基于水平集的转炉出钢图像钢渣检测方法  

Level set based slag detection method for BOF vessels tapping images

作　　者：江俊[1] 李培玉[1] 唐任仲[1] 

机构地区：[1]浙江大学机械与能源工程学院,浙江杭州310027

出　　处：《浙江大学学报（工学版）》2011年第2期222-227,共6页Journal of Zhejiang University：Engineering Science

摘　　要：针对转炉出钢红外灰度图像具有噪声大、物体间灰度差较小和边缘模糊等特点,难以实现区域分割及下渣检测的问题,提出一种基于水平集的转炉出钢图像钢渣检测方法.根据钢水与钢渣的轮廓分布特点对Chan-vese活动轮廓模型进行改进,构造双轮廓曲线的演化函数.利用水平集方法求解模型,并通过最小化双轮廓曲线的演化函数来获得分割区域及其灰度特征值.最后通过分割区域的灰度特征值识别钢水与钢渣及计算钢渣含量,完成钢渣检测.实验表明,该方法具有较高的抗噪性和准确性,能很好地应用于红外热成像转炉出钢下渣检测.Infrared images of basic oxygen furnace（BOF） vessels tapping stream always had harsh noise,small grayscale difference between target objects and blur edges.It＇s difficult to achieve the target region segmentation and slag detection.In order to solve these problems,a level set based slag detection method for BOF vessels tapping images was put forward.First,according to distribution characteristics of molten steel and slag contours,the active contour model of Chan-vese were improved to construct a double-contours of curve evolution function.Then using the level set method to solve model,the segmentation of the region and its gray value could be obtained easily by minimizing double-contours of curve evolution.Finally,in terms of regions＇ gray values,molten steel and slag could be identified,and slag ratio be calculated.Experiment shows the high adaptability,noise immunity and accuracy of our method,which can be impressive usefully implemented in Infrared thermal imaging slag detection for BOF vessels tapping.

关 键 词：转炉出钢 几何活动轮廓 水平集方法 下渣检测 

分 类 号：TF713.7[冶金工程—钢铁冶金]