A Novel Smart Agricultural Management System using Hybrid Convolution-based Deep Learning Model with Multitask Classification

Abstract

Agriculture is essential to the Indian economy. Landowners, who make up 58% of the population, depend on it to make their livelihoods. Every single day, farmers encounter a variety of pest-related difficulties. It is difficult to diagnose the disease manually using human resources. It has an apparent impact on the agricultural sector. The investigation intended to employ innovative methods to identify insects and disorders in the agriculture field. The Internet of Things (IoT) is adding new dimensions to the intelligent agricultural industry. This allows the individual to gather information from farms in the actual moment and send it to faraway locations for analysis. Automatic illness diagnosis is achievable using sensor information and photographic evidence collected in the field. Furthermore, IoT-based intelligent irrigation management tools can aid in improving water supply efficiency in an accurate agricultural environment. Existing efforts gather images from IoT devices for pest identification and categorization, but their precision isn't good enough. Therefore, in this work, an accurate and effective IoT-based smart agriculture management scheme is introduced. In the beginning, from the IoT sensors, the necessary soil and environmental data, field images, and crop and plant images are garnered. In the first phase, the crop yield prediction stage is performed by employing the soil and environmental data and the crop images. Here, the collected images are given as input to the recommended Hybrid Convolution (1D-2D)-based EfficientnetB7 (HCENetB7) technique for predicting the crop yields. In the second phase, the same technique as HCENetB7 is employed for detecting plant disease and pest detection using plant leaf images and soil and environmental data. Whilst in the last stage, smart irrigation is predicted by utilizing the designed HCENetB7 approach using the soil and environmental data and field images. Finally, the numerical analysis is conducted on the recommended smart agriculture management scheme by comparing it with the conventional techniques to ensure the efficiency of the presented scheme.