Real-time target tracking algorithm based on compressed sensing

Abstract

The invention belongs to the vision tracking field, relates to a real-time target tracking algorithm based on compressed sensing and solves problems of low speed, low robustness and inaccurate target tracking existing in a target tracking algorithm in the prior art. The algorithm comprises steps that a tth image is acquired and is grayed, the compressed sensing theory is utilized, a quite sparse random matrix is utilized to map extracted high-dimensional multi-scale image characteristics to a low-dimensional subspace to acquire low-dimensional image characteristics, the characteristics are classified through utilizing a classifier, and a sample with the largest classifier reaction value is selected as a tracking target. The algorithm is advantaged in that disadvantages in the prior art are solved, not only can high robustness and high accuracy under various interference factors be realized, but also an average image processing speed reaches 35 frames per second, the timeliness requirement is realized, and the method is of important reality significance.

Description

technical field [0001] The invention belongs to the field of visual tracking and relates to a real-time target tracking algorithm based on compressed sensing. Background technique [0002] Object tracking is a very popular research topic in the field of computer vision because of its significance in vehicle navigation, traffic monitoring, and human-computer interaction. Although the subject of object tracking has been studied for decades and many tracking algorithms have been proposed, it is still a very challenging problem. Because the target appearance is disturbed by various factors, such as illumination changes, pose changes, complete or partial occlusions, and sudden movements, etc. Therefore, it is a challenging problem to develop a high-performance tracking system under the interference of the aforementioned factors. [0003] Currently, online tracking algorithms often take observation samples from recent frames to update the appearance model. Although such algorit...

AI Technical Summary

Problems solved by technology

Although such an algorithm has achieved a certain degree of success, there are still many problems that have not been solved

First, these adaptive appearance models depend on data, but there is not enough and a large amount of data for the online tracking algorithm to learn at the beginning. Second, during the self-learning process, some uncalibrated samples will be Adding it in reduces the performance of the apparent model, so online tracking algorithms often encounter drift problems

Images