車画像の特徴抽出のためのワイヤーフレーム化処理

概要

論文の詳細を見る
A method of extracting features for making a wire-frame (line drawing) that is constructed by eight surfaces from an image data of a car is introduced. This method is constructed by the following three steps : In the first step, two surfaces are extracted from the silhouette of the car using degree of change in angles of straight lines that are observed from the outline of the car image. In the second step, a surface is extracted by brightness gradation of pixels in the inside of the silhouette. And in the third step, three surfaces, that are watched barely and can not be observed accuracy, are estimated using relationship of the already extracted lines. The system that can extract the wire-frame from a miniature car image and can calculate thirty kinds of feature values is constructed. As the result of experiments using 35 input images by nine kinds of miniature cars, 19 images are succeeded in extracting correct wire-frame of the cars. This method will be expected to provide some good features for recognition of a kind of cars/car names.
2000-01-31