導入非監督式學習方式建立交通違規偵測系統(2/3)

The main purpose of this project is to establish an intelligent traffic rule violation detection system. The imaging device of this system uses the monitor set up on the road. This system can detect whether the traffic rules are broken, the detected events include whether the vehicle runs a red light, whether the vehicle turns left illegally, whether the vehicle crosses a double yellow line or a double white line, whether the vehicle parks illegally on the roadside, and whether the vehicle stops in the area exclusive for the motorcycle waiting for a green light, etc. Besides, this system has also focused on whether the vehicle gives way to pedestrians while they are crossing the street. Therefore, the tasks to be completed in this project include vehicle detection, pedestrian detection, license plate detection and recognition, vehicle tracking, and vehicle trajectory analysis. The deep learning models are used to detect long-term vehicle trajectories and identify vehicles that have violated the regulations. We develop the related algorithms detecting violation of traffic rules by finding out the correlation between the objects on the road. Hopefully, this system can help law enforcement personnel reduce their workload and make more effective use of human resources, and enforce the traffic laws.