ビデオフレームからの人物領域の高精度抽出ならびに複数の領域情報を用いた自動追跡

概要

論文の詳細を見る
This paper presents a human tracking algorithm considering the illumination change and the occlusion of the objects.This algorithm enables to isolate a person in the indoor and outdoor environment even with the abrupt changesof the illumination. In order to isolate the person, each moving object is detected by using the variable inter-frame subtractionmethod and the using moving vector of each moving objects. It is considered that the movement of a person is in a straight-line motion within a short time interval. From thisview point, the moving person can be from other moving objects such as trees, cars, and animals. In the tracking process, each person is traced using moving vector and the texture information. In general, trackingproblems, occlusion is the most serious problem. To overcome this problem in this paper, We prepared many variable-size templates based on this approach are prepared to solve the occlusion problem As a result illumination change,appearance changes are size changes were all experimentally solved in each case. The experimental results using 2265 video frame for a single person for show indoor environment, 92% trackingaccuracy. For the outdoor environment, 88.5% of tracking accuracy was obtained for plural target. From these experimentalresults, it can be concluded that this method is widely applicable for using in an actual environment.
2012-04-30