农业机械学报2018，Vol.49Issue(3)：196-204,9.DOI:10.6041/j.issn.1000-1298.2018.03.024
基于TOPSIS和BP神经网络的高标准农田综合识别
Multi-characteristic Comprehensive Recognition of Weil-facilitied Farmland Based on TOPSIS and BP Neural Network
摘要
Abstract
China puts forward the major strategic deployment of constructing well-facilitied farmland vigorously to improve the overall production capacity of farmland and adapt to the development of agricultural modernization.The recognition of well-facilitied farmland is foundation of site selection before constructing and evaluation after constructing.The well-facilitied farmland was understood from the point of view of production demand and recognized based on the evaluation of farmland comprehensive quality.Firstly,the characteristics of farmland comprehensive quality was analyzed from a lot of angles,such.as background condition,spatial shape,construction level,ecological protection and so on,by fusing the multi-source data and taking the farmland patches as the basic units.The description system of farmland comprehensive quality was built by using five characteristics,including soil productivity,land contiguous,field shape,road accessibility and ecological protection.Secondly,it assumed that these five characteristics were the same important for farmland comprehensive quality,so the weights were all made as 0.20 and the preliminary evaluation results were got by TOPSIS method.Thirdly,the true-value samples were acquired by using the combined method of preliminary evaluation results and man-machine interactive optimization.The man-machine interactive optimization was achieved by spatial overlay between the preliminary evaluation results and the farmland utilization grade from the farmland-grading work in China.And then BP neural network was used to fix the feature weights.Fourthly,the final accurate comprehensive quality evaluation results were got and the recognition of the well-facilitied farmland was achieved.Finally,Daan City in Jilin Province was taken as the study area.The research results showed that the accuracy of the method to evaluate farmland comprehensive quality was above 96％,basing on the multi-characteristic description system.The well-facilitied farmland was widely distributed in the study area.The well-facilitied farmland mainly concentrated in northeast,north,edge of northwest and part of the southern region.These regions had the advantage of agricultural modernization,such as concentrated farmland,villages,roads and forest.The well-facilitied farmland which was registered with the law and the prospective high-quality farmland which was not registered with the law were both recognized effectively.The above result had strong consistency on the spatial distribution with the preliminary evaluation results,but the former refined the comprehensive quality results of partial farmland based on the relative importance of each characteristic.The research result can provide scientific reference and technical support for regulation,protection and construction of wellfacilitied farmland.关键词
高标准农田/综合识别/多源数据/TOPSIS/BP神经网络Key words
well-facilitied farmland/comprehensive recognition/multi-source data/TOPSIS/BP neural network分类
管理科学引用本文复制引用
吕雅慧,郧文聚,张超,朱德海,杨建宇,陈英义..基于TOPSIS和BP神经网络的高标准农田综合识别[J].农业机械学报,2018,49(3):196-204,9.基金项目
国家高技术研究发展计划(863计划)项目(2013AA10230103) （863计划）
