Deep learning based condition monitoring of road traffic for enhanced transportation routing

The efficient management of road traffic is crucial for enhancing transportation routing and improving overall traffic flow. However, the conventional methods can not accurately analyze real-time traffic data and do not provide valuable insights for effective transportation routing decisions. So, this work proposes a deep learning-based approach for road traffic condition monitoring networks (RTCM-Net) for various illumination conditions. Initially, the fuzzy block-based histogram equalization (FBHE) method enhances the colour properties of the input image, which improves the low-light conditions, haze removal, illumination condition balancing, and colour balancing. The proposed approach leverages deep learning techniques, specifically convolutional time domain neural networks (CTDNN), to learn and extract meaningful features from road traffic data. By training the CTDNN model on a large-scale dataset comprising historical traffic patterns, the system can effectively capture complex traffic conditions and identify anomalies or congestion in real time. Finally, the RTCM-Net is capable of classifying the high, low, dense traffic, fire attack, and accident classes from the input images. The proposed RTCM-Net achieved high accuracy at 99.51%, sensitivity at 98.55%, specificity at 98.92%, F-measure at 99.98%, precision at 99.42%, Matthews Correlation Coefficient (MCC) at 99.72%, Dice as 98.62%, and Jaccard as 99.48% scores, indicating its effectiveness in classifying and monitoring road traffic conditions, which are higher than traditional approaches.