Image processing and behavior planning for robot-rat interaction

In this paper, we proposed an automated video processing system to replace the traditional manual annotation, and to improve the adaptivity of the rat-like robot to autonomously interact with rats. The feature parameters of rats, such as body length, body area, circularity, body bend angle, locomotion speed, etc., are extracted based on image processing. These parameters are integrated as the input feature vector of Artificial Neural Network (ANN) and Support Vector Machine (SVM) classification methods respectively. Preliminary experiments reveal that the rearing, grooming and rotating actions could be recognized with extremely high rate (more than 90% by SVM and more than 80% by ANN). Furthermore, SVM needs less training computational cost than ANN. Therefore, SVM is superior to ANN for the behavior recognition of rats. By using the SVM-based recognition system, the behavior of the robot is generated adaptive to the rat behavior for different interactions.