I'm working with multi-temporal datasets from Landsat (Landsat 8 & Landsat 7 & Landsat 5) and I tried to extract impervious areas, more exactly urban areas, but I can't obtained results very close to reality (accuracy was approximately 80 % by using reference point in Google Earth). I used supervised classification: SAM & Maximum Likelihood, then logical classification (between SAVI + MNDWI + NDBI) but I can't obtained very good results. I used bands: Red, Green, Blue, NIR, MIR, SWIR 2. How I can extract urban areas from different Landsat image? What methods would you me recommend ?
Classifying urban areas from Landsat data is a common practice and usually yields accurate results. To improve your accuracy I would reassess your training data. As a rule-of-thumb: 1) the more samples the better, 2) samples should be distributed evenly throughout the scene. There are numerous studies on just this topic:
- Extraction of urban built-up land features from Landsat imagery...
- An assessment of landsat MSS and TM data for urban and near-urban land-cover digital classification
- Land cover classification and change analysis of the Twin Cities (Minnesota) Metropolitan Area by multitemporal Landsat remote sensing
- Urban classification using full spectral information of Landsat ETM+ imagery in Marion County, Indiana
- Use of impervious surface in urban land-use classification