Real Time Motion Detection for Traffic Analysis Using Computer Vision

Background

All traffic control systems contain sensors or is done manually. The violations detected by the sensors are accurate. But there are few errors in using sensors. Since they require to be placed at the place of violation detection, they also need to be protected. This increases the cost of hardware requirements. In the proposed method, all the violations are detected using only a surveillance camera. The current system focuses on only one violation and thus require many such systems to be integrated to detect multiple violations. In the method stated in (Zheng et al., 2013) it requires aerial image of the place and thus increases the dependency on many factors. These sensors will be placed on top of the area covering the entire area. Detecting violations using multiple trajectories will cause significant calculations to predict the type of violations. This will increase the software requirements to use different models. Traffic violation need not always be detected using models or other machine learning solutions, they can also be detected by using normal methods which detects violations using non-machine learning approach. Many of the speed detection system requires two sensors that are placed in two different places and once the vehicle crosses both the sensors it calculates the speed by using the distance between the two sensors and the time taken by the vehicle to cross the particular area between the sensors. These sensors can measure speed from one direction only.