Estimating the Positions and Postures of Non-Rigid Objects Lacking Sufficient Features based on the Stick and Ellipse Model

We propose a method for tracking non-rigid objects, using an object model generated automatically from a set of sample images. Our model consists of multiple sticks and ellipses which represent the skeleton and the areas of an object, respectively. Because appearance features have to be extracted, previous methods cannot estimate the whole area and posture for 2-D image of a non-rigid object lacking sufficient characteristic features (e.g., texture patterns, shape and so on) to be detected easily. With the proposed model, on the other hand, our method can work well because (1) each component of the model can fit each rigid part of a non-rigid object and (2) the reliability of each component is evaluated. To confirm the effectiveness of the proposed method, we conducted several experiments with goldfish. The tracking system automatically generated a model of the goldfish, and could then track goldfish even when they were partially occluded.