Testing Flight Paths for Collecting 3D LADAR Imagery of Inconspicuous Targets

A pilot study was performed to examine flight paths for an airborne foliage-penetrating Laser Detection and Ranging (LADAR) three-dimensional (3D) imaging system. Such systems form 3D images based on time-of-flight of laser photons, some of which pass through gaps in foliage or other partial obscurants such as camouflage nets. Hence the 3D image will contain partial information about any objects behind or underneath such obscurants. The obscurant can be removed from the image by only keeping points within some appropriate range, leaving a partial image of the hidden objects which may include targets of interest. An improved overall image can be formed by combining images taken from several different viewpoints, using knowledge of the LADAR sensor's location at each viewpoint. In this study, we compared the overall 3D images obtained from an airborne LADAR system during different flight paths. The scene for each flight path consisted of a four-wheel-drive vehicle placed in a section of a eucalypt forest. Models for the vehicle and the individual trees were created in the 3D modelling software Maya, and exported as point clouds to be used in the general-purpose analysis software MATLAB. The overall forest scene was then assembled from the individual trees. The amount of light penetrating the foliage was determined at three different locations within the scene. On average, results were found to roughly agree with the prediction that light penetration scales with the sine of the angle from horizontal. Formation of LADAR images was modelled by determining the set of points in the scene that had a direct line-of-sight to the airborne sensor. This low-fidelity approach was taken because the aim of the study was to investigate ways of tasking the sensor system, rather than using the model as part of a hardware design process or a testbed for data processing algorithms. Since there is already a random effect due to line-of-sight through the forest canopy, it was decided that further random effects leading to false returns or missed returns would unnecessarily complicate the comparison between results from different flight paths. The flight paths in this study were intended to keep the sensor footprint directed at a known or assumed target area on the ground. One type of path was an arc centred at this location, with the sensor constantly directed sideways at an appropriate elevation. The other type of path went straight past the target area, with the sensor needing to be constantly redirected. Different scan spacings along the path were also investigated. It was assumed that the unobscured view from each flight path would be sufficient for the surveillance or reconnaissance task. So for each flight path, the Hausdorff distance was calculated as a measure of the difference between the 3D image of the target taken through the foliage and the corresponding unobscured image from the same flight path. No significant difference was found between results for the straight and arc paths, but the 3D images were closest to the unobscured views for the smallest scan spacing.