The spatial resolution of remotely sensed imagery has improved considerably during the last few years and will increase dramatically in the near future due to the imminent launch of the new generation very high-resolution sensors. For urban applications in particular, with the spatial properties of the new sensors it will be possible to recognise, not only a generic texture window with specific urban characteristics, but also to detect in detail the objects that constitute the "urban theme". However, the improvement in the spatial resolution may result in a decrease of the accuracy of automatic classification techniques, if only the standard multi-spectral analysis procedures are applied. In this chapter a per-segment segmentation procedure is presented, based on the gray-scale geodesic morphological transformation and has been successfully utilised to detect built-up objects using only the 5 m spatial resolution panchromatic data of the IRS1-C satellite. The imagery is subsequently classified on a per-segment basis using a multi-layer perceptron neural network classifier.
展开▼