Abstract: The carbon content in BOF can be characterized by the boundary curvature of flame image, so it is important to calculate the boundary curvature accurately for the end point forecast based on the method of flame image analysis. A calculation method of boundary curvature based on differential chain code is proposed. The boundary curvature formula of flame image based on differential chain code is defined. For the effectiveness of the formula, the segmentation, tagging, binarization, morphology edge extraction and other pre-treatment are done to the original images. the flame boundary is rebuilt by polygon approximation aimed to overcome the shortcomings of curvature calculation using differential chain code. Experimental results show that this method has a better distinguish characteristics of the different stages of blowing than the methods of circular or image skeleton curvature measurement. Also this method has a good calculation speed, so it has a high practical value.

Key words: flame image, morphology boundary, differential chain code, curvature, circular, image skeleton

摘要： 转炉火焰图像边缘弯曲程度能表征炉内铁水碳含量的多少，准确计算曲线弯曲度，对火焰图像判断转炉吹炼终点有重要的意义。提出一种基于差分链码曲率的边界弯曲度计算方法，定义了基于差分链码的火焰边界曲率计算公式;为了保证曲率公式的有效性，对火焰原始图像作分割、标记、二值化、形态学边界提取等预处理;利用多边形近似方法重建了火焰边界的二值图像，有效克服了边界小尺度拐角对曲率公式的影响。实验结果表明，该方法与边界圆形度及火焰骨架曲率计算的方法相比，所提取出的特征值能明显区分吹炼的不同阶段，且计算速度快，具有较高的实用价值。

关键词: 火焰图像, 形态学边缘, 差分链码, 曲率, 圆形度, 图像骨架