Steel Water Continuous Casting Slag Detection Method Based On Image Recognition

With the rapid development of China's steel industry, higher requirements have set on the quality of steel production and the degree of automation. The subject investigated is the process of bale slag detection in the continuous casting production workshop of Wuhan Iron and Steel Plant. According to the poor working environment, importance in production, through plenty of theoretical research and on-site analysis, a new slag detection system of molten steel casting is presented, which has real-time slag detection and off-line image analysis functions. Specific designs are also achieved.Combined with the environment of steel casting scene and the actual demand, the system adopted the image enhancement program and the feature extraction program. Contrast enhancement, image smoothing and image sharpening used in the slag detection system were theoretically elaborated in detail. Meanwhile, the feature extraction technique based on color, texture and shape features was minutely analyzed. Finally, the different image analysis algorithms were given and the image analysis experiments were done by using on-site photos.The slag detection system based on image recognition comprised two main components: the industrial-camera-based image acquisition system and the PC-based slag detection software. Between image acquisition system and PC software, Ethernet was used to gather control parameters and transmit video data. The system collected and transmitted the video images of molten steel surface of pouring box through the camera. The detection software was used to judge whether the bale slag through timing crawls, enhancement, different methods of feature extraction and analysis video images. Meanwhile, using the off-line image analysis function of the software analyzed the preserved image to optimize the parameters of the decision-making module.Paper introduced the slag detection software implementation on the PC side, and pointed out the software design techniques from multi-thread control, file access operation and image analysis method. Then, it analyzed the software modules, the realization of subsystems, mutual visit between modules and cache. In the Last, it displayed the main flow chart and effect map of the software.Finally, the paper made a conclusion about what have done and future work prospects.