Mango leaf disease identification and classification using a CNN architecture optimized by crossover-based levy flight distribution algorithm

• Published in: Neural computing & applications Vol. 34; no. 9; pp. 7311 - 7324
• Main Authors: Prabu, M., Chelliah, Balika J.
• Format: Journal Article
• Language: English
• Published: London Springer London 01.05.2022; Springer Nature B.V
• Subjects: Algorithms; Artificial Intelligence; Artificial neural networks; CAI; Classification; Computational Biology/Bioinformatics; Computational Science and Engineering; Computer assisted instruction; Computer Science; Crossovers; Data Mining and Knowledge Discovery; Diagnosis; Disease; Feature selection; Image classification; Image Processing and Computer Vision; Machine learning; Mangoes; Medical imaging; Original Article; Performance evaluation; Plant diseases; Probability and Statistics in Computer Science; Support vector machines; Crossover-based levy flight distribution; Mango leaves; MobileNetV2; Support vector machine; Convolutional neural network
• ISSN: 0941-0643; 1433-3058
• DOI: 10.1007/s00521-021-06726-9

Mango leaf diseases have a negative impact on mango quality and yield. It is difficult to make an accurate diagnosis of mango leaf disease diagnosis with the naked eye. A lot of computer-aided and machine learning techniques have recently been used by researchers for the classification of mango leaf diseases. However, it has been reported that these approaches have some limitations to their performance which can be attributed to problems due to higher feature dimensionality, overfitting, computational complexity, and lack of feature qualities. To overcome these issues, we proposed a novel framework for mango leaves disease classification. The images were taken from Andhra Pradesh, the largest mango cultivating land in India. The proposed framework is categorized into four stages: data preparation stage, feature selection stage, learning and classification stage, and the performance evaluation stage. We selected 380 images from the categories of healthy and diseased (Mango Anthracnose, Bacterial black spot, and Sooty mold). Different data augmentation techniques are applied to prevent overfitting and improve generalization. Next, a convolutional neural network with crossover-based levy flight distribution is applied for better feature selection. Further, the pre-trained MobileNetV2 model is used for the learning stage and leaves diseases classification is done via support vector machine at the final stage of the MobileNetV2 model. The experimental results demonstrate superior classification performances over other state-of-art methods.