TOPOGRAPHIC NORMALIZATION USING HYPERION IMAGE

Mohamad Izwan bin Ghazali 3 SGS

1. Introduction High spatial resolution remote sensing images in rugged areas often have great terrain effects, so besides atmospheric correction, topographic correction is also a necessary step of pre-process in the remote sensing application for rugged areas. developing topographic correction technique of remote sensing imagery has great importance in forestry survey, resources exploration and other interrelated fields. So far, the scientists in the field of remote sensing have developed many methods of topographic correction, such as cosine correction, C correction, Minarets correction (Smith, 1980; Teillet, 1982, 1986; Colby, 1991; Meyer, 1993 et al.). The three methods are STS (sun-surface-sensor) methods, which have an assumption (Scott, 2005) that a pixel in rugged terrain can be considered as a declining plane to handle, and all the illumination correction is based on the assumption. 2. Objective To perform topographic normalization prior to remotely sensed image classification 3. Data This lab exercise will use the following image/ancillary data: a. Hyperion Hyperspectral Image over Pasoh Forest Research b. Topography Map of Pasoh Forest Reserve Image Processing Software : ERDAS 9.2 AUTOCAD map 2012 ENVI 4.5

Page | 1

4. Study Area Study area is area Pasoh, Negeri Sembilan. Hyperion image that being used in this study. This image already done with geometric correction

Figure 1. Hyperion Image

5. Methodology The methodology used in study is as shown in figure 2 Digitizing Using topographic map to create contour using AutoCad

Digital Elevation Model

Develop DEM using ERDAS software

Topographic Normalize Compare and combine Hyperion image with DEM

Figure 2. Flow chart of Methodology

Page | 2

6. Digital Elevation Model Digital Elevation Models are data files that contain the elevation of the terrain over a specified area, usually at a fixed grid interval over the "Bare Earth". The intervals between each of the grid points will always be referenced to some geographical coordinate system. This is usually either latitude-longitude or UTM (Universal Transverse Mercator) coordinate systems. The closer together the grid points are located, the more detailed the information will be in the file. The details of the peaks and valleys in the terrain will be better modeled with small grid spacing than when the grid intervals are very large. Elevations other than at the specific grid point locations are not contained in the file. As a result peak points and valley points not coincident with the grid will not be recorded in the file. Quality DEM products are measured by how accurate the elevation is at each pixel and how accurately the morphology is presented. Several factors are important for quality of DEM-derived products: a. b. c. d. e. f. Terrain roughness Sampling density (elevation data collection method) Grid resolution or pixel size Interpolation algorithm Vertical resolution Terrain analysis algorithm

Common uses of DEMs include: a. b. c. d. e. f. g. Extracting terrain parameters Modeling water flow or mass movement (for example, landslides) Creation of relief maps Rendering of 3D visualizations Creation of physical models (including raised-relief maps) Rectification of aerial photography or satellite imagery Reduction (terrain correction) of gravity measurements (gravimetry, physical geodesy) h. Terrain analyses in geomorphology and physical geography 7. Topographic Normalization The correction of illumination variations is referred to as Topographic Normalization or Topographic Correction. Techniques are grouped into two major categories, (1) band ratios and (2) modeling of illumination conditions. Band ratio techniques assume that the spectral response is distorted (increased or decreased) in the same way across all bands. Therefore, the quotient between them will compensate for topographic effects. Their drawback is the loss of spectral resolution. Techniques under group (2) model illumination to compute the flat-normalized Page | 3

radiance of each pixel. They are grouped into two additional subcategories, Lambertian and non-Lambertian, depending on whether they assume a Lambertian or nonlambertian surface behavior. A Lambertian surface is presumed to be a perfectly diffuse reflector, appearing equally bright from all viewing directions (Ekstrand, 1996). Therefore a Lambertian surface is presumed to be a perfectly dif- fuse reflector, appearing equally bright from all viewing directions (Ekstrand, 1996). Therefore, a Lambertian correction function attempts to correct only for differences in illumination caused by the orientation of the surface (Jones et al., 1988).The incidence angle (i) is the angle between the surface normal and the solar beam, and may be calculated at equation 1 (Smith et al., 1980). When the sun is not zenith, correction of the radiance of a projected horizontal surface would be achieved by the equation.

Page | 4

8. Result and analysis

Contour DEM (Digital Elevation Model)

Page | 5

Topograpic Nomalization

Classification Unsupervised (K-Mean) Hyperion Image Topographic nomalization

Using topograpic normalization (lambertian model) shown more accurate classification before topograpic normalization.

Page | 6

9. Conclusion After topograpic normalization, combination between DEM and classification, we can make land cover map more accurate because this technique can cover hill area because when satellite take some image not covered by sun light, so DN can appear dark on image. Using this technique can predict DN for lost data on image

Reference Comparison of Lambertian and non-Lambertian topographic normalization algorithms: A case study in Gelibolu, Turkey , Mufit Cetin A Model Of Topographic Correction And Reflectance Retrieval For Optical Satellite Data In Forested Areas, Cheng Wei, Tian Qingjiu, Wang Liming Influence of different topographic correction strategieson mountain vegetation classification accuracy in the Lancang Watershed, China. Zhiming Zhang, Robert De Wulf, Lieven P. C. Verbeke, Frieke M. B. Van Coillie, Eva M. De Clercq, and Xiaokun Ou

Page | 7