Extracting spatial information from aerial video imagery for monitoring riparian areas

Abstract

[Abstract]: There are many different techniques being used in remote sensing to capture data for natural resource management. Aerial video mapping is a relatively new technique that is gaining popularity because of its non invasiveness, relative cost-effectiveness and timeliness. The Queensland Murray Darling Committee (QMDC) has collected video footage of rivers within its catchment area to facilitate their river management activities. It endeavours to ascertain the usefulness and reliability of information provided by aerial video mapping technology for riparian management. The aim of this project was to develop object-oriented image processing techniques and GIS based techniques for extracting riparian area parameters from aerial video imagery. Specifically, the objectives were to a) use traditional image processing techniques to extract the identified riparian parameters; b) identify and test objectoriented image processing techniques that may be suitable for mapping the selected riparian variables; and c) assess the accuracy of the results generated from both the traditional per-pixel and object-oriented image processing techniques. Four images were extracted from the aerial video footage. Each image represented a dominant land cover/use type (i.e. agriculture, urban, pasture and forest). For each image, a set of classes representing various riparian parameters were created. These were then used for classifying the images using the maximum likelihood algorithm in ERDAS IMAGINE 9.1, and the object-oriented classification techniques in Definiens Professional 5. The object-oriented approach achieved results with accuracies ranging from 90% up to 97% while the pixel-based approach managed accuracies ranging from 69% up to 82%. The data was found to have two major limitations. It had only three spectral bands, red, green and blue. Accurate measurements could not be made from the imagery because it was collected at an oblique angle.