Image Classification Toolbox

Features:

• Performs Classification using several algorithms based on the selected seed areas.

• Quantifies the area and other parameters.

Steps:

1.      Open a hyperspectral file through the Main Interface.

2.      Navigate to Toolboxes → Classifications. A new window will pop-up

3.      Select an algorithm from the ALGORITHM SELECTION panel. Currently available options are:

a.      Spectral Matching - supports only one class.

b.      Spectral Angle Mapping (default) - supports several classes (areas of interest).

Spectral Angle Mapping 

4.      Select Spectral Angle Mapping.

5.      Click Add Class and select a region of interest. Click Classify to perform classification. Adjust the output by specifying Threshold Parameters Alpha and Beta Values.

Tip: Start by changing the Beta Value 0.5 units up and down. Click Classify to perform mapping.

You can select many regions by clicking Add Class again and selecting another region.

Note: the colors in the classified image are assigned randomly. To change the colors click Classify again without changing any other parameters. If you need to select a different region or add another class after you performed classification, click Reset and repeat the steps.

To copy results to Excel, click Export Data to Excel. Alternatively, highlight the results with Ctrl+A and copy (Ctrl+C). Paste the results to Excel or other software.

Spectral Matching

1.      Select Spectral Matching (only one class can be selected).

2.      Click Add Class and select a region of interest. Click Classify to perform classification. A new pop-up window along with a dialog will appear. The left image will show a new map where each pixel is measured against the average spectrum from the selected region of interest. The lower Score (dark blue in the default color scheme) corresponds to the higher similarity. The plot on the right shows the histogram Number of Pixels vs. Score.

3.      From the histogram, identify the first feature in the histogram, (i.e., the first peak) and put the value into the dialog window. Click OK to visualize the resulting image to the right. The segmented image can be viewed in different look-up tables (LUTs) that could be selected from the Color Toolbar menu. The color schemes show the similarity between objects of the image but in different colors. Very close (matched) objects will have the same colors, slightly different will be colored differently. 

4.      To copy results to Excel, click Export Data to Excel. Alternatively, select the results with Ctrl+A and copy (Ctrl+C), then paste the results to Excel or other software.

Additional Information: