A Hybrid of Plant Leaf Disease and Soil Moisture Prediction in Agriculture Using Data Mining Techniques

摘要

This paper focuses on developing novel technologies for monitoring the plant health, evaluating the water content level for different plants by analyzing the plant and soil images respectively. Initially, plants and soil images are captured through digital camera with a required resolution. The innerdistance shape context based descriptor and geometrical descriptors are used for extracting the shape and geometric features from the plant images. In addition, texture and color features are extracted from the soil images. By using the contour features of the plant images, the plant type is identified through the botanical plant species dictionary. The leaf diseases are predicted by Transductive Support Vector Machine classification. The causes for the specific plant disease are identified based on the Latent Dirichlet Allocation and Artificial Neural Network classification technique through the features of soil images and the diseased plant images. The obtained results are broadcasted to the cultivators through mobile phones by means of text messages on a daily and seasonal basis with possible suggestions of preventive measures.