Target detection tracking method based on feature fusion and environment self-adaption

Abstract

The invention relates to a target detection tracking method based on feature fusion and environment self-adaption, and the method comprises the following steps of: 1, preprocessing video data by using a pre-classifier, and dividing the video data into a simple frame and a complex frame; 2, extracting a single feature for the simple frame; for the complex frame, extracting HoG, Haar and CN features to carry out multi-direction feature fusion; and 3, based on a sample screening mechanism of motion prediction, fitting a motion trail of the target by using a least square method, taking the motion trail as a parameter of weight calculation, and re-evaluating a sample generated by cyclic sampling to obtain a more accurate weight. Under the condition that the tracking speed and precision are guaranteed, the method has better robustness when resisting target deformation, fast movement and scale change, and especially has more advantages in complex backgrounds and long-term sequence processing. The complexity of system calculation can be reduced as much as possible under the condition that the tracking accuracy is not influenced, and the tracking speed of the algorithm is improved.

Description

technical field [0001] The invention relates to the technical field of computer vision target tracking, in particular to a target detection and tracking method based on feature fusion and environment self-adaptation. Background technique [0002] Video object tracking has always been a key research issue in the field of computer vision, and it has applications in many fields such as public transportation, intelligent surveillance, intelligent security, and human-computer interaction. With the continuous expansion of the application range and the continuous improvement of the level of science and technology, people's research on the field of target tracking is also getting more and more in-depth. [0003] Target tracking refers to accurately tracking the target area in continuous video frames. However, many current target tracking algorithms are not well suited to target tracking tasks in multiple scenarios. Currently, relatively mature tracking algorithms are generally It c...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to provide a target detection and tracking method based on feature fusion and environment adaptation. On the basis of the KCF algorithm, an adaptive classifier and a color moment feature descriptor are introduced to solve the problem of target factors in the tracking process of the KCF tracker. Tracking failure problems caused by factors such as severe occlusion, similar target interference and out of view, to achieve fast and accurate tracking of the target, with good robustness

Images