Research on Multi-Feature Fusion Target Tracking Algorithm
Aiming at the challenges of occlusion,target deformation,complex scenes and other challenges in the target tracking process,a high-precision visual target tracking algorithm combined with a memory mechanism is proposed.The algorithm first initializes the target and extracts the target's HOG feature and texture features and CN features construct a feature model to make up for the inaccurate tracking problem caused by a single feature.Perform related filtering operations on the model and the con-structed filter to obtain the maximum response position of the target search area.In the subsequent video frames,a three-layer ro-tating circle memory model containing instantaneous,short-term and long-term memory spaces is used to perform the target tem-plate.The update effectively enhances the tracking ability of the algorithm.The OTB50,TC128,OTB100 experimental bench-mark data set shows that compared with other algorithms mentioned in the article,this algorithm can track the target more accu-rately,which has important research value.
NETL
NSTL
万方数据
