基于GF-2遥感影像的博湖县监督分类方法对比研究

Comparative study on land use classification of Bohu County based on GF-2 remote sensing image

土地利用类型对一个地区各方面的发展都具有一定的影响,对土地类型的了解有助于更加准确把握发展的动向。本文以新疆维吾尔自治区博湖县的一部分作为研究区,选取GF-2遥感影像作为数据源,采用监督分类中的马氏距离法、最大似然法、最小距离法、随机森林法、光谱信息散度法五种分类方法,对博湖县的地物进行分类与识别,通过混淆矩阵法计算总体分类精度,利用Kappa系数对分类结果进行精度验证。结果表明,随机森林法在五种分类中呈现出最好的结果,分类效果最好,其次是最大似然法。这说明随机森林法能够更好地发挥GF-2遥感数据在土地利用分类中的实际应用潜力,且有效提高了各种土地利用类别的分类精度。

The type of land use has a certain impact on the development of all aspects of a region.The understanding of land types helps us to grasp the development trend more accurately.In this paper,apart of Bohu County in Xinjiang Uygur Autonomous Region is used as a research area,and GF-2 remote sensing image is selected as the data source.The five classification methods of Mahal’s distance method,maximum likelihood method,minimum distance method,random forest method and spectral information divergence method in supervised classification are used to classify and identify the features of Bohu County,and finally pass the confusion matrix.The method calculates the overall classification accuracy and the Kappa coefficient to verify the accuracy of the classification results.The results show that the random forest method shows the best results in the five classifications,and the classification effect is the best,followed by the maximum likelihood method.It is indicated that the random forest method can better exert the practical application potential of GF-2 remote sensing data in land use classification,and effectively improve the classification accuracy of various land use categories.