Intelligent Detection of Object's Anomalies for Road Surveilance Cameras

Nowadays intelligent data processing tools are widely applying into cities infrastructure in order to make citizent's life safer and convenient. One of the most important part of it is video surveillance, which plays important role in many traffic flow monitoring applications. Recent advances in hardware, computer vision and machine learning techniques makes it possible to solve different tasks for applications in real time without human, including recognition of different objects, anomalies and incidents detection. In this paper, we analyzed the impact of image pre-processing technique such as grayscaling on our AI based object detection and tracking model. We also used machine learning algorithms to detect object's frequent trajectories and anomalies on real time video from road surveillance cameras. Based on this, we suggested an approach to detect violations of legal trajectories by vehicles or pedes trians.