Plant Health Report Through Advanced Convolution Neural Network Methodology

This paper proposes an ideal approach to detect the micro (B, Cl, Cu and Fe), macronutrients (nitrogen phosphorus and potassium) deficiency and disease detection in plants for effective development of the plant health monitoring system for farm fields. It is possible to obtain the pattern of a specific disease and the characteristics of a deficiency using the Convolution Neural Network (CNN) algorithm, which employs approximately 128 filters; this enables to identify 128 specific features hence providing accuracy of 92%, using the dataset of 50 thousand diseased plants from plant village and 1.8 thousand nutrient deficient plants from other sources were analysed to provide greater accuracy in detection, by using large dataset it has improved the quality and accuracy of the disease detection in plants using Keras and TensorFlow for image processing. By doing so, the mobile device is capable of identifying the disease with higher efficiency and is able to suggest the measures that farmers can take to avoid the pest infection and diseases that have been identified in their plants, to grow a healthy plant for high yield. The disease detection is done using the classifier present in the cloud, and hence, it gets a regular update with little change in the hardware so as to reduce the investment cost.