ABSTRACT:Soil roughness represents fine-scale surface geometry which figures in many geophysical models. While static photogrammetric techniques (terrestrial images and laser scanning) have been recently proposed as a new source for deriving roughness heights, there is still need to overcome acquisition scale and viewing geometry issues. By contrast to the static techniques, images taken from unmanned aerial vehicles (UAV) can maintain near-nadir looking geometry over scales of several agricultural fields. This paper presents a pilot study on high-resolution, soil roughness reconstruction and assessment from UAV images over an agricultural plot. As a reference method, terrestrial laser scanning (TLS) was applied on a 10 m x 1.5 m subplot. The UAV images were self-calibrated and oriented within a bundle adjustment, and processed further up to a dense-matched digital surface model (DSM). The analysis of the UAV-and TLS-DSMs were performed in the spatial domain based on the surface autocorrelation function and the correlation length, and in the frequency domain based on the roughness spectrum and the surface fractal dimension (spectral slope). The TLS-and UAV-DSM differences were found to be under ±1 cm, while the UAV DSM showed a systematic pattern below this scale, which was explained by weakly tied sub-blocks of the bundle block. The results also confirmed that the existing TLS methods leads to roughness assessment up to 5 mm resolution. However, for our UAV data, this was not possible to achieve, though it was shown that for spatial scales of 12 cm and larger, both methods appear to be usable. Additionally, this paper suggests a method to propagate measurement errors to the correlation length.

KEY WORDS: multi-sensor calibration, oblique aerial imagery, tie point decimation, bundle block adjustment, affine-invariant feature point ABSTRACT:Aerial multi-camera platforms typically incorporate a nadir-looking camera accompanied by further cameras that provide oblique views, potentially resulting in utmost coverage, redundancy, and accuracy even on vertical surfaces. However, issues have remained unresolved with the orientation and calibration of the resulting imagery, to two of which we present feasible solutions. First, as standard feature point descriptors used for the automated matching of homologous points are only invariant to the geometric variations of translation, rotation, and scale, they are not invariant to general changes in perspective. While the deviations from local 2D-similarity transforms may be negligible for corresponding surface patches in vertical views of flat land, they become evident at vertical surfaces, and in oblique views in general. Usage of such similarity-invariant descriptors thus limits the amount of tie points that stabilize the orientation and calibration of oblique views and cameras. To alleviate this problem, we present the positive impact on image connectivity of using a quasi affine-invariant descriptor. Second, no matter which hard-and software are used, at some point, the number of unknowns of a bundle block may be too large to be handled. With multi-camera platforms, these limits are reached even sooner. Adjustment of sub-blocks is sub-optimal, as it complicates data management, and hinders self-calibration. Simply discarding unreliable tie points of low manifold is not an option either, because these points are needed at the block borders and in poorly textured areas. As a remedy, we present a straight-forward method how to considerably reduce the number of tie points and hence unknowns before bundle block adjustment, while preserving orientation and calibration quality.