Road Traffic Anomaly Detection (RTAD) plays an important role in Intelligent Transportation Systems (ITS) by enabling the rapid identification of abnormal situations, such as road accidents, hazardous road conditions, and traffic congestion. An efficient ensemble-based RTAD framework, named TriNet-RTAD, is proposed based on the combination of three different deep convolutional neural networks: ResNet50, ResNet101, and EfficientNetB0. The proposed framework integrates heterogeneous CNN architectures using a lightweight probability-averaging fusion strategy to enhance detection robustness while maintaining real-time feasibility. The results revealed that the TriNet-RTAD model outperformed all other models on both datasets, with accuracies of 96.18% and 98.05% for the Traffic-Net and RAD datasets, respectively. The characteristics of the dataset and the performance of the proposed model were discussed to highlight its efficiency. The results show that the proposed efficient ensemble-based RTAD model can provide a consistent solution for RTAD and is suitable for intelligent traffic surveillance systems.
Sathish et al. (Fri,) studied this question.