CHLAC based vision sensing method for bicycle rider detection to avoid confusing similar shape pedestrian

This paper proposes robust bicycle rider detection method from video sequence. The proposed method does not confuse rider with similar shape pedestrian such as bicycle pusher. There are a large number of accidents happened while riding a bicycle. To prevent these accidents, bicycle rider detector is required for traffic monitoring camera system. Thus, many researches have been done in this filed. However, previous works make no mention of discriminating bicycle rider from pusher, to say nothing of bicycle rider detection. In order to realize the detection method to avoid confusing such similar shape pedestrian, the authors utilize CHLAC (Cubic Higher-order Local Auto-Correlation). The proposed method can detect human automatically and also recognize bicycle rider. In an experiment using video sequence, bicycle rider detection rate can be achieved 80.23%. Experimental results prove effectiveness of the proposed method.