Introduction to Pix4D

Introduction:

Pix4D is a stand alone product for Microsoft Windows and is a premier software for those looking to do photogrammetry. This software is used most often to analyze unmanned aerial vehicle (UAV) data. It can do many functions including creating orthomosaics and construction point clouds. Pix4D is incredibly easy to use because of its user friendly automated tools. It obviously can't be picked up by just anyone, background knowledge on the topic is necessary to make sure everything is done correctly. 

Look at Step 1 (before starting a project). What is the overlap needed for Pix4D to process imagery?

     Terrain type is the determining factor for the amount of overlap. For most instances that don't have snow, water, farm fields, forests, or other difficult or highly vegetated terrain, 75% frontal overlap and 60% side overlap is the lowest recommendation.

What if the user is flying over sand/snow, or uniform fields?

     Dense vegetation, forest terrains, agricultural fields, or snowy areas require at least 85% frontal overlap and a 75% side overlap. Rivers and lakes can be reconstructed but need land in every picture, oceans however cannot be reconstructed.

What is Rapid Check?

     Rapid check is a processing method in Pix4D that is faster but produces a less quality product. Resolution is reduced and accuracy is lowered, sometimes resulting in incomplete results. This is only recommended for field processing to check that the dataset was collected properly.

Can Pix4D process multiple flights? What does the pilot need to maintain if so?

     Pix4D can process multiple flights only if the the flights have the same height so the images can be processed correctly.

Can Pix4D process oblique images? What type of data do you need if so?

     Yes it can, Oblique imagery is used to reconstruct 3D objects. In order to do so however, the construct to be reconstructed needs to be flown around with the camera at a 45 degree angle. Two more flights are required, at higher heights and harsher camera angles.

Are GCPs necessary for Pix4D? When are they highly recommended?

     Ground control points are unnecessary for Pix4D; however, they are recommended in certain cases because they improve the accuracy of the project. They are nearly necessary when processing images without geolocation. Without GCPs the final model will not be scaled, georeferenced, or oriented correctly.

What is the quality report?

     The quality report is a summary of processed data. After running the processes a quality report is automatically generated and displayed. To find it again click the Process tab at the top of Pix4D and select "Quality Report..." in the drop down menu. The quality report will tell the user the number of calibrated images, the difference between optimized and initial cameras, the georeferencing used, and the matches of each calibrated image. The quality report is necessary to look at to make sure the data is acceptable to use.

Methods:

Starting a new project, first click start new project under the project tab. Name the project and chose the workspace, this is important so all the files are saved automatically in the correct folder. Select the data that needs processing, this will likely be many images from different flights. Verify the coordinate system, then choose a coordinate system that the final project will be in. Lastly chose the type of project that is to be completed.

After all the parameters have been set up the initial processing can begin. The initial processing can take some time depending on the amount of data. It is important to run the initial processing alone before other processes to ensure that the data will work well by checking the quality report afterwards. Figure A below shows the information from the initial processing as well as the name of the project and the time and date it was processed. It also shows the quality check, another part of the quality report. The quality check has 5 criteria, Images, Dataset, Camera Optimization, Matching, and Georeferencing. Green Check marks are good, Yellow Warning signs show caution that the criteria may not be up to par, and Red Warning signs are not good. The median key points per image was met, the dataset has 100% of images enabled, and the median matches per calibrated images were met. For georeferencing, there are no 3D GCPs, this is fine. And the one red warning for Camera Optimization is fine, Dr. Hupy assured us that for this project the camera optimization does not need to be its usual below 5%.

Figure A, the summary and quality check for the initial processing of the Litchfield Mine Data.

Next by selecting the "Processing" side tab, deselect "Initial Processing" and select "Point Cloud and Mesh" and "DSM, Orthomosaic and Index". Then press start. After running the two processes a quality report should be generated again. See Figure B, it will show the are that was processed and show the overlap of the entire area. It has a spectrum of red (bad, 1 image) to green (good, 5+ images). Areas of less overlap can potentially be slightly distorted. It is in good practice to collect enough images, especially on the edges to make sure a good amount of overlap is there to preserve the study are quality.

Figure B, almost all of the image has 5 or more overlapping pictures this is good, and not much of a worry that the edges may be slightly distorted.

Further down in the quality report lies the methods and data regarding the DSM, Orthomosaic and Index Details. IDW interpolation was used.

Figure C shows information regarding the DSM, Orthomosaic, and Index Details.

Once all the processes have been run, an image is generated and it can be found in the "Volumes" side tab. This 3D image can be manipulated using the tools across the top of Pix4D. It can be viewed directly from the top down, or the trackball navigation mode can be used to look at it from literally any angle. Figure D below shows the Litch field Mine image that was processed.

Figure D is a 3D image of Litchfield Mine made in Pix4D

In Pix4D volumes can be calculated of the sand hills that can be seen in the image above. With the image in top down mode, click new volumes and digitize around a hill. Left clicking places a vertices and right clicking places a final vertices to complete the polygon. Then click compute, the volume and area of the hill will be calculated. Figure E below is what the digitized hill base should look like.

Figure E, Digitized hill base.

Figure F, shows the hill from the side after the Area and Volume have been computed.

The next step is to export the volume data. Above the data on the left hand side is a box with an arrow pointing up and to the right. Click this and select the volumes to be exported, then click export and save the image as a shape file so that it can be opened in ESRI ArcMap. The hill calculated had a total volume of 779.33 cubic meters with a buffer of 15.40 cubic meters. This volume could be more exact if GCPs had been placed, lowering the margin of error.

The last thing to do was to create maps. In ArcMap the DSM and Mosaic rasters that were created when the Pix4D processes were run were opened. From there the necessary manipulation was done to create a viewer friendly Digital Surface Model map and a Mosaic Map. The maps can be seen in the results section below.

Results:

Figure G, Digital Surface Model of Litchfield Mine

Figure H is a Red, Green, Blue Mosaic of the Litchfield Mine.

Conclusions:

All in all, Pix4D is a ridiculously user friendly software that made myself and my classmates, who know very little about processing UAV data, very easy. All of the tools used in this lab were easy to understand and if questions arose, it was easy to find answers using the help button. As stated this by the professor this lab did not dive into the full extent of this software as we didn't even place GCPs for the lab, and we only computed one hill. It is very effective for new users and it would be fun and challenging to use this software more with new UAV data.

Grassy Knoll Assignment

Introduction:

This lab was designed to teach the user how to use a dual frequency survey GPS to collect data and create continuous rasters in ArcMap with the data collected. The lab was centered right in the middle of campus and used a familiar land feature to make the students comfortable with the equipment.

Study Area:

Figure A: the study area denoted by the red rectangle in the middle of UWEC campus.

Methods:

Using the dual frequency survey grade GPS on a small grassy knoll on campus mall, the group was able to get 19 data points (Figure B below) The students broke up in to groups of 2 so each person could learn how to use the equipment effectively. Each group gathered at least 2 GPS points that had X,Y,Z data. This is done by first leveling the tall GPS post with its 3 legs. Then gather the point on the handheld touch screen GPS monitor. The class gathered data points at random, meaning the class used a stratified sampling method. The data was compiled into a text file and given to the class by our professor. Individually each student had to normalize the data. After having an X, Y, and Z column it was time to import the table to ArcMap. From there a feature class was created from the XY data ready to be interpolated. 5 rasters were created for the 5 types of interpolation, IDW, Kriging, Natural Neighbor, Spline, and TIN. From there the rasters were then put into ArcScene to be viewed as 3 dimensional.

Figure B: Dr. Hupy teaching us the ins and outs of the survey grade GPS

Results/ Discussion:

The Interpolations ran smoothly without a hiccup, and were translated successfully in ArcScene.

Figure C: IDW interpolation

The IDW interpolation put too much emphasis on the low points and high points giving the almost level parts divots and the top of the hill a point. Neither of which exist in real life.

Figure D: Kriging Interpolation

The kriging gave a very accurate representation of the hill, nice and steady slope and even shows where the slope is most drastic.

Figure E: Natural Neighbor Interpolation

Nearest neighbor was great because, not the best, however it did use the outer most points to make the shape of the actual area we were dealing with and not a rectangle.

Figure F: Spline Interpolation

Spline worked quite well in accurately showing how the hill wasn't perfectly even. It showed the wave of the base of the hill nicely.

Figure G: TIN Interpolation

The TIN was great as it used triangles to show depth and variation. It really shows the slope better than the others, however it makes it seem more drastic and puts a point at the top of the hill.

Conclusion:

This field assignment was great in teaching the class how to really use a survey grade GPS, as we had tried to do in the first week but failed due to time problems. These skills can be transferred to the real world easily because a GPS unit like this is quite common for gathering data. The technology worked very well and the rain didn't stop our class from learning well.