SOM-Based Classification Method for Moving Object in Traffic Video

We do research on the problem of moving object classification. Our aim is to classify moving objects of traffic scene videos into pedestrians, bicycles and vehicles. The self-organizing feature map (SOM) is an unsupervised learning algorithm, which is developed by simulating the signal processing of human brain, has the advantage of simple principle and self organization, and has been used in many fields. This paper applies SOM combined with K-means to moving objects in traffic video, constructs a system including four parts, and proposes an improved method to obtain initial background when using subtraction method to do motion detection and a tracking method based on bidirectional comparison of centroid to track moving objects. Experimental results demonstrate the effectiveness and robustness of the proposed approach.