Lab Instructions
Lab 5, due 8:00 am October 7, 2006
EES5053: Remote Sensing, Earth and Environmental Science, UTSA
Student Name: ______
Digital Image Processing III: Supervised Classification
Objective: In this lab, you will compute all bands’ radiance, stack individual bands together, and classify the image using supervised classification. Finally, you will use the classified image to examine the flooding area in 2002 summer flooding events using the landsat images on two dates in the Gonzales County (at the Guadalupe River Basin).
1. Preparation:
(1). Create your own folders:
c:\EES5053\studentname\Lab5, storing your work and final lab report.
(2). Connect to the server
\\129.115.25.240\XIE_misc\EES5053\Lab5
(3). Setup the ENVI data and output directory for lab5 as we did in last lab. Load Mask_DOC_020708.tif image as RGB742.
2. Background
On June 30 - July 8, 2002, a tropic storm hit the central and south-central parts of Texas. During the 8 days, the storm fell as much as 35 inches of rainfall, with heaviest depths occurring in the Texas Hill Country northwest of San Antonio. The floods caused twelve deaths and damage to about 48,000 homes. Nearly 250 flood rescue calls were reported, more than 130 roads were closed, and thousands of homes and businesses lost electrical power and telephone service.
Because of the cloud impact, this lab use the subset of two-date images, one is just after the flooding, July 8 2002, and the other is on the day when the flooding completely retreated on Nov 13, 2002. Through comparing the water body on two dates through supervised classification, we can calculate the flooding area in the Gonzales County of the Guadalupe River Basin. Meanwhile, we can also compute other land cover change during the two dates, such as forest, non-forest vegetation, and bare land, etc. Because of the boundary impact, this method is not an accurate method for patchy land cover change, like deforestation, etc. As to the large flooding area, this method is reliable to some degree. If you are interested, you can examine the entire flooding area in the San Antonio River or other river Basin, even in the entire state, for your class project. I am very happy to help you to make further improvement.
3. Calculate the Radiance
In the lab 4, you calculate the radiance and reflectance of band 3 and band 4. In this lab, you are required to compute the radiance of band 1, band 2 and band 7 using band math tool based on equation (1), table 1, and fig 1 as you did in lab 4. Output to lab5 and name them as Radn_189_b1.tif, Radn_189_b2.tif, Radn_189_b7.tif, respectively.
(1)
Table 1 Detector gain offset Abs Calib?
------
1 | 15 0.775686 -6.20000 FALSE
2 | 12 0.795686 -6.39999 FALSE
3 | 8 0.619216 -5.00000 FALSE
4 | 7 0.965490 -5.10001 FALSE
5 | 14 0.125725 -0.99999 FALSE
6 | 8 0.066823 0.000000 FALSE
7 | 10 0.043726 -0.35000 FALSE
8 | 27 0.971765 -4.70000 FALSE
9 | 8 0.037059 3.200000 FALSE
Figure 1. Compute Radiance
4. Stack layers
After you computed the radiance, they are in individual bands, and we must stack them together as one image file for the image classification.
Close all files on the Available band list, and then open radiance that you computed in the order of band7, band 5, band 3, band2, and band 1.
Click Basic Tools->Layer Stacking->Import file,
Select the five radiance bands (make sure that the order is b1, b2, b3, b4, and b7 from the top to bottom), click ok. (if the radiance b3 and b4 have not opened yet, you need to open them from your lab4 directory.
Select the 5 bands again as in figure 2.
Output the result to file in lab5 as Radn_020708.tif. Click ok.
All the five band radiance has been stacked together in the order of b1 to b7. Later, we will use the radiance to classify the image.
Close all files on the Available band list, open image Mask_DOC_021113.tif as RGB742. I already made the atmospheric correction and cloud mask for this image.
Do the same operations (compute radiance and stack radiance together) for image Mask_DOC_021113.tif. Save individual radiance band as Radn_317_b(1-4, 7).tif. Save the stacked radiance as Radn_021113.tif.
Figure.2 Stacking layers
5. Training sites
In this lab, we will make four training site, water, forest, nonforest-vegetation, and bare land using ROI. Each type of training site must contain as much as possible polygons representing different situations. The final classification will have 5 classes, one for unclassified.
Open image Radn_020708.tif as RGB742. Click Basic tools->Region of Interest (ROI)->ROI Tool. First training water body (red), then draw forest on the dark green pixels (green), then draw nonforest vegetation (blue), finally draw bare land training site (yellow). When you draw the training site, each class must include at least 20 polygons under different condition in the entire image. Then save the ROI as train_site_020708.roi.
6. Supervised Classification
Click Classification->Supervised->Maximum Likelihood. Then select Radn_020708.tif as the input file. Select the four classes of ROI you trained in the 5th step (see figure 3). Input Probability Threshold as 0.1, then output the result to Memory. Then click Ok.
After it’s done, upload the Class memory to a new window. Then link it with RGB image and compare your classified image with raw RGB742 image. If you are not satisfied with your image, you can modify your training site or change your probability threshold and then re- do the classification until it seems good to you. Finally save the final class results and compute the statistic of the class image.
7. Repeat the step 5 and 7 for image Radn_021113.tif.
Questions (1): Show part of both class images (including Legend/Mapkey) like figure 4. You also can modify the color for different classes.
Questions (2): Compute statistic results of classified images, and compute the area of each class (note: area=pixel number * 30*30 m2).
Questions (3): Compare the area/pixels of four classes: water body, forest, nonforest-vegetation, and bare land, explain the land cover change, in particular the water body change, which is the flooded area in Gonzales County.
Figure 3 Supervised classification.
Figure 4.
1