Orthomosaics Aerial photography for Kakadu National Park

Brief description

This data set consists of .tif files of true colour orthomosaics for expansive areas of mangroves in Kakadu National Park in Australia's Northern Territory.

The orthomosaics were generated from 68 stereo pairs of true colour aerial photographs acquired in 1991 in the lower reaches of the East Alligator, West Alligator, South Alligator and Wildman Rivers and Field Island, Kakadu National Park, Northern Australia (Mitchell et al., 2007). The photographs were taken at a flying height of 13,000 ft (3,960 m) using a Wild CR10, a standard photogrammetric camera with a frame size of 230 x 230 mm. The focal length was 152 mm. The photographs were scanned by Airesearch (Darwin) with a photogrammetric scanner to generate digital images with a pixel resolution between 12 and 15 mm. The orthomosaics have a spatial resolution of 1 m, cover an area of approximately 742 km² and a coastal distance of 86 km.

These orthomosaics were co-registered using ground control points identified from 1:100,000 digital topographic maps with a Universal Transverse Mercator (UTM), and subsequently co-registered to LiDAR data acquired over the same region in 2011.

Lineage

Data Creation
Overview: Digitizing points from 1:100 000 scale maps implied an accuracy of 0.5 mm on the map and hence 50 m on the ground. Based on the topographic map scale, the horizontal and vertical (absolute) accuracies of the ground control points (GCPs) were estimated at 50 m and 2 m respectively. The overall integrity of the registration was high as numerous tie points were located in the overlapping regions to assist in stabilizing adjoining images.
Methods: Following collection of GCPs, the Kakadu National Park (KNP) photographs were orientated using standard collinearity equations (Jones, 1982) and resampled to 0.37 m spatial resolution. The output from the transformation gave residuals (root mean square errors; RMSE) to all point measurements in image coordinates. These residuals were checked and, where necessary, new points were taken or existing ones were relocated until the residuals were less than one pixel, thus ensuring that the relative positioning within the block was of the order of one pixel (or 0.37 m) on the ground. The residuals on the GCPs were checked to ensure that they were within the predicted horizontal (i.e. 50 m) and vertical accuracy (i.e. 2 m).

Notes