Improving Blur Image Detection on Traffic Signs for Autonomous Car Using Adaptive Wiener Filter

Abstract

Autonomous cars are innovations in artificial intelligence technology that allow vehicles to operate autonomously without human intervention. These systems' critical features are the ability to detect and recognize traffic signs, which plays a crucial role in safe and efficient road navigation. A significant challenge in developing this system is image blurring caused by camera movement on the vehicle, which can lead to errors in traffic sign detection and interpretation. To address this issue, this research proposes using an Adaptive Wiener Filter integrated with YOLOv5 to improve blurred image quality and increase traffic sign detection accuracy. This study uses 600 traffic sign image data as a test dataset, with data collection conducted at car speeds varying from 40 km/hour to 60 km/hour. The test results show that applying the Adaptive Wiener Filter significantly improves the image quality, with an average PSNR value reaching 28.03 dB and the detection accuracy increasing from 91.66% to 98.34 %. This improvement shows that the proposed method effectively copes with various vehicle speed conditions.

Access to Document

Other files and links

• Link to publication in Scopus

Fingerprint