Assessment of hazard tree/snag detection using drone-based, multi-spectral sensors

Abstract

Snags are an integral component of forest ecosystems as they provide habitat for a number of different species and add complexity to vertical forest structure. However, snags also may pose as potential hazards to people and property. Efficient and effective methods to locate and assess snags/hazard trees holds value to resource and conservation managers. This study aimed to assess the feasibility of using drone-based, multi-spectral sensors for detecting snags/hazard trees. The methods used in the study included an autonomous drone flight over the study areas, orthomosaic processing, object-based image analysis (OBIA), an accuracy assessment, and a field ground truth. The results provided sufficient evidence of drone-based, multi-spectral sensors being effective at detecting snags/hazard trees. However, the methods used in this study were found to only be accurate at detecting high quality/hazard snags. Segmentation parameters had a significant impact on the degree of quality/hazard of snag that the algorithm could detect. The orthomosaic classification was considered as highly accurate with an overall accuracy of 93.4%. Resource and conservation managers can effectively use the methods from this study for a variety of applications that aim to promote biodiversity and/or minimize public hazards.