MAPPING AND ESTIMATION OF CROP HEALTH USING UNMANNED AERIAL VEHICLE

Abstract

Zero hunger is the second goal of the sustainable development goals (SDGs). The production of food is facing challenges worldwide from crop pests and diseases, which are capable of adversely affecting a large range of crops, and resulting in major yield losses. It is thus imperative for farmers to detect promptly, and treat crop health issues as precise and timely detection, monitoring and mapping of crop pest and diseases is essential for zero hunger, especially in Africa where hunger and scarcity have reached a terrifying stage. In improving the economic growth of the country and maintaining the sustainability of the agricultural systems, there is need for precision agriculture. The concept of this research is to develop a low cost geospatial automated system for assessing crop health using available technological tools such as the visible light unmanned aerial vehicle (UAV). The study area for this research is located between Federal University of Technology main campus and Garatu village and it measures about 21 hectares. Images were acquired from visible light camera (RGB) with 1.0cm resolution, following site plan design, pre-marking and establishment of GCPs. The image processing incorporates three basic stages: Initial processing, generation of point cloud mesh and, lastly, digital terrain model (DSM), orthophoto and index map generation by implementing the following algorithms: Scale invariant feature transformation (SIFT), Bundle block adjustment (BBA), and Structure from motion (SFM), respectively. Different vegetation index maps (NDVI, VDVI, NGRDI, NExG maps) were developed from the orthophoto to estimate the healthiness of the crop samples collected from the field. The extracted values of the sample crops ranges differently for various vegetation index as follows: NDVI value ranges between 0.32 to 0.96, the VDVI ranges from -0.25 to 0.63, NGRDI ranges from -0.23 to 0.43, NExG ranges from -0.19 to 0.42 and VARI ranges from -0.82 to 0.86. The crop samples exhibit a relatively fair healthy characteristic in the Vegetation indices, but the most optimal and reliable amongst them is the NDVI and VDVI. Further statistical analysis was carried out in correlating the NDVI value of each of the crops with the respective tissue test result of the same crops. This analysis makes it evident that results from photogrammetry can be more reliable in estimating cassava and rice’s health amongst others, as there is a 100% chance of obtaining the same results.