KCF target tracking method integrating motion information detection and Radon transformation

Abstract

The invention discloses a KCF target tracking method integrating motion information detection and Radon transformation. The method comprises the steps: extracting HOG features from an input current frame image, and extracting moment features through Radon transformation; predicting an approximate range of a moving target by combining an optical flow method and an inter-frame difference method; inputting HOG features, training a KCF position filter, and predicting the center point position of a target area; inputting moment characteristics and training a KCF scale filter, and predicting the scale of the target; determining the accurate position of the target by combining the central point position of the target area and the scale of the target; performing adaptive strategy updating on the model; and repeating the above steps until the video frame tracking is finished, and determining the target to be tracked. According to the method, an optical flow method, an inter-frame difference method, Radon transformation and a related filter are combined, so that the problems of poor real-time performance and reduced tracking precision and success rate of a KCF algorithm when a target is shielded, the scale is changed and the ambient light is changed are solved.

Description

technical field [0001] The invention belongs to the technical field of visual target tracking, in particular to a KCF target tracking method combining motion information detection and Radon transformation. [0002] technical background [0003] As an important research content in the field of computer vision, target tracking has been widely used in the fields of video surveillance, intelligent transportation, human-computer interaction and UAV cooperation. Scholars at home and abroad have done a lot of research on this and made great progress. However, how to ensure the real-time performance, accuracy and robustness of the algorithm is still a difficult point in the research of visual target tracking for complex problems such as scale changes, occlusion and environmental lighting changes. [0004] Visual object tracking can be divided into generative and discriminative algorithms according to the object's appearance model expression. The generative tracking algorithm mainly...

AI Technical Summary

Problems solved by technology

Relevant scholars use HOG (Histogram of Oriented Gradients), SIFT (Scale Invariant Feature Transform) and CN (Color Name) features on the basis of correlation filtering algorithms to improve the representation ability of features and the iteration speed of algorithms, and the tracking speed is fast. The lower tracking accuracy is better, but for complex tracking problems in the case of target scale changes, occlusions, and environmental lighting changes, the tracking effect is not ideal due to the incompleteness of the extracted target appearance information.

M Danelljan et al. achieved scale adaptation by building a scale pyramid and training a scale filter, but because it essentially uses an exhaustive search strategy, the scale estimation link is added, and the amount of calculation increases, so the tracking speed and performance are not very good. ideal

Images