Using Image Processing for Plant Health Monitoring

Abstract

The problem farmers face in knowing the health status of the plants during the growth period has limited the yield produced, informed application of pesticide and sustainable agriculture. The challenge to better plant health monitoring is traceable to the methods used in the monitoring process. The aim of this project was to investigate the application of image processing in solving plant health monitoring problems. The specific objectives that the project sought to address include; identification of spectral features for plant health state classification, developing of a model for plant health classification, developing image processing segmentations based on the models for plant health classification and developing a prototype for the plant health monitoring system. The intention was to have the data collected easily, understood, and interpreted for early mitigation plans on the onset of a change of plant health status. In this project, two approaches were used where an unmanned aerial vehicle platform was used for carrying cameras to capture aerial images on specific coordinates. The images were then processed through an algorithm that was developed to identify Normalized Difference Vegetation Index (NDVI) data to generate the required data for development of the plant health monitoring prototype system. This was a continuous monitoring system as the plants were monitored through the planting to harvesting. The plant health monitoring system prototype using image processing was developed. The acquired aerial images were processed from the planting to harvesting and a graph produced for tracking the health status of the plants. The testing and evaluation findings revealed that a variation in graph pattern would indicate the onset of a change in health status. The evaluation of the system showed that the system can improve plant health monitoring for farmers as well as provide effective ways in determining plant health status.