Online graphics recognition system

Abstract

Online graphics recognition system (OGRS) can provide a natural, convenient, and efficient way for users to input rigid and regular shapes or graphic objects, such as triangles, rectangles, ellipses, straight lines, arrowheads and so on. All what users have to do is quickly drawing their sketchy shapes in single or multiple strokes, and the system will automatically convert the strokes into the user-intended rigid shape based on the shape similarity and the time constraint of the sketchy line. Two main stages are involved in the system, one is stroke segmentation and the other is composite objects matching. For the stroke segmentation stage, Liu Yin in our research group proposed a new dynamic programming (DP) framework, which employed a penalty function to eliminate the requirement of knowing the number or templates of the segments in traditional dynamic programming frameworks. In most situations, a single stroke contains less than 100 points and experiment shows that the new DP framework can give an accurate result and respond in real-time. However, in extreme cases, when strokes containing more than 200 points are processed, the response time is relatively slow. In order to further improve its performance, I proposed another DP framework, which was also based on penalty function. Experiments show that my improved DP framework is efficient and accurate. Both of the two penalty-based DP frameworks have predefined parameters and I developed a Parameter-Optimization system to find the best parameter set for them in the parameter space. For composite objects matching, I conducted a complete experiment to test the accuracy of an algorithm proposed in [25] and in addition, developed an algorithm based on linear programming to match two acyclic graphs in polynomial time. An experiment was also conducted to show the correctness of the algorithm.