基于改进Otsu算法的原油蒸馏塔金属腐蚀小目标检测仿真  

Simulation of small target detection of metal corrosion in crude oil distillation tower based on improved Otsu algorithm

作　　者：李英波[1] 刘凤花[1] 李娜 LI Yingbo;LIU Fenghua;LI Na(Tianjin Petroleum Vocational and Technical College,Tianjin 301607,China)

机构地区：[1]天津石油职业技术学院,天津301607

出　　处：《金属功能材料》2025年第1期87-91,共5页Metallic Functional Materials

摘　　要：受到光照条件以及背景复杂度等多种因素的影响,金属腐蚀区域与背景区域混合,待检测区域较大,导致腐蚀检测质量不佳,信噪比较高,对此,提出基于改进Otsu算法的原油蒸馏塔金属腐蚀小目标检测方法。采用二维函数,对图像亮度进行描述,结合双边滤波算法提取出光照分量,引入伽马因子以及亮度均值,对光照分量进行校正。在原有分割标准的基础上,加入颜色特征与以及纹理特征参数,结合类间方差构建出分割阈值,从而实现金属腐蚀区域与背景区域的分离处理。将金属区域分割结果划分为不同的子单元,结合疑似腐蚀检验系数对每个子单元进行判断,通过迭代,更新腐蚀区域聚类中心,结合光照分量,输出腐蚀区域检测结果。仿真结果表明,该方法应用后,金属腐蚀图像处理信噪比更高,可以在每个单元下识别出重度腐蚀区域,并具备更为精准的检测效果。Due to various factors such as lighting conditions and background complexity,the metal corrosion area is mixed with the background area,resulting in a large area to be detected,which leads to poor corrosion detection quality and high signal-to-noise ratio.Therefore,an improved Otsu algorithm based small target detection method for metal corrosion in crude oil distillation towers is proposed.Using a two-dimensional function to describe the brightness of the image,combined with bilateral filtering algorithm to extract the illumination component,introducing gamma factor and brightness mean to correct the illumination component.On the basis of the original segmentation standards,color features and texture feature parameters are added,combined with inter class variance to construct a segmentation threshold,thereby achieving the separation processing of metal corrosion areas and background areas.Divide the metal area segmentation results into different subunits,judge each subunit based on the suspected corrosion inspection coefficient,update the corrosion area cluster center through iteration,and output the corrosion area detection results based on the light component.The simulation results show that the metal corrosion image processing is higher,which can identify the severe corrosion area in each unit,and have more accurate detection effect.

关 键 词：蒸馏塔 金属腐蚀图像 检测方法 OSTU算法 聚类中心 分割阈值 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]