Setting up for Image Processing

Scan the milkweed leaves on a flat bed scanner with a white background. A piece of plain white paper works well. Save the scanned image as a .TIF image. Make sure to keep track of the leaves and which plant they come from.

Open ArcView GIS and load the Image Analysis extension (accessible from the pulldown menu: File -> Extensions. Be sure to check mark the "Image Analysis" option. Press "ok".)

With a blank view, press the "Add data" button. Navigate to the folder where your .TIF images are located. Reset your data type to: Image Analysis Data (not Image Data). Select your image and press "ok".

Display the leaf image by pressing the small gray box near the image title.

Make the leaf image active, by pressing once on its title. The leaf title box will appear slightly embossed.

Identify the Damage

We'll now make a new data layer, by asking the GIS to create a new classification based upon the color variations in the image. Go to the "Image Analysis" pulldown menu and select the option "Categorize". A small window will open asking how many categories (of color variation) you would like. We suggest using 5 classes. Be sure to press "ok".

The computer will create a new data layer, taking the existing colors of your milkweed leaf and dividing them into the five most significant colors/hues. In most cases, damaged and non-damaged leaf regions will be assigned a different color class. Background colors and other variations will be assigned the remaining three classes (these are to be ignored). If your leaf damage is not separated from the undamaged regions, re-run the "Categorization" and increase the number of color classes designated. Be sure to turn on the newly created data layer, by pressing the small gray box near the categorization title.

Some Milkweed plants may have insect damage (evidenced by holes in the leaf), be sure this area is not counted in the Damaged or Non-damaged class. We are only interested in effects from Ground Level Ozone. Similarly, Milkweed occasionally has damage due to fungus. Typically, the fungus is a substantially different color than damage due to Ozone. If the fungus is grouped into either the Damaged or Non-damaged class, simply re-run the Categorization, making sure to set more classes. In essence, this will separate the fungus by providing it's own color classification. You may, inadvertently, break the Non-damaged colors into several classes. Be sure to account for this when computing percentage area as described below.

When processing images, it's important to understand that these pictures are represented by hundreds, if not millions of tiny digital "paint buckets" set side by side. These paint buckets are called pixels and can each contain only one color. To compute the damage, all we need to do is count the number of total paint buckets (pixels) representing the leaf and divide by the number of paint buckets (pixels) representing damaged areas to the leaf. To calculate the percent, we will divide:

Percent damage = damage pixel count / total leaf pixels

More importantly, by looking in the column called count, you will find the number of pixels that fall into the various classes of Common Milkweed. Add your Damaged and Non-damaged class counts together to get the total number of pixels in the milkweed leaf (such as 50822 + 14075 = 64897). Note the number of pixels in the non-damaged class.

Divide the number of pixels in the non-damaged class by the total number

(50822/64897 = 78.3% damage)

Repeat this for every leaf from each milkweed plant. Calculate the average ozone damage for each milkweed plant in the sample. Enter the average percent damage for each of the 10 milkweed plants on the data submission page for this project.