Traffic surveillance systems generate vast amounts of data regarding road conditions every second. Monitoring this data manually is labor-intensive and prone to human error. To enhance traffic monitoring and control, we propose utilizing Deep Learning Convolutional Neural Networks (DLCNN). This approach involves preprocessing traffic surveillance data to create a robust training dataset. We develop the Traffic-Net by adapting and retraining an existing network specifically for traffic applications using our self-established dataset. This Traffic-Net can effectively detect regional traffic patterns in large-scale implementations and is suitable for widespread deployment. Furthermore, the DLCNN is employed to predict traffic statuses, including dense traffic, low traffic, accidents, and fire incidents based on test samples. Simulation results demonstrate that the proposed DLCNN outperforms existing models, showcasing its potential for improving traffic management systems.