An Imaging Camera Anomaly Detection System Based on Optical Flow

In the manufacturing industry, automation is crucial for increasing the efficiency of production processes. The robot-based automation is being used in all tasks required in the factory such as transport, processing, inspection and testing. Robot vision, in which a camera is mounted on a robot arm, is actively used in automated inspection. However, during automated inspection, not only the object to be inspected but also the robot vision itself may be damaged, stopped or collided. Therefore, the anomaly detection is required to deal with these problems. In recent research works, for automatic inspection is used deep learning to recognise the image of the object. However, sometime it is not possible to properly get the imaging target due to camera malfunction or dust on the lens. In this paper, we propose a camera anomaly detection system based on optical flow to detect anomalies such as camera failures and object adhesion in lens. The experimental results show that the proposed system can detect anomalies based on optical flow.