Monday, August 4, 2008 - 4:40 PM

COS 8-10: The Digital Transferscope: A new software tool for the analysis of digital aerial photographs for land cover change

Stephen H. Roxburgh, University of New South Wales and CSIRO Forest Biosciences and Michael C. Fallon, University of New South Wales.


The analysis of aerial photographs has a long history in environmental management, particularly for vegetation mapping and for quantifying land cover change. Aerial photographs are unique in providing landscape-scale images at a high resolution (often < 1m), with photographic records often extending over many decades.

The amount of pre-processing required before aerial photographs can be used effectively depends upon the types of data required, and their accuracy. For vegetation mapping, or other purposes requiring knowledge of the precise geographic locations of features within the photographs, ortho-photographic rectification is necessary. This involves transforming the photograph to match an existing coordinate system, by e.g. correcting the image for the influence of camera lens distortion, terrain relief, and camera tilt. Full ortho-photographic rectification is, however, expensive and time-consuming.

In contrast, a number of researchers have developed simpler, more cost-effective methods to extract coarser level information from aerial photographs, without the need for full ortho-rectification. These methods are usually based on statistically sub-sampling the image, e.g. by manually overlaying a transparent grid of sampling points. For such studies a Zoom Transferscope is often used. This is an optical instrument with a system of adjustable zoom lenses and prisms to align and overlay two projected images.


The Digital Transferscope (TDT) software described here was developed as a digital solution to, and extension of, the above procedure. Aerial photographs are first digitally scanned at a high resolution, and are then co-registered relative to one another within the software using a combination of user-defined control points and photographic analysis algorithms. A downhill simplex search algorithm is used to automate the rectification process, through the iterative application of either affine or projective image transformations. To allow analyses to be conducted on a standard desktop PC, novel algorithms had to be developed to facilitate the manipulation of very large images (typically greater than 350MB per photo, and 10,000 x 10,000 pixels resolution). Once rectified, a user-defined sampling-grid overlay is added, allowing vegetation cover to be sampled interactively using a number of methods, e.g. at a number of points in the landscape (either on a grid, or located at random), or by recording grid-cell cover dominance. The software has been used extensively for the analysis of land cover change in Australian alpine ecosystems, and examples from this work will be used to illustrate the tool and its operation.