Next-gen traffic rule violation detection using optimum feature extraction techniques on highway and toll tax using Raspberry-pi hardware

Across the globe, vehicle collision on roads results in the death/disabilities of people. Moreover, it results in substantial monetary burden to the concerened people and other stakeholdes. Generally, the accidents take place due to ignorance while crossing the lane and use of electronic gadgets. Government is spending a lot of money to create awareness and encourage people to follow traffic rules. Over the last two decades, significant reserach has been carrried out in traffic management system. Generally, sensor based methods are utilized to track traffic violations. These methods need appropriate infrastructure. In this work, authors have proposed a machine-vision based method to recognize the traffic rule(s) violators on highways and at toll tax plazas with the help of some important descriptors of the images and classification algorithms. This paper presents a feature extraction based system for lane and traffic rule voiation detection and tracking using low cost Raspberry Pi hardware.The experimental work suggest that, Grab cut and Hough transform techniques performed better on test image dataset to identify vehicle lane on highways. Further, combination of RootSIFT with Flann-index matcher gives superior results (accuracy of 95.3%) as compared to other feature extraction and matchers on the given dataset for detection of traffic rule violation and tracking of vehicles. The average computation time of 0.13s for the obtained results. Further, Haarcascade algorithm was used to detect mobile phone usage while riding vehicle and achieved 91% accuracy on collected datset on Raspberry pi 2(B) hardware and further vehicles detected in traffic rule violation undergoes for license plate detection and challan generation to penalize the on defaulters.