The focus of this study is to develop a flexible and quantitative method in place of the expertise to identify the types of tire marks in the photo for further accident analysis. The artificial neuro-network model is an available method to formulate the tire-marks problems influenced by so many factors. The study starts by collecting all kind of tire marks, with the general non-metric camera, on the roadway surface, and obtaining their corresponding digitalised images by the scanner. Thus kind of imaged tire-marks can be represented by it's regularity changes of the gray-levels in the tire-mark area. After training, by the back-forward propagation method, and validating the model, the weights on each link can be determined and the tire-marks identifications model constructed. For the covering abstract see ITRD E116488.
Abstract