Regional Detection of Traffic Congestion Using in a Large-Scale Surveillance System via Deep Residual Traffic Net

With the growing demand for Smart cities applications, traffic control and its management have huge demand and a highly interested research area. Surveillance images and videos can be monitored effectively to identify traffic congestions. There is existing research available on traffic signal controls through image processing and various machine learning methods. Traffic prediction through surveillance camera videos, images are very interesting as it can update with live data for users. The proposed work detects traffic prediction based on multiclass problems. There are four classes considered for this proposal are heavy traffic, less Traffic, accident prediction and fire accident prediction. As a result, the suggested approach outperforms existing systems that rely mostly on binary categorization. For image training and detection, the suggested work uses a single deep learning technique, Convolutional Neural Network (CNN). With low maintenance, the proposed system can be used for large-scale traffic surveillance systems. The proposed system attained the best accuracy of 80% for 20 epoch training with four detection classes, according to the results of the experiment.