Traffic Signs Classification using CNN (Terrain Vision

Traffic Signs Classification using CNN (Terrain Vision)

Shalini Kumari

Department of CSE

SOET, CMR university

Bangalore, India

Shalini.hajipura@gmail.com

Gowda Mahalaxmi Kumara

Department of CSE

SOET, CMR university

Bangalore, India

gowdaruchita60@gmail.com

M Hari Teja

Department of CSE

SOET, CMR university

Bangalore, India

hariteja.arg@gmail.com

Indhu S

Department of CSE

SOET, CMR university

Bangalore, India

indhugs2004@gmail.com

ABSTRACT – Traffic sign classification in forest regions define rare challenges due to environmental situations such as partial occlusions, faded signage, varying illumination, and background interference. This study proposes a Convolutional Neural Network (CNN) based model optimized for recognizing traffic signs in such terrains. A specialized dataset comprising real-world forest-region traffic signs is used, supplemented with data augmentation techniques to enhance model generalization. The proposed CNN architecture is designed to balance accuracy and computational performance, doing it suitable for real-time applications in autonomous and assistive driving systems. Past research results demonstrate that the model achieves high classification accuracy, outperforming traditional machine learning approaches. The results help to the development of intelligent transportation systems tailored for rural and forested areas, improving route safety and navigation in challenging environments.

Keywords—Traffic sign classification, convolutional neural network, deep learning, forest terrain, intelligent transportation systems