基于高光谱数据的土壤含盐量BP神经网络模型研究

运用Hyperion数据,以黑龙江省大庆市某一实验区为例,进行了对土壤含盐量的定量研究,将BP神经网络模型(Back Propagation Network)应用到高光谱数据对研究地区土壤含盐量的反演中。通过对隐含层的传递函数、输出层的传递函数、训练算法的优化组合以及最适合隐层节点数量,得到最优的BP神经网络模型,实现了土壤含盐量的反演。对高光谱数据反演土壤含盐量采用BP神经网络具有一定的指导意义。     

Bistatic scatterometer monitoring of soil moisture by microwave remote sensing at X-band

Soil moisture is important for irrigation planning, as well as in forecast the risk of flash floods, or the occurrence of fog. Measurement of scattering coefficient σ0 of a bare soil was performed by changing the soil moisture content by using the X-band (9.5 GHz) scatter meter. These experiments were conducted over a range of incidence angle from 20° to 70° at step of 5° for both the vertical–vertical (VV) and horizontal–horizontal (HH) polarization. The emissivity is observed by measuring the reflectivity from the microwave system. Results show a good angular variation of emissitivity in both the polarization with soil moisture. An idea was developed by using incidence angles as a modulating factor for retrieving the soil moisture at X-band.     

Modeling of tomato drying using artificial neural network

This study involves experimental works on drying of tomatoes in a tray dryer covering different variables like power of heater and air flow velocity. The data are modeled using artificial neural network and empirical mathematical equations. The results were compared with experimental data and it was found that the predictions of the artificial neural network model fit the experimental data more accurately in comparison to the various mathematical equations.     

人工神经网络用于异噻唑啉酮的定量结构-活性关系研究

采用误差反传前向人工神经网络(artificial neural network,ANN)建立了21种2-(4-取代-苯基)-3-异噻唑啉酮类化合物的结构与其抗菌活性之间的定量关系模型(ANN模型),以21种3-异噻唑啉酮类化合物的量子化学参数和拓扑指数作为输入、抗菌活性作为输出,所构建网络模型的交叉检验相关系数为0.991 6、标准偏差为0.080 1、残差绝对值≤0.221,应用于外部预测集,预测集相关系数为0.973 1;而多元线性回归(multiple linearregression,MLR)法模型的相关系数为0.841 8、标准偏差为0.303 9、残差绝对值≤0.636。结果表明,ANN模型获得了比MLR模型更好的拟合效果。     

Predictions of apple bruise volume using artificial neural network

Bruise damage is a major cause of fruit quality loss. Bruises occur under dynamic and static loading when stress induced in the fruit exceeds the failure stress of the fruit tissue. In this article the potential of an artificial neural network (ANN) technique has evaluated as an alternative method for the prediction of apple bruise volume. Neural bruise estimation models were constructed to calculate Golden Delicious apple bruise volume with respect to fruit properties. The neural models were built based upon impact force and impact energy as the main input parameters including fruit curvature radius, temperature and acoustical stiffness. Optimal parameters for the network were selected via a trial and error procedure on the available data. In this paper, the performance of Basic Backpropagation (BB) training algorithm was also compared with Backpropagation with Declining Learning Rate Factor algorithm (BDLRF). It was found that BDLRF has a better performance for the prediction of apple bruise volume. It is concluded that ANN represents a promising tool for predicting apple bruise volume in comparison to regression model.     

Artificial neural network for the estimation of soil moisture and surface roughness

Estimation of the soil moisture and soil roughness by using microwave data with less complex and fast method is a significant area of research today. For this purpose an Artificial Neural Network (ANN) based algorithm is used and tested in present study. The ANN model is calibrated and tested with the experimentally obtained data by using X-band scatterometer for different field roughness 3.78, 1.83 and 1.63 cm and at fixed value of soil moisture 22.8%. The measurement of scattering coefficient was carried out over a range of incidence angle from 20° to 70° by 5° steps for both the HH (horizontal transmitter and horizontal receiver) and VV (vertical transmitter and vertical receiver) polarization. Two training algorithm of Feed Forward Backpropagation neural network namely Levenberg-Marquardt (TRAINLM) and Gradient-Descent (TRAINGD) were used for analysis. The performance of the ANN models with different algorithm is evaluated by comparing the direct measured value of soil roughness and soil moisture with the soil roughness and soil moisture estimated by the ANN. Our work suggests that ANN model with training algorithm (TRAINLM) is more suitable for the soil moisture and surface roughness prediction in comparison to (TRAINGD) and ANN modeling may be the promising alternative for the soil moisture and surface roughness estimation. The main advantage of the ANN approach for the surface roughness and soil moisture estimation is its potential for world wide reporting.     

神经网络专家系统开发工具

研制了一个建造神经网络专家系统的开发工具.实现了建造神经网络专家系统的知识获取、并行推理机制、知识库及人机界面等主要功能模块.提高了神经网络专家系统的建造速度,使其更加实用化.     

基于人工神经网络的林分直径分布预测

以马尾松人工林为研究对象,用人工神经网络建模技术构建了林分直径分布预测模型。经训练和优选,得到的理想模型结构为3∶6∶6∶1,训练误差指标为0.000281,总体拟合准确度为98%。模型对82块训练标准地的累积频率拟合准确度最大为100%,最小为95%,平均为98%;频率拟合准确度最大为96%,最小为75%,平均为87%。模型对18块检验标准地的累积频率预测准确度最大为99%,最小为97%,平均为98%;频率预测准确度最大为96%,最小为76%,平均为88%。所建模型具有很好的拟合效果和很强的预测能力,可用于10～30年生马尾松人工林。研究结果证明,人工神经网络技术可以作为有效的林分直径分布预测技术。     

基于HJ卫星数据的土壤含水量反演及其旱情预测

干旱是影响中国农业生产的重要自然灾害之一。为确定温度植被干旱指数(TVDI)法在苏北地区干旱监测中的适用性,本试验构建基于HJ卫星数据的NDVI-Ts特征空间,提取TVDI并结合实测数据将遥感指数转化为土壤相对湿度。结果显示:特征空间构建时,考虑NDVI0.2区间可提高干、湿边的拟合精度;TVDI与各层土壤含水量均有一定的相关性,其中与10~20 cm土层土壤含水量的相关系数达-0.649~-0.854(P0.01)。结合同期降水数据,可认为基于HJ卫星数据的TVDI法对苏北地区旱情具有较好的监测效果。     

人工神经网络在遥感图像森林植被分类中的应用

应用人工神经网络模型对陆地卫星TM多光谱图像进行了森林植被分类的研究 ,共选取了 8种主要植被类型 ,重点是研究在不同背景条件下存在同谱异物现象的云杉、油松和落叶松等针叶林树种的分类方法 .所采用的网络模型为 3层误差后向传播神经网络模型 ,鉴于贺兰山自然植被垂直带谱明显 ,利用误差后向传播网络模型的并行分布式结构 ,研究中引入高程数据作为一个独立波段与 3个多光谱波段一起直接进行分类 ,取得了很好效果 .该方法与常规的最大似然法相比 ,存在同谱异物现象的云杉、油松和落叶松的分类精度平均提高了 2 7 5个百分点 .对存在同物异谱现象的阔叶林的分类精度也有一定程度的提高 .     

基于神经网络的宏观农业生产预测模型的研究

为探索宏观农业生产系统预测的新方法，构建了基于人工神经网络的预测模型，利用1994-2003年的气象、经济、生产、投入、技术、价格各方面的数据对我国粮食生产进行了拟合分析，并预测了2004年粮食总产，预测的结果为46125．46万t。结果表明，与灰色系统相比，本文建立的模型具有90％以上的拟合精度，模型具有容错能力、联想能力和学习能力．可以用来尝试解决农业生产系统预测问题。     

基于人工神经网络的粮食产量预测模型

研究了人工神经网络在经济预测中的应用问题,探讨利用人工神经网络进行农业粮食产量预测的方法。提出一种基于多层前馈BP神经网络的农业粮食产量预测模型,可以得到影响粮食产量的主要因子和粮食产量之间的非线性映射关系。并通过实例验证了神经网络模型的预测精度明显高于线性回归模型的预测精度。     

Retrieval of data sotred in a neutron moisture meter using a portable microcomputer

A scheme is described to retrieve data accummulated in the solid-state memory of a neutron moisture meter and place them in interim storage on a portable microcomputer ready for transmission to a central computer. The criteria for choosing the microcomputer are discussed in relation to the data transmission characteristics of the microprocessor of a particular neutron meter and a VAX 11/750 central computer. A program CPNDUMP, written in Epson Microsoft BASIC, is described that ensures error-free transmission of data from the neutron meter to an Epson HX-20 microcomputer in the field, and during its transfer to magnetic tape. This program and the HX-20 have proved robust and reliable over three seasons of field use. The structure of CPNDUMP and the considerations governing its design are of general utility in establishing effective communications between many similar instrument-microprocessors and microcomputers.     

基于BP神经网络的太湖典型农田土壤水分动态模拟

收集太湖典型农田2010年10—12月和2011年3—6月2个时间段的逐日气象资料和土壤水分资料,运用BP(back propagation)神经网络和缺省因子分析法确定影响该地区土壤水分动态的主要气象因子(降水量、蒸发量、平均气温和平均地表温度以及平均风速),以这些主要影响因子作为输入变量建立该地区土壤水分动态模拟的BP神经网络模型。利用100组实测样本对神经网络进行训练,用剩余的64组实测样本进行检验。结果表明:0～14 cm和14～33 cm土壤含水量模拟的平均相对误差(MARE)最大为0.062 9,均方根误差(RMSE)最大为1.764,不同土壤层次的训练样本和检验样本的精度(PA)都在0.87以上。因此,BP神经网络用于太湖典型农田的土壤水分动态模拟是可行的。     

人工神经网络对果蝇鸣声的分类识别

为了利用昆虫鸣声对昆虫进行种间或种下分类,对实验室环境下同种2个不同品系黑腹果蝇的飞行翅振鸣声进行了采集、分析,提取鸣声信号特征参数,并利用人工神经网络对采集的果蝇鸣声信号进行分类识别。结果表明,2个品系果蝇鸣声的基频均为236.86 Hz,有多个谐频,频率范围为0~4000 Hz,重叠较大;所建立的人工神经网络对种内不同品系果蝇鸣声的正确识别率均在75%以上,识别效果很好。研究结果为果蝇种下分类提供了新的方法和依据。     

基于人工神经网络的农用地分等研究 Ⅱ.前置数据优化处理

对基于BP 人工神经网络的农用地分等中出现的原始数据标准化和训练样本的类型典型性问题,提出用数据分类标准化和模糊系统聚类分析建立训练样本集的方法解决.实证研究表明,这两种方法结合应用能充分发挥BP ANN分类器的功效,并获得高精度的结果.     

基于BP神经网络的土壤养分空间插值

以广东省增城市为研究对象,采集全市内200个土壤样点,利用BP神经网络插值方法对研究区土壤的氮和磷进行空间插值预测,将插值结果与土壤样点实测值进行对比,得到预测数据的误差均方根.结果表明,BP神经网络的插值精度比克里格高,在样点较少的情况下,BP神经网络的插值结果克服了克里格插值方法的平滑效应.BP神经网络对插值的样本数据的分布类型没有要求,比传统插值方法有更强的泛化能力,是一种可替代的插值方法.     

西北太平洋柔鱼BP神经网络渔场预报模型比较研究

柔鱼是西北太平洋的重要经济种类。研究根据1995-2001年7-11月采集的鱿钓生产数据以及相对应的海洋环境因子数据,包括经纬度、表温(SST)和海平面高度距平(SSHA),分别以单位捕捞努力量渔获量(CPUE)和捕捞努力量作为中心渔场指标,采用BP神经网络方法,以经纬度、海洋环境因子作为输入因子,分别以CPUE和捕捞努力量作为输出因子,采用4-3-1和4-2-1两种模型,共4种方案对西北太平洋柔鱼渔场进行预报,并以拟合残差最小的模型作为最优预报模型。分析结果显示,7-11月各月中心渔场预报模型均以4-3-1模型为最优,但7、8月最优预报模型以捕捞努力量为输出的4-3-1模型,9、10、11月最优预报模型以CPUE为输出的4-3-1模型,总体平均误差以捕捞努力量为输出的4-3-1模型为最小。研究认为,CPUE和捕捞努力量作为中心渔场预报指标有差异,以捕捞努力量为输出的4-3-1模型较合适作为柔鱼渔场预报模型。     

Classification of two different rough soil surfaces by using microwave X-band data through support vector machine (SVM)

Support vector machine (SVM) model is employed and tested for the soil surface roughness classification. SVM is calibrated (trained) and tested with the experimentally obtained data. The experimentally data is obtained by using X-band (9.5 GHz) scatterometer for two soil surface roughness 3.78 cm and 1.83 cm at constant soil surface moisture equal to 22.80%. The measurement of the scattering coefficient was carried out over a range of incidence angle from 20° to 70° at the step of 5° for both the HH and vv polarization. The performance of the SVM model is evaluated from the outcome classification result on trained data set and test data set. Radial Gaussian kernel function results 100% correct Classification and identification of soil surface roughness both in training and validation phase. SVM is a proficient technique for soil surface roughness classification by such experimentation and have numerous of advantages over artificial neural network (ANN) based approaches and other theoretical approaches as its less complexity and less time consumption ability.     

Predictions of viscoelastic behavior of pomegranate using artificial neural network and Maxwell model

Stress relaxation is one of the defined tests to characterize the viscoelastic properties of food and agricultural materials. Stress relaxation data are very important because they provide useful and valuable information such as fruit firmness and ripening, food processing and predicting changes in the material during mechanical loading. Viscoelastic behavior of some varieties of pomegranate that are cultivated in Iran has been studied in current research. For this purpose, stress relaxation test was conducted with three cultivars of pomegranate (Ardestani, Shishekap and Malas) for three sizes (small, medium and large). In this article the potential of artificial neural network (ANN) technique is evaluated as an alternative method for Maxwell model to predict the viscoelastic behavior of pomegranate. Neural stress relaxation models were constructed to describe stress relaxation behavior of pomegranate with respect to time. The neural models were built based upon relaxation time as input network and stress relaxation as output network. The results revealed that both ANN model and Maxwell model have high capability of producing accurate and reliable predictions for stress.