Monday, March 13, 2017

Lab 6: Construction of a point cloud data set, true orthomosaic, and digital surface model using Pix4D software

Introduction

This lab worked with the premier software for creating point clouds, Pix4D.  Previously captured images were imported into the software and they were processed into a point cloud and an animated fly over.

Basic Overview of Pix4D

This software is easy to use and is user friendly by having the Pix4D user guide readily available online.  The Pix4D mapper software is an image processing software that operates by finding thousands of common points between images.  A characteristic point found in an image is called a keypoint and when two keypoints found on two different images are identified as the same, they are called matched keypoints. The group of matched keypoints will form a singular 3D point.  The more overlap that exists between two images the more matched keypoints there will be increasing the accuracy of the 3D points created.        

  • Common Questions:
    • What is the overlap needed for Pix4D to process imagery?
      • The ideal image overlap depends on the type of terrain and the objects to be reconstructed.  The generally recommended overlap for most cases is to have at least 75% frontal overlap with 60% side overlap.  An example of a correct flight path with proper overlap can be seen in Figure 1. 
Figure 1. Model of a correct flight path that can be taken to 
acquire correct overlap from the Pix4D user guide.
    • What if the user is flying over sand/snow, or uniform fields?
      • If the user is flying over snow, sand, or a large uniform area, the overlap will need to be increased because in these cases, there is very little visual content due to the high uniformity.  
      • A high overlap of at least 85% frontal overlap and at least 70% side overlap must be used.  Also the exposure settings should be set to get as much contrast as possible in the image.   
    • What is Rapid Check?
      • Rapid check is a program that runs inside of Pix4D.  Rapid check is an alternative initial processing system that prioritizes speed over accuracy.  Imagery is processed very quickly but the accuracy is relatively low.   
    • Can Pix4D process multiple flights? What does the pilot need to maintain if so?
      • Pix4D can process images taken from multiple flights, it just needs to be made sure that there is enough overlap between the two images, and that the two images were taken under the same conditions.  Same conditions include, weather, cloud cover, sun direction, and that the terrain is generally the same with no new buildings or developments.  In Figure 2. the difference between two flight paths having enough overlap and not enough overlap can be seen.   
      • A pilot in order for multiple flights to be processed needs to make sure spatial resolution is maintained between the multiple flights by maintaining the same flight height between the flights.
Figure 2. Models of the correct and incorrect amount of 
overlap needed between two flights for Pix4D processing to occur.  
Model taken from the Pix4D user guide.

    • Can Pix4D process oblique images? What type of data do you need if so?
      • Pix4D can process high quality point clouds for oblique images.  Oblique imagery are aerial photographs taken with a 45 degree angle formed with the ground.  This type of imagery is helpful in image reconstruction of buildings because building imagery provides not just images of the tops of objects, but also of the sides of the objects.  This type of image processing requires a specific image acquisition plan.  For an example working with reconstructing a building, the building should be flown around a first time with a camera angle of 45 degrees.  The building should be flown around a second and third time should be flown at an increasing flight height while decreasing the camera angle with each round.  Enough overlap should be insured with a 5 to 10 degree change between each image.  Larger images require images taken at less degrees of change.  The Orthoplane tool needs to be utilized when processing oblique images.  
Figure 3. An example of how images should be acquired for 
oblique image processing of a building for Pix4D software. 

    • Are GCPs necessary for Pix4D? When are they highly recommended?
      • In an area of interest, GCPs are points of known coordinates.  These coordinates are obtained by using traditional surveying methods or other sources like LiDAR or other maps of the area.  GCPs are highly recommended, however they are not required for the use of Pix4Dmapper.  The accuracy of projects is increased when GCPs are used and GCPs can be used to check the accuracy of points on the map.     
    • What is the quality report?
      • The quality report is automatically generated for each output created by Pix4D.  The quality report gives an assessment of the image quality of the output of the software.  It gives a general quality check checking the images, dataset, camera optimization, matching, and georeferencing.  It gives a preview as to what the outputs will appear like before densification.  The quality report gives an image that shows were each image was taken in reference to the images rendering a flight path of the images.  It also gives more data quality measurements and data to help the user better understand the quality of their data.    
Using the Software

This demo with Pix4Dmapper was completed with imagery that was already taken for the Litchfield mine site.  First the Pix4Dmapper Pro software was opened and to create a new project was selected.  In the New Project window the name of the project and location for the new project to save to was set.  What the new project is named is what the folder that is created will be named where all of the outputs of the program will be saved.  Next was selected at the bottom of the window, and in the next screen all images to be processed are imported.  After selecting next again, the next window prompts to select image properties.  The Pix4D software looks at the exif file to read the metadata to render the proper coordinate system associated with the data if the images are geotagged.  The camera model is also detected from the metadata.  Both of these parameters can be changed if necessary.  While changing the camera model, the shutter model should also be checked to make sure it is in the correct setting.  In the next window, which is the Select Output Coordinate System window, the default is usually accepted.  The next window lets the user select the processing options template.  There are a variety to choose from like, 3D Maps, 3D Models, Ag Multispectral, and more.  The best one should be selected for you project needs.  Typically the 3D Maps option is used.  Finish can now be selected at the bottom of the window.  

The Map View Screen will be brought in and the overall layout of the pattern of the flight can be seen.  This is a good preliminary check of the data quality the user is working with.  If the data points appear in an organized manner with a linear, precise layout the data quality can be expected to be good, but if the data appears to be scattered and having no sort of pattern, the chances are that there is not enough overlap between images and the data quality rendered will be poor.  For the example being completed the data is very organized and linear so good data quality can be expected.  In the processing toolbar at the bottom of the window the Point Cloud and Mesh and the DSM, Orthomosaic, and index should be unchecked with only Initial Processing checked.  Processing options can also be customized if desired.  After reviewing this and making any changes start can be selected to begin initial processing.  This takes time, but once complete an initial quality report is created, Figure 4.  The quality report give an assessment of the image quality, for this example the image quality is good and there are no concerns.  The quality report includes a preview of the data, Figure 5, a diagram showing the initial image positions, Figure 6, and more data accuracy analysis. Also the data will appear in the window like it does in Figure 7.  When processing is complete all steps will appear green and if there is an error, it will appear red.    

For the final step, processing will be ran again, but this time Initial Processing will be unchecked and Point Cloud and Mesh and DSM, Orthomosaic, and Index should both be checked.  After selecting start there is a long wait period.  Once Pix4D is complete your data should appear like it does in Figure 8.  Further data interpolation can be conducted, an example includes creating an animation. The output DSM and orthomosaic can be found in folders located where the save parameters were set to.  The folders appear the way they do in Figure 9.                         

Figure 4. Quality report for the Litchfield mine site data.


Figure 5. Preview of data within the quality report.

Figure 6. Initial image positions layout, can see the flight path appeared like in the quality report. 


Figure 7. How the Litchfield mine data appears after running initial processing.  All of the steps are complete and the processing is done because the steps all appear green.  There are no errors because none of the steps appear red. 

Figure 8.  How the Litchfield mine data appears after running the point cloud and mesh, and DSM, orthomosaic, and index processing.  

Figure 9. What the output folders appear like after completing all of the processing steps.  They will be saved in the designated location in a folder that is named your project title. 


Discussion

The orthographic imagery and DEM created from using the Pix4D software were used to create the maps in Figure 10 and Figure 11.

In Figure 10 it can be seen that at the Litchfield mine site there are many piles of what appears to be a light colored gravel.  There is minimal vegetation surrounding the mine site.  There are wide paths located between the piles of mined gravel allowing for vehicles to enter the mine site to export the mined resources.
Figure 10.  A orthomosaic image of the Litchfield mine in Eau Claire County, WI created by Pix4D software. 


In Figure 11 it can be seen that the Litchfield mine is of relatively low elevations with the exception of small patches of higher elevation where the mined resources are piled.  The two highest elevations on the east and south edges of the DSM are not parts of the mine operation.  The highest pile of mined resources is located in the mid-east portion of the mine.  The largest pile is located in the central north of the mine.     

Figure 11.  A DSM of the Litchfield mine in Eau Claire County, WI created by Pix4D software. 


An animation for the Litchfield mine was also created.  The animation "flies" you over the area of interest, giving a full 3D view of the study area.


Video 1. Animation of the Litchfield mine site.

Final Critique of Pix4D software

Pix4D software contains a very thorough and capable set of tools.  In terms of creating a DSM, an orthomosaic, and an animation, this software is very good and makes quality outputs.  This software can have many applications and be useful across fields and disciplines.  The software is easy to use and the user manual is easy to read.  Still yet to experiment with the GCP tool set but if that operates as well as the tools and processes used in this lab, this software will be quite impressive.  This is a great tool set to learn and the applications are endless.