Aerial Image Processing; Problems and Successes

Remote Sensing of Vegetation and Soil

Chris Banman (Dec. 2001)

Abstract

Vegetation indices have been in use for many years and are of increasing importance in the field of remote sensing. This project discusses how plants and soil respond to contact with light energy and how images associated with remotely sensed soil and vegetation can be quantified and viewed using the Ratio Vegetation Index (RVI). Several 4 band images stored in ERDAS Imagine .img format were analyzed utilizing several different Geographic Information Software (GIS) and Geographic Image Processing (GIP) packages to eventually arrive at the final result. Agricultural fields and surrounding features were analyzed to compare the ratio of the Near Infrared Reflectance (NIR) to the Visible Red Reflectance at different times during the year. As a general characteristic, active vegetation will absorb a high percentage of the available visible red wavelengths which are used in the photosynthesis process, while they will tend to reflect most of the harmful NIR radiation. Soil, water, and manmade features tend to have a more static response to the above wavelengths bands throughout the year.

Contents

Soil and Vegetation Characteristics

Soil has properties which cause the ratio of Near-Infrared and Visible Red reflectance to plot in a linear fashion. For very homogenous soils, this line may be very short and for soils that vary in type and moisture content, this line may be stretched significantly more. This means that as light energy comes into contact with the individual soil particles, each particle will reflect a proportionately similar amount of Near-Infrared and Red wavelengths. The following illustration shows how the soil line of a bare field may plot when looking at the reflectance values of NIR vs. Red.


Vegetation also has distinct properties when exposed to light energy. The reflective and absorptive properties of vegetation with respect to red and near-infrared radiation are much different then that of soil. Active vegetation uses light energy available in the atmosphere as an ingredient in the photosynthesis process. The plant uses water from the soil, CO2 from the air, and light energy to produce sugar molecules used by the plant. The by-product is oxygen molecules which are released into the air. For this process to take place, certain wavelengths of light are absorbed by the plant and used as fuel for the process. As light strikes the leafy surfaces of the plant three things can occur. The light may either be reflected back into the atmosphere, it may be absorbed by the plant, or it may be transmitted through the plant onto other lower lying surfaces. In most cases a combination of all three things happen. The light that is transmitted through the leaf will then come into contact with whatever lies beneath it. If it is another leaf, a combination of the above three circumstances will again occur. With this in mind, it is easy to see how light energy that is not reflected but transmitted by the first leaf layer may eventually be reflected back into the atmosphere by lower leaf levels. In the same fashion, some wavelengths that are transmitted through the first layer may eventually be absorbed by lower leaf layers. As the biomass of the vegetation increases over the soil, the amount of light reaching the soil decreases. Intense radiation can be harmful to vegetation and therefore plants have adapted to absorb only what is necessary, and reflect the longer more harmful rays (Jensen, 2000). As the vegetation biomass increases, more and more visible red light is absorbed by the plants for photosynthetic purposes, while more and more near-infrared energy that can be damaging to the plant is reflected.

Numerous indices have been developed to quantify how vegetation responds to light energy as it grows. In fact there over 50 vegetation indices that exist. The first developed of these indicies is the Ratio Vegetation Index (RVI) and is over 30 years old (Leblon). The RVI is defined as: RVI = NIR/red and simply divides the near infrared reflectance values by the visible red reflectance values. The below illustration shows in green where pixels representing active vegetation may fall on a scatter-plot using the RVI.

Pixels falling at the peak of the curve indicate the most dense canopy, while pixels near the soil line indicate little to no present vegetation. All areas between indicate a combination of soil and vegetation. Using a ratio such as this puts the data in a format that is easier to view and also removes factors which affect both band widths equally. Ideally the soil and vegetation in the index should be easily distinguishable. In most vegetation indices, visible red, and near infrared are used. These are chosen because soil can be readily distinguished from vegetation, visible red corresponds to chlorophyll light absorption, and the NIR is a good indicator of vegetative biomass (Leblon). In short, these two bands contain the most significant information dealing with the plant canopy.

Aerial Imagery and Processing

Several four band aerial photographs were processed and analyzed using the RVI. The photos were taken over specific agricultural fields in southwestern Iowa at different times over a two year period in an effort to get accurate and quality remotely sensed images. The images were georefrenced and stored in ERDAS.img format. For the purpose of this project, the images were brought into ArcView, presented as different single band images and compositions, and exported as JPEG files for further processing. Using various GIP softwares the images were cropped, resized, and converted for processing in Idrisi32. Once the selected portion of each image was cropped they were converted to 8-bit format. This assigns each separate pixel a discrete value in the range from 0-255 based on the original reflectance intensity. Images were then imported into Idrisi as JPEG files and converted to raster files. In order to export the data associated with each pixel location and corresponding reflectance intensity value, the file had to be further converted to a vector file. At this point the data could be exported as a table in ACSII XYZ format. This provides each pixel x and y location along with the value (scale from 0 - 255) associated with it. This process was done for various photos in both the red and NIR band widths. 10,000 pixel values were examined and plotted for each image. Click on the links below to view the processed images and scatter plots.

Year 1 Image 1

Year 1 Image 2

Year 2 Image 1

Year 2 Image 2

Year 2 Image 3

Discussion

The area analyzed in the first year's images shows primarily a homogenous portion of one field as well as small sections of road and an adjacent field. From the natural color image of Year 1 Image 2, areas of standing water and developing vegetation can be seen. By importing the pixel values into Microsoft Excel and plotting them on a scatter plot, the various features seen in the image can be distinguished by their ratio of NIR to Red.

You will notice in the Year 1 Image 2, there is a great deal of saturation at the high end of the sensor values (255) for both the red and infrared reflectance. This phenomenon is likely a combination of extremely high reflectance of some objects in the scene (road, water, and vegetation in adjacent field) as well as the various processing steps used to get the data into spreadsheet format.

A different section was chosen for analysis in the images from the second year. Here nearly equivalent portions of two separate adjacent fields were analyzed. In year 2, it is evident that different crops were planted in the two fields. From the initial photo (Year 2 Image 1), a nice soil line was plotted for both fields. Later in the growing season the northern field had increased vegetative growth earlier than the southern field. This is easily seen in the false color composite image as well as on the scatter plot of the two fields. Finally, towards the end of the growing season, the active vegetation in the northern field had died off while vegetation in the southern field was growing.

Using the ratio vegetation index gives general information about the vegetative properties that are occurring in a given area. From this relatively simple index, many other indices have been created to give more detailed and organized information regarding the vegetative vigor of areas from remotely sensed imagery. Using the values derived from the RVI, most features present it the various scenes could be accurately identified as well as a general indication of the vegetation biomass occuring in certain areas. Appling information acquired from the RVI in other more specific vegetation indices may provide more detailed information specific to vegetative health and growth stage.

References

Anonymous. How a Vegetation Index Works. URL
http://www.uswcl.ars.ag.gov/epd/remsen/vi/VIWorks.htm.

Jensen, J.R., 2000. Remote sensing of the Environment, An Earth Resource Perspective. Chapter 9. Prentice-Hall. Upper Saddle River, NJ.

Leblon, B. Soil and Vegetation Optical Properties. The Remote Sensing Core Curriculum. URL
http://research.umbc.edu/~tbenja1/leblon/module9.html.

Vonder, O. 1998. Vegetation - Applications of present and future optical remote sensing satellite sensors. URL
http://137.224.135.82/cgi/projects/bcrs/multisensor/report1/4.htm#s4_2.

This project was created for Remote Sensing (ES771 Remote Sensing) completed at Emporia State University instructed by Dr. James Aber.