Deep neural network based Rider-Cuckoo Search Algorithm for plant disease detection

• Published in: The Artificial intelligence review Vol. 53; no. 7; pp. 4993 - 5018
• Main Authors: Cristin, R., Kumar, B. Santhosh, Priya, C., Karthick, K.
• Format: Journal Article
• Language: English
• Published: Dordrecht Springer Netherlands 01.10.2020; Springer; Springer Nature B.V
• Subjects: Agriculture; Algorithms; Artificial Intelligence; Artificial neural networks; Belief networks; Clustering; Computer Science; Feature extraction; Histograms; Image processing; Image processing systems; Image segmentation; Mathematical optimization; Medical imaging; Medical imaging equipment; Medical research; Medicine, Experimental; Neural networks; Optimization; Plant diseases; Search algorithms; Texture; Piecewise fuzzy C-means clustering; Deep belief network; Texture features; Plant disease; Classification
• ISSN: 0269-2821; 1573-7462
• DOI: 10.1007/s10462-020-09813-w

Agriculture is the main source of wealth, and its contribution is essential to humans. However, several obstacles faced by the farmers are due to different kinds of plant diseases. The determination and anticipation of plant diseases are the major concerns and should be considered for maximizing productivity. This paper proposes an effective image processing method for plant disease identification. In this research, the input image is subjected to the pre-processing phase for removing the noise and artifacts present in the image. After obtaining the pre-processed image, it is subjected to the segmentation phase for obtaining the segments using piecewise fuzzy C-means clustering (piFCM). Each segment undergoes a feature extraction phase in which the texture features are extracted, which involves information gain, histogram of oriented gradients (HOG), and entropy. The obtained texture features are subjected to the classification phase, which uses the deep belief network (DBN). Here, the proposed Rider-CSA is employed for training the DBN. The proposed Rider-CSA is designed by integrating the rider optimization algorithm (ROA) and Cuckoo Search (CS). The experimental results proved that the proposed Rider-CSA-DBN outperformed other existing methods with maximal accuracy of 0.877, sensitivity of 0.862, and the specificity of 0.877, respectively.