Automatic Robotic Crop Disease Detection and Pesticide Dispenser Using Machine Learning

Vijayakumar Ponnusamy, C. Amrith, S.S. Akhash, S. Sanoj and Shashank Srikant

Agriculture diseases are the main factors in reducing the yield. To enhance the productivity, it is needed that the crops are supplied with required nutrients in correct quantities, sufficient sunlight and other environmental conditions. The identification of disease symptoms at early stages of crop growth could reduce the loss. The Identification using the conventional naked eye method is a laborious task and is not reliable, and the usage of Automated Pre-Trained machines would drastically improve the efficiency. This paper deals development of such automatic detection of plant disease by training a convolutional neural network model to classify the leaves of Pepper Bell plant into healthy and unhealthy in real time. A line follower robot is designed to travel across the farm for collecting real time image and apply machine learning algorithm of convolutional neural network to classify the leaves into healthy and unhealthy in real time. The line follower robot also equipped with a spray system to spray pesticide on affected areas. Deep Learning algorithms are applied on the dataset containing 2475 sample images (both healthy and unhealthy) of Pepper Bell leaves. The proposed mechanism archives accuracy 93.5% to classify the leaves.

Volume 11 | Issue 11

Pages: 119-125

DOI: 10.5373/JARDCS/V11I11/20193176