A methodological investigation into numerical analysis of data from aerial photographs for vegetation survey

Grainger, John Edward (1977) A methodological investigation into numerical analysis of data from aerial photographs for vegetation survey. University of Southampton, Doctoral Thesis.

Abstract

This investigation constituted an attempt to forge a methodological link between ground and aerial survey practice for large scale vegetation surveys, specifically with relation to objective data collection and sorting techniques.The work involved an evaluation of the use of physiognom-c categories to serve as vegetation descriptors as it was considered that the vegetative features which govern its tonal and textural appearance on air photographs are primarily physiognomic in nature. The results of a pilot survey in the New Forest showed that a classification of the vegetation based on a set of plant features of organ status (eco-organs) gave promising ecological results when compared to a 'control' classification based on traditional floristic data. Modifications to the original eco-organ field list were made subsequently to improve categorisation and to include descriptors of gross features of the vegetative scene. Refinements to the ground survey techniques were also made with regard to sample location and size; these were incorporated into the methodology for the main survey of a more extensive area of the New Forest. Floristic and eco-organ data were used to categorise the vegetation and the results of this survey represented those vegetation 'classes Prof interest' again which the analyses of the air photographic data were assessed.Alternative methods for automatic data extraction from black and white aerial photographs were investigated; the results favoured a method involving equidensitometry and this was facilitated by the use of a Quantimet 720.Two approaches to the classification of photographic grey scale data were evaluated. The unsupervised classification approach produced site groups which showed only slight resemblance Lc those from the ground survey data, though this was enhancea by 'extenaln;' the image groupings. In the supervised approach, allocations ci the image sites to the tonal image signatures of the ground survey central sites again produced site groupings showing only slignt correspondence with those from the ground survey analyses. However the correlations were considerably improved through the Introduction of a simple 'threshoiding, device into the allocation procedure.The main reasons for the limited success of this attempt to '_ink the ground and photographic data analyses were recognised as (i; the large element of noise in the data matrices involved and (ii.) the inadequacy of simple tonal measurements to fully characterise image signatures.Although a large number of aspects obviously require further investigation the study in general has indicated those lines for future development which could effectively combine the two survey practices into an integrated methodology.

This record has no associated files available for download.

Contributors

Download statistics