IoT- Deep Learning based Prediction of Amount of Pesticides and Diseases in Fruits
D Devi, Akshaya Anand, S Sophia, M. Karpagam, S. Uma Maheswari
Abstract
The use of pesticides, steroids and fertilizers has tremendously increased the negative effects caused to the people in terms of health. Harmful pesticides enter into the human body through fruits and vegetables, so an optimal solution is needed to recognize the disease and the pesticides in the fruits the common man is consuming. Hardware and a software design are done to obtain an accurate and a real time output. In this paper, a prototype of the system is developed with the use of four sensors, (temperature, gas, pH and moisture), Arduino microcontroller and a Wi-Fi module to get the information about the presence of pesticides. The maximum level of pesticides that is accepted legally to be consumed by animals and humans is given by MRL. If a fruit is detected to belong in a range above or below the MRL then it is said to contain pesticides. Through IoT, the pesticide content and the values obtained from each sensors are stored in the Cloud server MATLAB ThikSpeak. Coming to the software design, CNN and SVM algorithms are chosen and the image of the fruit is diagnosed by them. Two algorithms are mainly used to compare the accuracy produced by both and to select the most accurate between the two. Deep Learning process is performed on the image of the fruit and the disease affected in the fruit is identified and later stored in the Cloud server. The information about the disease in fruits and the pesticide value in fruits, the harmful effect caused by it, are sent to the cloud, which is then processed and sent to the application present in the consumer's smart phone which is developed in HTML5, thereby a real time regular monitoring is possible.