KCF long-term gesture tracking method integrated with SLIC algorithm

Abstract

The invention discloses a KCF long-term gesture tracking method integrated with an SLIC algorithm, and the method comprises the steps: 1) building a gesture training data set, extracting and carryingout the offline training of an SVM model of a super-pixel block, and obtaining a coarse classification model of gesture detection; 2) constructing a foreground-background dictionary, and designing a similarity function of a KNN algorithm by combining the FHOG feature and the CN feature so as to finish fine classification of gesture detection; 3) obtaining a gesture detection model through the coarse classification model of gesture detection and the fine classification of gesture detection, and detecting a target by using the gesture detection model to obtain a detection box of a target gesture; 4) estimating a rectangular frame of the most suitable target gesture by using the designed target scale estimator; and 5) designing a confidence function, and determining whether the current tracking result is credible to realize gesture tracking by comparing the similarity between the current frame tracking result and the previous frame tracking result. The method has the advantages of low algorithm complexity, high tracking precision and high robustness, and is suitable for real-time application occasions.

Description

technical field [0001] The present invention relates to gesture recognition technology, more specifically to a KCF long-term gesture tracking method integrated with SLIC algorithm. Background technique [0002] Gesture recognition technology has always been a research hotspot, and gesture tracking is an important part of gesture recognition technology. Gesture tracking is generally classified into two categories, one is short-term tracking, which only considers the movement tracking of the target within a short period of time, such as KCF, DSST, MOSSE and other algorithms; the other is long-term tracking, which can be used for a long time can better track the target. [0003] The KCF target tracking algorithm is a discriminative correlation filtering algorithm. This type of method generally trains a target detector during the tracking process, uses the target detector to detect whether the predicted position of the next frame is a target, and then uses the new detection Th...

AI Technical Summary

Problems solved by technology

[0005] 1. The KCF algorithm relies on the circulatory matrix and its initialization matrix and cannot be adaptively changed. Therefore, the KCF algorithm is not ideal for multi-scale target tracking;

[0006] 2. The KCF algorithm lacks the ability to track high-speed moving targets and targets in low frame rates. The reason is that the target displacement between adjacent frames is too large, which exceeds the search range of the KCF algorithm;

[0007] 3. It is difficult for the KCF algorithm to continue tracking the target after the target is blocked for several frames

Images