Real-Time Traffic Sign Detection and Recognition System for Assistive Driving

• Main Authors: Santos, Adonis, Abu, Patricia Angela R, Oppus, Carlos, Reyes, Rosula SJ
• Format: text
• Published: Archīum Ateneo 2020
• Subjects: Traffic sign detection and recognition; embedded system; computer vision; machine learning; convolutional neural network; Computer Engineering; Electrical and Computer Engineering
• Online Access: https://archium.ateneo.edu/ecce-faculty-pubs/84; https://astesj.com/v05/i04/p71/
• Institution: Ateneo De Manila University

Road traffic accidents are primarily caused by drivers error. Safer roads infrastructure and facilities like traffic signs and signals are built to aid drivers on the road. But several factors affect the awareness of drivers to traffic signs including visual complexity, environmental condition, and poor drivers education. This led to the development of different ADAs like TSDR that enhances vehicle system. More complex algorithms are implemented for improvement but this affects the performance of a real-time system. This study implements a real-time traffic sign detection and recognition system with voice alert using Python. It aims to establish the proper trade-off between accuracy and speed in the design of the system. Four pre-processing and object detection methods in different color spaces are evaluated for efficient, accurate, and fast segmentation of the region of interest. In the recognition phase, ten classification algorithms are implemented and evaluated to determine which will provide the best performance in both accuracy and processing speed for traffic sign recognition. This study has determined that Shadow and Highlight Invariant Method for the pre-processing and color segmentation stage provided the best trade-off between detection success rate (77.05%) and processing speed (31.2ms). Convolutional Neural Network for the recognition stage not only provided the best trade-off between classification accuracy (92.97%) and processing speed (7.81ms) but also has the best performance even with lesser number of training data. Embedded system implementation utilized Nvidia Jetson Nano with interface Waveshare IMX219-77 camera, Nvidia 7” LCD and generic speaker and programmed in Python with OpenCV, sci-kit learn and Pytorch libraries. It is capable of running at an adaptive frame rate from 8-12 frames per second with no detection and down to approximately 1 frame per second when there is a traffic sign detected.