Abstract
One of the problems in image recognition is how to acquire the 3-dimensional range data on the designated regions in the scene to extract the features, such as 3-dimensional shape and orientation of an object in order to reduce the processing quantity of information and simplify the recognition procedures. This paper proposes a method in which the visual feedback principle is introduced into the procedure of acquiring the 3-dimensional range data on the specified area in the scene. This method is based on the image recognition technique and the active measurement of 3-dimensional range data. The inputs of the visual feedback control system are referred to as the picture cells in a specified window (attention window) on the image. The attention window is determined according to the pattern recognition results. The outputs are referred to as the positions of the laser spot in the scene. As soon as the position of a picture cell in the attention window is selected as the input, the spot of the laser range finder will automatically approach the point corresponding to the input in the scene until the output position on the image overlaps onto the input position. Then, the 3-dimensional coordinate of the point in the scene is acquired according to the triangulation. In this way, all the 3-dimensional range data of the specified area in the scene are input consecutively for shape recognition and for extraction of object features.
