摘 要：本文采用面向对象的影像分析方法对Lansat8、GF1以及GF2影像中的围网养殖区及其它地物类别进行信息提取，在多尺度分割的基础上，利用选取地物特征的方法进行细化分类，同时基于矢量样本点进行精度评价和比较，分析尺度效应，选取最优的分割尺度。同时，用传统的基于像元的信息提取方法对Lansat8、GF1以及GF2影像进行分类，主要采用监督分类和非监督分类方法，并将这两种分类结果进行比较。实验结果表明：本文在对围网养殖区进行提取时，应用面向对象方法时，Landsat8、GF1以及GF2最优分割尺度总体分类精度分别为50。2563%、61。7982%和82。2139%。而用基于像元对影像中地物类别进行分类时，Landsat8和GF1监督分类的总分类精度分别是为40。1428%和70。1763%，非监督分类的总分类精度分别为67。5141%和72。3942%。综上所述，基于像元分类方法的分类总体信息提取的精度普遍较低。但对于中低分辨率的遥感影像，面向对象的方法并不适用，总分类精度并不是很高，在基于像元的基础上，可以改进分类方法，进一步提高分类精度。从本文的实验结果也可以看出，在精度评价以及分类结果来看，本文认为，面向对象的方法在提取高分辨率影像的时候总体比基于像元的分类方法好。所以进行围网养殖区的相关识别与监测，还是应该选取高分辨率的遥感影像。79434

毕业论文关键词：基于像元，面向对象，最优分割尺度，围网养殖区，高分影像

Abstract：In this thesis, we extract enclosure culture and other types of ground objects from images of Lansat8, GF1 and GF2 based on object-oriented classification method。 On the basis of multi-scale segmentation, using the method of selecting features of objects to refine the classification。 Then based on the vector sample point to evaluate the accuracy, and comparing the classification accuracy of the multi-scale segmentation to analysis scale effect and select the optimal segmentation scale。 At the same time, my paper also used the pixel-based method。 Taking the unsupervised classification with the supervised classification, and then comparing the results of these two kinds of classification。 The experimental result shows that the overall classification accuracy of the optimal segmentation scale of Landsat8,GF1 and GF2 are 50。2563%,61。7982% and 82。2139%%。 And when using the pixel based on the classification of objects in the image classification, the overall classification accuracy of the supervised classification of Landsat8 and GF1 are 40。1428% and 70。1763%; the overall classification accuracy of the unsupervised classification of Landsat8 and GF1 are 67。5141% and 72。3942%, Stated thus, classification based on pixel-based method is generally low accuracy of the overall information extraction。 But for low resolution of remote sensing image, the object oriented method is not suitable, and the overall classification accuracy is not very high。 From the experimental results can also be seen and in terms of accuracy evaluation and classification results, this paper argues that in the extraction of high resolution remote sensing image, the result of the oriented object method is better than the pixel classification method。 We should choose high resolution remote sensing image to distinguish and monitor the enclosure culture area。

Keywords：Pixel based, object-oriented, optimal split-scale, scale effect, Enclosure culture, High Resolution Satellite Images

目  录

1  前  言 5

2  国内外研究现状 5

3  研究内容及其技术路线 6

3。1  研究内容