Face Tracking Algorithm of Power Warehouse Based on Multi-Feature Fusion Particle Filter

Abstract

In order to improve the tracking effect of video objects in power warehouse, a particle filtering target tracking algorithm based on multi-feature fusion is proposed, which combines the nonlinear and non-Gaussian characteristics of state model and observation model in video objects tracking. Firstly, an adaptive method for selecting target color histogram is proposed to obtain more accurate target color features. Then, Local Ternary Patterns (LTP) texture features are introduced into the tracking algorithm. Based on the LTP texture features, the LTP key texture model is constructed to further enhance the target key texture information. Finally, the color feature and texture feature are used to represent the target, and the two features are fused into the framework of particle filtering, and the weight of the particle is calculated by using the fused information to reduce the influence of the algorithm on the target deformation and complex environment; at the same time, the idea of weighted classification is adopted to assign different weights according to the different contributions of different features to the classification effect, and larger weights are assigned to the reliability features to improve the classification accuracy. The experimental results show that the proposed algorithm can not only track the face effectively, but also get better tracking effect when the color of the target is disturbed, which can be applied to face tracking task in power warehouse.