A divide-and-conquer approach to contour extraction and invariant feature analysis

A novel method to process and analyse spatial information is presented in this paper. The spatial information, such as geometric shapes, object boundaries and trajectories, is extracted as a discrete sequence of points from images obtained through remote sensing. The algorithm generates contour point sequences and then uses a scale invariant analysis to extract invariant arc features. These arc features are subsequently used for object identification and recognition, as well as for image matching. The algorithm detects corner-like features in the presence of low curvature, sharp noise and discretization (spatial quantization), typical for images obtained by aerial photography, digital map scanning, satellite imaging or other remote sensing image acquisition techniques. The resulting feature vectors can be used for stable and robust object feature analysis and object detection. Experimental analysis confirms the efficiency and robustness of this method, using dataset consisting of varied shapes with considerable noise and ambiguity. The method allows not only stable feature detection, but also general shape analysis through identifying convexity, linearity and curvature properties.