Multiple cues integration for part-based adaptive workpieces detection

Conventional object detection algorithms usually suffer from illumination changes and partial occlusions. In this paper, we proposed a novel detection algorithm that integrates multiple cues with adjustable weights to robustly detect objects in various illumination conditions. To further cope with partial occlusions, each cue detector is decomposed into several individual part detectors whose weights are adaptively adjusted. Experimental results demonstrate the efficiency and robustness of the proposed algorithm.