Design And Implementation Of Continuous Casting Pouring Anomaly Detection System Based On Intelligent Vision

During the continuous casting of steel,if the molten steel level of the crystallizer fluctuates abnormally,the quality of the product will be affected,or,the safety of equipment and personnel will be threatened.At the same time,the actions of the production personnel are also closely related to the safety of the entire production line.Although most traditional industrial production lines have visual sensors,they are not equipped with efficient back-end image analysis systems.In response to the above problems,this paper proposes a set of design and implementation scheme for anomaly detection system during the continuous casting of steel based on intelligent vision.For continuous casting pouring liquid level detection,this system uses a method of liquid level anomaly detection from the perspective of computer vision.The method combines image technology and dynamic modeling ideas,uses a set of parameters to measure the safety of the current pouring liquid level,and alarms when abnormal.For the action recognition of production personnel,this system uses the skeleton modeling and graph convolution techniques in deep learning to first detect the key points of the human skeleton in continuous video,generate a graph structure,and then cooperate with the neural network to determine the action category.In order to fully realize and verify the above-mentioned functions,the main body of this system integrates a common monitoring platform framework to complete the development work.It also completes the collection and segmentation of high-definition video signals in a combination of software and hardware,then sends them into two functional modules to reduce the burden of the algorithm on processing image signals.Multi-stage test shows that the liquid level method can accurately capture the anomalies such as steel leakage and steel overflow during continuous casting.The system can cooperate with the existing safety system in the factory.The abnormal motion recognition function can give a result of the worker motion in the image,and assist the monitoring personnel to measure the safety factor.The entire system has application value and upgrade space.