The Digital Agricultural Revolution. Группа авторов
Читать онлайн книгу.
target="_blank" rel="nofollow" href="#ulink_9f6f29eb-f130-54dd-bf72-283073e43f3f">Figure 2.1 Architecture of artificial neural network (original figure).
2.2.3 Types and Suitability of Neural Networks
The artificial neural networks are usually selected based on the mathematical functions and output parameters. Among the different types of artificial neural networks, some of the most important kinds of the neural networks are discussed in this section.
A feed-forward neural network (FFNN) is an artificial neural network and is one of the simple type of neural network. In which, the input data travels in only one direction no loop or cycle formation. In an FFNN, every neuron (perceptron) in one layer is connected with each node in the immediate layer. As a result, each and every node is fully connected. This systematic arrangement of FFNN generates output by output layer. The number of hidden layers may arrange in between input and output layers and do not have a connection with the outer environment. These neural networks may or may not have a hidden layer. Common applications are pattern recognition, speech recognition, data compression, computer vision, and so on. If an FFN network uses more than one hidden layer, it is called a deep feed-forward network. By adding more hidden layers, overfitting will be reduced and improved generalization. The synaptic operation order in a hidden neuron, the ANNs were classified as first order, second, third, or higher order [14]. The back loops are absent in the FFN network. To reduce the error value in prediction, the back propagation algorithm may be used. The weights between the input hidden and output layers can be adjusted by using back propagation algorithm through learning rate and momentum. Then, the error value is back propagated from output layer to the input layer [15]. Back propagation algorithm was adopting widely for forecasting problems with the networks [16, 17]. Radial Basis Network (RBN)s behave as feed forward networks using other functions (radial basis function) to activate the network. Radial Basis Networks determine the gap between the generated output and target output. The logistic (sigmoid) function produces output value between 0 and 1. An RBN cannot be used for continuous value. Radial basis function considers the distance from the center to a point. The main advantage of RB neural networks is universal approximation and faster learning rate.
Recurrent neural network (RNN) uses previous information in current iterations. The principle of the RNN is keeping output and feeding the output back to input layer to help in estimating the output of the layer. In this type of neural network, hidden layers every neuron receives an input with a specific time delay. Recurrent neural network is required more time with low computational speed. These types can be used in time series anomaly detection, speech recognition, speech synthesis, and robot control. Long-/ short-term memory networks (LSTM) use a memory cell that processes data in RNNs. Gate recurrent units (GRUs) are different from LSTMs with similar models and produce equally better results. Extreme learning machines (ELMs) determine the output weights by choosing hidden nodes randomly. Extreme learning machine networks learn the output weights in only one step, and assigned weights are never updated. These algorithms work faster than the other general NN algorithms.
Convolution neural networks (CNN) are primarily used for image classification, image clustering, time series forecasting and object recognition. Deconvolutional networks (DCN) are CNNs that work in anopposite process. The major drawbacks of conventional neural networks are low learning rate and all the parameters tuning iteratively. In Hopfield network (HN), every neuron in network is directly connected with other neurons. The HNs are used to save memories and patterns and also applied in optimization problems. A Kohonen neural network (KN) also known as self-organizing maps is an unsupervised algorithm. This is very useful for multidimensional scattered data. It gives output in one or two dimensions only, so it is treated as a method of dimensionality reduction. This self-organizing process has different phases. A small weight is initialized to each neuron. In the second phase, the close neuron is the “winning neuron,” and the other neurons are connected to the winning neuron. The KN networks use more competitive learning than error correction learning.
2.3 Application of Neural Networks in Agriculture
2.3.1 Potential Applications of Neural Networks in Agriculture
Various ANN techniques are suitable to use in agriculture. The widely accepted neural network includes radial basisfunction neural network (RBFNN), backpropagation neural network (BNN), convolutional neural network (CNN), recurrent neural network (RNN), and Adaptive Neuro-Fuzzy Inference System (ANFIS) for different applications. These NNs are also used in combination with remote sensing classification, rice mapping [17, 18]. Rice mapping is done by combining MS Sentinel 2 and Sentinel 1 SAR band using the LSTM network and dynamic time warping distance approach [19]. Detailed crop classification and mapping is the area for high scope of NN application [20, 21]. A ANN-based model combining a CNN and generative adversarial network (GAN) is used for crop classification with advantages of few data requirement and high flexibility [22]. A CNN-based approach with fused MODIS NDVI data, multitemporal Landsat, and synthetic phenological variable classifies a land accurately into small patches [23]. The methodology for crop classification using deep learning models was framed [24]. They proposed CNN model from multiple Landsat and Sentinel-1 images. Cropland was classified by a LSTMRNN model using Landsat imagery with high accuracy [25]. Wheat production was forecasted and compared the performance of different training methods using an ANN model [26].
Plant identification is done using deep learning neural networks and showed good performance [27, 28]. Deep learning algorithm is used for mango fruit identification and fruit classification [29]. Support vector regression (SVR) and genetic algorithm (GA)–based BPNN were combined to develop a spectral model for evaluating cultivated land quality based on MODIS-GPPs on late rice phenology [30]. SVM and ANN combined to predict crop with accuracy of 86.80% in India adopting the factors like temperature, rainfall, relative humidity, soil, and so on [31]. Five remote sensing (RS) data-derived parameters like VI, slope, vegetation dryness index, temperature, patch-fractal dimension, and road accessibility were used, and they found that cultivated land quality was significantly and nonlinearly correlated [32].
Machine learning methods were also used in pest classification of crops [33, 34]. The CNNs were used to identify the plant pests from leafs [35, 36], weed classification [37], crop quality evaluation [38], and field pest classification. Many agricultural-related research surveys were done to provide comprehensive results [39]. An ANN model used was for soil moisture estimation using temperature of soil, temperature of atmosphere, and RH. Artificial neural networks were adopted to predict the production of biofuel from cow manure and agricultural wastes at high accuracy [40]. The study helped in identifying favorable conditions to predict the behavior of biofuel production in a short time [41].
2.3.2 Significance of Neural Networks in Crop Yield Prediction
The factors that responsible for crop yields, like soil type, climate variables, water application, and crop management, are nonlinear and complex. Traditional statistic applications lead inaccurate results in yield estimation. There are several studies reported about crop yield estimation using neural networks in response to climate, soil, genotypes, and crop management practices. A design to train an ANN to forecast the soybean production demand in Brazil was framed by adopting a nonlinear autoregressive solution [42]. They concluded an increase of about 26.5% for 2017 compared with 2016. An ANN is used to predict soybean yield and production and to compare with time series analysis [43]. A crop yield model corresponding to soil-related parameters was developed [44] by training a BP neural network. Different vegetation indices and plant density were analyzed from UAV to analyze grain yield of corn crop by using an NN model [45].
Maize yield estimated using time series data of different satellites and also radar with a neural network with an R2 of 0.69 [46]. Chlorophyll data were used to estimate productivity of corn crop [47] resulting in an r2 of 0.73. Vegetation indices and crop height were correlated with maize yield to predict maize yield using neural networks [48, 49]. Corn grain yield was calculated with RS-based plant density, canopy cover, and VI using neural networks [50]. Sugarcane yield was estimated using feed forward and back propagation