Audience: Pilots, Mission Planners, Post Processing
The senseFly S110 RGB sensor has been a reliable part of our UAS processes. Coupled with the eBee RTK, the S110 sensor has proven capable when used with proper lighting and meteorological conditions.
In late 2016, senseFly introduced their new S.O.D.A. sensor (~$1,800.00) for the eBee platform. This sensor has some material improvements over the S110 RGB and this brief is a summary of the initial findings on the operational use of the SODA sensor and integrating it into the business processes.
This high level overview is provided as guidance mainly for our pilots, mission planners and photogrammetrist, but should be considered by all team members. Part 1 of this series is focused on the hardware changes, how the changes impact mission planning, and comparing the actual raw imagery. Part 2 of the series will examine impact on post processing. Since Pix4D is recommended by senseFly, we will use Pix4D and compare the outputs and any differences in quality.
The S.O.D.A. sensor arrived with its own mount kit, ready to snap into the eBee… except that it requires a few mods to the eBee chassis. The camera boasts a 1 inch 20MP sensor and a new lens. Interaction with the sensor is accomplished via the eMotion 3 (not supported in eMoton 2) software as there is no battery power, screen or other mechanisms by which to configure the S.O.D.A. directly.
Once mounted in the eBee it becomes immediately obvious that the vertical displacement of the sensor system is larger than the S110. The sensor lens with the lens protector attached hangs near the belly of the eBee fuselage, as opposed to the S110’s retractable lens. While still protected and above the fuselage skid plate, it is awfully close to the ground when landing. This is most noticeable when handling the eBee during preflight, as users can easily touch this area, pushing the S.O.D.A. assembly up and unlocking the magnets. Not to mention the need to wipe off the fingerprints from the lens glass. DO NOT FORGET TO REMOVE THE LENS CAP BEFORE LAUNCH!
Connecting the sensor assembly to the eBee uses the same right angle, micro USB connector as the other senseFly sensors use. However, because of the height of the USB connector and the thickness of the sensor kit, there is perceived a slight upward pressure on the kit top panel. While it is not enough pressure to free the magnets that bond the two surfaces, there is an uneasy feeling that any jolt could dislodge the sensor kit. Use caution when handling on the ground and verify before launch that the sensor assembly is secure. This could have been addressed by senseFly in the design of the connector by simply reducing the height of the USB connector, or perhaps notching the upper canopy. Perhaps they will ship out a new cable with a shortened connector in the future to reduce this handicap.
Perhaps the most annoying aspect of this sensor assembly design is that the SD card slot is located in such a way that removal of the SD card from the sensor body requires the sensor be ‘popped’ out of the mount. It is unknown how long the mounts’ plastic legs will tolerate this.
All said the S.O.D.A. is very tight, light, no batteries to remove preflight and nothing to configure outside of the flight line software. With no moving parts (retracting lens) and sealed ports, it should prove much more rugged in dusty and sandy environments. Would have given them an A for design except for what feels like a design fault with the USB connector issue and the SD Card slot accessibility.
For the gear heads, listed below are the specifications for the two sensors we are comparing. The S110 RGB settings are customized for use with the eBee platform and in flight may not allow full use of all the camera options, for example, it uses fixed focal length of 5mm.
In this brief’s (Part 1) testing of the new S.O.D.A. sensor, comparisons will be made with the S110 RGB in the following areas: flight planning, image quality.
The mission target is approximately a 12-acre sized tract of land that contains field and forest, with the forest largely defoliated deciduous trees at the time of this flight. Both sensors will be flown using an eBee RTK platform using eMotion 3. Below is a table (Table 2) specifying the flight profile characteristics of each mission.
Note: The S110 will be flown twice, the first flight at a comparable GSD as the S.O.D.A. flight and the second flight at a comparable altitude as the S.O.D.A. flight. This will allow us to compare two sets of performance parameters.
Meteorological conditions at the time of flight were sunny, clear skies, winds at 7 knots with occasional gusts to 11 knots (A Bluebird day). The first mission was launched at approximately 1153 hrs (GMT -5) with each subsequent mission flown immediately after recovery and reprogramming. The readers may notice the movement of the shadows due to the overall elapsed time of all missions.
Note that because of the lens geometry differences in the two sensors, the altitudes are dramatically different to achieve the same GSD.
Prior to post processing our imagery in Pix4D, let’s compare the actual imagery resolution. Obviously the images from the different sensors were taken on separate flights so we will not have images that share an exact position as could be achieved with a ground test. However, the flight paths of the two missions with the same GSD are very close, thus we can find images that are similar in sensor position and image content. This will sufficiently allow comparative matches in image quality due both sensor position and relative solar angle.
In the first comparison below (Image 1), the structure containing a green roof will provide great contrast to the earth tones that surround it. In the uncorrected images below, notice that the S110 image has reached color saturation on the left (sunny) side of the roof, whereas the S.O.D.A. image is able to bring a more consistent exposure across the entire surface (see black rectangle). The S.O.D.A. image also better balances the shadow side (see red ellipse). This example clearly is a positive for the S.O.D.A. sensor in how it handles white balance, saturation and contrast.
In our second set of images for comparison (Image 2), an area that contains a light colored sandy surface (black ellipse) is examined. In the image from the S.O.D.A. sensor the surface detail is better visible. The light surface areas are not bleached out, or over exposed, yielding significant details for potential matching. The image from the S110 shows these areas as over exposed and yielding very little detail for matching purposes.
Analyzing the surrounding areas present, the S.O.D.A. image has better color balance and more even saturation as opposed to the S110 image.
And finally, the concrete object in these images (blue arrow) shows complete over exposure by both sensors. However, there is far greater detail and object sharpness present in the S.O.D.A. image as evidenced in the surrounding features.
In our third set of images for comparison (Image 3), an area that contains light colored sandy surface (black ellipse) is examined. These areas seem to be very similar in quality by both sensors. Again, in the image from the S.O.D.A. sensor the surface detail is better.
The color quality in this shot is similar with both sensors. The light surface areas are not bleached out, or over exposed, yielding significant details for potential matching.
Observing the object in the upper right (blue arrow) the S.O.D.A. image clearly has the better focus and renders much sharper definition. This higher focus of both the surface features and objects will prove beneficial to Pix4D when matching during post processing.
Reviewing the above images, it is evident that the S.O.D.A. has a much better balanced sensor than in the stock S110 camera. The S.O.D.A. has shown superior management of white balance, color balance and saturation, and exposure for aerial work. At least in this sample collection. The S.O.D.A. sensor also shows better focus throughout the range of captured images on this project.
Takeaway / Conclusion
The S.O.D.A. sensor has a different lens profile than does the S110 RGB. The narrower FOV of the S.O.D.A. will cause flight characteristics changes in the flight profile of the eBee, specifically the eBee will have to fly higher with the S.O.D.A. sensor payload to achieve the same GSD as it would with the S110. Or stated in reverse, the S.O.D.A. sensor allows for greater GSD at the same flight altitude as the S110.
This has advantages and disadvantages for pilots. If flying an area with taller obstacles, then the S.O.D.A. sensor allows for higher flight altitudes to maintain obstacle clearance and still achieve great GSD. However, if the goal is to have the tree canopy resolved in Pix4D, then it will be impossible to fly the S.O.D.A. sensor as the selected payload and not bust the FAA 400ft AGL limit. So pilots may wish to keep the S110 sensor in the inventory at job sites.
Also, the S.O.D.A. sensor proves a more efficient payload as the flight times are reduced compared to the S110 for capturing the same target area.
In Part 2 of this brief, the post processing of the imagery captured will be explored and the Point Cloud and Orthomosaic quality will be compared.