Analysis of Edge Detection for Road Lanes through Hardware Implementation

Abstract

Recent years have witnessed the rapid advancements and commercial success of autonomous vehicles due to the developments in Image Processing, Machine Learning and AI. This has attracted researchers from both academia and industry to conduct various research on driverless or self-driving cars. One of the significant aspects of self-driving vehicles is the detection of lanes through Edge Detection to correct the vehicle from inadvertent road departure. As real-time Edge detection is a fundamental step of Complex Image processing that handles lane detection and autonomous driving, it is crucial for engineers to select optimum edge detection technique. This paper, presents a practical approach to compare various edge detection algorithms to figure out the most ideal method for lane detection, considering processing time and efficiency. Though, multiple research works are available regarding edge detection method, all of these comparisons are either software simulation based or was not focused primarily on the task of lane detection. Thus, creating an absence of hardware-based practical implementation and comparison of the edge detection methods. In this work, we have discussed the working of four most commonly used edge detection technologies: Sobel, Roberts, Laplacian and Canny, followed by the results from implementing them on a scaled autonomous electric car using Raspberry Pi, Arduino and Raspberry Pi Camera Module, under similar test conditions. Furthermore, a detailed literature review is also presented in this article.