating, inhibition of the antibody concentration, promotion, crossover, variation and other types of operation, so that novel particles can be generated on the premise that the effectiveness of particles with high weighted values is maintained, thereby improving the diversity of particle sets and the number of effective particles. The novel particles can better express the true state of the target, thereby enhancing the robustness and effectiveness of the intelligent video tracking and reducing the risk of mistaken tracking.

Description

technical field [0001] The invention relates to the technical field of intelligent video processing, in particular to an intelligent video tracking method based on immune genetic particle filter. Background technique [0002] The traditional video surveillance system can only obtain clues through the playback of video images after the occurrence of hazardous events, which cannot fundamentally reduce the occurrence of hazardous events. The intelligent video surveillance system can issue early warnings before harmful events occur, so as to prevent problems before they happen. In the intelligent video surveillance system, the intelligent analysis system is the most critical component. It currently has two architectures, namely embedded video analysis and pure software video analysis. Regardless of the architecture, the intelligent video analysis software must complete the target All functions such as detection, tracking, classification and rule definition. After the target is...

AI Technical Summary

Problems solved by technology

Under normal circumstances, it is difficult to improve the observation model greatly, so the research on the problem of eliminating sample poverty has attracted more people's attention

The "sample poverty" problem stems from the resampling process. Due to the excessive copying of high-weight particles, the number of effective particles decreases after resampling. In this way, after several recursive calculations, the effective particles are almost exhausted, and the particles The information capacity of the set will be severely reduced, it is difficult to reflect the real state statistical characteristics, and it is easy to cause a decrease in the tracking rate of the video target and false tracking.

Images