Detection of plant leaf nutrients using convolutional neural network based Internet of Things data acquisition

In this paper, the study detects the nutritional deficiencies from these leaves using Internet of Thing (IoT) based image acquisition and nutrition analyser devices. The former captures the color of the leaf and the latter helps in finding the nutrients in each zone based on the image captured by the device. The study uses an improved convolutional neural network to detect automatically the nutrients present in a leaf. The type of leaf is considered from the plants including coriander, tomato, pepper, chili, etc. The Convolutional Neural Network (CNN) is used to extract the patterns of leaf images from the data capturing IoT devices and nutrition analyser device. The system stores and process the data in cloud, where the CNN integrated in Virtual Machines enables the process of input data and process it and sends the report to the authority. A total of 3000 images are collected out of various disorders in five different plants. A 5 fold cross-validation is conducted on training and testing dataset. The system is tested in terms of accuracy, sensitivity, specificity, f-measure, geometric mean and percentage error. The comparison made with existing models shows an improved detection accuracy by CNN than other deep learning models.