Component 1 – Remote Sensing

Remote Sensing Pre/Processing

Preprocessing and processing maybe required by the user before an image can be analyzed. The OhioView program supplies scenes that have cloud cover less than 30% and all have been georectified. Vegetation land cover changes from season to season. Therefore land cover’s spectral responses will change according to phenological cycles as well as other factors. For instance, examples include agricultural fields being fallow or dormant during the winter months, grass prairies will be in bloom during the warmer seasonal months, and impervious surfaces will not vary greatly. These details can be used for advanced classifications. For initial classifications a two season multitemporal stacked image has been utilized. An image from early spring and an image from late summer. The stacked image for this study gives a total of 14 bands of data ranging the electromagnetic spectrum.

 

In order to gain maximum detail from the satellite imagery a radiometric enhancement was applied using ERDAS Imagine software v8.6. 

Below are two histograms. The left details the spectral response of a single Landsat image. The histogram on the right details the same pixel spectral response of the multitemporal two season stacked radiometrically enhanced image. 

 

       

 
 

Band Resolutions

 

Landsat 7 Enhanced Thematic Mapper

Band                   Spectral Resolution

   1                                  .45 - .515

   2                                 .525 - .605

   3                                 . .63 - .69

   4                                    .75 - .9

   5                                1.55 – 1.75

   6                                10.4 – 12.5

   7                                2.08 – 2.35
 
Band Resolutions

 

 

 

 

 

 

 

 

 

 

 

 

 

Radiometirc Enhancement

Atmospheric Haze Reduction

For multi-spectral images, this method is based on the

Tasseled Cap transformation which yields a component

 that correlates with haze. This component is removed

 and the image is transformed back into RGB space.

For panchromatic images, an inverse point spread

convolution is used.

 

Right graphic displays enhancement model.

 

 

 

Below - Image on left is not manipulated. Image on right is multitemporal stacked image with radiometric enhancement applied. Notice the different level of detail, this will increase classification capabilities.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 


Remote Sensing Methodology 

The ERDAS Imagine Expert Classifier has two main elements; the Knowledge Engineer and the Knowledge Classifier. The Knowledge Engineer provides methodology for users with advanced information and experience to define variables, rules, and classifying interests to design a hierarchical decision tree and knowledge database. The Knowledge Classifier provides methodology to utilize the knowledge database created by the user and Engineer.

 

Previous attempts at classifying wetland types provides confirmed accurate training sites that can be utilized. Using the inquirer cursor function and signatures editor precise pixel values and signatures can be extracted for an Area Of Interest (AOI). With the hierarchical decision tree a hypothesis can be created with rules defining variables. The Knowledge Engineer feature allows the user to define nearly every aspect of the image.  

 

After a subset of the AOI was complete, the signature editor and inquirer curser tool packages were used. Known wetlands class types could then be investigated and pixel responses examined. Complete detail can be extracted and ranges of spectral responses developed.

 

 

GPS

 

 

 

The image to the left displays GPS locations over an aerial photo within Irwin Prairie in Lucas County. GPS points were collected for groundtruthing and accuracy assessments.

 

 
 
 

 

 

 

 

 

 

 

 

 


Future Directions

To improve on the classified image further adjustments can be made. The Current Agricultural Use Values (CAUV) program details all areas in the county enrolled in the agricultural registration program. Using this data all CAUV areas can be pulled out of the model increasing the accuracy. Additionally images will be stacked for greater detail and other classification techniques will be executed. Additionally training sites will be investigated for complete knowledge database development. Higher end integration with a GIS model will continue for optimal mapping capabilities and a statistically significant analysis shall be conducted. 

 

 

Component 2 - GIS

 

The conceptual model above represents the bulk of the GIS integration on the project. A comprehensive wetlands related database for the entire study area is being constructed. The combination of GIS coverages with the image classifications will provide optimal information. Model simulations and analysis are under development and all model parameters are being investigated. An IMS will be placed online shortly for public access.