Detection of Lane and Speed Breaker Warning System for Autonomous Vehicles using Machine Learning Algorithm

Abstract

Vehicle’s camera-based lane and speed breaker detection and the drastic updation in modern technology needs the transportation to be automated and self-driven. It depends on the outcome of detection algorithms. But, the computation and pre-processing of these algorithms directly affects the prediction of objects for the vehicle to drive autonomously. Thus, to make an efficient detection algorithm, machine learning and computer vision algorithms are utilized in our work for the tasks of the lane and speed breaker detection for an autonomous vehicle. The lane detection is performed by using OpenCV for gradient and color thresholds; and Fully Convolutional Network (FCN). The FCN uses the pixels of the lane image to find the region of interest (the lane) irrespective of the environmental conditions. The speed breaker system is detected by using YOLOv5 custom object detection with the help of diverse augmented datasets. For real time object detection, YOLO algorithm uses neural networks. Speed and accuracy is the important feature of this algorithm. Experiments are carried out using collection of dataset for lane (1795) and speed breaker (1870) detection, and the accuracy is increased quantitatively (Normal, Shadowed and Multiple speed breaker are 78%, 65% and 75% respectively).