LABORATORY - ANALYSIS OF SCENE SPECTRAL REFLECTANCES AND VEGETATION INDEX
Assignment
3.1 - Vegetation indices:
(a) Generate NDVI images from the CASI images on both dates as follows: (1) the standard NDVI using reflectance in one band in the near-infrared region and one in the red region; (2) the modified NDVI using reflectance along the red edge, specifically one band near 750 nm and one near 710 nm. The band and corresponding wavelength information can be found in the spectral data excel file.
Tools in PCI Geomatica: Tools -> Algorithm library -> Image Processing -> Image Operations -> ARI: Image Channel Arithmetic. In the "Input Params 1" panel, set "Autoscaling mode" to OFF.
(b) Set a reflectance threshold to classify the standard NDVI images into "vegetation", "non-vegetation" classes. What's your expectation and how well does the binary classification result consistent with your expectation? Give examples to justify your answer. You may highlight some area in the image and show them in the .word report.
Tools in PCI Geomatica: Tools -> Algorithm library -> Image Processing -> Image Operations -> THR: Thresholding Image to Bitmap. In the "Input Params 1" panel, set "Threshold Maximum" as your selected reflectance threshold.
(c) Comment on any similarity and difference between the standard NDVI and modified NDVI images derived from the data on both dates (i.e. for June 13 data, compare the standard NDVI and modified NDVI images; and for June 26 data, compare the standard NDVI and modified NDVI images). For example, comment on which NDVI image provides more variations within vegetation pixels etc.
3.2 - Reflectance spectra of targets derived from the IFC-2 June 26 CASI image:
(a) Which target is characterized by the greatest spatial variability? Suggest some potential reasons for why this target exhibit significant spatial variability in reflectance? (Hint: consider the general standard derivation values for each target)
(b) Focus only on the targets #1, #2 and #4. If this image was acquired with a monochrome camera working only to reflectance at 544 nm, and the standard deviations of your spectral data apply, which of the targets could be distinguished and which could not. Explain your reason(s). Can the targets be distinguished if the monochrome camera was working to reflectance at 785nm?
(c) One common technique in spectral classification in image processing is to use cluster analysis of the n-dimensional spectral space. Using some statistical measure one can then determine whether targets are statistically separable using any number of bands. A simple visualization of this is to plot the reflectance of the target at one wavelength versus the reflectance at another wavelength and use the standard deviation at each of the wavelengths to determine whether the uncertainty in the two reflectance's for a specific target make them separable or not. Choose reflectance at ~680nm and reflectance at ~780nm and plot all the 6 targets and determine which is/are separable using these two wavelengths for separation.
(d) Broadly classify the cover types of each target (vegetation and non-vegetation) based on the observed spectra. Provide reasons for your selection (i.e. specific spectral features).
3.3 - Reflectance spectra of targets derived from both IFC-1 June 13 and IFC-2 June 26 CASI image
(a) Plot the reflectance spectra on June 13 and June 26 together (i.e., within a figure) for each target and comment on the seasonal change of this target.
(b) Based on the changes in the reflectance spectra and the growth cycle of the three crops (corn, spring wheat, and soybean), identify the vegetation targets in IFC-2.
Attachment:- Assignment Files.rar