Detection of traffic signal by adaptive approach and shape constraints

Thresholding plays a vital role in image segmentation. A wrong selection of threshold may lead to improper segmentation. Adaptive approach will be good enough for the selection of threshold in many occasions. This paper proposes an iterative approach for segmentation of traffic signal based on the prior knowledge about the size of the traffic signals. Region of Interest is found based on a tentative threshold and the threshold is varied according to the previously computed dimension of the identified object. Initially the input colour image is smoothed by applying morphological opening operation. The path information of the contour is used to find the coordinates of the clipping window. The image statistics of the ROI, viz. geometric area, pixel population and perimeter are used to extract the traffic signals from the input image. Here, the input colour image is divided into three channels and the threshold is applied to the red channel to extract the red as well as amber signal and green channel is processed to extract the green signal. The method is tested with 50 images and found successful.