Lab 4: An introduction to Pix4D

Overview:

This exercise focused on using Pix4D, a premier software used to process UAS data. The software can perform a wide variety of functions, however for this assignment only a few basic functions were used including volume analysis and video fly-through. 

Before any project can be started, UAS data from the field is needed. This includes orthorectified imagery and ground control points. The higher the quality of the data, the better the project will be. Additionally, Pix4D requires a certain level of overlap between images. The recommended values are 75% front overlap and 60% side overlap. If flying over snow, sand or uniform fields these values should be increased to 85% frontal overlap and 60% side overlap.It is also recommended that the camera platform is kept at a constant height as much as possible.

When in the field, the Pix4D rapid check processing system can be used. This processes images more rapidly to determine if sufficient coverage was obtained, but sacrifices accuracy to do so.

The Pix4D software can process data in a variety of ways. For example, data from multiple flights may be processed together provided that both datasets were collected from the same height, in similar weather conditions and have enough overlap. The software can also process oblique imagery, but that is beyond the scope of this exercise.

When processing data in Pix4D, ground control points are not required, but can help with georeferencing and the accuracy of the final product. 

The final feature users should be aware of is the quality report. This is a report generated by the software after processing data to give an overview of what happened, including errors in processing and processing parameters. 

Using Pix4D software

Two techniques that can be performed with Pix4D include volume measurement and video capture. The volume measurement tool can be found under the volume tab. Once a project is loaded, control points can be placed around whatever object is being measured. In this exercise, we measured the volume of a large gravel pit.

The second function, video capture, can be accessed from the ray cloud tab. This function allows the user to create a video where the camera moves through the project scene. To do this, the camera can be adjusted and waypoints are created along a route for the different camera views. The camera then moves through this route. An example may be seen below in figure one. 

 

Figure 1

Video capture from Pix4D

Creating a Map from Pix4D Software

The final section of this exercise involved making two maps in ArcMap using data obtained by Dr. Joseph Hupy from the University of Wisconsin-Eau Claire. ArcScene was also used in creating these maps. 

Figure 2
Map of Ortho-Mosaic created from UAS imagery

The first map created was of an Ortho-Mosaic that was generated by Dr. Hupy. The file was edited in ArcScene to eliminate the black background and uploaded into ArcMap. One thing that was noted was that the image quality changed after uploading the file into ArcMap. The original file may be seen below.

Figure 3
Ortho-Mosaic image prior to being uploaded to ArcMap

The second map created is of a digital surface model, or DSM of the mine site. A DSM shows ground surface features as well as elevation, which is often obtained from Lidar data. A hillshade was taken from the DSM using the hillshade tool. This was then made partially transparent and overlaid on the DSM to better show surface features. To display elevation, a blue-red color ramp was chosen with blue representing low elevations and red representing high elevations. One thing to note is that many areas of "high elevation" displayed on the map are actually just trees.

Figure 4
DSM map of mine site

Summary and Conclusion

Pix4D is capable of performing numerous functions, and this introduction exercise just scratched the surface. The main objective of this lab was to learn how to enter data into a Pix4D project and perform basic processing operations. These basic functions are very important to have a good grasp on since they will be used again more in depth in future UAS labs.

Creating a Navigation Map

Objective and Background

The objective of this exercise was to design a navigation map of the Priory, an area of land owned by the University of Wisconsin-Eau Claire, for a future course assignment. A secondary objective was to gain a better understanding of how projections and coordinate systems may effect navigation.

A coordinate system is a sort of grid that uses a series of numbers to determine a point in space. For example, on earth location is determined using latitude and longitude values.Different coordinate systems may be used depending on what is being mapped, what projection is being used and other factors.These maps use two different coordinate systems

WGS-1984 Web Mercator Auxilary Shpere- This coordinate system has become the standard for web mapping applications and is the system utilized by Google Maps. This coordinate system has many advantages and disadvantages that are associated with the traditional Mercator projection.

NAD1927-UTM zone 15N- This coordinate system is part of the Universal Transverse Mercator or UTM system that breaks the earth into 60 north-south zones, each covering six degrees of latitude. Each zone can be mapped with a specific transverse Mercator projection with very little distortion.

Methods:

There were a few key terms that needed to be understood prior to creating the navigation grid

1. Define Projection- This tool changes the information about the current projection, however it does not change the projection itself. You basically give it a new name. The project tool is need to actually change the projection.

2. Project/Project Raster- These two tools are used to actually change the map projection in the map document. Project Raster is obviously used for rasters and the Project tool is used for vector data.

3. Contour- This tool was used to create contour lines in the map document. This was done using elevation information from Lidar data of taken from the study area. In these maps, each contour line represents a 5 foot change in elevation.

After creating the contour lines, the next step was to create a navigation grid using ArcMap software. The grid can be created under data frame properties, which can be accessed by right clicking within the data frame. Our first map created was a navigation map that used a gradicule grid. The grid lines were spaced one second apart and the decimal degree labels were given four decimal points. 

The second map used a measured grid with each grid line being 50 meters apart. Instead of using decimal degrees, this units on this map stand for how many meters north of the equator and west of the prime meridian a point is. The final step to creating the maps was adding all of the necessary map elements

Results

  

Figure 1
Priory Navigation Map in Decimal Degrees

For the navigation map I choose to have my contour interval set at 5 feet, therefore every yellow line represents a change in elevation of roughly five feet. The grid spacing of every one second is ideal because the map is not too cluttered however there are still plenty of coordinates to navigate by. For this maps background I choose to use an areal image of the study area with about 70% transparency. This provided some visual reference to the map without making it appear too cluttered.

Figure 2
UTM Priory Map

This map is similar in design to the first map, however it uses a different coordinate system. This graph has a grid with spacings every 50 meters. Because this map does not utilize decimal degrees, degrees minutes and seconds are not measured. This will be the main map used for navigating during the lab exercise. One design error I would correct is the map background. I used two separate black and white images with about 70% transparency and there is a significant amount of overlap between the two images. In the future I would perform a mosaic of the two images in a program such as Erdas Imagine.

Conclusions: 
Conclusions that may be drawn from this assignment include the fact that making navigation maps can be quite challenging. The cartographer needs to understand which coordinate systems and projections to use in order to insure that the map remains accurate. There is a fine balance between the map being too cluttered or hard to read and being to vague and impossible to navigate by.

Sandbox Visualization

Overview and Objective:

This lab is a continuation of the first lab that involved creating a survey grid to survey a sandbox. The objective for this lab was to use the data that was collected in the survey to create a 3D model of the sandbox environment. The X,Y and Z values were compiled into a spreadsheet; a process known as data normalization. Each X,Y and Z value in a row corresponds to 1 point on the survey grid. A secondary objective was to gain an understanding of different interpolation methods and use different methods to create each model. Interpolation is the process of a computer program calculated the estimated value of areas in between points on a grid.

Fig 1.0
Normalized Data in Excel Spreadsheet

Methods:

In this lab five interpolation methods were used to create a 3D model using ArcMap and ArcScene software. The methods used are as follows. The descriptions are based on those provided by ESRI


Spline- The Spline interpolation method uses a mathematical function to estimate values in-between given points and provide a smooth surface. The spline method is best used when a data set has a large number of sample points. Because a spline surface passes directly through the points, the more points there are the smoother the surface will be.

Kriging-This interpolation method uses an advanced equation to investigate the spatial correlation between data points to estimate a surface. Kriging works well with scattered data.

Natural Neighbor- Finds closest subset of input samples to a point and applys weights based on proportionate areas to interpolate a value. An advantage of this method is that it can work with equally or irregularly spaced data

IDW- Also known as inverse distance weighted, estimates cell values by averaging values of sample points in neighboring cells. IDW works better with closely grouped data.

TIN-Interpolates by triangulating sets of vertices. A surface is generated from non-overlapping triangles. More triangles are generated in areas with more surface variation, such as a slope that changes in elevation.

After collecting the data, the values were placed into a spreadsheet, which may be seen in figure one. Each X,Y and Z value represents one point. This Excel document was then placed imported a geodatabase in Arc GIS software. Then, using the add XY data function, a grid of the data points was generated. Interpolation tools were then used on the grid to generate a terrain model. These were transformed 3-D models using the Z-coordinate values with Arc Scene software. Elevation surfaces were set to float on a custom surface. The models were oriented with north being at the top of the image. To represent scale I drew a straight line at the base of the model created a label that displayed the overall length. It is important to include the scale and orientation because when doing an experiment or exercise the process needs to be repeatable  Once the scenes were generated they were exported as JPEG images. These images were then loaded into an Adobe Illustrator document along with map elements from ArcMap to create a final map.

Results

The end result of this exercise was a map showing the 3-D terrain model for each interpolation method. These may be seen below.

Kriging

The Kriging method did not do an adequate job  modeling the terrain. The major features are shown, however several smaller depressions are not shown in this model but are in models generated using different interpolation methods. On a positive note, the surface generated by the Kriging method looks more realistic.

Figure 2
Map with Kriging Interpolation

Natural Neighbor

I believe that the natural neighbor interpolation does a nice job of representing the contours of the sandbox surface. Smaller ridges and depressions may be seen which do not appear in other models. The smooth surface in between features is accurate as to what the sandbox actually looked like. 

Figure 3
Surface Model using Natural Neighbor Interpolation

IDW

The IDW interpolation method is not the best method to use. Because of the way the points are weighted, small bumps in the sand box look peaked. The depressions on the surface are more difficult to make out as well with this model.

Figure 4
Model using IDW interpolation

Spline

The spline interpolation method proved very useful in showing small changes in the terrain. For example, in the center of the map there are three depressions that form a triangle. The depression at the top is actually two smaller depressions, which was harder to make out in the other models. One aspect of this model that should be changed would be the color ramp. The contrast between red and blue accurately shows changes in elevation, however the highest points and lowest points are both represented as black, which may be confusing

Figure 5
Surface Model Using Spline Interpolation

T.I.N

The Tin model did a great job visualizing changes in elevation. This can especially be seen on the slope in the northeast corner. one downside to this map is that it looks more computer generated and not as "pretty" as the other models.

Figure 5
T.I.N model

Overall the Spline and Natural neighbor interpolation methods produced the best terrain models. All of the models included a very small amount of variance between what was actually in the sandbox and what was generated in the model. This could be due to measurement error. The initial exercise was done in single digit temperatures and it's possible this contributed to human error when measuring.

Conclusion

The overarching goal of this exercise was to understand the fundamentals of data creation by gathering data in the field and importing the data into a geodatabase in order to generate a model. The survey we performed was very similar to professional surveys in that we divided an area of land into smaller, equally sized sections to map it. The differences include the scale of the project and the tools used. Because we were in a sandbox, hand measuring tools were fine. In the field professional survey equipment needs to be used in order to insure accuracy. The grid pattern we used is ideal for relatively flat land. This can be seen in the landscape of the rural mid-western United States. In other areas, such as the east coast, the different terrain was not conducive to a grid like survey system. 

Interpolation may be used in many different ways besides those in this exercise.  For example, bi-linear interpolation is used in remote sensing to resample images.

Understanding Survey Grids

Background and Objectives:

The objective of this exercise was to gain a better understanding of survey grids and construct a survey grid in a sandbox landscape. A secondary objective was to understand how different sampling techniques may be used in field research. Spatial sampling involves determining how measurements will be taken across a given study area. Three sampling types were discussed

Random: Random samples are taken across the study area in order to reduce bias

Systematic: Samples are taken at systemic intervals throughout the whole study area

Stratified: Samples are taken within small groups to portray a portion of the whole study area

This exercise was completed outdoors between 1:00 and 4:00 pm with a temperature of 3 degrees Fahrenheit

Methods:

To construct our survey, the following items were used

1. Meter Stick
2. Thumb Tacks
3. Measuring Tape
4. String

To begin, the perimeter of the sand box was measured using the measuring tap. Each side was found to be 114 centimeters in length. A systematic sampling method was choosen, and each side was divided into 19 separate 6 centimeter segments. Thumb Tacks were used to mark off each six centimeter segment. String was then woven in between the thumb tacks to create a grid. An X and Y axis was assigned to the grid, and the X/Y coordinates for each section were recorded in a spreadsheet.

The next step involved measuring the height above "sea level" to obtain a Z coordinate for each section. For the purposes of this assignment, sea level was considered to be the string that made up the grid, with features below the string being below sea level. Using a meter stick, this height was measured to the nearest half centimeter and recorded for each section and entered into a spreadsheet.

Results:

Our survey consisted of 361 individual data points. According to the data obtained, the average height of our sandbox landscape was approximately 6 centimeters below sea level. The highest point just made contact with the string and was recorded to be at sea level. The standard deviation is 2.38 meaning that the vast majority of measurements fall within 2.38 centimeters of the average.