| Soil and water conservation monitoring is the basis of finding out the development andthe effect of water and soil conservation. The remote sensing monitoring method of soil andwater conservation is an effective mean to improve the monitoring precision of theinformation of soil and water conservation measures and realize real-time, rapid, quantitative,a wide range of dynamic soil and water conservation monitoring. And it is the key point ofmanagement and research on soil and water conservation information and business. Due tothe terrain complexity of the area with serious soil erosion and the diversify of remote sensingimage feature of soil and water conservation practices, there is lots of difficulties of soil andwater conservation practices information extraction using remote sensing images.Object-oriented information extraction method can comprehensively consider the groundobject’ spectrum feature, structure feature, texture feature etc, ananlysis remote sensing imagein multi-scale, use soft classifier designed on the basis of fuzzy mathematical theory forclassification and is closer to human cognitive way of things. So this paper will explore theapplicability of using Object-oriented method to extract soil and water conservation practices.Based on the above considerations, Ma Zhuang basin of Yan’an City was taken as researcharea in this paper, and the SPOT5remote sensing image and the DEM of this area were takenas data source. First of all, Grey-Level Co-occurence Matrix and Grey-Level Co-occurenceMatrix of Color Image were analysed and contrasted in this paper. And then, texture featureswas extracted from remote sensing images. Secondly, research on object-oriented informationextraction of soil and water conservation measures and object-oriented information extractionof soil and water conservation measures combined with texture characteristics was carried on.The conclusions are as follows:1. Using the method of Gray Level Co Occurrence Matrix(GLCM) and Color Gray LevelCo Occurrence Matrix to extract texture features, conclusions can be drawn by systematiclyanalyzing and comparing these features:The average of four directions(0°,45°,90°,135°) canbe used to describe texture feature in the corresponding direction. The texture feature is morerobust using3~9pixels as the distance parameter. Texture extraction results are little affectedby grayscale, so the grayscale can be compressed to level16without affecting the extractionresults. The texture feature is more robust when the dimension of the window is larger than50 pixel. CGLCM is more superior than GLCM in analyzing and conveying texture results.2. Two kinds of object-oriented soil and water conservation measures were contrasted inthis paper, the results are as follows:(1) If object-oriented method is adopted to improve theinformation extraction of soil and water conservation measures, the user accuracy andproduction accuracy of Vegetation measure will be greater than70%. And, for the two kindsof engineering measures which are dam land and terrace, the user accuracy and productionaccuracy will be greater than60%. The overall accuracy of classification result is77.34%.The Kappa coefficient is71.96%. Very good quality classification standard is arrived in.(2)When using CGLCM to extract texture features and the Object-Oriented method to extract theinformation of soil and water conservation measures, the user accuracy and productionaccuracy of vegetation measures is above80%, the accuracy is more than70%in dammedfield and terraces measures, overall accuracy is86.19%, Kappa coefficient reaches88.26%,this is a very excellent classification quality standard.(3) Soil and water conservationmeasures can be extracted with Object-Oriented method, and in combination with texturefeature using CGLCM, the extraction accuracy can be greatly increased, the overall accuracycan be increased by8.85%, Kappa coefficient increased by10.3%. |
PDF Full Text Request