Detecting clandestine graves
Year: 2020
Type: dissertation
Abstract: By far, hyperspectral imaging, as a tool for grave detection, has only been tested in warm, moist environment where the vegetation re-growth is relatively rapid. In earlier studies, differences in the vegetation re-growth over a grave and other soil disturbances were the key factor why a grave could be spectrally distinguished from respective soil disturbances. However, the performance of hyperspectral imaging for distinguishing graves from other soil disturbance has not, until now, been studied in dry, unproductive environment. \n \nThe spectral reflectance of a grave containing a rabbit and a respective pit filled with soil was measured with hyperspectral camera immediately, one week and three weeks after burial. The spectral characteristics and differences between the grave and the control, and further, in respect to their surroundings, were determined by comparing the mean spectra and by subjecting the images to the unsupervised ISODATA classification. ISODATA classifier assigned the image pixels in each image to six spectrally distinctive classes: grave, control pit, vegetation, stones, and white and grey reference target. This suggested that the grave and the control, as well as the grave and the control and the surrounding vegetation, deviated spectrally from each other during the entire monitoring period. Subtle spectral differences, mainly in the wavelength region above 750 nm, were observed in the comparison between the mean spectra of the grave and the control. However, given the subtleness of the observed differences and the standard deviation of the reflectance values, especially for the grave, it is possible that the differences were caused purely by the large variation of values, rather than true differences in spectral characteristics. \n \nFurthermore, this thesis presents a feasible workflow for hyperspectral image acquisition and analysis, specifically designed for distinguishing a grave from other soil disturbances. The suggested workflow enables wider application of hyperspectral imaging for grave detection, since it requires no special expertise in remote sensing. Wider application of the method would further support defining the environment specific advantages and limitations of the method, necessary for concluding the true potential of hyperspectral imaging for grave detection.
Source: Aaltodoc (Aalto University)
Author: Ruotsala, Anni-Helena
Institution: Aalto University
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7009062338
