Quality assessment of CHRIS/PROBA image and recommendation for land cover classification

Abstract

Quality assessment of satellite data forms an important part for any land use classification process and as a result every user should know the quality of the image used for preparation of a map. In the present day scenario, data from small experimental satellites are being used a lot for preparation of maps for land and water resources application. To assess the quality of these products, accuracy is in general the true value of the quantity that is being measured. An accuracy statement enables the end user to have a first hand knowledge about the cost involvement and the algorithms to be used for data calibration. The present study involved the i) quality assessment of nadir view CHRIS data (raw and destriped) and ii) recommendation for mapping land cover units. The images were processed and destriped using 1) DIELMO destriping method and 2) ESSC destriping algorithm using HDF Clean. The study looks into the procedure followed by the above two methods of destriping the raw data and their outcome. But the question that arises whether these destriping processes are at all effective in generation of landuse / landcover map? A comparative qualitative assessment of the raw image and the two images destriped by two different methods, were carried out. A quantitative analysis of the three images were also undertaken to find out whether the enhancement of the images by way of destriping helps in the classification process. The analysis was undertaken by i) statistical approach and ii) classification of the three images using unsupervised classification technique. A comparative study of the classified images generated from the raw data and data cleaned by ESSC algorithm and DIELMO was also carried out. This study on the basis of the quantitative and qualitative analysis recommends the best images for landcover classification.

More information

Identifiers

Catalogue record

Export record