UAV for Segmentation and Object Detection: A Systematic Review Using PRISMA Method

Abstract The purpose of this literature review study is to examine the extent of the object prediction and detection approaches carried out by previous researchers and identify research gaps. The YOLO and CNN algorithms are used to group articles using several open source tools to process data in the form of images. The databases used include data taken from Google Scholar, indexed by IEEE and Elsevier using the keywords “UAV”, “CNN”, “YOLO” and “remote sensing.” The data used in this study were taken from the last 5 years and only included articles in Q1 and Q2 journals. The results of this study were mostly articles using the YOLO algorithm (80%), followed by articles using the CNN algorithm and machine learning. Of the selected articles, 87% discussed object detection, followed by detection and clustering. The results showed that several optimizations of the YOLO algorithm are important in object prediction and detection research, such as maximizing image quality, adding parameters, detecting edges, and adding other companion algorithms for better results. In-depth analysis using NDVI and GNDVI is still rarely discussed in research in the agricultural sector, from all the journals obtained in this study only 2 journals were obtained that used NDVI in in-depth analysis of color differences. This study identified that the next research gap is that there has been no research that discusses the specifications of pest detection and detection of disease spread using remote sensing-based objects, especially drones, and the need to use NDVI or GNDVI in detecting plant health levels and even ready-to-harvest plants.