Autonomous vehicles have the potential to improve safety by eliminating human error in driving, as well as providing mobility to those who cannot safely drive. Such vehicles do require new technology to monitor their environment and ensure that they are operating safely. One such technology that will be necessary is the ability of the vehicle to recognize traffic situations and traffic signs. This can be accomplished by an appropriate implementation of edge detection methods. In this paper, we compare three different edge detection methods: Canny method, Sobel method and Zhang method. This comparison was conducted on both still pictures and on a video. When analyzing the video, which was taken on a clear day with an undamaged and clearly visible stop sign, all three methods performed equally well; the time at which the stop sign was identified, based on the edge map, was the same. The purpose of this comparison is to evaluate the performance of each of the three methods, in the context of the problem of identifying traffic signs. The methods are compared on still images of a stop sign under various conditions, in addition to the single video comparison. Based on the still image comparison, we conclude that Zhang’s method (linear prediction) generates the best edge map, particularly when the images include snow, ice, rain or other factors and even at night vision.

This content is only available via PDF.
You do not currently have access to this content.