Object detection on deformable surfaces using local feature sets

Object detection is one of the important tasks in many computer vision applications. Especially, in vision-guided robotics, numerous pattern matching algorithms are used for detecting and localizing randomly oriented objects in pick-and-place systems. Feature matching algorithms such as SIFT, SURF etc. are widely employed where geometric pattern matching algorithms often fail when objects lack contours and edges. Even though these algorithms give reliable results under extreme conditions of scene clutter and occlusion, they usually fail when there are multiple instances of same object and object shape deformation problems in the scene. In this paper, we present an approach which uses SURF feature sets consisting of local neighbor features for matching and hierarchical clustering for estimating object center. Using extracted local neighbor features and their descriptors, our algorithm finds more number of true-positive matches among features and improves the detection in the case of deformation and multiple instances. Experimental results show the effectiveness of the algorithm.