Identification of rice diseases using deep convolutional neural networks

Abstract

The automatic identification and diagnosis of rice diseases are highly desired in the field of agricultural information. Deep learning is a hot research topic in pattern recognition and machine learning at present, it can effectively solve these problems in vegetable pathology. In this study, we propose a novel rice diseases identification method based on deep convolutional neural networks (CNNs) techniques. Using a dataset of 500 natural images of diseased and healthy rice leaves and stems captured from rice experimental field, CNNs are trained to identify 10 common rice diseases. Under the 10-fold cross-validation strategy, the proposed CNNs-based model achieves an accuracy of 95.48%. This accuracy is much higher than conventional machine learning model. The simulation results for the identification of rice diseases show the feasibility and effectiveness of the proposed method.

Introduction

Rice is one of the most important food crops in China and even in the world. Rice diseases have a devastating effect on rice production. Also, it is a major threat to food security. Thus, the diagnosis and identification of rice diseases play a very important role in ensuring the high yield, high quality and high efficiency of rice. The traditional method of detecting rice diseases requires lots of experts’ experience and knowledge. With the development of computer and internet technology, farmers can search the rice diseases images database or consult the plant pathologists to judge rice diseases remotely. The disadvantage is that not only the judgement is easy to be wrong, but also the efficiency is low.

In order to improve the accuracy and rapidity of the diagnosis results, many researchers have studied the automated rice diseases diagnosis based on pattern recognition and machine learning. Such as, using pattern recognition techniques [25], support vector machine [16], digital image processing techniques [2] and computer vision[1]. Meanwhile, these advanced techniques are not only applied to the diagnosis of rice diseases, but also to other crops, such as wheat [20], maize [34], cotton [29], tomato [5], etc.

The past 10 years have witnessed successful applications of deep convolutional neural networks (CNNs) in diverse fields including image classification [6], [21], video classification [18], traffic sign recognition [17] and human action recognition [15], etc. The study of CNNs and related research have therefore gained persistent research interest since the early 1990s, see [10], [12] and the references therein. During this period, Hinton and co-workers have done widely meaningful and fundamental research on deep neural network to improve algorithm performance and optimize architecture, see e.g. [3], [23], [24], [27], and the references cited. A number of research have been made for CNNs to improve the original architecture of Krizhevsky et al. [21]. For example, Zeiler and Fergus [33] used stochastic pooling for regulation of CNNs and Simonyan and Zisserman [31] proposed very deep CNNs to 16–19 layers, which achieved the state-of-art accuracy on ImageNet Large-Scale Visual Recognition Challenge (ILSVRC) [26]. Especially, in March 2016, AlphaGo beat Lee Sedol in a five-game match, the first time a computer Go program has beaten a 9-dan professional, and it was a major milestone in artificial intelligence research [30], [32]. LeCun et al. pointed out that deep convolutional neural networks have brought about breakthroughs in processing images, video, speech and audio in a survey paper published in Nature in 2015 [22].

Recently, several studies on automated plant disease diagnosis have been conducted using deep learning techniques [19], [35]. Kawasaki et al. proposed a system based on CNNs to recognize cucumber leaf disease; it achieves an accuracy of 94.9% in distinguishing among melon yellow spot virus, zucchini yellow mosaic virus and a non-diseased class. Sanyal et al. have identified rice disease of brown spot and blast diseases by using color texture of rice leaf images [28].

As we all know that the machine learning technology is essential in the intelligent diagnosis of crop diseases. However, there exist some defects and difficulties in the above research, involving the recognition rate is higher for specific samples under certain circumstances. In the diagnosis model, some parameters are not optimal, the convergence speed of training algorithm is slow, and it is easy to fall into local minima, and so on.

So far, no research has been published which explores deep convolutional neural networks for rice diseases identification. The goal of this research is to construct deep convolutional networks model to achieve fast and accurate automated recognition by using rice diseases images. The 10 common rice diseases include rice blast (RB), rice false smut (RFS), rice brown spot (RBS), rice bakanae disease (RBD), rice sheath blight (RSHB), rice sheath rot (RSR), rice bacterial leaf blight (RBLB), rice bacterial sheath rot (RBSR), rice seeding blight (RSEB) and rice bacterial wilt (RBW).

The key motivation for developing the deep convolutional networks model for rice disease is to provide the farmers an easy-to-use system to detect early-stage infections by using common digital camera. Second, extracting effective features for identifying rice diseases is a critical but challenging task, and CNNs are highly expected to be automated feature learning from the raw inputs in a systematic way. Through the deep architecture, the learned features are deemed as the higher level abstract representation of low level raw rice disease images. In addition, to improve diagnostic results, the CNNs is regarded as one of the best classifications in pattern recognition tasks. So we develop the deep convolutional neural networks model for rice diseases identification in this paper.

In this paper, we present a novel rice diseases identification method based on deep convolutional neural networks. The proposed CNNs-based model can get higher classification ratio. On the basis of an analysis of structure and parameters in CNNs, the gradient-descent algorithm can be applied to train CNNs. A total of 500 rice leaves and stem images are pre-processed first, then the processed images are used to train CNNs. This CNNs model can improve the convergence speed while training the parameters in CNNs, and obtain a higher recognition accuracy than the conventional model. The main contribution of this paper is mainly twofold. (1) Convolutional neural network is first applied to the issue of the identification of rice diseases. Note that the proposed method can correctly and effectively recognize the 10 common rice diseases. (2) Experiment results show that the CNNs method can not only improve the convergence speed, but also obtain a higher recognition accuracy than other models: (1) standard BP algorithm; (2) support vector machine (SVM); (3) particle swarm optimization (PSO).

The advantages of the proposed CNNs are that images can be input directly to the model. First, sparse-auto encoding is used to learn the features from images. Second, we can classify images from a reduced data set applying convolution and pooling. The stochastic pooling method is applied in the proposed CNNs model. For it is simple enough to randomly select elements in feature map according to their probability values, that is, the elements with large probability are easy to be chosen. Unlike max-pooling, the maximum element always chosen, and mean-pooling, the average value of the elements is chosen. Also, it strengthened generalization ability of the proposed CNNs model. Third, softmax regression learning algorithm can solve multi-classification problem. Finally, we can distinguish between 10 common different rice diseases.

The remainder of this paper is organized as follows. In Section 2, we describe the architecture and learning algorithm of CNNs. Section 3 introduces the rice disease identification method. The application of the developed CNNs to the problem of rice diseases recognition is presented in Section 4. Conclusions are given in Section 5.