Overview:
The data collected by a UAS goes well beyond a pretty picture of what your house looks like from above. If gathered correctly, with the right sensor for the right job, your UAS data is the proverbial oyster waiting to be opened for you to exploit the value, or the proverbial sponge waiting to have every last drop of value squeezed out. It is what the operator and analyst know in terms of limits and potential that matters.
For lab this week, I made use of an online tutorial that walked me through classifying an aerial image to determine surface types. I was instructed to complete an ESRI tutorial and write down the appropriate steps it takes to classify surfaces and calculate surface area. In a later lab, I will apply what I did here to some multi-band UAS data.
Deliverables
Complete the ESRI tutorial 'Calculate Impervious Surfaces from Spectral Imagery'.
Create a document that provides step by step instructions on how to engage in object-based classification.
Create a proper map with all the needed cartographic elements.
Step by Step Object Based Classification:
Section 1: Extract Spectral Bands
1. In contents pane, select the .tif file layer you are using.
2. On the ribbon, go to the imagery tab. In the analysis group, select raster functions.
3. In raster functions pane, search for “extract bands” and select it.
4. In parameters tab for raster select your .tif file and confirm method is set to “band IDs”.
5. For combination delete existing text and use 4 1 3 (with spaces) or other combo as required. Missing band action set to best match.
6. In the general tab use custom name.
7. Click create new layer.
Section 2: Configure the classification Wizard
1. In the contents pane, select the extracted bands layer.
2. On the ribbon, on the imagery tab, in the image classification group, select the classification wizard.
3. Make sure classification method is set to supervised and the classification type is set to object based.
4. For classification schema click the drop down and select “use default schema”
5. Confirm the output location is set to your database
6. Click next
Section 3: Segment the image
1. For spectral detail replace the default value to 8 or other as required.
2. For spatial detail, replace the default with 2 or required.
3. For minimum segment size in pixels confirm the value is set to 20.
4. Confirm show segment boundaries only is unchecked.
5. Click next.
6. In contents pane, right click the preview segmented layer and choose zoom to layer.
7. Save project.
Section 4: Classify the imagery and create training samples
1. In the image classification wizard pane right click each of the default classes and click remove class.
2. Right click “NLCD2011” or the name of the data you are using and select add new class.
3. In add new class window, for name make it “impervious”. For value, type 20 and for color choose gray 30% or whatever other color you want. Click ok.
4. Right-click the same data base (NLCD2011) and add new class but this time name it pervious with a value of 40 and a color of quetzal green. Other other parameters you’d like.
5. Right-click the impervious parent class and add new class name gray roofs (or whatever other type of surface you want to classify) give it a value of 21 and a color of your choice. Click ok.
6. Time to train the data. Click the gray roofs class and click the polygon button
7. Draw a polygon over the surface you want to classify as gray roofs. Double click to finish the polygon.
8. Add more samples to the class.
9. Click the first row of sample and press shift and select all of the samples.
10. Above the list click the collapse button (looks like a merge sign).
11. Right click impervious and pervious to add new subclasses to the parent classes.
12. Add training samples to the new subclasses created and follow the same steps as above.
13. Once you have all the training samples collected click the save button at the top of the training samples manager pane.
14. Click next
Section 5: Classify the Image
1. In the image classification wizard pane, in the train page of the wizard, confirm classifier is set to support vector machine.
2. For maximum number of samples per class, type 0.
3. Click run.
4. If satisfied with the preview click next. To the Classify page.
5. For output classified dataset change the name to “classified_[name of choice]”
6. Leave other parameters unchanged and click run. To the merge classes page.
7. For each class in the new class column, choose either pervious or impervious.
8. Click next.
Section 6: Reclassify errors
1. The final page is the reclassifier page.
2. On the ribbon click appearance, use the swipe tool to compare the original image to the preview class.
3. In the wizard click reclassify within a region.
4. In the remap classes section, confirm that current class is set to Any. For new class, choose pervious or the class you’d like to change an area to.
5. Drag a polygon around the portion of a region you’d like to change. It will automatically change to the class of your choice you selected above.
6. Reclassify the image as needed.
7. Zoom to the full extent of the data.
8. In the image classification wizard, for final classified dataset type the name you’d like and include the .tif extension.
9. Click run and when it completes click finish.
10. Save project.
Results:
Fig. 1: Louisville neighborhood image classification map layout created with the ESRI tutorial.
Comments