Lab 7

Introduction:

The main goal of this laboratory exercise is to develop my skills in performing key photogrammetric tasks on aerial photographs and satellite images. Specifically the lab is designed to train me in understanding the mathematics behind the calculation of photographic scales, measurement of areas and perimeters of features, and calculating relief displacement. Moreover this lab is indented to introduce me to stereoscopy and performing orthorectification on satellite images. At the end of this lab exercise, I will be in a position to perform diverse photogrammetric tasks.

Methods:

                The lab was broken up into three parts. Part one had us calculating scales, measurements, and relief displacement.  We had to calculate the scale of an image by only being given two points on the image and the distance between them. To do this we had to take that distance and the distance on the screen as measured by a ruler and run an equation to find the representative scale. Another image that we had to calculate the scale for was where we were only given the altitude of the aircraft that took the photo, the focal length of the lens, and the elevation of Eau Claire. This was simple division after we converted everything to inches.  We were also introduced to the measurement tools in Erdas and some different ways we can calculate area and perimeter.

                The second part of the lab was creating a stereoscopic images in Erdas. This was very interesting and to complete the task we needed an image and a DEM of that image. The result of the tool we ran was an anaglyph image. The process of making that anaglyph image was just entering the files into the tool and running it. By using polaroid glasses we could view the image as if it was in 3D.

                The final and longest part of the lab was orthoreticifying two images together in Erdas to be used for triangulation and other analysis that requires two images. The roots of the tool that we ran (which is called LPS or Lecia Photogrammetric Suite) is setting it up for the two images that we will input. We had to verify that the images were collected in a polynomial based SPOT pushbroom technique to help figure out the geometry of the tool. Then we added the correct coordinate systems for the images and then began the select point measurement tool. In this tool we have our two different images next to each other with different views of the images that get more and more detailed. We simply had to place one Ground Control Point (GCP) on the first image and then find the same location so that we could place it on the second image. After placing two GCPs we were able to Automatically drive the x,y coordinates from the first image to the second. As we kept adding points, the automatic point generation got closer and more accurate to the first image. We started just entering in the coordinates for the first image and then having the next point driven to the second image of where we verified where the point was.

                The next step was to set the vertical reference source by using a DEM file for Palm Springs. It was the same technique as used in the first step but this time it also updated the Z elevation for the image. This process was very quick and led into the next step which was automatically tieing points together and resampling the triangulation and ortho images. This consisted of many adjustments made to the output images that we were about to create and after they were finished processing we were able to bring them back into an Erdas viewer to see how accurate the images were rectified together.

Results:

Anaglyph image that can be seen in 3D when using polaroid lenses, can you see it?

Final Orthorectification image pair that I created.

 

Summary:

                This lab consisted of many valuable tools that we learned from finding the scale of an image and how to measure it, to making our own 3D image, to orthorectifying images together. These tools can be used in a variety of ways for a wide range of potential careers and by having the ability to say that I can  run these image processing tools, I am able to market myself in a very experienced way.

Sources:

National Agriculture Imagery Program (NAIP) images are from

   United States Department of Agriculture, 2005.

Digital Elevation Model (DEM) for Eau Claire, WI is from United States Department of

   Agriculture Natural Resources Conservation Service, 2010.

Spot satellite images are from

    Erdas Imagine, 2009.

Digital elevation model (DEM) for Palm Spring, CA is from

    Erdas Imagine, 2009.

National Aerial Photography Program (NAPP) 2 meter images are from

    Erdas Imagine, 2009.

 

Lab 6

Goal and Background:

This lab is suppose to expose us to a very important image preprocessing exercise known as geometric correction. The lab is structured to develop our skills on the two major types of geometric correction that are normally performed on satellite images as part of the preprocessing activates prior to the extraction of biophysical and sociocultural information from satellite images.

Methods:

To properly correct the geometry of our images we worked within Erdas Imagine. We set the geometric model (which is a multispectral tool) which required us to bring in our two separate images. From there we accepted the defaults under the Polynomial Model Properties window and started adding our Ground Control Points (GCPs). We entered four for the first pair of images and 12 for the second pair. We needed a minimum of three for the first pair and 10 for the second pair which is determined by what order polynomial they were. Once the points were added we had to go back through and edit the points so that our Root Means Square Error Total was below a predetermined value. That is how we can measure the accuracy of our GCPs in relation to the input image. This was all that needed to be done besides exporting the resampled images as an .IMG so that we could view the geometrically corrected image.

Results:

Our first image that we corrected using 4 GCPs and the 1st order polynomial.

Our second image that we corrected using 12 GCPs and the 3rd order polynomial.

 

 

Conclusion:

Geometric Correction is an essential image preprocessing tool that we need if we are to do further analysis with our images. Without it we would not be able to calculate the brightness values from the different bands that we received from the satellite. By using spatial interpolation we can be very certain that our images are corrected and are ready for a multitude of analysis’s that can now be run.

 

Lab 5

Goal and Background:
The main goal of this lab is to learn basic LIDAR data structure and processing. We will work with processing of various surface and terrain models and creating intensity images and other derivative products from point cloud. This will all be very useful knowledge since LIDAR is really growing exponentially as an subset of remote sensing.

Methods:
To start our lab we first had to import all the LAS data. We first viewed the dataset in ERDAS and then opened ArcMap so we could generate our own LAS dataset. This was an easy task as we just needed to select all the LAS files and import them together. We then ran statistics so we could get our Min Z and Max Z. Then we needed to assign the dataset a XY coordinate system as well as a vertical coordinate system, both of which were in feet. Now we were able to look at a variety of filters such as elevation, aspect, slope, and contour.

We then went to make our DSM and DTM as well as hillshades for both. The DSM and DTM were made through the LAS Dataset to Raster tool. Once we had each of those rasters we could create a hillshade from them that would give us a good interpretation of the surface and terrain from shading.

Now we moved to creating an intensity image. This involved the same steps and tool but this time we switched the field to INTESITY. The output image was really dark so we had to export it as a TIFF so we could view it in ERDAS.

Results:

Hillshade created from the Digital Terrain Model raster.

Intensity image viewed from ERDAS Imagine.

Conclusion:
This lab was a great learning experience as we got our first exposure to LIDAR point data which is a very powerful dataset. From this dataset we can create many different rasters images that we could then analyze even further. Some examples would be Hillshade, Slope, and Aspect. This could then be used as tools when interpreting a landscape or while studying it. The most complex part of all of this is the actual data collection and once that is done and the vendor has went through all the data, LIDAR point data can be used for many many things.

Lab 4

Goal and Background:

                The goal of this lab is to introduce students to the following topics broken up into 7 different parts. Part one was Image subsetting where we created an area of interest using two different methods. Part two was image fusion where we were tasked to create a higher spatial resolution in order to optimize the image for interpretation purposes. Part three we performed techniques to enhance the radiometric quality for our images. Part four introduced us to linking Erdas software with Google Earth. Part five familiarized us with how to resample an images pixels in order to perform analysis. Part six is where we mosaicked images for the first time for ease of viewing. Part seven was the most intricate topic where we took a look at binary change detection of Eau Claire and Chippewa Counties. 

The data for this lab was given to us in several different folders which each pertain to a separate topic. To complete this lab we used Erdas Imagine and ArcGIS software. 

Methods:

            Part 1:

                        To create an image subset we were introduced to two different methods. The first which we worked on in section 1, was by the use of an inquire box. This was relatively easy as we just had to drag the box around our AOI (area of interest) and then use the Subset and Chip tool. This tool then created the subset image. In the second section we brought in a shapefile that we could then use the same Subset and Chip tool to create our subset image.

            Part 2:

                        Now we were introduced to image fusioning. This is done by pan sharpening through the resolution merge tool. We had to import our high resolution panchromatic file and import our multispectral file into the tool. We then resampled using the Nearest Neighbor Technique.

            Part 3:

                        This was really neat, to show how to radiometrically enhance our images, we performed a haze reduction. This was very simple as all we had to do was import the image and then save the output and run the tool.

            Part 4:

                        I thought I knew everything about Google Earth before but now Erdas has recently developed a version of the software where we can bring an image from Erdas and view it in Google Earth to better help us understand what is going on in the photo instead of trying to read the radiometric outputs. This was simple as we just had to link Erdas with Google Earth using the GeoEye Satellite and then sync Google Earth to one of our viewers. We then are able to view them simultaneously just like we would any other two images. 

            Part 5:

                        The first step in resampling was to view the ever important metadata in order to get the original pixel size. We then could go under the Raster tools and then to Resample Pixel Size. We accepted most of the defaults after this and selected the nearest neighbor method. Then we selected the Bilinear Interpolation method afterwards.

            Part 6:

                        To image mosaic we selected our two images and imported them. However, when we imported them we had to make sure that we selected on the Select Layers to Add toolbar that we wanted to add Multiple Images in Virtual Mosaic. We then made the background transparent and repeated the same processes for the next image. Now that we have added the two images we needed used the Mosaic Express for the first mosaic section. For the second section we used MosaicPro which allowed use to compute the Active Area of the image. The next step was to set the overlap function which we kept at Overlay.

            Part 7:

                        The final part of our lab was to check out image differencing or Binary Change Detection. To do this we needed two images from different Temporal Resolutions but same Spectral and Spatial Resolutions. Once we had those imported we accessed the Two Input Operators interface where we choose that we wanted to see the difference between the infrared bands of these images. We then accessed the Histogram to find the mean and standard deviation of which we then added and subtracted to find the upper and lower limits of the change/ no change threshold. Now that we have accomplished this we moved onto mapping the change in pixels by using a spatial modeler. By using the equation “ΔBVijk = BVijk(1) – BVijk(2) + c” we were able to construct a model in Erdas that would show us the binary change in the pixels as long as it was outside our upper and lower limits. Then through a series of other equations we were able to export our data to ArcMap so we could present it in a spatial fashion.

           

Results:

             Part 1:

Section 1: Making an image subset using an Inquire Box

           

Section 2: Making an image subset using a shapefile

  

              Part 2: 

                         The pan sharpened image has better resolution. I wouldn’t say it has much better resolution but it is absolutely better than the input reflective image.You can see individual streets with the pan-sharpened image instead of just a blur of streets with the input image.

             Part 3:

                         

                      While reducing the haze from the input image I noticed that the haze was rather eliminated from the image. Instead of the haze there is now a dark shade wherever the haze was. This area seems to have a little less quality than the surrounding areas but at least the haze is gone.

              Part 4:

 

                        You are able to see in better detail through google earth then you are using another image which allows you to identify more objects. This can make a remote sensing experts job alot easier.

              Part 5:
                        

                         There does not seem to be much different between the two resampled images. The nearest neighbor resampling took some of the other pixels and found pixels around it that were similar and grouped them together. This to me did not seem to change the resolution that much as far as distinguishing features. Im sure it may have made the file size a little smaller though. The bilinear interpolation took the other pixels and broke them into groups of smaller pixels. In some ways, this gives a more detailed view of the image but overall I would say it made the image too busy and did not make it any easier to distinguish features.

              Part 6:

                          For the image mosaic part, we broke it up into two different sections. In section one there is not a smooth color transition between the two images in the output. The input has a relatively smooth transition but the outputs transition is very abrupt. This makes it harder to study along the boundaries.

Section one image mosaic output.

 

                     For section two I would say that the reason for the difference between the two mosaics is that the second one computed the active area which got rid of the overlap between the two and smoothly mosaicked the two together. The actual boundary itself is difficult as the three bands along the boundary are not smoothly transitioned but on either side of the boundary we can see a very smooth transition and can actually see differences between the two images since they are the same color scheme.    

Section two image mosaic output

 

                   Part 7:

         

                               In the Binary Change Detection part of this lab I found that over 20 years the areas that changed are really not that close to urban centers. I am using Eau Claire and Menominee as examples of urban centers and I do not see any change that is relatively close. This surprises me especially with the addition of highway 53 bypass, but I guess that was two small for the sensors to pick up. I assuming the reason why it’s not near urban areas is because it’s the rural areas that have changed so much in the last 20 years and those areas have changed on a large scale so the sensors could pick it up.

 

Final Map of the 20 year change with red depicting change.

 

Histogram showing where the upper and lower limits of change were assigned. The limits are show by the black triangles.

 

Sources:

               Earth Resources Observation and Science Center

               United States Geological Survey. 

               Shapefile is from Mastering ArcGIS 6th edition Database by Maribeth Price, McGraw Hill. 2014.