Obliques and Merge for 3D model Construction

Introduction

This weeks lab is to work with oblique imagery.  This has been different than the past 11 weeks because the previous were all on nadir. On nadir means that the camera is facing directly straight down at the ground.  Whereas, oblique imagery is at an angle compared to the ground surface (Figure 1).  Oblique images are used in the geospatial marketplace to determine the height of different structures.  It can also be used to get a larger photo compared to the size of a nadir image.  It also gives the user the ability to measure the top of objects and the sides as well.  Taking oblique imagery is also used to create 3D models because to process the image it needs to have many different angles of the object.

Figure 1: A is on Nadir and B is off nadir or oblique. 

The study areas that are used are a bulldozer at the Litchfield mine, a Toyota Tundra in a parking lot, and finally a small building next to a track and field course both located at South Middle School.  The bulldozer images basically had sand surrounding the vehicle that needs to be annotated out.  The Toyota Tundra had grass and the sky that needed to be annotated, while the building had the sky, grass and the track that needs to be annotated to create a model.  Each of these sets of data seem to be taken during mid-day based on the shadows, except for the small building which may have been done later in the day due to the larger appearance of shadows.

Methods

To start this lab create a new project in Pix4D and set up a default folder to save all of the work in.  This will need to be done three different times for the bulldozer, the truck, and the building.  Once the folder is created use the "add director" tool to add all the images for the current project.  Click next, and change the camera to linear shutter. click next a final time and make sure to select the 3D Model template, and then hit finish to open up the main screen (Figure 2). at this point the location and images show up on the map. The pattern shows a circle around the object which looks like this because the goal is to make a model, and not like the previous labs where the goal was to create a DSM or orthomosaic which is why the pattern is a circle.  From here go to processing options and go to advance on the ray cloud to make sure the annotation is checked.  Close that window and un-check 2 and 3 from the processing bar at the bottom and start the initial processing.

Figure 2: 3D - Model Template

Now, after the initial processing is done the thrilling annotation part can begin.  To accomplish this next time on the left side of the screen, click camera -> calibrated cameras -> and click the first image.  now on the right side of the screen the picture selected will appear in the selection window like in Figure 3.  Now in Figure 3 where the mouse is hovering, the "Image Annotation" needs to be selected.  This will allow the user the ability to start highlighting the areas that are desired to be annotated.  This will create a purple like color on the image that will be the mask.  Once the image is fully covered in the areas needed it will look like the image in Figure 3.  Next, hit apply and move on to the next image and repeat this processes for as many images as needed.  After about 4 images go to Process -> Reoptimize to check the progress. The quality increases with more annotations due to the removal of unwanted pixels in the model, and it also increases with the use of different images at different angles.

Figure 3: Showing the annotation window and how it should look after annotating an image.

Next, un-select the processing step 1 and check the 2 processing option.  Start the processing.  once this is finished the annotation will be updated in the ray cloud and a fly-by can be created, which is most logical way to show off the newly created model.  These videos can be seen below.  For the other two sets of data the workflow will be the same to create the different sets.

Results

The first video shows the bulldozer that was not annotated.

This is the video that shows the bulldozer after the annotation was done.

Next, this is the Shed by South Middle School.

Finally, the annotated truck.

Discussion

Oblique imagery makes calculating heights and sizes easier with the ability to know the different angles and the height being flown at.  It also helps to create 3D models using the Pix4D software. Using Pix4D to annotate the images is a tedious task due to the way it selects pixels that cannot be easily predicted or quickly highlighted. There is not much of a difference between the first two images.  There seems to be more taken out of the second video, but it honestly does not make that large of a difference.  The process seems to need to do most of the images to get the best product out of this process.  The other annotated videos seem to be the same way.  It took out some of the background, but not nearly enough to make it noticeable.  For the models made above the truck seemed to turn out the best.  This most likely due to the fact that it had an easier boundary for the truck vs the ground/other objects.  To get the best results for this type of project the annotation should be done at many different angles.