DEEP CNN FOR SMART MOBILITY: AN AI-BASED TRAFFIC FLOW ANALYZER FOR ADAPTIVE URBAN INFRASTRUCTURE

Abstract

This research introduces a novel hybrid deep learning-based system for intelligent road traffic 
monitoring, aimed at improving transportation safety through accurate identification and 
classification of road incidents. The system integrates multiple machine learning models—
 including a Random Forest Classifier, a Deep Neural Network (DNN), and a hybrid model 
that combines a Convolutional Neural Network (CNN) with an Extra Trees Classifier 
(ETC)—to detect and classify incidents such as accidents, dense traffic, fire, obstacles, 
smoke, and sparse traffic. A comprehensive dataset of road scene images is preprocessed, 
divided into training and testing sets, and used to train these models. Among them, the hybrid 
CNN+ETC model achieved the highest accuracy of 95.14%, significantly outperforming the 
others. The system is equipped with an intuitive graphical user interface (GUI) that allows 
users to upload datasets, preprocess images, train models, and perform incident predictions on 
test images. It also offers visualization tools such as accuracy and loss graphs, along with 
confusion matrices for performance evaluation. The model has shown precise prediction 
capabilities, correctly identifying scenarios like smoke, accidents, and heavy traffic. This 
application not only demonstrates strong practical relevance for real-world use but also 
represents a promising step forward in advancing real-time traffic incident detection and 
smart transportation management systems.