講演情報
[4O4-IS-2b-05]A New Proposal for Road Sign Recognition in Adverse Weather
yudai yahiro1, KIYOMI ASAI1, YUDAI INAHUKU1, 〇TAKENOBU SAWAI1 (1. IWASAKI GAKUEN)
work-in-progress
キーワード：
Autonomous Driving、AI、Traffic sign
This study aims to prevent the misidentification or omission of road signs under adverse weather conditions, such as rain, fog, and snow, where image recognition accuracy typically degrades significantly. In contrast to conventional methods that rely solely on deep learning-based visual recognition, this research proposes a complementary algorithm that integrates spatial information from Geographic Information Systems (GIS) as a "secondary sensor."
Specifically, the developed system cross-references real-time image data captured by on-board cameras with GIS location data synchronized with the vehicle's current coordinates. This integration allows for the statistical optimization of the deep learning model's recognition rates. The experimental results demonstrate that the proposed method maintains high-precision detection even in low-visibility environments. By overcoming the limitations of systems dependent exclusively on visual information, this approach provides a robust solution for the realization of all-weather advanced driver assistance systems (ADAS).
Specifically, the developed system cross-references real-time image data captured by on-board cameras with GIS location data synchronized with the vehicle's current coordinates. This integration allows for the statistical optimization of the deep learning model's recognition rates. The experimental results demonstrate that the proposed method maintains high-precision detection even in low-visibility environments. By overcoming the limitations of systems dependent exclusively on visual information, this approach provides a robust solution for the realization of all-weather advanced driver assistance systems (ADAS).