基于二级LTP纹理颜色直方图的动物实时跟踪

目标表征方法是目标跟踪算法当中的关键环节.近来,目标跟踪技术在动物的跟踪以及其他农业领域当中得到广泛应用.但是,只对区域进行一次纹理特征的提取往往不够丰富.因此,对原始提取LTP纹理特征的方法进行改进,提出二级提取LTP特征融合的目标表征方法.通过对原始图像进行相同LTP特征提取,得到目标区域的多级LTP图像特征,将HSV模型中受光照影响较小的H分量作为颜色特征向量,来建立每一级特征相应的直方图.根据嵌入到Mean Shift当中的二级纹理颜色模型来进行动物跟踪.与传统的基于核的跟踪算法进行对比,此方法显著提高动物目标跟踪的鲁棒性.

Animals Real time Tracking Algorithm Based on Joint Second level LTP Texture and Color Histogram

Target representation method plays a critical role in the object tracking algorithm.Recently,object tracking techniques are widely applied in the field of animal tracking and other agricultural issues.However,that will be not enough if we extract the texture features only once.Therefore,the original LTP is improved by using the method of second-level LTP to represent the target.Second-level LTP texture is obtained by extracting the same LTP features from the original image.Then,the Hcomponent which can avoid the illumination changes is employed to establish the second-level LTP texture and color histogram.Finally,the Joint multi-level LTP texture and color model is embedded into the mean shift tracking framework to conduct the tracking process.Compared with the traditional kernel-based object tracking,the proposed method can improve the robustness of theanimal object tracking remarkably.