Design Of Multi-Sensor Data Fusion Based Billet Surface Temperature Measurement System

Billet surface temperature is one of the most important parameters in the steel industry. It can improve the billet quality, reduce the consumption of energy, and be used to implement closed-loop optimization control. This thesis detects the position between the secondary cooling zone export and the straightening machine, because of the adverse circumstances and oxide skin, single sensor is not accurate and stable enough. Based on the former measurement system, this thesis designs a paneled CCD and infrared thermometers based fusion temperature measurement system, it mainly study the image pre-procession, different scale data registration, location methods of moving object and fusion method design.Firstly, the preprocessing of the image collected by area array CCD camera in the fusion temperature measurement system is studied. This thesis proposes the area median filtering method by analyzing noise characteristics to deal with the billet image disturbed by noise, it calibrates the CCD array with two-point method for the non-uniform problem of CCD camera imaging. For the geometric distortion problem of billet image, the pinhole camera model is used to calibrate video camera, then image geometric distortion is corrected and the calibration results is verified, the mechanism model and image data fusion based temperature field restore method is then proposed.Then, an in-depth research for multi-sensor fusion temperature is studied. To guide the infrared thermometer measuring in oxide-free regions, the best measurement location selection method combination of the regional average temperature, position and variance is proposed. By analyzing the fluctuations in temperature curve, the data matching algorithm based on feature vector for the automatic registration problem under conditions of no external positioning device sensor data in different scales is put forward. For the high matching error rate problem of target location in video streaming caused by factors such as information-poor targeting, geometric distortion and image intensity, maximum entropy is used as a template region, and a positioning information entropy vector algorithm in light of the characteristics region is proposed, which is also verified by experiment. According to the results of registration and target locating, a static and a dynamic multi-sensor fusion based measurement system are designed, which combines both the accuracy and stability effectively.Finally, after the disposal of the billet images acquired from a steel company, the method and software designed by this thesis is found every effective, the fluctuation range of the curve is controlled within5℃.