This research aims to identify rice diseases, namely Leaf blast, Brown spot, Healthy and Hispa. The purpose of this research is to utilize deep convolutional neural network (DCNN) with support vector machine (SVM), DCNN with artificial neural network (ANN) and DCNN with long short-term memory (LSTM). To enhance the performance of LSTM further, the research includes particle swarm optimization, artificial fish swarm optimization (AFSO) and efficient artificial fish swarm optimization (EAFSO) to identify optimal weights. This research also compares the proposed technique results with a conventional feature extraction approaches like texture, discrete wavelet transforms and color histogram with SVM, ANN and LSTM. The results exhibit the superiority of proposed DCNN-LSTM (EAFSO) technique over other techniques. The proposed technique EAFSO associates DCNN-LSTM identifies the rice diseases with 97.5% accuracy, which is better than DCNN-SVM and DCNN-ANN.

本研究旨在鉴定水稻病，即叶瘟，褐斑病，健康病和Hispa。这项研究的目的是利用带支持向量机（SVM）的深度卷积神经网络（DCNN），带人工神经网络的DCNN（ANN）和具有长短期记忆的DCNN（LSTM）。为了进一步提高LSTM的性能，该研究包括粒子群优化，人工鱼群优化（AFSO）和有效的人工鱼群优化（EAFSO）来确定最佳权重。这项研究还将提议的技术结果与传统的特征提取方法（如纹理，离散小波变换和具有SVM，ANN和LSTM的颜色直方图）进行了比较。结果表明，提出的DCNN-LSTM（EAFSO）技术优于其他技术。