A Video Vehicle Tracking Method Based on Compressed Sensing

Abstract

The invention discloses a video vehicle tracking method based on compressed sensing. The method uses the compressibility of the original signal, that is, the sparsity, and uses a non-correlated random measurement matrix to compress the signal without losing or damaging the original signal. information, and then achieve object tracking by building a classifier and updating the classifier in real time. The tracking algorithm based on compressed sensing does not process the redundant information in the original image features, and directly obtains the compressed features from the feature space sampling, and then classifies the compressed features to achieve target tracking, which has the advantages of small amount of calculation and high tracking efficiency.

Description

technical field [0001] The invention relates to the field of visual tracking, in particular to a video vehicle tracking method based on compressed sensing. Background technique [0002] In the field of intelligent transportation, the video-based vehicle tracking system is an important component module to realize vehicle behavior analysis and vehicle assisted driving and other technologies. Video vehicle tracking technology refers to a series of steps such as image processing of the collected video image sequence containing the target vehicle. According to the characteristics of the target vehicle in the video, an appropriate algorithm is selected, and in the next frame of the continuous video or image sequence Find the target vehicle for further analysis of the target vehicle. The efficiency of video-based vehicle tracking depends on the performance of the tracking algorithm. Therefore, the research on video-based vehicle tracking algorithm has important academic value and...

AI Technical Summary

Problems solved by technology

[0003] At present, the commonly used vehicle video tracking methods include vehicle detection and tracking algorithm combined with offline boost and online TLD, video vehicle tracking algorithm combined with Mean-shift and Kalman (Kalman) particle filter, using Kalman filter to predict vehicle position and using distance and Color information-assisted vehicle tracking algorithm, centroid tracking algorithm, region tracking algorithm and semi-automatic starting region growth algorithm, these algorithms can achieve good vehicle tracking results in practical applications, but there are still some problems, such as shadows, similarity Target mixing and blurred video quality all affect the tracking algorithm to track the target vehicle stably in the video sequence. Therefore, algorithms that can achieve robust, accurate, and stable tracking of the target vehicle in complex environments still need to be studied

Images