Analysis of lane detection techniques using openCV

Lane detection is one of the most challenging problems in machine vision and still has not been fully accomplished because of the highly sensitive nature of computer vision methods. Computer vision depends on various ambient factors. External illumination conditions, camera and captured image quality etc. effect machine vision performance. Lane detection faces all these challenges as well as those due to loss of visibility, types of roads, road structure, road texture and other obstacles like trees, passing vehicles and their shadows. There are several lane detection methods having their own working principles and backgrounds, merits and demerits. We have used Receiver Operating Characteristic curve (referred to as ROC hereafter) and Detection Error Trade-off curve (referred to as DET hereafter) which establish the accuracy of computer vision methods. We have studied and analyzed several lane detection methods. The performance of two methods has been analyzed and compared using standard computer vision performance evaluation methods and it was found that method based on Canny edge detection was better than the other one based on Sobel operator. Its performance was established and presented better results when plotted using DET and ROC.