引用本文: | 王梅凯,赵成萍,李博,等.基于ATI-BP的土壤墒情反演研究[J].灌溉排水学报,0,():-. |
| wangmeikai,zhaochengping,libo,et al.基于ATI-BP的土壤墒情反演研究[J].灌溉排水学报,0,():-. |
|
摘要: |
土壤墒情的监测对农作物生长、水资源的合理分配起着至关重要的作用。【目的】为提高遥感墒情的反演精度和通用性,【方法】本文利用MODIS(MOD021KM)数据建立一种基于修正表观热惯量的墒情反演模型,在此基础上,结合BP人工神经网络(BPNN)进行协同反演。首先针对单一表观热惯量反演土壤墒情的局限性问题,选取EVI和ATI共同作为墒情反演的评价指标,并利用实测土壤墒情作为验证指标,将原始遥感影像预处理后计算出评价指标,并实地测量25组10cm深度土壤墒情值;然后以评价指标作为输入层,以验证指标作为输出层,构建BPNN土壤墒情反演模型。【结果】本文运用该方法,以均方误差作为衡量标准,选取都江堰灌区作为研究对象有较好的反演效果,模型的评价指标与10cm深度的土壤墒情有较好的相关性,综合植被覆盖和表观热惯量的因素反演灌区土壤墒情,最终反演均方误差为0.0039,相较于线性、对数和幂函数,其相对误差分别减小了66.9%,81.1%和74.9%。【结论】以该模型进行土壤墒情反演其精度有了明显的提高。依赖遥感影像和部分实测数据,使得该方法在大面积墒情反演的研究上有较好的参考价值。 |
关键词: MODIS;BP神经网络;ATI;EVI;都江堰灌区;墒情反演 |
DOI: |
分类号:S |
基金项目:国家自然科学(项目批准号:U1933123) |
|
Soil Moisture Inversion Based on ATI-BP |
wangmeikai1, zhaochengping1, libo2, zhouxinzhi1
|
1.Institute of intelligent control, sichuan university;2.Joint laboratory of water conservancy informatization, sichuan university
|
Abstract: |
Soil moisture monitoring plays an important role in crop growth and rational distribution of water resources.【Objective】 to improve the accuracy and universality of remote sensing moisture inversion,【Method】in this paper, MODIS(MOD021KM) data was used to establish a moisture inversion model based on modified apparent thermal inertia. On this basis, BP artificial neural network (BPNN) was combined for collaborative inversion.Firstly, considering the limitation of soil moisture inversion with a single apparent thermal inertia, EVI and ATI were selected as the evaluation indexes of soil moisture inversion, and the measured soil moisture was used as the validation indexes. After the pretreatment of the original remote sensing images, the evaluation indexes were calculated, and 25 groups of soil moisture at 10cm depth were measured on the spot.Then, the BPNN soil moisture inversion model was constructed with the evaluation index as the input layer and the validation index as the output layer.【Results】 this paper USES the method of mean square error as the measure, selected as the research object in dujiangyan irrigation area has good inversion results, the model of evaluation index and 10 cm depth of soil moisture has good correlation, the comprehensive factors of vegetation coverage and apparent thermal inertia inversion of soil moisture in the irrigation area, the final inversion mean square error is 0.0039, compared with the linear and logarithmic and power function, and the relative errors were reduced by 66.9%, 81.1% and 74.9%.【Conclusion】the accuracy of soil moisture inversion with this model has been significantly improved.Depending on remote sensing image and partial measured data, this method has a good reference value in the study of inversion of large area moisture content. |
Key words: MODIS;BP Neural Network;ATI;EVI;Dujiangyan Irrigation Area;Soil Moisture Inversion |