Target Scale Adaptive Tracking Method Based on Correlation Filtering and Color Detection

Abstract

The target scale adaptive tracking method of the present invention comprises the following steps: step 1 reads the first frame picture, determines the target position and size, sets the target area and the background area; step 2 obtains sample images of different scales, calculates the target area O and The color histogram of the background area B, to obtain the probability of the target sample; step 3 extract the hog feature and grayscale feature of the image, train the position filter and scale filter, and obtain the corresponding filter template; step 4 read in the next frame picture, Update the current position of the target; step 5, update the target scale according to the current position of the target; step 6, judge whether the current frame tracking is normal, and perform corresponding operations; step 7, according to the judgment result, substitute the target detection in the next frame until the tracking task ends. Through the different responses to the target image and the background image, the target and the background are distinguished, and the background information of the target is effectively used to improve the accuracy of the target in occlusion, deformation and other scenes.

Description

technical field [0001] The invention relates to the technical fields of image processing and machine vision, in particular to a target scale adaptive tracking method based on correlation filtering and color detection. Background technique [0002] The main way for human beings to obtain information is through vision. With the research and development of computer technology, people began to use computers to simulate the functions of human vision, instead of human eyes and brains to perceive, explain and understand the scene environment. The discipline of computer vision was born. Computer vision is a popular subject in the field of artificial intelligence. It integrates research methods and achievements in the fields of signal processing, pattern recognition, applied mathematics, and neurophysiology. It is a comprehensive subject. In the field of computer vision, video target tracking is a basic problem. It uses various methods to estimate the state of the target in continuo...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to overcome the shortcomings of the above-mentioned prior art, and mainly aim at the problems of too little background information utilization around the target, the scale cannot be accurately estimated, and the long-term tracking is easy to fail. The adaptive tracking method is specifically implemented by the following technical solutions:

Images