Urban Feature Extraction From High-resolution Images
- Paper number
IAC-07-B1.I.12
- Author
Dr. Mukund Rao, Navayuga Spatial Technologies Pvt. Ltd., India
- Coauthor
Dr. BSP Rao, Andhra University College of Engineering, India
- Coauthor
Prof. Ian Masser, United Kingdom
- Coauthor
Dr. K. Kasturirangan, Indian Space Research Organisation (ISRO), India
- Year
2007
- Abstract
Images, from satellites and aerial platforms, have long been employed as a tool in urban planning and management. Indeed, this form of remote sensing is still extensively used today and benefits from very sophisticated digital image-processing techniques. As far as urban features are concerned, the most important features calling attention are landuse classes (to finest detail), building outlines and road features - these three sets of features embody most of the needs of Urban Mapping – which is a major application area of high resolution satellite images. With high resolution images spectral separability is just one parameter – even as shape and pattern become important. Further, extreme “scene variance” due to high resolution geometry render spectral separation limited in high resolution images – especially so when it becomes impossible to characterize spectral signatures for classes – represented by difference in materials (roofs of Asbestos against roofs of Wood or Cement; though all are buildings). Automated methods of urban information extraction from satellite images are drawing attention – especially with the availability of high resolution images from satellites. In particular, the process of delineation of urban features and entities in the image is said to be the need off the hour. Methods of incorporating more parametric classification are subjects of intense research. This paper looks at addressing this issue and researches on new techniques of segmentation and fuzzy-classification and characterizing urban features from these high resolution images.
- Abstract document