X-t時空間画像を用いた歩行者の計数 : 画像通信システム : テレコンファレンスおよび画像一般

概要

論文の詳細を見る
This paper presents a pedestrian counting system which emploies x-t spacetime image processing. Counting moving objects needs three tasks; object extraction, object separation, and object tracking. Object tracking has been one of the difficult problems in motion image processing. By using the fact that an object usually forms a closed region in the spacetime image representation, we can avoid the problem of object tracking. A two dimensional spacetime image forms the skewed projection of objects when the image is gathered with a slit which is placed angular to the object position and velocity. The skewed projection generally contains the information of object shape and velocity. Count of pedestrians can be easily computed from the number of regions and the 0th, 1st, and 2nd order moments of each region. The sign of the 2nd order of moments is used to discriminate the direction of walking. The detail of system description and the experimental result are also shown.
社団法人映像情報メディア学会の論文
1990-09-13