Comparing airborne laser scanning-imagery fusion methods based on geometric accuracy in forested areas

摘要

In this study, different methods for fusing airborne laser scanning (ALS) and multispectral imagery were compared. The improvement observed when computing the position of individual ALS returns in the original unrectified scenes (back-projecting ALS) was contrasted against the geometric errors observed in many orthorectified images. Results showed that back-projecting the ALS is the most accurate technique for fusing high-resolution datasets. Potential sources of systematic errors were studied for all techniques, finding the nadir angle as the most influencing one. The capacity of back-projecting ALS and true orthorectification for reducing the tree leaning problem found in traditional orthophotos was evaluated. Modelled tree lean correlated with geometric errors even while individual factors alone did not. The technique used for true orthorectification solved tree leaning near nadir, but it was affected by the variance of the kriging model, resulting in a low overall accuracy. In view of the results observed in back-projected ALS, the spatial resolution of the imagery was regarded as its main source of uncertainty. The convenience of adding a correction of the effect of the atmospheric refraction and the Earth's curvature was also discussed. Even though the magnitude of those corrections was low, they succeeded in avoiding nadir angle-dependent systematic errors.