ビデオフレームからの人物領域の高精度抽出ならびにその自動追跡

概要

論文の詳細を見る
In this paper a highly accurate extraction of moving individuality from video frames is presented. With regarding to lighting conditions in a natural environment and occlusion of objects, it is very difficult to isolate individuality from the background only using color and movement information. To improve extraction accuracy, a new algorithm must be developed. The vector information and the weighted moving average of the objects is adopted in this paper to increase the identifying accuracy of objects from a background. To identify the objects template matching is used; however, the diamond search method is utilized to improve the isolating accuracy of the objects. From the experimental results, more than 86% of correct extraction accuracy was obtained in an outdoor environment.
2011-10-30