The Evaluation of Cascade Object Detector in Recognizing Different Samples of Road Signs
Shahad J. Shahbaz, Ali A. D. Al-Zuky, Fatin E. M. Al-Obaidi · 2022
In road-sign detection, several problems are involved like variations in perspective, illumination, occlusion, motion blur, and weatherworn deterioration of signs. The goal of the current work is to assess the quality of the cascade technique in detecting and recognizing road signs during the daytime in Baghdad City under different circumstances such as vehicle speed, environment, etc..., and determine the best range for the threshold value. The method consists of two main stages; labeling and training to detect and identify the true and false targets upon a set of captured images. The proposed threshold range has been implemented through cascade software and investigated for solving the problem's vision of road sign deterioration. The precision value for all used road signs obeys a polynomial relationship in precision-threshold variation with the highest value recorded for speed-40 followed by speed-80, cross, and speed-60 road sign at the final. The conducted experiments have confirmed the proposed software operability and allow recommendation for use in practice for solving the problems of road sign diagnosis and automatic classification.