Detecting Red-Light Runners (RLR) and Speeding Violation through Video Capture

Abstract

The rise in the number of vehicles entails a higher chance of increase in traffic accidents. In that case, it is necessary to catch traffic violators to lessen accidents and guarantee safety in the road. Studies show that the leading cause of road collisions is traffic violations such as red-light violation and speeding. The proponents present a method for detecting vehicles in video clips using a single deep neural network called Single Shot Detector (SSD). Since this system detects violators faster than the human eyes, it makes the responsibilities of officers more effective and efficient. Traffic violation detection system helps to recognize the most common traffic violations whether it is over-speeding or red-light-running. In this work, the researchers employ a camera to acquire data from an intersection. The data shall be fed to the prototype to compare the results of three different convolutional neural network (CNN) features by applying it to the base network of SSD in detecting traffic violations in captured videos. Based on the results, the prototype can achieve up to 100 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup> accuracy on all feature extractors for red light runner (RLR) and 92.1% accuracy for speeding with the consideration of the best set-up.