关键词: 局部模型, 超像素, 推土机距离（EMD）相似性度量, Snake模型

Abstract: The traditional tracking methods make moving target easily drift, even result into the loss of the moving target under the complex background. Focusing on these problems, it proposes a object contour tracking algorithm based on Locally Model Matching （LMM）. It uses super-pixel combining with the EMD similarity measure to build locally features model. It carries on locally model matching. The Snake model combines particle filter which extracts the target contour to achieve contour tracking accurately. The experimental results indicate that the proposed moving target tracking method maintaines a higher success rate that is under the conditions of the target deformation, partial occlusion and complex background. Compared with other moving target contour tracking methods, the proposed algorithm possesses accurateness and robustness.

Key words: locally model, super-pixel, Earth Mover’s Distance（EMD） similarity measure, Snake model