人工神经网络在土壤质量监测中的应用

发布时间：2018-09-04 19:36

【摘要】：土壤质量是土壤特性的综合反映是土壤生产力的最直接反应,它是揭示土壤条件动态的最敏感的指标,能体现人类活动对土壤环境的影响。近年来由于人口的持续增长和经济的快速发展使人们对粮食的需求日益增加,加上土地资源退化和后备土地资源的稀缺,使得我国人地关系进一步紧张,土壤质量监测预测对于缓解人地关系促进土地资源的合理利用和实现农业的可持续发展具有重要意义。人工神经网络是为了进行信息处理而建立的一种数学模型,它之所以被称为神经网络是因为这一数学模型所应用的是一种类似大脑神经突触的联接结构。BP神经网络是人工神经网络中应用范围最广的一种算法,本文以江西千烟洲生态试验田为例,结合BP神经网络建立土壤质量评价和监测的模型,为土壤质量监测提供一个新的方法,期望为农田生产力的提高和土壤资源的合理利用提供一个可靠的科学依据,实现农业可持续发展。本文结合MATLAB工具箱对BP神经网络土壤质量监测模型主要进行了以下几点优化:首先是专家训练样本,本文基于全国第二次土壤普查土壤的养分分级标准结合研究区土壤养分特性建立专家样本,并依据一定算法对专家样本进行了百组数据扩充。基于已建的专家样本对不同等级划分再次细化分级建立新专家样本,实现了对研究区土壤质量的细化研究;其次是本文采用trainlm算法对样本数据训练,trainlm算法的收敛速度较快,同时还具有自学习性和数据动量调整的特点并在实际训练中得到验证,符合预定的误差训练目标要求;本文在激励函数的选择上结合土壤质量仿真结果评价等级范围不大于1的特点,选择采样连续可导的Sigmoid型函数,并对输出结果误差进行比较;最后在隐含层数和节点数的选择上,结合输入数据和输出结果特点,先确定出入层和输出层结点个数,在结合节点选取的几个常用公式分别将不同的节点数带入网络中试验,确定选取一个隐含层7个节点的三层BP网络结构,并在数据训练和仿真结果中表现出了较高的收敛速度和误差精度。仿真结果表明本文建立的BP网络模型在土壤质量监测和评价上与期望结果接近,网络误差精度较高,对土质的科学管理和合理利用具有一定的指导意义。
[Abstract]:Soil quality is the most direct response of soil productivity and the most sensitive index to reveal the dynamics of soil conditions. It can reflect the influence of human activities on soil environment. In recent years, due to the continuous growth of population and the rapid development of economy, the increasing demand for food, coupled with the degradation of land resources and the scarcity of reserve land resources, have further strained the relations between people and land in our country. Soil quality monitoring and forecasting is of great significance to alleviate the relationship between man and land to promote the rational use of land resources and to realize the sustainable development of agriculture. Artificial neural network is a mathematical model for information processing. It is called a neural network because the mathematical model uses a connective structure similar to the synapse of the brain. BP neural network is the most widely used algorithm in artificial neural networks. Taking Qianyanzhou ecological experimental field in Jiangxi province as an example, a model of soil quality evaluation and monitoring is established by combining with BP neural network, which provides a new method for soil quality monitoring. It is expected to provide a reliable scientific basis for the improvement of farmland productivity and the rational utilization of soil resources to realize the sustainable development of agriculture. Based on MATLAB toolbox, the soil quality monitoring model based on BP neural network is optimized as follows: first, expert training sample. Based on the soil nutrient classification standards of the second national soil census and the characteristics of soil nutrients in the study area, an expert sample was established, and a hundred groups of expert samples were expanded according to a certain algorithm. Based on the established expert samples, a new expert sample is established to refine the soil quality in the study area. Secondly, the trainlm algorithm is used to train the sample data and the convergence speed of the trainlm algorithm is fast. At the same time, it has the characteristics of self-learning and data momentum adjustment, and is verified in actual training, which meets the requirements of the target of error training. In this paper, according to the characteristics of soil quality simulation results evaluation grade is not greater than 1, the Sigmoid type function of sampling continuous derivation is selected, and the error of output result is compared. Finally, in the selection of hidden layer number and node number, combined with the characteristics of input data and output results, the number of nodes in and out layer is determined first. A three-layer BP network structure with seven nodes in the hidden layer is selected, and the convergence speed and error accuracy are high in the data training and simulation results. The simulation results show that the BP neural network model established in this paper is close to the expected results in soil quality monitoring and evaluation, and the accuracy of network error is high, which has certain guiding significance for the scientific management and rational utilization of soil quality.
【学位授予单位】：浙江海洋学院
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：S158;TP183

【相似文献】