船舶涡轮机叶片细小裂痕视觉显著性检测研究
Research on visual significance detection of small cracks in ship turbine blades
张麟华 1王煜1
作者信息
- 1. 太原工业学院计算机工程系,山西太原 030008
- 折叠
摘要
为精准掌握涡轮机叶片的运行状态,判断其是否需要进行更换,提出船舶涡轮机叶片细小裂痕视觉显著性检测方法.利用改进同态滤波算法平滑叶片CCD图像,增强图像的对比度以及均衡度,基于谱残差视觉注意模型提取显著图,通过线性迭代聚类算法分割获取的显著图,设定判断阈值,确定最终的船舶涡轮机叶片微小裂痕区域,完成船舶涡轮机叶片细小裂痕检测.测试结果显示,该方法具备较好的应用效果,可显著提升图像整体均匀度;显著图的提取效果较好,平均绝对误差至均在0.021以下,可靠确定船舶涡轮机叶片微小裂痕区域.
Abstract
To accurately grasp the operating status of turbine blades and determine whether they need to be replaced,a visual significance detection method for small cracks in ship turbine blades is proposed.Using an improved homomorphic filtering algorithm to smooth the blade CCD image,enhance the contrast and balance of the image,extract saliency maps based on spectral residual visual attention model,segment the saliency maps obtained through linear iterative clustering al-gorithm,set the judgment threshold,determine the final small crack area of the ship turbine blade,and complete the detec-tion of small cracks in the ship turbine blade.The test results show that this method has good application effects and can sig-nificantly improve the overall uniformity of the image;The extraction effect of saliency maps is good,with an average abso-lute error below 0.021,which reliably determines the small crack areas of ship turbine blades.
关键词
船舶涡轮机/叶片细小裂痕/视觉显著性检测/图像处理Key words
ship turbine/small cracks in the leaves/visual saliency detection/image processing引用本文复制引用
基金项目
山西省科技厅重点研发项目(201903D121171)
出版年
2024
NETL
NSTL
万方数据