Comparison of pixel and object-based classification methods in the evaluation of unmanned aerial vehicle and high resolution satellite data

Remote sensing technologies have been used frequently in recent years as a result of developments in both satellite, aircraft and ground based systems as well as computer software and hardware. In the light of these technological developments, unmanned aerial vehicles, previously used for safety purposes, have been also used for civilian purposes for many reasons. The reason for the widespread use of unmanned aerial vehicles is not to risk human life. Besides, in comparison with manned aircraft, unmanned aerial vehicles have more important advantages such as low cost and high manoeuvre ability. Nowadays, classification studies, which provide analysis of images obtained from remote sensing data in natural sciences studies, have gained a different dimension with these technological developments. In this context, in addition to pixel-based classification processes, object-based classification method is frequently used in image analysis processes using high resolution data. The main data used in this study conducted on the central campus of Akdeniz University is the current data obtained from the visible and near infrared region of the electromagnetic spectrum mounted on unmanned aerial vehicle (UAV) and the high resolution WorldView-4 (WV4) satellite data of the study area. In this context, the study area was analysed with pixel-based and object-based classification methods using UAV and high resolution WV4 satellite data. The study consists of two basic stages. These are the stages of data collection and preparation and classification. In the first stage of the study, recently unmanned aerial vehicle and satellite data were obtained. At this stage, the data was prepared by using the image pre-treatment and image enrichment techniques. In the second stage, these data were analysed with different classification techniques. As a result of this study, the usage types of the lands in the area were analysed separately according to different data using pixel-based classification and a new approach known as object-based classification method, and the results were compared and it was revealed which method was more successful. As a result of the findings, while the highest overall accuracy value with pixel-based classification was found 54.92% using WV4, 75.40% value was found with object-based classification using unmanned aerial vehicle data.