Moving target detection and tracking method based on embedded system

Abstract

The invention discloses a moving target detection and tracking method based on an embedded system. The method comprises the steps that first, pixel-level and frame-level secondary analysis is performed, and a background model is updated; background subtraction is used, and then morphological filtering is adopted to obtain a moving foreground; if multiple moving objects are available through judgment, and a tracking target is selected manually; samples are selected according to particle weights at the previous moment, and the number of selected particles with large weights is greater than the number of selected particles with small weights; a particle set is propagated through a dynamic model; an MS is utilized to optimize part of the particles, then similarity is judged, and if the MS is more approximate to a target model, the optimized particles are merged with standard sampled particles; color features and movement features are adopted to observe the similarity between a target possible state and a target real state represented by each particle; and whether model update needs to be performed is judged according to the judgment of whether a similarity coefficient between average state output of the particles and the target model is greater than a threshold value. Through the method, sampling efficiency is greatly improved, and the method is suitable for real-time tracking on the embedded system.

Description

technical field [0001] The invention relates to the technical field of target tracking, in particular to a moving target detection and tracking method based on an embedded system. Background technique [0002] Object tracking is widely used in intelligent monitoring, human-computer interaction, intelligent transportation, robot vision and other fields. The research on target tracking technology has experienced more than 30 years of development, and has made great progress. Researchers at home and abroad have proposed many mature algorithms. But on the one hand, the actual target tracking system needs to face a wide variety of real-world environments, such as noise, shadows, background interference, deformation, occlusion, etc. On the other hand, the hardware and software resources of the embedded system are limited, so how to realize an accurate and high real-time tracking algorithm is still a contradiction. So far, the unification of robustness, accuracy, and real-time pe...

AI Technical Summary

Problems solved by technology

However, this type of algorithm needs to train a large number of samples in order to improve the accuracy, the algorithm complexity is high, and it is difficult to guarantee real-time performance

Images