Lab 4: Miscellaneous Image Functions
Introduction & Goal
The objective of this laboratory exercise is to help students develop skills in Erdas Imagine 2020 software that include image preprocessing, enhancing image spatial resolution, delineate any study area (of interest) from a larger satellite image scene, mosaic multiple image scenes, and construct a simple graphical model for remote sensing analytic use. Eventually, students will be prepared to utilize these skills of image processing, interpretation, delineation, and enhancement with experience upon completing this lab.
Methods & Results
Utilizing the data provided for this lab, the first skill learned is subsetting images to create an area of interest (AOI) of a chosen study area. This is meant to introduce students to use a shapefile to delineate/subset an AOI out of a larger satellite image scene. To create the AOI, we needed to input the TM image of Eau Claire taken in 2011 (.img) into the Erdas Imagine Viewer. To create our area of interest file, we need to input a shapefile of Eau Claire and Chippewa counties to subset the counties. We selected the shapefile and paste the AOI file to the viewer to overlay the TM image of Eau Claire. In order to "crop" the satellite image scene, utilize the tool Raster and Subset & Chip to bring in the subset window. Here we are able to list the input and output file, and voila! A subsetted image of Eau Claire and Chippewa Counties.
Figure 1: Subsetted image of Eau Claire and Chippewa Counties utilizing Raster and Subset & Chip Tool
The next section we touched on was Image Fusion in order to create a higher spatial resolution image from a coarser resolution image. By doing this, it optimizes the image spatial resolution for visual interpretation purposes. The 15 meters panchromatic image was used to 'pan-sharpen' the 30 meters reflective image of Eau Claire and Chippewa counties. The 'pan-sharpen' menu to get to the 'Resolution Merge' tool can be found when activating the Raster toolbar. After inputting the parameters for the image and running the resolution merge model, the 'pan-sharpen' image can be inserted into the viewer.
Figure 2: Utilized the Resolution Merge Model to sharpen the image.
In order to enhance and clarify the image even further, simple radiometric enhancement techniques were used to enhance image spectral and radiometric quality. To do this, activate the Raster tab and select Radiometric --> Haze Reduction in the drop down menu. After accepting the default parameters, the Haze Reduced image can be inserted into the viewer. The image is clearer and it demonstrates how much haze collects on the images and how important it is to utilize these tools to "sharpen" or clean up the images for surveying use.
Figure 3: Haze Reduction tool utilized to clarify the image.
A recent development Erdas Imagine Software is utilizing Google Earth. This procedure can be helpful for collection training data for image classification and these images from a GeoEye high resolution satellite are most recent. By clicking on the Google Earth tab, there are many options, however, Connect to Google Earth, Sync GE to View, and Link GE to view are the main tools to look for. By synchronizing both windows, it is interesting to compare and contrast the corresponding effects on google earth and the image.
Resampling is the process of changing the size of pixels and can be done to reduce or increase the size of pixels depending on the type of analytical need at the time. In this lab, decreasing the size of pixels (resampling up) which has an effect on image clarity. By activating the Raster tab, Spatial and Resample Pixel Size is accessible. For this particular exercise, the output (square) cells were changed from 30x30m to 15x15m. Two resample methods were used to run the model and create an output image: Nearest Neighbor and Bilinear Interpolation. There is a distinct difference between the original image and the resampled image. Naturally, the resampled image looks clear and more detailed with the 15x15 m pixel.
Figure 4:Resampled image utilizing the Bilinear Interpolation and Nearest Neighbor method.
Image mosaic is performed when a study area is larger than the spatial extent of one satellite image scene. In this part of the lab, two adjacent satellite scene images were used to demonstrate this method. By activating the Raster tab, use the Multiple Images in Virtual Mosaic tab. After making sure the background is transparent and the image will fit to frame, the image loads into the viewer. After repeating this process for the next photo, use the Mosaic Express from the Mosaic drop down menu in the Raster toolbox. Insert the images into the program to be mosaicked, and after accepting the parameters, a Mosaic-output image will be ready to load into the viewer. Image mosaic with the use of Mosaic pro is different. This method is more advanced and includes color correcting for a seamless transition between the two satellite images. After adding the images into the program, highlight your image and click the Image Area Options tab and then click Compute Active Area. When the photos look like they've been stacked by hand in the viewer, it is time to synchronize the radiometric properties at rthe area of intersection of both images. Using Histogram Matching for the color correction tool, click on Set to open the dialog box. Then select Overlap Areas, and keep in mind we only want to match the histograms of the images to preserve the brightness values for the other areas of the image. After clicking ok, export the file in the Output Image Options window and accept all paramaters, click on the Set Overlap Function and select "Overlay", then clcik on the Processs tab in the MosaicPro window to Run Mosaic.
Figure 6: Mosaic Pro used to image mosaic and focuses more on the color correction transition between the two satellite images to show a smoother transition.
Binary change detection (or image differencing) is utilized to estimate the brightness values of pixels that has changed in Eau Claire from August 1991 to August 2011 with this equation:
Open Model Maker with Erdas Image by clicking on Toolbox on the top bar. The functions in model maker are simple and include two input raster objects for both 1991 and 2011 images, one function object to calculate the differences between the two images, and one output raster object for the difference image. For the function object, find the mean and the standard deviation in the image histogram to determine the change of the threshold (mean + (3 x standard deviation). After inputting the data, Run the model by clicking the red 'lightning bolt'. Repeat similar steps to then eventually receive an image that demonstrates what brightness values have changed.
Comments
Post a Comment