Abstract—As the growth of automation of daily life careers increases, An intelligent vehicle driving system became an essential necessity nowadays. One of its functions is the ability of recognizing the traffic lights and detecting their status. In this paper, a system for monitoring the traffic light status is presented, by passing the video frames through shape and color classifiers. The shape classifier is based on the extracted edges as the classifier parameter. It uses Multilayer Feed forward Neural Network for classification. Alternately, the shape classifier can be implemented using template matching where the traffic light is detected using its perspectives ratio and the matching with prepared traffic light templates is accomplished. The algorithm tries to find the area of interest where the traffic light may be found. This is done based on SUSAN reliable corner detection procedure, because noise tends to cause false corners. A feature of the SUSAN corner detector is its ability to find corners by direct analysis of the image intensities rather than their derivatives, thus removing the need for smoothing to reduce noise. This makes it faster than most of the other corner detectors like Harris corner detection algorithm and it also provide better localization of the corners regardless of the mask size used. The color classifier uses a range of the red color histogram to recognize the red color of the traffic light. The experimental results show that the proposed system has high performance both in shape and color detection.
Index Terms—Intelligent driving systems, image processing, image matching, neural network, artificial intelligent.
Khalid A. Al-Shalfan is with College of Computer & Information Sciences, AlImam University, Kingdom of Saudi Arabia (e-mail: kshalfan@gmail.com).
Cite: Khalid A. Al-Shalfan, "Auto Vehicle Driving Assistance," International Journal of Machine Learning and Computing vol.5, no. 5, pp. 392-398, 2015.