Methodological Framework for Developing a UAV Based Dataset to Validate EO Based Fractional Vegetation Cover Estimates

Kurzfassung

Validating fractional vegetation cover (fCover) at a large scale is challenging because so far in situ measurement availability is often very local, limited and fragmented. Furthermore, existing methodologies show limitations in producing consistent fractional vegetation cover validation data from RGB UAV imagery across the extremely diverse surface appearances encountered in agricultural fields. The proposed methodology addresses this gap by employing an ensemble approach that integrates the complementary strengths of image segmentation and classification techniques. This framework is further supported by derived surface parameters, resulting in a unified processing chain. At the same time, the validation approach is cost and time efficient, easy to perform and works regardless regional characteristics. Taken advantages of the ESA CaRMA campaign, the DLR collected ~15.000 UAV RGB images between 03th of March until 31st of October (~weekly) under a variety of illumination conditions, covering a complete vegetation cycle in ~87 fields. By explicitly choosing a lightweight consumer UAV, no expert knowledge for operating the UAV is necessary, neither it require maintains or calibration of the camera system. It can be transported in a backpack to maximalize fast, cost-effective data acquisitions to create a comprehensive dataset for the validation of EO based fractional vegetation cover products. This study, described the developed process to design a comprehensive dataset and outlines the methodology for validating fractional vegetation cover products