A MODIFIED EXOSKELETON AND A HAUSDORFF DISTANCE MATCHING ALGORITHM FOR SHAPE-BASED OBJECT RECOGNITION

摘要

Recently object recognition is a crucial process in many image retrieval systems, computer vision, and image analysis. This paper presents a simple but powerful algorithm for shape-based object recognition using a Hausdorff distance matching algorithm and a modified exoskeleton (mES) as an object representation. The mES is a special type of skeleton, which is one of the most widely used object representations, that is less sensitive to distortion caused by rotation and noise. The proposed Hausdorff distance matching, algorithm measures the similarity between two objects based on the Hausdorffdistance and the modified exoskeleton function. The performance of the proposed matching algorithm for 2D binary object recognition is compared with the skeleton matching algorithm(SMA) presented in [5]. The experimental results reveal that the proposed method achieved a 92.94% recognition rate, whereas the SMA achieved a 89.10% recognition rate.