Vehicle Speed Estimation and Tracking Using Deep Learning and Computer Vision

This paper proposes an intelligent traffic surveillance system that detects the vehicle and its speed, colour, direction, and type of vehicle using computer vision and deep learning. This information can be used to find the traffic violator using automatic number plate recognition. Research shows that over-speeding accounts for 60% of total accidents in India, which raises a serious concern. The proposed approach uses TensorFlow object detection API for vehicle detection, cumulative Vehicle counting, and colour detection of the vehicle using colour histogram integrated with the KNN machine learning algorithm in a real-time environment and a robust approach using deep learning and computer vision for speed estimation and direction detection. This study will effectively monitor traffic usage and help officials track, detect, and lay a floor plan to effectively stop speeding and wrong-side driving vehicles from getting into accidents. This paper proposed an efficient and robust approach for detecting moving vehicles along with their speed and other attributes. The proposed approach can be integrated with a pre-installed traffic monitoring camera system without significant adjustments.