Shape- and texture-based 1-D image processing algorithm for real-time stop sign road inventory data collection

Accurate and timely road inventory data is important for supporting the planning, design, construction, and maintenance of a variety of transportation facilities. Road inventory data includes various traffic signs, such as stop signs, speed limit signs, lane numbers, pavement widths, and others. Georgia Department of Transportation (GDOT) is currently developing a real-time road inventory data collection system that uses image processing techniques to improve data quality and operation efficiency for existing road inventory data collection operations. This article presents a simple and feasible image processing technique using a shape- and texture-based one-dimension (1-D) stop sign detection algorithm to provide good detectability and computation time. This algorithm includes three components: color segmentation, Region of Interest (ROI) extraction, and stop sign validation. First, selective color segmentation is used to process stop sign images. Second, a column/row splitting strategy is used to produce a stop sign candidate (ROI) matrix. Finally, two one-dimensional correlation functions are decomposed to speed up the computation of stop sign validation. The algorithm was written with a C program and tested on a Pentium III 1.1 GB computer using the actual stop sign images with an image size of 400 by 300 pixels. The average processing time was 20 ms for each image with a standard deviation less than 3 ms. The results of a test program show that out of 106 images, the program correctly detected 101 images. The computation times for different image sizes are presented and show that this algorithm can provide competitive computation time when image size becomes larger. The results also show that this algorithm is especially good at detecting stop signs that are tilted in all directions or partially occluded because of the use of the shape-and-texture-based 1-D correlation validation algorithm. The characteristics of this algorithm for detecting stop signs are summarized and discussed.