Multiple object detection for smart TV shopping video using point to point feature based SURF method

摘要

The paper presents a novel mechanism to enhance the consumer experience of TV shopping. A customer's experience can be enhanced, if he/she could interact with the smart TV to purchase interesting items leading to TV-enabled shopping. For this, a method is proposed which could track and identify the multiple objects in video sequence using surf features. Most of the present multiple object detection methods rely on stable camera location, slow moving objects or same type of objects multiple times. However, in general for TV shopping, this is not the case. For this reason, the system proposed is trained to detect multiple types of objects in a single video. Proposed mechanism will perform well, if the camera is either moving, object movement is fast or the presence of multiple type of objects in a single frame. Also, the experimental results clearly demonstrate the high precision of results leading to low error rate.