Deep learning approach for recognition and classification of yield affecting paddy crop stresses using field images

Abstract

On-time recognition and early control of the stresses in the paddy crops at the booting growth stage is the key to prevent qualitative and quantitative loss of agricultural yield. The conventional paddy crop stress recognition and classification activities invariably rely on human experts identifying visual symptoms as a means of categorization. This process is admittedly subjective and error-prone, which in turn can lead to incorrect actions being taken in stress management decisions. The work presented in this paper aims to design a deep convolutional neural network (DCNN) framework for automatic recognition and classification of various biotic and abiotic paddy crop stresses using the field images. The work has adopted the pre-trained VGG-16 CNN model for the automatic classification of stressed paddy crop images captured during the booting growth stage. The trained models achieve an average accuracy of 92.89% on the held-out dataset, demonstrating the technical feasibility of using the deep learning approach utilizing 30,000 field images of 5 different paddy crop varieties with 12 different stress categories (including healthy/normal). The proposed work finds applications in developing the decision support systems and mobile applications for automating the field crop and resource management practices.

Keywords

• Paddy crop
• Biotic and abiotic stress
• CNN
• VGG-16