Detection and Recognition of Obscured Traffic Signs During Vehicle Movement

The present study proposes an algorithm to extract obscured traffic sign information from road driving images and fuse it with the vehicle movement process by integrating the vehicle speed. Given the low recognition rate of obscured traffic signs, the proposed algorithm utilizes a traditional color-...
Ausführliche Beschreibung

The present study proposes an algorithm to extract obscured traffic sign information from road driving images and fuse it with the vehicle movement process by integrating the vehicle speed. Given the low recognition rate of obscured traffic signs, the proposed algorithm utilizes a traditional color-shape recognition approach to extract potential traffic sign regions. The fusion of traffic sign information from multiple consecutive frames into a single frame enhances the completeness and accuracy of the traffic sign information, bringing it closer to the original characteristics of the traffic sign. The experimental results demonstrate that fusing traffic sign information from the first and third frames improves the template matching similarity by 15.2%. And successfully identifies traffic signs that cannot be recognized by the YOLOV4 and YOLOV8 convolutional neural network when driving at vehicle speeds of <inline-formula< <tex-math notation="LaTeX"<$18\left ({\mathrm {km}/\mathrm {h} }\right)$ </tex-math<</inline-formula<, <inline-formula< <tex-math notation="LaTeX"<$36\left ({\mathrm {km}/\mathrm {h} }\right)$ </tex-math<</inline-formula<, and <inline-formula< <tex-math notation="LaTeX"<$54\left ({\mathrm {km}/\mathrm {h} }\right)$ </tex-math<</inline-formula<. The findings highlight that the proposed algorithm can effectively fuse obscured traffic sign information during vehicle movement to obtain traffic signs with more similar geometric features to the original traffic signs. Ausführliche Beschreibung