A Method to Incorporate Uncertainty in the Classification of Remote Sensing Images

The authors analyze in this paper whether the introduction of the uncertainty associated to the classification of surface elements in the classification of landscape units can improve the results accuracy. To this end, a hybrid classification method is developed, incorporating uncertainty information in the automatic classification of very high spatial resolution multispectral satellite images to obtain a map of landscape units. The developed classification methodology includes the following steps: 1) a soft pixel-based classification; 2) computation of the classification uncertainty; 3) image segmentation; and 4) object classification based on decision rules. The first step of the proposed methodology is the soft pixel-based classification performed with the maximum likelihood classifier, aiming to identify the surface elements (e.g., tree crown, shade, bare soil, buildings). Subsequently the posterior probabilities are computed to all pixels of the image. This information enables the computation of the classification uncertainty. An image segmentation is then made to obtain image-objects. The classification of the resulting objects into landscape units is performed considering a set of decision rules that incorporate the probabilities assigned to the several classes at each pixel and the degree of uncertainty associated to these assignments. The proposed methodology was tested on the classification of an IKONOS satellite image. The accuracy of the classification was computed using an error probabilistic matrix. The comparison between the results obtained with the proposed approach and those obtained without considering the classification uncertainty revealed a considerable improvement in the classification accuracy. This shows that the information about uncertainty can be valuable when taking decisions and can actually increase the accuracy of the classification results.