基于GF-1影像的渭-库绿洲外围土壤含盐量定量反演研究

发布时间：2018-03-04 13:37

本文选题：GF-遥感影像　切入点：BP神经网络　出处：《中国农村水利水电》2017年02期 　论文类型：期刊论文

【摘要】：为探讨国产GF-1卫星影像在干旱区土壤盐渍化监测中的适用性,以渭-库绿洲外围荒漠交错带为研究对象,利用BP神经网络和RBF神经网络2种建模算法,以GF-1影像的4个波段的反射率及影像提取的归一化差异植被指数(NDVI)、差值植被指数(DVI)、土壤调节植被指数(SAVI)、盐度指数(SI1、SI2、SI-T)共10个指标构建土壤含盐量反演模型。结果表明:在2种算法中,BP神经网络模型预测精度最高,R2为0.818,RMSE为0.194;发现利用植被指数更能提高模型的预测精度;利用BP神经网络预测模型反演研究区的土壤含盐量,发现预测情况与研究区实际情况相符,说明利用GF-1数据结合BP神经网络构建的反演模型适用于监测研究区土壤盐渍化问题。
[Abstract]:In order to study the applicability of domestic GF-1 satellite images in soil salinization monitoring in arid areas, two modeling algorithms, BP neural network and RBF neural network, were used to study the desert ecotone around Wei-ku oasis. Based on the reflectivity of four bands of GF-1 image and the normalized difference vegetation index (NDVI) extracted from the image, the difference vegetation index (DVI), the soil regulation vegetation index (Savi) and the salinity index (SI 1 / SI 2 / SI-T), a soil salt content inversion model was constructed. In the two algorithms, the prediction accuracy of BP neural network model is the highest (R ~ 2 = 0.818) and RMSE is 0.194. It is found that vegetation index can improve the prediction accuracy of the model. The prediction model of BP neural network is used to invert the soil salt content in the study area, and it is found that the prediction is consistent with the actual situation of the study area, which indicates that the inversion model based on GF-1 data combined with BP neural network is suitable for monitoring soil salinization in the study area.
【作者单位】：新疆大学资源与环境科学学院;绿洲生态教育部重点实验室;
【基金】：国家自然科学基金项目(U1303381,4126090,41161063) 教育部长江学者计划创新团队计划(IRT1180) 自治区科技支疆项目(201504051064) 自治区重点实验室专项基金(2014KL005) 高分辨率对地观测重大专项(民用部分)(95-Y40B02-9001-13/15-03-01)
【分类号】：S156.41;S127

【相似文献】