| 摘要 |
PURPOSE:To shorten the operating time without lowering detecting accuracy by reducing an image before the detection of the body recognition and the position attitude of the image, recognizing an object body, detecting the position attitude, setting a local region on the object body, and detecting the body recognition and the position attitude for the local region set in the image before the reduction of the image. CONSTITUTION:The image is inputted from a two-dimensional camera 5 into an image processor 6 with standard picture elements. The inputted picture elements are reduced step by step. Preprocessing is performed with the reduced image, and the characteristic value is extracted. Then, model base matching is performed, and body recognition and position detection are performed. The body position/attitude data as the result of matching are used, the image is expanded to the image of standard resolution before the reduction and a local region is set. The preprocessing and matching of the image are performed for the local region again, and the detailed position detection of the object body is performed. The local region includes the characteristic value of the body in the standard image and the characteristic value of a model. Therefore, the same detecting accuracy as that in the matching for all picture elements is obtained, and the time is shortened. |
