Create Vegetation Products from Hyperspectral Imagery

来源: https://vis-webcontent.s3.amazonaws.com/quickguides/Spectral+Analysis/Create+Vegetation+Products/index.html

Lesson content

Lesson 1 of 1

Create Vegetation Products From Hyperspectral Imagery

In this quick guide, you will use the Vegetation Analysis Workflow to:

•

Mask out non-vegetation features. * •

Create a classification image showing the extent and vigor of vegetation. * •

Create a classification image showing the distribution of crop stress.

Open an AVIRIS Hyperspectral Image

1

Select File > Open from the Menu bar. An Open dialog appears. 2. 2

Go to the "hyperspectral" directory in your ENVI installation path.

Windows: C:\Program Files\NV5\ENVIxx\data\hyperspectral (xx is the version number)
Linux: /user/local/NV5/envixx/data/hyperspectral
Mac: /Applications/NV5/envixx/data/hyperspectral 3. 3

Select the file AVIRISReflectanceSubset.dat and click Open. This is an AVIRIS hyperspectral image of an agricultural area in central California.

This image was corrected for atmospheric effects using the FLAASH® tool in ENVI. Pixel values represent surface reflectance; they range from 0.0 to 1.0. This is ideally how images should be preprocessed(opens in a new tab) before creating narrowband spectral indices. Refer to the Mitigate Atmospheric Effects and Scale and Mask Reflectance Images quick guides for more information.

The Vegetation Analysis Workflow

1

In the Toolbox, expand the Workflow toolbox and double-click on the Vegetation Analysis Workflow. The workflow begins with the Select Data panel. The Input Raster field is populated with the AVIRIS file name.

2

Click the Next button to proceed to the Mask from Spectral Index panel.

Mask Out Non-Vegetation Features

The next two steps of the workflow guide you through the process of masking out non-vegetation features. Examples include water, roads, man-made objects, bare soil, and fallowed fields. In the Mask from Spectral Index panel, you can choose any vegetation index that highlights vegetation in general. In the subsequent "thresholding" step, you interactively set a threshold on the VI image. Pixel values below the set threshold will be masked out and set to values of "No Data."

1

In the Index drop-down list, keep the default selection of Normalized Difference Vegetation Index. 2. 2

Enable the Preview option. An NDVI image is displayed. It effectively highlights a wide range of different vegetation. Most of the non-vegetation features are very dark. Thus, it will be relatively simple to set a threshold on this NDVI image.

3

Click the Next button to proceed to the Threshold Normalized Difference Vegetation Index panel. ENVI automatically determines a suitable threshold. In this case, the threshold value is 0.418. Pixels above this value are colored red in the NDVI image and histogram. Pixels not highlighted in red will be masked out.

With the current threshold value, some agricultural fields will be omitted. Let's reduce the threshold to include more of these fields.

4

Click and drag the red vertical bar in the histogram toward the left, to a value of 0.31. Look for the red "Data" value above the histogram.More fields are highlighted in red.

This threshold does a reasonable job of omitting dormant crops, roads, and man-made objects.

5

Click the Next button. ENVI creates and displays a "Mask_Raster" layer. This is a version of the AVIRIS image with non-vegetation features masked out and set to values of "No Data." 2. 6

In the Layer Manager, uncheck the Binary_Greater*.* and AVIRISReflectanceSubset.dat layers to hide them. Only the "Mask_Raster" layer is displayed. White areas indicate "No Data." These pixels will not be used in subsequent steps.

Next, you will create some vegetation products.

Create a Vegetation Delineation Image

The Vegetation Analysis Product panel of the workflow is displayed. The Method drop-down list lets you select what vegetation product to create. The default selection is Vegetation Delineation. The Index drop-down list lets you select a VI for determining vegetation delineation. The default selection is Normalized Difference Vegetation Index.

1

Enable the Preview option. A preview of the vegetation delineation image is displayed. It is based on NDVI.

The Class Colors/Ranges/Names fields are populated with a default set of classes. Each class corresponds to a range of NDVI values, shown as arrays in the Class Ranges field. The following table summarizes the classes:

Class Name	Class Range (NDVI Value)	Class Color
No Veg	-1 to 0.25	Brown
Sparse Veg	0.25 to 0.5	Yellow
Moderate Veg	0.5 to 0.7	Light green
Dense Veg	0.7 to 1	Dark green

You can optionally define your own classes and colors, although you do not have to do so for this quick guide.

2

Knowing that NDVI values can become saturated as Leaf Area Index (LAI) increases, you could experiment with a different VI. Since you are using a hyperspectral image, click the Index drop-down list and select Red Edge Normalized Difference Vegetation Index. This VI differs from the NDVI by using bands along the red edge, instead of the main absorption and reflectance peaks. It is more effective at detecting vegetation stress than NDVI. In this example, it does not show nearly as much moderate and dense vegetation as NDVI.

When you are satisfied with the results, you can optionally click the Next button to create an output raster and finish the workflow. For this quick guide, however, you will create another vegetation product without leaving the workflow.

Create an Agricultural Stress Image

Next, you will create a classification image that shows the distribution of crop stress. The Agricultural Stress Classification option is intended specifically for use on agricultural land with a focus on growth efficiency. Dry or dying crops do not efficiently use nitrogen and light, which manifest as "high stress," whereas a crop showing healthy, productive vegetation indicates low stress.

1

Click the Method drop-down list and select Agricultural Stress Classification. 2. 2

Enable the Preview option. A preview of the agricultural stress classification image is displayed. The Layer Manager lists the classes and their colors. They cannot be modified.

Agricultural stress classification is determined by three categories of VIs. Click the links below to read more about each type in the ENVI Help.

•

Broadband(opens in a new tab) and narrowband(opens in a new tab) greenness * •

Canopy water(opens in a new tab) or nitrogen(opens in a new tab) content * •

Light use efficiency(opens in a new tab) or leaf pigments(opens in a new tab)

3

Set the Minimum Valid Greenness Value to 0.2. 2. 4

Experiment with different VIs in the Greeness Index, Canopy Water or Nitrogen Index, and Light Use Efficiency or Leaf Pigment Index drop-down lists.

Analyze the results with some caution. Results from different crop types are not directly comparable. Because different crop types have different phenological cycles, canopy closure characteristics, and vegetative properties, the signals within and between crop types may be confused. Ideally, you should compare the results using the same crop type from different images. Another option is to create a mask to separate crop varieties, and process each crop type individually.

5

Click the Method drop-down list and notice two additional options that are meant for forests and other non-agricultural areas:

•

Fire Fuel Classification: Fire fuel mapping(opens in a new tab) can be useful for forest planners, as well as local governments attempting to mitigate fire risks within the rapidly growing forest/urban interface. High fire fuel distributions contain dry or dying plant material, which contains less water, whereas low fire fuel distributions typically consist of lush, green plants. * •

Forest Health: Forest health mapping(opens in a new tab) is useful for detecting pest and blight conditions in a forest, and it is useful in assessing areas of timber harvest. A forest exhibiting low stress conditions is usually made up of healthy vegetation, whereas a forest under high stress conditions shows signs of dry or dying plant material, very dense or sparse canopy, and inefficient light use.

6

Optional: Click the Next button to proceed to the Export Final Result panel. Here, you can create an output classification image for later use. 2. 7

This concludes the exercise.

In summary, both multispectral and hyperspectral images can be used to create Vegetation Delineation products. However, Agricultural Stress Classification, Fire Fuel Classification, and Forest Health Classification use narrowband VIs, so they are best used with hyperspectral images. The high spectral resolution of imaging spectrometers allows specific wavelengths to be leveraged when analyzing vegetation properties such as greenness, leaf pigments, canopy water content, and light use efficiency.

Your input is important to us, please take a few moments to fill out ourQuick Guide Feedback(opens in a new tab)form.