Detection of vegetation indexes and species diversity of Soğanlı Botanical Park (Bursa) tree species with the use of multispectral cameras

Unmanned Aerial Vehicles (UAVs), which have recently gained popularity with developing technology, have begun to be used to obtain aerial photographs in forestry studies. Vegetation can be detected by analyzing the images obtained thanks to multispectral cameras on UAVs. The data obtained from these images are processed through orthophotos. Various analyzes are performed on the images using different band combinations and information about the health of the vegetation is obtained. These analyzes are usually made using reflectance values and vegetation indices used to determine different characteristics of the vegetation. These methods have become an important tool in vegetation monitoring and management in the field of forestry. Bursa Soğanlı Botanical Park is included in the green belt project and provides various benefits to the city. The focus of this study is the potential to use images obtained from multispectral cameras mounted on Unmanned Aerial Vehicles (UAVs) to detect vegetation types in Bursa Soğanlı Botanical Park. In the study carried out over Soğanlı Botanical Park, flights were made using two different UAVs and two different multispectral cameras. During these flights, the number of flights, flight altitude and flight boarding rates were calculated. Then, the images obtained with UAVs were transferred to the computer environment and processed in Agisoft Metashape software to create multispectral orthophotos and vegetation indices. The created multispectral orthophotos and plant indices were saved in ".tiff" format and segmentation and object-based classification were performed in eCognition Developer 9 software. During this classification process, local control points were used. To measure the accuracy of the resulting classified images, a plant species map was created using the ArcToolbox module in the ArcGIS software. With these methods, a detailed map of the vegetation and areas in the park was created and its accuracy was measured. As a result of the study, 9 classes were created: maple, plane, linden, building, wetlands, agricultural areas, other areas, other forest areas and grass areas. While the kappa value calculated in the study was 87%, the accuracy rate on a class basis was 83.33% for maple, 96% for buildings, 70% for plane trees, 93% for other areas, 73% for other forests, 80% for linden, and 80% for grassy areas. It has been determined as 93%, 100% for wetlands and 96% for agricultural areas. This proves that plant species can be detected accurately with the help of multispectral cameras integrated on the UAV.