摘要
在目标跟踪过程中,图像匹配技术是跟踪至关重要的环节,直接影响跟踪的效果。针对图像匹配算法中传统块匹配的搜索框和匹配准则问题提出了相应的改进。首先,采用并行粒子滤波算法对图像匹配中搜索框的位置和大小进行改进。其次,采用基于时空域信息的条件随机场模型以及CRF最大似然系数,对目前主流的依赖颜色信息的Bhattacharyya系数匹配准则进行改进。实验结果表明该算法不仅在匹配速度上有所提升,而且大幅减少了对目标颜色和形状的依赖,在匹配精度上也有了大幅提升,能更好的处理目标和背景颜色相似等复杂问题。
Image matching technology is the vital section of object tracking which is directly affected by the result of image matching.This paper proposes improved algorithms on search region and match principle in block matching process.Firstly, based on CRF model and CRF maximum likelihood parameter,a new algorithm of image matching is presented to improve the traditional Bhattacharyya parameter principle.Secondly,applying the distributed particle filter algorithm,the position and range of searching region is improved.
出处
《工业控制计算机》
2012年第9期87-89,共3页
Industrial Control Computer
