Traffic sign recognition system (TraSRes)

• Main Author: De Guzman, Steven Edgar G.
• Format: text
• Language: English
• Published: Animo Repository 2006
• Subjects: Signal processing—Digital techniques; Visual communication—Digital techniques; Traffic signs and signals; Computer Sciences
• Online Access: https://animorepository.dlsu.edu.ph/etd_bachelors/4914
• Institution: De La Salle University

To protect passengers and pedestrians, and to increase the possibility of autonomous vehicle navigation, a vehicle may be guided with minimal human intervention using automated vision-based traffic sign recognition. However, existing studies, addressing only specific aspects of the solution, must be improved. Hence, Traffic Sign Recognition System (TraSReS) is a system that detects and recognizes traffic signs from afar while being invariant to lighting condition, perspective distortion, and partial occlusions, thereby not limiting the application to a fully controlled environment only. Edge and colour information are used to detect potential traffic signs. To increase the probability of proper pattern recognition, the perspective distortion of a potential traffic sign is corrected while following the established aspect ratio and the detected symbol is resized afterwards. A comparative analysis on two pattern recognition techniques is performed. Tests are conducted on each of the detection and recognition processes using both artificial images and real-world images. The success rate of the red colour detection is 27.5591%, and the success rate of border detection is 89.7436%. The success rate of symbol detection is 100%. All the false positive cases encountered in the detection processes are rejected in the succeeding processes, and the overall success rate of the all detection processes as a whole is 100%. In the two pattern recognition methods studied, success rates of 70.4762% and 41.9048% are obtained. The average time for processing an input image is 90.5753 seconds. In the study, Digital Signal Processing is applied to establish a foundation of a highly useful traffic sign recognition system and to explore its applications in computer vision."