粒子滤波框架下多特征融合的视频目标跟踪方法

An Algorithm of Video Target Tracking Based on Particle Filter Framework and Multiple Features

针对单一目标特征在复杂场景下难以实现有效的跟踪问题,提出了一种边缘纹理与颜色特征相融合的新方法。将Sobel算子与局部二值模式算子相结合,得到一种新的边缘纹理SLBP(Sobel Local Binary Pattern)特征提取方法,并与HSV(Hue,Saturation,Value)颜色特征融合应用于粒子滤波框架的视频目标跟踪。实验结果表明:本文提出的SLBP+HSV特征融合方法能够克服视频中光照变化、目标遮挡等复杂背景影响的问题,提高跟踪的精确度。

To improve the performance of object tracking under complex environment based on single object feature, a new particle filter method base on the fusion of edge, texture and color feature is presented.Combined the Sobel-operator and the Lo-cal Binary Pattern, a new feature extraction method named SLBP （Sobel Local Binary Pattern） is proposed.SLBP contains the texture feature and the edge feature at the same time.The result shows that the proposed algorithm based on SLBP＋HSV can over-come the complex environmental impact, such as occlusion and illumination change.The algorithm improves the tracking accura-cy.