基于视觉特征提取算法的超临界通流特性评估

Evaluation of Supercritical Flow Characteristics Based on Visual Feature Extraction Algorithm

为判断焊瘤对超临界机组金属管道液体流通的阻碍程度,提出了基于逆透视变换的管道通流能力分析算法。该算法设计语义分割网络从整张金属管道射线检测图片中提取焊缝投影;通过二维图像滤波和最小二乘椭圆拟合方法从焊缝投影中提取通流面投影;选取特征点计算单应性矩阵,通过逆透视变换还原投影前通流面形状;在通流面中使用最大熵阈值分割方法提取焊瘤边界,计算瘤径比,分析焊瘤影响。实验结果表明:算法可以根据射线检测图片精确评估焊瘤对管道流通能力的影响,为实现焊瘤检测自动化提供了有效的解决途径。

To evaluate the obstruction degree of weld to the flow of metal pipe outside the supercritical unit,a flow capacity analysis algorithm based on inverse perspective mapping(IPM)is proposed.Semantic segmentation algorithm is designed to extract welding zone from the radiographic image of metal pipe.Two-dimensional image filtering and the least square ellipse fitting algorithms are employed to extract the projection area of the flow surface.Feature points are selected to calculate the homographic matrix and the flow surface before projection is restored through the IPM.The maximum entropy threshold is used to extract the edge of weld within the flow surface and the weld diameter ratio is calculated to analyze the influence of the weld.The experimental results demonstrate that the algorithm could accurately assess the influence of weld on the flow capacity of pipe according to the radiographic inspection images,providing a feasible solution for the automation of weld inspection.