AUTOMATIC INCIDENT DETECTION: WRONG-WAY VEHICLE DETECTION USING IMAGE PROCESSING

This paper presents a new function added in an automatic incident detection system. It concerns an algorithm based on image processing able to detect the wrong-way vehicle on the motorway. The algorithm is working in most outdoor conditions (day, night, rain). The principle of the method consists in analysis of the moving vehicles. Each vehicle is tracked in the image and its trajectory is then built. The trajectory is a good criteria to qualify the circulating vehicle direction but in practice, parts of the trajectory include some error due to physical parameters (noise, deformation of objects due to the perspective view, etc.) or to natural phenomena (shadows, glints). A proposed paper presents the following points: a survey of the main difficulties meted in the motion detection field and automatic incident detection; the principle of the method developed by the INRETS; the advantages and drawbacks of the method; and results obtained in field experimentation.