Expert Systems with Applications: An International Journal
| Hi-index | 0.00 |
Land cover changes are producing profound impacts on global biodiversity, terrestrial carbon stocks, soil fertility and erosion. Manually detecting land cover changes using satellite images is a big hassle. Automation of change detection process overcomes the difficulty of manual detection. Automatic change detection methods require the images obtained at different times by satellite, are comparable in terms of radiometric characteristics. Relative Radiometric Normalization (RRN) process is used to prepare multitemporal image data sets for the detection of spectral changes associated with phenomena such as land cover change. A variety of image normalization methods, such as Haze Correction (HC), Minimum-Maximum (MM), Mean-Standard deviation (MS), Pseudo-Invariant Features (PIF), Dark and Bright set (DB), Simple Regression (SR), and No-Change (NC) set determined from scattergrams, are introduced which have been tested either with Landsat TM data, MSS data or both. In this paper, existing methods are tested to adopt for normalizing currently available high-resolution multispectral satellite images on different dates from Resourcesat1 LISS III sensor, which gives drastic change in spatial resolution and difference of available multispectral bands. Some improvements are introduced to get better results. The normalized results are compared in terms of visual inspection and statistical analysis. Key Words: Radiometric Normalization, Satellite Image Preprocessing