摘要
图像检索匹配是在给定的图库中找出与查询图相匹配的图像或者无匹配图像。首先,对红外图像进行预处理后用Sobel算子与图像作卷积,得到图像边缘强度图;然后提取图像边缘强度图的灰度共生矩阵作为图像的特征;最后用改进的Manhattan距离方法计算图像特征向量间的距离,并根据一定阈值找出相匹配的图像或无匹配图像。本文用变电站红外图像进行的实验结果表明,本文算法能准确找出相匹配的图像,具有可行性。
Image retrieval, by extracting image feature, identifies similar images to the query one. Image match is based on image retrieval, it finds out the images which match the query one. First, get the difference approximation image with infrared image preprocessing and image convolution with the Sobel operator. Then extract GLCM（Gray Level Co-Occurrence Matrix） of the difference approximation as image features. At last, calculate image similarity using an improved Manhattan distance method, and find match images based on a certain threshold. The results of this experiment performed by the infrared images of the power substation show that the proposed algorithm can accurately identify the matched images.
作者
朱琪
赵飞
潘丽杰
徐本亮
ZHU Qi;ZHAO Fei;PAN Li-jie;XU Ben-liang(Yangtze Delta Region Institute of Tsinghua University Zhejiang,Department of Advanced Manufacturing Technology,Jiaxing Zhejiang 314000)
出处
《数字技术与应用》
2018年第7期129-131,共3页
Digital Technology & Application
