Traffic Sign Detection and Classification using Deep Learning Algorithms with Advanced Preprocessing Techniques

View at publisher Summarise with VIRAT
Authors: A. Mary Jenifer, R. Balamanigandan
Cited by: 0
References: 20

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Autonomous vehicles and intelligent transportation systems must rapidly and accurately identify traffic signs. This work outlines a thorough pipeline for traffic sign recognition and classification. The model employs a pre-trained ResNet50 as its foundational architecture for feature extraction. The proposed approach utilises a multi-tiered preprocessing framework that integrates Contrast Limited Adaptive Histogram Equalisation (CLAHE), Fast Non-Local Means (FNLM) denoising, and normalisation to improve image quality and reduce the effects of noise and variable lighting conditions. Utilising CSV-based annotations, regions of interest (ROIs) corresponding to traffic signs are automatically extracted and incorporated into datasets for each category. This results in enhanced accuracy in testing and training. The dataset employed for training and evaluating the model consists of nine unique kinds of traffic indicators. They analyse its performance by analysing the F1-score, accuracy, precision, and recall. Multiple visualisations illustrate the effectiveness of the preprocessing pipeline and the model's capacity to detect signals in difficult conditions. The proposed technique effectively identifies traffic signs in real-world scenarios due to its high classification accuracy and practical efficacy.

Fields of Study & Subjects

Advanced Neural Network Applications Autonomous Vehicle Technology and Safety Computer Science and Engineering

Publication details

Journal / source
International Conference on Image Processing and Capsule Networks (ICIPCN)
Pages
258–262
Publication date
27 January 2026
Language
EN