GEMs Review

Overview:
The Geo localization and Mosaicing System or GEMs for short is a precision agriculture multispectral sensor payload that be used on many different UAS platforms. This sensor was designed to capture RGB, NIR, and NDVI imagery in NADIR. With the purchase of the GEMs hardware, you also receive the GEMs software package as well. This software allows you to process the imagery that you just took and automatically receive orthomosaiced RBG, NIR, and NDVI imagery.

Workflow:
To achieve the final product imagery, there are multiple steps that need to be taken. First, you must mount the sensor to the UAS platform. Whether this be a fixed wing or a multicopter, one important thing to note, the positive charge must be connected to the positive connector. Not doing this on a conventional power system could result in the failure of the hardware or a fire. Once you have the sensor hooked up you must insert a SanDisk extreme 32gb usb jump drive. This is how the GEMs stores its data when in flight.

Now you must consider what the Ground Sampling Distance (GSD) is as well as the Pixel Resolution. This is an important aspect of collecting imagery because you don't want your pixel size to be so large that it doesn't even distinguish separate features that you are trying to study. The GEMs has a GSD of 5.1cm at 400 feet or 2.5cm at 200 feet. The pixel resolution is 1.3MP for both RGB and Mono which comes out to be 1280x1024. Something else that you must take into consideration are all the parameters of mission planning. All these different parameters all relate the quality of the data, but also the efficiency of the platform to accomplish the task at hand. Here are some of the parameters of the GEMs:

Image Sensor resolution: 1280 x 960 pixels

Sensor dimensions (active area): 4.8 x 3.6 mm

Pixel Size: 3.75 x 3.75 μm 

Horizontal Field of View: 34.622 degrees

Vertical Field of View: 26.314 degrees

Focal Length: 7.70 mm

After the flight has been flown, the data exports in a certain folder structure with all the different imagery in their own folder as well as the flight data. These are self proclaimed orthomosaic photos. Now the difference between orthomosaic and georeferenced is that orthomosaics use the photos geometry versus georeferenced photos use the GPS coordinates of concurring points between two different images that are overlapped. Orthomosaics are the best since they are geometrically correct and they take topography and elevation into effect.
 
Once you are ready to run the software it is fairly easy. You simply go into the GEMs software and run an NDVI initialization with the images you took. You then can generate mosaics from them while also computing NDVI, using the default color map, and performing fine alignment. GEMs then gives you the option to export to Pix4D for further processing within there software. What this does is pretty much gets your images ready to be processed in the format that Pix4D wants. Another option is to export to powerOFground which is a cloud based image processing platform that also analyzes your images geospatially. Once you are done with all your other software exports you can look at the five images that GEMs exported for you. Those five are RBG Fine, NDVI Mono Fine, Mono Fine, NDVI FC1, and NDVI FC2 (Figure 1).

Figure 1: These are the five exported images that you receive from the GEMs software along with their pixels values.

You can see from the figure above that you receive different values for each image. For instance between the two NDVI FC images, one has a color scheme that shows yellow or orange as healthy vegetation while the other shows healthy vegetation as green. Which makes more sense (Green)? You can also see that the RGB Fine doesn't have values associated with it. That's because this is purely so that we can have high resolution imagery that is better than something that we would get from an ESRI Basemap (Figure 2).

Figure 2: Comparison of the GEMs RGB Fine image versus an ESRI Basemap image. If you had the raw images files for both of these you can see that the quality difference is quite apparent.

Critique:
Overall I like the GEMs hardware. I like that is has the ability to collect all five of those images at the same time which can dramatically cut down on flight time if you needed to go back through and collect more. The parameters that are set on the system allow for respectable photo quality although it
can always be improved. The system is quite simple to set up on the chosen UAS platform and can be used with a fixed wing or a multicopter. This gives the GEMs an advantage because it can be that one sensor that you HAVE to have because of its wide variety of applications. Not only can it produce high quality images but also NDVI's for precision agriculture.

The GEMs software gets some points for me because it is really easy to run. All you do is click a few buttons to run the initialization and select your photos and your off. This makes it practical for the farmer that is just getting into precision agriculture and wants to use this technology without having to go through some intense training like other softwares would make you do. However, even though it is easy to use, the end product is not up to par. If you look at the Figure 1 above you can see that in most of the images there are mosaicking issues. There are streaks of off color lines that run through the images that can totally throw off your values. This makes for a rather poor end product. I will say that the RGB Fine image is definitely better than what we would get from an ESRI Basemap and that in of itself is worth its weight.

To wrap up I would say that the GEMs is a sensor that can be very applicable but you need to know your needs first. What is your goal? What are your standards and how much money do you have? All of these should go into consideration before you purchase this product because depending on your standards you may be disappointed. If your standards are low and you just need a quick representation of what your fields look like, this sensor will do that job. You will get a high quality RGB image as well as a good representation of how healthy your vegetation is. If that is your need, then the GEMs is your want.

Obliques for 3D model construction

Introduction:
This activity was our first transition from taking imagery in NADIR format to taking them in oblique format whether that be from high oblique(you cant see the horizon) or low oblique(you can see the horizon). To demonstrate oblique imagery we are going to be producing a 3D representation of a pavilion at the soccer fields. Most of this activity is centered around the data collection and the processing will come at a later date.

Study Area:
Our study area was once again located at the Eau Claire soccer fields across from the universities Bollinger Fields (Figure 1). The actual feature to be mapped was a pavilion located in the middle of the soccer fields (Figure 2). We conducted the study on October 7th at 4pm where there was hardly any winds at all and some cirrus clouds with some wispy icy mares tails on them.



Figure 1: Map showing the soccer fields (Study Area) in relation to the University



Figure 2: Map depicting the pavilion that we took imagery of for our 3D model

Methods:
To introduce us to oblique imagery we took pictures from two separate platforms. The first being the Iris multicopter. The Iris flew in a corkscrew motion so that it took pictures at eye level and then in ascending altitude with the camera angled down towards the building. After the corkscrew was finished at about a height of 26 meters, we did a couple crisscross passes so that we got all of the different angles from the roof. One thing to note is that we were taking these photos with a GoPro. GoPro's do not have GPS associated with their photos which can cause some problems with other image gathering but for this activity it works just fine since the GoPro has such a wide lense. Our professor Joe Hupy made it a point that we take note of the different cameras and their abilities because they all go toward proper misson planning since they all have different uses (Figure 3).


Figure 3:  Professor Hupy explaining the pros and cons of the GoPro

The second mulitcopter we flew was the phantom. We did the same procedure with the Phantom as we did with the Iris. The camera that was on the phantom come standard with the platform and DOES have GPS associated with it. The mission was planned through the Mission Planner software and needed different parameters to be set such as altitude, circle radius, number of turns, and number of corkscrews. This software made the actual flight quite easy to run through, but afterwards we decided we wanted more pictures from eye level so we all got the opportunity to manually fly the phantom (which is quite easy due to its self correction). Some of us flew it via the camera on the IPad (Figure 4) while others walked around the pavilion with the Phantom (Figure 5).

Figure 4: Photo of myself flying the phantom via the IPad

 

Figure 5: Photo of Michael Bomber flying the Iris while Professor Hupy shows where he wants the imagery to be captured
 

Discussion:
This form of data collection is very applicable, especially when a client would want a 3D representation of a feature. Otherwise, you could use oblique imagery to capture a rock face or perhaps a soil profile. I would say that the actual collection of the imagery was a different experience than what we were used to because this time we had the UAS right in front of us the whole time and we knew somewhat what the imagery was going to look like. Also, it would be very difficult to manually take imagery from a NADIR perspective (Figure 6), but taking it manually from an oblique perspective was quite simple (Figure 7 and Figure 8). I would say a similarity between the two collections is that you still want good overlap between your rows of photos to ensure good quality when processing the imagery.

Figure 6: A image taken from NADIR from a previous activity

Figure 7: An image taken at a high oblique angle

Figure 8: An image taken from a low oblique angle

Conclusion:
The is a definite difference between NADIR and Oblique imagery and they both have different uses in the UAS world. One can be good for a large area of interest especially when you are looking at creating a DSM or an orthomosaic. The other is excellent for creating models of vertical structures that do not get much attention from the NADIR collection method since it is taking the photos from straight up, all it would get is the roof or poorly represented snapshot of one side of the structure. With oblique imagery however, you can take advantage of a low flying multicopter that can also gain altitude for some high oblique shots as well and together you can process them into a 3D model which could prove to be very helpful for someone who maybe wants to assess the structural integrity of a building.

Gathering Ground Control Points

Introduction:
For this lab activity we headed out to a prairie area south of South Middle School for some lessons on Ground Control Points (GCPs). GCPs are basically a ground control that your UAS can pick up to help with the accuracy of your data. What we did to practice using GCPs and to learn more about them was setting up several points and then using various Global Positioning Systems (GPSs) so we could test the accuracy of each of the devices.

Study Area:
Our area of interest was two small drainage ponds south of South Middle School (Figure 1). The South Middle School community garden was just to the North of our study area and we set up home base at the Northwest corner (Figure 2). We conducted this small study on September 30th from 4-7pm.

Figure 1: South Middle School is located south of Highway 12 and west of Highway 93

Figure 2: Map showing where our home base and study area were in relation to South Middle School

Methods:
The first step that we took was getting all the GCPs ready. Our GCPs consisted of a 4x4 black and white material that we could easily lay down over an area and stake it. Before we set up our 6 GCPs we first had to logically think about where we should put them. In order for your GCPs to be useful at all, you need to have at least three so that the computer can triangulate their position. There was a trail that went around one pond so that made for a nice flat area to lay our GCPs out. We set 5 of the GCPs outside the perimeter of our study area as well as one that was a little more inside the perimeter. The reason why you want some inside your study area is to lessen distortion of the accuracy of the GCPs. If you only have them around the perimeter then the center of your study area may be distorted.

Once all the GCPs were set up we used five different GPSs and tested each of their accuracies, The five we used were as follows and their approximate price listed beside them.
Dual Frequency Survey Grade GPS-$18k
Bad Elf GNSS Surveyor GPS-$600
Bad Elf GPS-$125
Garmin GPS-Less than $100
A smart phone- Free if ya have one!
  As with many things in the geospatial industry, it is up to us to decide what kind of accuracy we are going to need. That is where this activity is useful because it can help us decide if we want and/or need to spend 18 grand on a survey grade GPS that has centimeter accuracy or if we need something that we can get at an everyday Gander Mountain. Yes with the Dual Frequency we will achieve the absolute best, but depending on our work environment it may not be plausible to carry that big heavy thing into the bush and quite frankly, most of us do not have 18k to be able to afford that system. Nonetheless we moved forward and took points with each system so we could compare them later (Figure 3). One point of interest about GCPs in the use of mobile devices in collecting GCPs. This is a bad idea and many people continue to use this method. Mobile devices are excellent for taking you somewhere on a map or texting someone but not for collecting GPS points as a GCP. There is too much interference with these machines and they cannot be trusted, they will harm your data and make it inaccurate. The Dual Frequency on the other hand is very accurate, so much so that it needs to be leveled as shown in Figure 4.

Figure 3: Myself using the Dual Frequency Survey grade GPS to take a point of the 6th GCP

Figure 4: Level on the Dual Frequency GPS

Once we had points from every single GPS, we came back to home base where we set up for a multicopter flight. This flight was made possible by group 1 (preflight checks and planning) and brought to you by Mission Planner (Mission planning software)(Figure 5). Professor Joe Hupy was the Pilot in Command and we watched the Matrix make its grid pattern over the pond (Figure 6). Michael Bomber safely landed the Matrix as the whole class watched (Figure 7).

Figure 5: Mission Planning software showing the statistics of the mission we flew and the flight path

Figure 6:  Professor Pierson, Michael Bomber, and Professor Hupy taking their eyes off the Matrix. NEVER DO THIS!!

Figure 7: Michael Bomber ensuring that the Matrix lands safely

Results/Discussion:
This exercise was able to show us the basics of GCPs and how they are implanted in the real world. As far as discussing which GPS was the best for this exercise, I would pick the Dual Frequency. I would like to note that if we did not already have one in stock at the university then I would not have picked this GPS but since we did and since the terrain was stable and easy to walk around, it was not cumbersome to use that big heaping GPS to collect those points. Was it overkill in terms of accuracy? Maybe, but you can never go wrong with better accuracy if all it involved was just a little more work. Now if I had to pick a backup favorite I would have picked the Garmin GPS since as shown in Figure 8 below, it had the next best accuracy and it was not cumbersome at all. The Garmin is a hand held device that was just as easy to use as the Bad Elfs, but during this test, the Garmin came out on top for the handheld devices. I don't not know why but the Bad Elfs underperformed in my opinion and I do not see much of a difference between the survey Bad Elf and the regular Bad Elf. I am also puzzled as to the accuracy of most of the GPSs in general, it seems as though most of these do not even have meter accuracy!

Figure 8: Map showing the location of all the GPS points during out GCP test.