A rat walking behavior classification by body length measurement

To study rat behavior have been playing an important role in psychology, medical science and brain science. Open-field test such as holeboard model is a popular experiment to analyze rat behavior. Rat behaviors such as walking, rearing and head dip are usually considered. These behaviors are observed and recorded by human that, obviously, included human errors. Commercial products have limitation for identifying rat behaviors. In this paper, we proposed a new method for classifying a walking behavior in Holeboard model test based on length of rat's body. Webcam is used to record data. The camera is installed over the models. The proposed method consists of three main processes. The first step is a background modeling; K-mean clustering technique is adapted to reconstruct the background. Second step, rat is extracted by means of background subtraction. Third step is an ellipse fitting by least square method. Then a length of rat's body is calculated for classifying rat behaviors. To test performance of the proposed method, classification accuracy is considered. 500 frames from five image sequence data sets are used. Based on pilot test, criterion of rat's body length for classifying walking behavior is 31 pixels. If the length of rat's body is greater than 31, it is indicated as rat's walking behavior, in the other hand, it is others behaviors. Accuracy of the proposed method is 72.52%. The result shows that the proposed method is satisfactory and able to be improved for higher performance. An advantage of the proposed method is that it is developed for recording rat behavior from a distance and classifying rat's walking behavior which decreases effect to rat.