COMPARISON BETWEEN PIXEL BASED AND OBJECT BASED METHODS FOR ANALYSING HISTORICAL BUILDING FAÇADES

Abstract

Techniques of digital image processing give the possibility to detect damages, such as moisture or biological changes on surfaces of monuments in a non-destructive way. Traditional classification methods are all pixel-based and do not utilize the spatial and context information of an object and its surroundings, which has potential to further enhance digital image classification. In this study, we compare the application of pixel based classification and the object-oriented classification technique using multispectral photographs taken with a Fujifilm IS-Pro digital single lens reflex camera in order to detect and locate damage affecting biocalcarenite stone employed in the construction of the Mosque of Cordoba (Andalusia, Spain). The best results were achieved with object-oriented classification obtaining an overall classification accuracy of 96.55% and an excellent kappa statistic (94.40%). These results has shown that object based feature extraction can result in powerful solutions for the interpretation of high resolution data. It is proved that object-based approach to classify imagery is much better than the traditional classification techniques, such as pixel-based.