Detection and classification of moving Thai vehicles based on traffic engineering knowledge

This paper presents detection and classification of moving Thai vehicles based on traffic engineering knowledge. The proposed technique consists of two main parts as follows. The first part is the detection of moving vehicles using image tracking methods e.g. background and foreground (BG/FG) detection and blob tracking. Such methods can provide the values of vehicle features such as position, length (L) and width (W). The second part is the classification of Thai vehicles based on traffic engineering knowledge which is traffic management for not only controlling traffic lights on a crossroad but also calculating volume/capacity ratio and queue length. Therefore Thai vehicles normally can be separated into five groups i.e. first: bicycle, motorcycle and motor tricycle (Tuk-Tuk); second: passenger car, pickup, van and passenger pickup; third: six-wheel truck and mini bus; fourth: ten-wheel truck and big bus; fifth: eighteen-wheel truck and trailer. From above reasons, the second part uses the key features of size (W, L and W/L ratio) from each group which are applied to a decision-tree method for classifying Thai-vehicle groups. The result shows that the use of one input feature is sufficient for the differentiation between 4-group with an overall classification accuracy of 97.37%.