An Efficient and Robust Random Forest Algorithm for Crop Disease Detection

Nourishment security is a notable risk to crop diseases, however the testimonial stays unmanageable in various countries of the world due to the lack of proper foundation. This study is for productive crop organization in large areas using the key variables namely, texture, phenology, soil moisture, topographic vegetation, different satellite, and climatic data (precipitation and temperature). Since machine learning methodology in the sector of leaf-based image organization has displayed magnificent outcomes, an efficient Learning algorithm to find the impending disorder existing in plants on a massive scale, is used. In this system topographic and climate variables associated with spectral responses are compared and the near-infrared band is used with high spectral range (0.85 to 0.88m). The characteristic feature in image is obtained using Histogram of an Oriented Gradient (HOG). To evaluate RF models, a 20% independent dataset of training samples is used in addition to OOB data. The mean drop in accuracy and mean drop in Gini score are calculated. A comparative analysis is done on different Machine learning algorithms. The proposed RF model is efficient and robust algorithm obtaining an accuracy of 97.2% in detecting the disease to provide nourishment security.