OCCLUSIONS AS A GUIDE FOR PLANNING THE NEXT VIEW

To resolve the ambiguities that are caused by occlusions in images, we need to take sensor measurements from several different views. The ta sk addressed in this paper deals with a strategy for acquiring 3-D dat a of an unknown scene. We must first answer this question: What knowle dge is adequate to perform a specific task? Thinking in the spirit of purposive vision, to accomplish its task, a system does not need to un derstand the complete scene but must be able to recognize patterns and situations that are necessary for accomplishing the task. We have lim ited ourselves to range images obtained by a light stripe range finder . A priori knowledge given to the system is the knowledge of the senso r geometry. The foci of attention are occluded regions, i.e., only the scene at the borders of the occlusions is modeled to compute the next move. Since the system has knowledge of the sensor geometry, it can r esolve the appearance of occlusions by analyzing them. The problem of 3-D data acquisition is divided into two subproblems due to two types of occlusions. An occlusion arises either when the reflected laser lig ht does not reach the camera or when the directed laser light does not reach the scene surface. After taking the range image of a scene, the regions of no data due to the first kind of occlusion are extracted. The missing data are acquired by rotating the sensor system in the sca nning plane, which is defined by the first scan. After a complete imag e of the surface illuminated from the first scanning plane has been bu ilt, the regions of missing data due to the second kind of occlusions are located. Then, the directions of the next scanning planes for furt her 3-D data acquisition are computed.