基于颜色和纹理特征的立体车库锈蚀检测技术
Corrosion detection technology of stereo garage based on color and texture features
岳仁峰 1张嘉琦 2刘勇 1范学忠 1李琮琮 3孔令鑫3
作者信息
- 1. 山东爱普电气设备有限公司济南高新分公司,山东 济南 250107
- 2. 山东大学控制科学与工程学院,山东 济南 250061
- 3. 国网山东省电力公司电力科学研究院,山东 济南 250003
- 折叠
摘要
针对立体车库锈蚀检测的迫切需求,提出基于颜色和纹理特征的锈蚀检测新方法.利用高斯滤波和伽马变换解决锈蚀图片光照不均匀的问题.采用HSV(hue saturation value)色彩空间实现锈蚀的颜色特征筛选,提出基于灰度共生矩阵进行锈蚀纹理特征分析的方法,对锈蚀区域进行测量和形状分析.结合方向梯度直方图(histogram of oriented gradients,HOG)特征提取和支持向量机(support vector machine,SVM)算法实现了立体车库锈蚀检测.试验结果表明,该方法锈蚀识别准确率达到 93.19%,实现了立体车库锈蚀表面的视觉检测,大大减少了外部环境的干扰.
Abstract
A new method based on color and texture features was proposed to meet the urgent need of corrosion detection in stereoscopic garage.Gaussian filtering and gamma transform were used to solve the problem of uneven illumination of corrosion images.The color feature selection of corrosion was realized by using hue saturation value(HSV)color space,and the texture feature analysis method of corrosion based on gray level co-occurrence matrix was proposed to measure and analyze the shape of corrosion area.Combining histogram of oriented gradients(HOG)feature extraction and support vector machine(SVM),the corrosion detection of stereoscopic garage was realized.The experimental results showed that the accuracy of rust identification using this method reached 93.19%,achieving visual detection of rust on the surface of the stereo garage,significantly reducing external environmental interference.
关键词
锈蚀检测/图像处理/GrabCut算法/HSV色彩模型/机器学习Key words
corrosion detection/image processing/GrabCut algorithm/HSV color model/machine learning引用本文复制引用
基金项目
山东省重点研发计划(重大科技创新工程)(2021CXGC010301)
出版年
2024
NETL
NSTL
万方数据