As object detection and object tracking are two connected aspects of video surveillance, they are both regarded as crucial steps in computer vision for video surveillance, traffic analysis, and public safety. The first step to recognize objects in movies, which is necessary before moving on more difficult tasks like tracking. A strong programming model known as deep learning neural networks teaches to represent and abstract data like pictures, sounds, and text at various levels of abstraction. In this paper, a well-known deep learning network, Faster Regional Convolution Neural Network (Faster R-CNN), is applied to introduce the Object Detection and Tracking System (ODTS) for object detection and conventional object tracking algorithm for automatic detection and monitoring of unexpected events in Closed Circuit Televisions (CCTVs) in tunnels. To collect Bounding Box (BBox) results through Object Detection, ODTS receives a video frame in time as an input. Then it compares the bounding boxes of the present and previous video frames to assign a unique ID number to each moving and identified object. This technique allows it to be possible to follow a moving item in real time, which is difficult in object detection frameworks that follow conventional methods. After that, four accident videos featuring each accident were used to evaluate the ODTS-based Tunnel Closed Circuit Television (CCTV) Accident Detection System based on a trained deep learning model. Consequently, the system is able to identify every accident within ten seconds. The most significant aspect is that as the training dataset becomes more extensive, ODTS's detection capacity can be automatically enhanced without modifying the program codes. Accuracy, Recall and F1-Score are used parameters for performance analysis.