Sensor Planning for 3d Object Search Ksenia Shubina a Thesis Submitted to the Faculty of Graduate Studies in Partial Fulfilment of the Requirements for the Degree of Master of Science Graduate Program in Computer Science and Engineering

An elaboration on Ye’s object search strategy is presented [64]. Our search agent is implemented on a Pioneer 3 robot equipped with a BumbleBee stereo camera. Of particular concern in this thesis is the incorporation of an appearance-based recognition algorithm for object detection and a stereo camera for depth estimation and environment representation. The detection algorithm can handle either 2D objects in a 3D environment or 3D objects with a distinguishing 2D surface. Experiments were conducted to test the implementation of the search agent and determine its limitations with respect to lighting conditions, background and occlusion situations, presence of distractors and relative location of the camera and the target object. We found our detection algorithm to be successful in 85% of the cases when the target is within the detection range of the method, is not occluded and the angle of in-depth rotation is less than 45 degrees. The search strategy proved plausible when we incorporated the stereo depth sensor instead of the laser range finder, and a more sophisticated detection algorithm. In addition, we have shown iv that performance of the search agent is relatively insensitive to various lighting and background situations.