欢迎访问《中国农学通报》，
中国农学通报

中国农学通报 ›› 2010, Vol. 26 ›› Issue (12): 321-323.

• 资源 环境 生态 土壤 气象 • 上一篇    下一篇

基于多重滤波的高分辨率影像中线性特征提取

刘兴1,马雪梅1,肖燕1,李希峰2   

  • 收稿日期:2010-02-02 修回日期:2010-03-18 出版日期:2010-06-20 发布日期:2010-06-20
  • 基金资助:

    聊城大学大学生科技文化创新项目

Linear Features Extraction in the High-resolution Images Based on Multi-filter

Liu Xing1, Ma Xuemei1, Xiao Yan1, Li Xifeng2   

  • Received:2010-02-02 Revised:2010-03-18 Online:2010-06-20 Published:2010-06-20

PDF (PC)

861

摘要：

在高分辨率遥感影像中，线性特征提取一直是个难题，难主要表现在线性地物的边缘信息和噪音是一对矛盾体，边缘信息越明显，噪音就越强。本文利用高通滤波和统计滤波器，构建一个多重滤波，对IKONOS影像进行处理，通过三种不同高通滤波窗口对IKONOS影像处理结果的比对，得到一个较好的高通滤波处理窗口，处理后的IKONOS影像边界特征突出、区域模糊程度减小，目标地物大致轮廓已经区分出来，在此基础上对影像进行统计滤波处理，噪音明显减少，灰度分布更加集中和平滑，对高通滤波和统计滤波处理后的图像进行边缘检测，经过统计滤波后的图像较只进行高通滤波后的图像检测结果较好。实验结果表明：滤波后的影像噪音明显降低，信噪比明显提高，统计滤波后的影像，能够提高检测精度，表明了从高分辨率卫星影像上自动提取矢量信息的可行性，并为边缘检测的细化与跟踪处理提供了可靠的数据来源。

关键词: 拮抗细菌, 拮抗细菌, 桃褐腐病菌, 分离筛选, 16SrDNA, 生理生化反应

Abstract:

The information extraction of the linear features’ edge in the high-resolution remote sensing images has been a problem，which mostly from the contradiction formed by the edge information and noise. The more obvious edge information the stronger the noise will be. In this paper, the high-pass filter and statistical filter are used to build a multi-filter, processing the IKONOS image. Compared with the different processing results of IKONOS images by three different high-pass filters windows, we can choose the best window that make the IKONOS images boundary feature outstanding and reduce the region fuzzy degree. The approximately contour of target scalar can be distinguished. Based on this we processed the IKONOS images by statistical filtering. The images noises decrease significantly and gray distribution more lumped and smoothing. We can find the result which is processed by statistical filtering is better than the result which is only processed by high-pass filtering after the edge detection. The result shows that noise in the filtered image significantly was decreased and the SNR markedly improved. The images processed by statistical filtering could improve the accuracy of detection. It indicates that automatic extract vector information is feasible high-resolution remote sensing images. And it offers dependable information to edge detection thin and track process.