A FEATURE POINT EXTRACTION METHOD BY POISSON EQUATION
Recently, bio-robotics based on biology and robotics,and neuro-robotics based on neuroscience and robotics have been developed very fast.In this field, a lot of studies on experimental animals” behavior especially rodents” behavior like ”open-field” experiment setup was developed to analysis the active and immobility time,region interesting and exploring behavior of rodents.This paper makes use of a Poisson PDE multi-grid solver to compute distance transform.We show how this idea extends naturally to the feature points extracting from shape of experiment animals.Solution of Poisson equation results in a good distance transforming method that body and limbs of a shape will be separated apart to different level totally.With this distance information,algorithm is induced to extract head point, tail point of experiment animals like mice and rats.For noise and jitter reason in captured images, we put a Kalman filter to get more steady track of feature points.This approach also has a good computational performance in practice.
Poisson equation distance transform multigrid Kalman filter animal behavior
Weihua Zhang Jiliu Zhou Gang Lei Kun He Chaobang Gao Hao Zhong
School of Computer Science,Sichuan University, Chengdu, CHINA Digital Media Department, Sichuan Normal University, CHINA College of Information Science and Technology, Chengdu, CHINA Chengdu Dashuo Experimental Animal Co.LTD, Chengdu,CHINA
国内会议
银川
英文
305-308
2016-06-01（万方平台首次上网日期，不代表论文的发表时间）