SAVE TIME, GENERATE GREATER REVENUE WITH AERIAL TECHNOLOGY
The Autel EVO II Aircraft, coupled with the Emlid RS2 base station are a cost-effective combination for surveyors, construction survey, engineering, etc.
Real-Time Kinematic (RTK) corrections bring significant precision to unmanned aircraft and workflows, even to the point of achieving repeatable precision within 1cm of actual position in a localized dataset.
Adding RTK to an unmanned aircraft enables real-time correctional data to be sent to the aircraft, allowing the aircraft to write corrected information to the metadata captured in the aircraft’s camera/sensor system. This is achieved through the aircraft remote control/ground station controller receiving correctional information from either a network system (NTRIP) or a local base station sending correctional data to the controller. In either case, the corrected positional data is uploaded to the aircraft.
No corrections, no RTK, nearly 10′ from point
One of the greatest challenges/concerns with an RTK aircraft is to ensure the aircraft is receiving corrections throughout the entire flight. Over large areas where there may be pockets of RF interference from powerlines, trees, buildings, or other obstructions, it’s possible to have a few images without RTK corrections, particularly at the edges of the flight area. One remedy is to add a range extender such as the 4Hawks antenna system to the aircraft’s remote.
The 4Hawks Antenna system assists in ensuring RTK corrections are sent to the aircraft throughout the entire mission area.
RTK systems gain slight benefit from the antenna-only system (autonomous positioning) while adding a base station or NTRIP network correctional system.
No corrections, RTK antenna-only. 13″ from actual
With RTK corrections sent to the aircraft, the point and the aircraft projections are 1 cm from the point.
We used a DT Research 301 data collector with a Seco RTK rod and head to verify data from the aircraft’s indicated position, and have since used a Leica GS18i to verify the points.
Jeremy Kippen from the KukerRanken Las Vegas store uses the DTResearch 301 system as a data collector to capture points on the ground, while flying the Autel EVO II RTK aircraft
Incorporating an RTK aircraft into the construction site, survey, engineering project, and many other uses provides a safer, faster, cost-effective means of capturing precise data, no matter the scenario. Topos, DSM, DTM, DEM, orthos, pointclouds, extraction for surfaces, and many other deliverables can become significantly more efficient when proper training and implementation techniques are observed. At KukerRanken, we’re here to help with UAS program development, training from Part 107 to operational techniques, and post-processing best practices.
Contact one of our KukerRanken staff to gain access to datasets demonstrating RTK with drone workflows. We offer Pix4DMapper, Pix4DSurvey, Leica Infinity, Leica 3DR, and many other training programs.
A Deep Insider’s Look at a Rugged Terrain Mission to Investigate a Helicopter Crash with Drones
Crash site investigation with drones has emerged as a leading application for unmanned systems in public safety. Gathering data that can be used by investigators in a courtroom, however, requires careful mission planning. Here, sUAS expert and industry figure Douglas Spotted Eagle of KukerRanken provides a detailed insider’s view of a helicopter crash site investigation.
Unmanned aircraft have become proven assets during investigations, offering not only the ability to reconstruct a scene. When a high ground sampling distance (GSD) is used, the data may be deeply examined, allowing investigators to find evidence that may have not been seen for various reasons during a site walk-through.
Recently, David Martel, Brady Reisch and I were called upon to assist in multiple investigations where debris was scattered over a large area, and investigators could not safely traverse the areas where high speed impacts may have spread evidence over large rocky, uneven areas. In this particular case, a EuroStar 350 aircraft may have experienced a cable wrap around the tail rotor and boom, potentially pulling the tail boom toward the nose of the aircraft, causing a high speed rotation of the hull prior to impact. Debris was spread over a relatively contained area, with some evidence unfound.
“The helicopter was on its right side in mountainous densely forested desert terrain at an elevation of 6,741 ft mean sea level (MSL). The steel long line cable impacted the main rotor blades and was also entangled in the separated tail rotor. The tail rotor with one blade attached was 21 ft. from the main wreckage. Approximately 30 ft. of long line and one tail rotor blade were not located. The vertical stabilizer was 365 ft. from the main wreckage.”
With a missing tail rotor blade and the missing long line, unmanned aircraft were called in to provide a high resolution map of the rugged area/terrain, in hopes of locating the missing parts that may or may not aid in the crash investigation.
The terrain was difficult and unimproved, requiring four-wheel drive vehicles for access into the crash site. Due to rising terrain, we elected to launch/land the aircraft from the highest point relevant to the crash search area, which encompassed a total of approximately 70 acres.
Adding to the difficulty of finding missing parts was that the helicopter was partially covered in grey vinyl wrap, along with red and black vinyl wrap, having recently been wrapped for a trade show where the helicopter was displayed.
We arrived on scene armed with pre-loaded Google Earth overheads, and an idea of optimal locations to place seven Hoodman GCP discs, which would allow us to capture RTK points for accuracy, and Manual Tie Points once the images were loaded into Pix4D. We pre-planned the flight for an extremely high ground sampling distance (GSD) average of .4cm per pixel. Due to the mountainous terrain, this GSD would vary from the top to the bottom of the site. We planned to capture the impact location at various GSD for best image evaluation, averaging as tight as .2cmppx. Some of these images would be discarded for the final output, and used only for purposes of investigation.
Although the overall GSD was greater than necessary, the goal is to be able to zoom in very deep on heavily covered areas with the ability to determine the difference between rocks and potential evidence, enabling investigators to view the overall scene via a 3.5 GB GeoTiff in Google Earth, and refer back to the Pix4DMapper project once rendered/assembled.
The same scene minus initial marker points.
Although working directly in Pix4D provides the best in-depth view of each individual photo, the Google Earth overlay/geotiff enables a reasonably deep examination.
Using two of the recently released Autel EVO II Pro aircraft, we planned the missions so that one aircraft would manage North/South corridors while the other captured East/West corridors. Planning the mission in this manner allows for half the work time, while capturing the entire scene. This is the same method we used to capture the MGM festival grounds following the One October shooting in Las Vegas, Nevada. The primary difference is in the overall size, with the Pioche mission being nearly 70 acres, while the Las Vegas festival ground shooting area is under 20 acres in total.
Similar to the Las Vegas shooting scene, shadow distortion/scene corruption was a concern; flying two aircraft beginning at 11:00 a.m. and flying until 1:30 aided in avoiding issues with shadow.
Temporal and spatial offsets were employed to ensure that the EVO II Pro aircraft could not possibly collide, we set off at opposite sides of the area, at different points in time, with a few feet of vertical offset added in for an additional cushion of air between the EVO II. We programmed the missions to fly at a lower speed of 11 mph/16fps to ensure that the high GSD/low altitude images would be crisp and clean. It is possible to fly faster and complete the mission sooner, yet with the 3 hour travel time from Las Vegas to the crash site, we wanted to ensure everything was captured at its best possible resolution with no blur, streak, or otherwise challenged imagery. Overall, each aircraft emptied five batteries, with our batteries set to exchange notification at 30%.
Total mission running time was slightly over 2.5 hours per aircraft, with additional manual flight over the scene of impact requiring another 45 minutes of flight time to capture deep detail. We also captured imagery facing the telecommunications tower at the top of the mountain for line of sight reference, and images facing the last known landing area, again for visual reference to potential lines of sight.
By launching/landing from the highest point in the area to be mapped, we were able to avoid any signal loss across the heavily wooded area. To ensure VLOS was maintained at all times, FoxFury D3060’s were mounted and in strobing mode for both sets of missions (The FoxFury lighting kit is included with the Autel EVO II Pro and EVO II Dual Rugged Bundle kits).
Once an initial flight to check exposure/camera settings was performed, along with standard controllability checks and other pre-flight tasks, we sent the aircraft on their way.
Capturing over 6000 images, we checked image quality periodically to ensure consistency. Once the missions were complete, we drove to the site of impact to capture obliques of the specific area in order to create a more dense model/map of the actual impact site. We also manually flew a ravine running parallel to the point of impact to determine if any additional debris was found (we did find several small pieces of fuselage, tools assumed to be cast off at impact, and other debris.
The initial pointcloud took approximately 12 hours to render, generating a high-quality, highly dense initial cloud.
After laying in point controls, marking scale constraints as a check, and re-optimized the project in Pix4D, the second step was rendered to create the dense point cloud. We were stunned at the quality of the dense point cloud, given the large area.
The dense point cloud is ideal for purposes of measuring. Although this sort of site would typically benefit (visually) from texturing/placing the mesh, it was not necessary due to the high number of points and deep detail the combination of Pix4D and Autel EVO II Pro provided. This allowed us to select specific points where we believed points of evidence may be located, bringing up the high resolution images relevant to that area. Investigators were able to deep-dive into the area and locate small parts, none of which were relevant to better understanding the cause of the crash.
“The project generated 38,426,205 2D points and 13,712,897 3D points from a combination of nearly 7,000 images.”
Using this method of reviewing the site allows investigators to see more deeply, with ability to repeatedly examine areas, identify patterns from an overhead view, and safely search for additional evidence that may not be accessible by vehicle or foot. Literally every inch of the site may be gone over.
Further, using a variety of computer-aided search tools, investigators may plug in an application to search for specific color parameters. For example, much of the fuselage is red in color, allowing investigators to search for a specific range of red colors. Pieces of fuselage as small as 1” were discovered using this method. Bright white allowed for finding some items, while 0-16 level black allowed for finding other small objects such as stickers, toolbox, and oil cans.
Using a tool such as the DTResearch 301 to capture the RTK geolocation information, we also use the DTResearch ruggedized tablet as a localized pointcloud scan which may be tied into the Pix4Dmapper application. Capturing local scan data from a terrestrial perspective with GCP’s in the image allow for extremely deep detail in small environments. This is particularly valuable for construction sites or interior scans, along with uses for OIS, etc.
Primary Considerations When Capturing a Scene Twin
GSD. This is critical. There is a balance between altitude and propwash, with all necessary safety considerations. Vertical surfaces. In the event of an OIS where walls have been impacted, the ability to fly vertical surfaces and capture them with a consistent GSD will go a long way to creating a proper model. Shadow distortion. If the scene is very large, time will naturally fly by and so will the sun. In some conditions, it’s difficult to know the difference between burn marks and shadows. A bit of experience and experimentation will help manage this challenge.
Exposure. Checking exposure prior to the mission is very important, particularly if an application like Pix4Dreact isn’t available for rapid mapping to check the data on-site. Angle of sun/time of day. Of course, accidents, incidents, crime, and other scenes happen when they happen. However, if the scene allows for capture in the midday hours, grab the opportunity and be grateful. This is specifically the reason that our team developed night-time CSI/Datacapture, now copied by several training organizations across the country over recent years.
Overcapture. Too much overlap is significantly preferable to undercapture. Ortho and modeling software love images.
Obliques. Capture obliques whenever possible. Regardless of intended use, capture the angular views of a scene. When possible, combine with ground-level terrestrial imaging. Sometimes this may be best accomplished by walking the scene perimeter with the UA, capturing as the aircraft is walked. We recommend removing props in these situations to ensure everyone’s safety.
What happens when these points are put aside?
This is a capture of a scene brought to us for “repair,” as the pilot didn’t know what he didn’t know. Although we were able to pull a bit of a scene, the overexposure, too-high altitude/low GSD, and lack of obliques made this scene significantly less valuable than it might have been.
Not understanding the proper role or application of the UA in the capture process, the UA pilot created a scene that is difficult to accurately measure, lacking appropriate detail, and the overexposure creates difficulties laying in the mesh. While this scene is somewhat preserved as a twin, there is much detail missing where the equipment had the necessary specifications and components to capture a terrific twin. Pilot error cannot be fixed. Operating on the “FORD” principle, understanding that FOcus, exposuRe, and Distance (GSD) cannot be rectified/compensated for in post processing means it has to be captured properly the first time. The above scene can’t be properly brought to life due to gross pilot error.
“ALWAYS PUT THE AIRCRAFT OVER THE PRIMARY SCENE LOCATION TO CONFIRM EXPOSURE SETTINGS, KEEPING ISO AS LOW AS POSSIBLE. USE ISO 50-100 IN MOST OUTDOOR SCENARIOS TO OBTAIN THE BEST IMAGE. NEVER USE OVERSATURATED PHOTO SETTINGS OR LOG FORMATS FOR MAPPING.”
Ultimately, the primary responsibility is to go beyond a digital twin of the scene, but instead offer deep value to the investigator(s) which may enhance or accelerate their investigations. Regardless of whether it’s a crash scene, insurance capture, energy audit, or other mapping activity, understanding how to set up the mission, fly, process, and export the mission is paramount.
Capturing these sorts of scenes are not for the average run n’ gun 107 certificate holder. Although newer pilots may feel they are all things to all endeavors benefitting from UA, planning, strategy, and experience all play a role in ensuring qualified and quality captures occur. Pilots wanting to get into mapping should find themselves practicing with photogrammetry tools and flying the most challenging environments they can find in order to be best prepared for environmental, temporal, and spatial challenges that may accompany an accident scene. Discovery breeds experience when it’s cold and batteries expire faster, satellite challenges in an RTK or PPK environment, planning for overheated tablets/devices, managing long flight times on multi-battery missions, or when winds force a crabbing mission vs a head/tailwind mission. Learning to maintain GSD in wild terrain, or conducting operations amidst outside forces that influence the success or failure of a mission only comes through practice over time. Having a solid, tried and true risk mitigation/SMS program is crucial to success.
We were pleased to close out this highly successful mission, and be capable of delivering a 3.5 GB geotiff for overlay on Google Earth, while also being able to export the project for investigators to view at actual ground height, saving time, providing a safety net in rugged terrain, and a digital record/twin of the crash scene that may be used until the accident investigation is closed.
Pro Read: As a leak indicates that Autel Robotics may be the first to offer a 6/8K camera on a drone, UAS expert and industry leader Douglas Spotted Eagle dives in to what the advantages of 8k may be – and if the drone industry is ready to take advantage of them.
In 2004, Sony released the world’s first low-cost HD camera, known as the HVR-Z1U. The camera featured a standard 1/3” imager, squeezing 1440×1080 pixels (anamorphic/non-square) pixels on to the sensor. This was also the world’s first pro-sumer camera using the MPEG2 compression scheme, with a color sample of 4:2:0, and using a GOP method of frame progression, this new technology set the stage for much higher resolutions and eventually, greater frame rates.
It’s “father,” was the CineAlta HDWF900, which offered three 2/3” CCDs, which was the industry standard for filmmaking for several years, capturing big hits such as the “Star Wars Prequel Trilogy”, “Once Upon a Time in Mexico”, “Real Steel”, “Tomorrowland”, “Avatar”, “Spykids” (1 & 2), and so many others. The newer HDV format spawned from similar technology found in the HDWF900, and set the stage for extremely high end camera tech to trickle down into the pro-sumer space.
Overtime, camera engineers identified methods of co-siting more pixels on small imagers, binning pixels, or using other techniques to increase the capture resolution on small surfaces. Compression engineers have developed new compression schemes which brought forward AVC (h.263), MP4(h.264), and now HEVC/High Efficiency Video Codec(h.265), and still others soon to be revealed.
Which brings us to the present.
We have to roughly quadruplemegapixels to doubleresolution, so the jump from SD to HD makes sense, while the jump from HD to UHD/4K makes even more sense. Following that theme, jumping to 6K makes sense, while jumping to 8K is perfect theory, and nears the maximum of the human eye’s ability to resolve information.
At NAB 2018, Sony and Blackmagic Design both revealed 8K cameras and in that time frame others have followed suit.
During CommUAV and InterDrone, several folks asked for my opinion on 6 and 8K resolutions. Nearly all were shocked as I expressed enthusiasm for the format.
– “It’s impossible to edit.”
– “The files are huge.”
– “No computer can manage it.”
– “There is no where to show 8K footage.”
– “Human eyes can’t resolve that resolution unless sitting very far away from the screen.”
– “Data cards aren’t fast enough.”
And….so on.
These are all the same comments heard as we predicted the tempo of the camera industry transitioning from SD to HD, and from HD to 4K. In other words, we’ve been here before.
Video cameras are acquisition devices. For the same reasons major motion pictures are acquired at the highest possible resolutions, and for the same reasons photographers get very excited as resolutions on-camera increase, so should UAS photographers. Greater resolution doesn’t always mean higher grade images. Nor does larger sensor sizes increase quality of images. On the whole, higher resolution systems usually does translate into higher quality images.
Sensor sizes are somewhat important to this discussion, yet not entirely critical. The camera industry has been packing more and more pixels into the same physical space for nearly two decades, without the feared increase in noise. Additionally, better noise-sampling/reduction algorithms, particularly from OEM’s like Sony and Ambarella, have allowed far greater reduction in noise compared to the past. Cameras such as the Sony A7RIV and earlier offer nearly noise-free ISO of 32,000!
Sensor sizes vary of course, but we’ll find most UAS utilize the 1/2.3, or the 1” sensor. (Light Blue and Turquoise sizes respectively, as seen below).
“Imagine an UAS equipped with an 8K camera inspecting a communications tower. Resolution is high, so small specs of rust, pitting, spalling, or other damage which might be missed with lower resolutions or the human eye become apparent with a greater resolution.”
WHY DOES HIGHER RESOLUTION TRANSLATE TO BETTER FINISHED PRODUCT?
Generally, we’re downsampling video or photos to smaller delivery vehicles, for but one reason. In broadcast, 4:2:2 uncompressed color schemes were the grail. Yet, most UAS cameras capture a 4:2:0 color/chroma sample. However, a 4K capture, downsampled to 1080 at delivery, offers videographers the same “grail” color schema of 4:2:2!
As we move into 6 or 8K, similar results occur. 8K downconverted to HD offers a 4:4:4 color sample.
CROPPING
We gain the ability to crop for post editing/delivery to recompose images without fear of losing resolution. This means that although the aircraft may shoot a wide shot, the image may be recomposed to a tighter image in post, so long as the delivery is smaller than the source/acquisition capture. For example, shooting 4K for 1080 delivery means that up to 75% of the image may be cropped without resolution loss.
As the image above demonstrates, it’s quite possible to edit 8K HEVC streams on a newer laptop. Performance is not optimal without a great deal of RAM and a good video card, as HEVC requires a fair amount of horsepower to decode. The greater point, is that we can edit images with deep recomposition. Moreover, we have more pixels to work with, providing greater color correction, color timing, and depth/saturation.
For public safety, this is priceless. An 8K capture provides great ability to zoom/crop deeply into a scene and deliver much greater detail in HD or 4K delivery.
The same can be said for inspections, construction progress reports, etc. Users can capture at a high resolution and deliver in a lower resolution.
Another benefit of 6 and 8K resolutions is the increase in dynamic range. While small sensors only provide a small increase in dynamic range, a small increase is preferable to no increase.
To address other statements about 6K and 8K resolutions; They human eye has the ability to see around 40megapixels, age-dependent. 8K is approximately 33megapixels. However, the human eye doesn’t see equal resolutions across the surface. The center of our eye sees approximately 8megapixels, where the outer edges are not as deep. High resolution does provide greater smoothing across the spectrum, therefore our eyes see smoother moving pictures.
BEYOND THE HUMAN EYE
Going well-beyond the human eye, higher resolutions are applicable to “computer vision,” benefiting mapping, 3D modeling, and other similar applications. Generally speaking, more pixels equals greater smoothness and geometry. As technology moves deeper into Artificial Intelligence, higher resolutions with more efficient codecs become yet even more important. Imagine an UAS equipped with an 8K camera inspecting a communications tower. Resolution is high, so small specs of rust or other damage which might be missed with lower resolutions or the human eye become more visible with a greater resolution. Now imagine that greater resolution providing input to an AI-aided inspection report that might notify the operator or manager of any problem. Our technology is moving beyond the resolution of the human eye for good reason.
DATA STORAGE
Files from a 6 or 8K camera are relatively small, particularly when compared to uncompressed 8K content (9.62TB per hour). Compression formats, known as “Codecs” have been improving for years, steadily moving forward. For example, when compressions first debuted in physical form, we saw Hollywood movies delivered on DVD. Then we saw HD delivered on Blu-ray. Delivery over disc formats is dead, and now we’ve moved through MPG2, AVC, AVCHD, H.264, and now H.265/HEVC. In the near future we’ll see yet even more compression schemes benefitting our workflows whether delivered via streaming or thumbdrive. VVC or “Versatile Video Codec”will be the next big thing in codecs for 8K, scheduled to launch early 2022.
Unconventional h.264 and H.265/HEVCare currently being used as delivery codecs for compressed 6 and 8K streams. 8K has been successfully broadcast (in testing environments) at rates as low as 35Mbps for VOD, while NHK has set the standard at 100Mbps for conventional delivery. Using these codecs, downconverting streams to view OTA/Over The Air to tablets, smartphones, or ground station controllers is already possible. It’s unlikely we’ll see 8K streaming from the UAS to the GSC.
U3 Datacards are certainly prepared for 6 and 8K resolutions/datastreams; compression is what makes this possible. The KenDao 8K and Insta 8K 360 cameras both are recording to U3 cards, available in the market today.
It will be some time before the average consumer will be seeing 8K on screens in their homes. However, 8K delivered for advertising, matching large format footage being shot on Weapon, Monstro, Helium or other camera formats may be less time-consuming when using 8K, even from smaller camera formats carried on an UAS (these cameras may easily be carried on heavy-lift UAS).
Professional UAS pilots will benefit greatly from 5, 6, or 8K cameras, and should not be shy about testing the format. Yes, it’s yet another paradigm shift in an always-fluid era of aerial and visual technology. There can be no doubt that these higher resolutions provide higher quality in any final product. Be prepared; 2020 is the year of 5, 6, and 8K cameras on the flying tripods we’re using for our professional and personal endeavors, and I for one, am looking forward to it with great enthusiasm.
Experts Tested 4 Different Drone Mapping Solutions for Crime Scene Investigation. Here’s What Happened.
At Commercial UAV Expo in Las Vegas, more than 300 drone industry professionals watched as experts tested four different drone mapping solutions for crime scene investigation at night.
Commercial UAV Expo brought UAS professionals, developers, manufacturers, first responders, and related industries under one roof for the first time in nearly two years. Due to the pandemic, the show was less attended than previous years, yet provided robust live demonstrations, night flight, daytime seminars, panels, and case studies for the relatively large audience. There was a strong buzz amongst the crowd about being at an in-person event, and experiencing face to face communication for the first time in many months.
In addition to the “Beyond the Cage” Live Drone Demo Day that launched Commercial UAV 2021, produced by Sundance Media Group, Wednesday night provided attendees with a glimpse of the Crime Scene Investigator tools function in the dark hours. Sundance Media Group developed this methodology several years ago at the request of a law enforcement agency and has been presenting this methodology at academies, colleges, universities, and tradeshows since 2017, with a variety of aircraft including DJI Mavic, Phantom 4, Yuneec H520, Skydio, and Autel EVO series (versions 1 and 2). All successfully output data, excepting Skydio, which struggles with brightly lit events in surrounding darkness.
Presented by FoxFury, Sundance Media Group, Autel, and Pix4D, this event also invited SkyeBrowse to participate in the demonstration, showing the effectiveness and speed of their application.
Testing Drone Mapping Solutions for Crime Scene Investigation: Setting the Scene
With a model covered in moulage, mock slit throat, and blood trail on the ground, the demonstration began with the multi-vendor team led by Brady Reisch, Bryan Worthen of Kuker-Ranken, Todd Henderson and Patrick Harris of SMG, and David Martel. The team placed four FoxFury T56 lighting systems at specific, measured points in the scene, supplemented by FoxFury NOW lanterns and Rugo lighting to fill in holes and eliminate shadows.
Douglas Spotted Eagle of SMG and KukerRanken emcee’d the event through the two flights.
Douglas Spotted Eagle addresses the crowd of 300 persons
SkyeBrowse had the first flight, with its one-button capture. Brady Reisch set up the mission, with input from the SkyeBrowse developer instructing the exposure levels of the camera for the SkyeBrowse video mission. Once the mission was completed, the photos were uploaded to the SkyeBrowse website, where results were found approximately 30 minutes following the flight.
Brady Reisch of KukerRanken sets up the Skybrowse mission with Bobby Ouyang of Skybrowse
The Autel EVO II Pro was programmed on-site for an automated Skybrowse mission and the demonstration began. The area is highly congested with palm trees and buildings enclosing the small rotunda in front of the Mirage Hotel Convention Center.
Brady Reisch flew the second EVO II mission manually, in much the same configuration as though the aircraft had flown a double-grid mission, supplemented by high-altitude orbit, coupled with manually captured orbit and select placements. Because of the crowd, time was a consideration. In an actual homicide scene, more low-placed images would have been captured.
Brady Reisch monitors time as Pix4DReact rapid-renders the scene (60 seconds)
The mission photos were uploaded to Pix4dReact on-scene and rendered while the audience observed, requiring approximately 60 seconds to output an ortho-rectified, 2D image, complete with evidence markers/tags, and PDF supplemental report output. Also loaded were the photo images into Pix4D and Leica Infinity software packages, to be rendered for 3D viewing once the show floor opened on Thursday. Pix4DReact is a two-dimensional, rapid-mapping solution, so there is no 3D view.
The four screen captures tell the rest of the story, and readers can determine for themselves what each software is capable of providing. One point of interest is that there were many claims of “guaranteed 1cm of precision regardless of flight area,” which has yet to be verified. The Kuker-Ranken team will be re-flying a mission with two separate GPS systems (Leica and Emlid) to verify the claims of precision.
Precision is Repeatable
Precision is repeatable. Accuracy is the degree of closeness to true value. Precision is the degree to which an instrument or process will repeat the same value. In other words, accuracy is the degree of veracity while precision is the degree of reproducibility. With a base station, NTRIP, Spydernet, PPK, or RTK workflow, precision is always the goal, well-beyond accuracy. This is a relatively new discussion in the use of unmanned aircraft, and although the topic seems simple enough, complexity holds challenges not easily dismissed by inexperience or lacking education and practice. We are fortunate to have a partner in Kuker-Ranken, providing precision tools to the survey, forensic, civil engineering, and AEC industries since 1928. The KR team includes PLS’, EIT, and other accredited precision professionals, rarely found in the UAS industry.
Precision is critical for surveyors, civil engineers, forensic analysts and investigators, construction sites, mapping, agriculture, and other verticals in the UAS industry, and this sort of scene is no exception. Being able to properly place a map or model into a coordinate is necessary for many professional pilots in the UAV field, and while this mission is not precise to coordinate, it is precise within itself, or in other words, measurements will be accurate in the image, while being imprecise to the overall location.
We’ll dive more deeply into precision in a future article. For purposes of this exercise, we’re more interested in accuracy of content in the scene, and all four outputs were similar in accuracy within the scene itself. In other words, distances, volumes, and angles may be measured point to point. Pix4DReact is not as accurate as the other three tools, as it’s not intended to be a deeply accurate application given speed of output.
Output Results of Drone Mapping Solutions
Output #1: SkyeBrowse (processing time, approximately 35 minutes)
Output #2: Pix4Dreact (processing time, approximately 1 minute)
Output #3: Pix4Dmapper (processing time, approximately 2.5 hours)
Output #4: Leica Infinity (processing time, approximately 2 hours, 50 minutes)
Selecting the Right Drone for Your Construction Business
Douglas Spotted Eagle and Brady Reisch headed into the field to collect aerial construction data over fourteen weeks with three different drones. Their goal was to determine which drone was best for the construction job site.
They used three popular aircraft for the comparisons and the results were pretty surprising.
Unmanned Aircraft (UA/Drones) have rapidly become a significant component of the modern construction industry workflow whether it’s for progress reporting, site planning, BIM, inventory control, safety awareness, structure inspection, topo’s, or other purposes. Site supervisors, architects, and stakeholders all benefit from the rapid output of accurate 2D/Ortho, or 3D models that may be used for purposes ranging from simple visualizations, progress reporting, stockpile calculations, DSM, contours, to more complex overlaying blue-prints in the As-Designed/As-Built or BIM process.
Choosing the right aerial asset/UA may be challenging, particularly as the marketing of many UA is focused on RTK built in (rarely accurate) PPK solutions and a many component workflow versus others that are single-step workflows. Decisions on aircraft choices will be made based on budget, accuracy requirements, speed to result, and overall reporting requirements.
On any site flown for BIM, input to AutoDesk or similar tools, having accurate ground control points (GCP) is required. GCP’s may be obtained from the site surveyor, county plat, or other official sources, and this is often the best method assuming that the ground control points may be identified via UA flight-captured images. Site supervisors may also capture their own points using common survey tools. Devices such as the DTResearch 301 RTK tablet may be used to augment accuracy, combining GPC location points from the air and on the ground. Failing these methods, site supervisors can capture their own points based on the specific needs of the site. These points may be calculated via traditional rover/base RTK systems, or using PPK, RTK, or PPP solutions, again being budget and time dependent. If centimeter (vs decimeter) accuracy is required, RTK or PPK are necessary.
Putting accuracy aside, image quality is gaining importance as stakeholders have become accustomed to photo-grade ortho or models. Oftentimes, these models are used to share growth with inspectors as well, which means having presentation-grade images may be critical. Image quality is high priority when generating pre-development topos, or simply illustrating a tract of land from all directions. In other words, a high-quality imaging sensor (camera) is a necessity. Some aircraft allow user-choice cameras, while many UA manufacturers are creating cameras specific to their aircraft design.
Turning to aircraft, we chose three popular aircraft for the comparisons:
Flying the site several times in various conditions, the same RTK capture points are used in all three mapping projects. The DTResearch 301 RTK system is used to capture GCP on-location, with Hoodman GCP kit as the on-ground GCP. The Hoodman SkyRuler system was also captured as a scale-constraint checkpoint.
This commercial site is small in size (1.64 acres), and one we were able to begin capturing prior to forms being laid, all the way to vertical installation.
Accuracy varied greatly with each aircraft system, particularly in elevation calculations. Deviations are from projected points vs the GCP points obtained through a surveyor’s RTK system. Overall (and to our surprise), the Autel EVO was most accurate with a deviation of:
x-5.112ft
y-47.827ft
z-16.541ft
The Yuneec H520/E90 combo was not far behind with a deviation of:
X-10.323ft
y-44.225ft
z-92.788ft
Finally, the DJI Phantom 4 presented deviations of:
x-1.95ft
y-45.565ft
z-140.626ft
All of these deviations are calculated and compensated for in Pix4DMapper, which is used to assemble all of these week-to-week projects. As 3D modelling was part of the comparison/goal, obliques were flown in addition to nadir captures. While manual settings are often essential for high quality maps and models, in the following images cameras were all set to automatic exposure, shutter, ISO.
It is important to remember that these are NOT corrected via network nor base station. This is autonomous flight, localized in Pix4D.
MODELS
AUTEL EVO (Original version)
YUNEEC H520/E90
PHANTOM 4 PRO
All aircraft models work well with Pix4DMapper, although at the time of this writing, Pix4D has not created lens profiles for the Autel EVO (they have indicated this feature should be available “soon”). We custom-sized the lens profile ourselves, based on information provided by Autel’s product managers. *as of 2.1.22, Pix4D has generated lens profiles for both Autel EVO and EVO II aircraft.
Orthos
AUTEL EVO
YUNEEC H520/E90
PHANTOM 4 PRO
Although image quality is subjective, our client and our team all agree the Autel EVO provides the best image quality and color of all aircraft, with all aircraft set to automatic exposure, shutters peed, and ISO of 100. This is a surprise, given the Autel is a ½.3 imager, vs the 1” rolling shutter of Yuneec and global shutter of the DJI aircraft. Based on internet forums, Autel is very well known for their camera parameters being impressive.
All flights are single-battery flights. This is important, as changing batteries offers different functions for the various aircraft. Using Yuneec and DJI products and their respective software applications, we are able to fly larger sites with proper battery management with the aircraft returning to launch point when a battery is depleted and resume a mission where it left off once a fresh/charged battery is inserted. The Autel mission planner currently does not support multi-battery missions (although we’re told it will soon do so).
There are a few aspects to this workflow that are appreciated and some that are not. For example, when flying Autel and Yuneec products, we’re able to act as responsible pilots operating under our area wide Class B authorization provided by the FAA. To fly the DJI Phantom, the aircraft requires a DJI-provided unlock that permits flights. It’s a small annoyance, yet if one shows up on a jobsite not anticipating an unlock, it can be tedious. In some instances, we are just on the edge and outside controlled airspace, yet DJI’s extremely conservative system still requires an unlock. Most times, the unlock is very fast; other times, it doesn’t happen at all.
All three aircraft are reasonably fast to deploy, and this is important when a LAANC request for a zero-altitude grid is a short window. Autel clearly wins the prize for rapid deployment, with the EVO taking approximately 30 seconds to launch from case-open to in-the-air. Mission planning may be managed prior to flight and uploaded once the UA has left the ground. We are experiencing much the same with the latest release of the EVO II 1” camera as well. We also appreciated the lack of drift and angle in relatively high winds (26mph+).
DJI is next fastest at approximately three minutes, (assuming propellers remain attached in the case), while the mission planning aspect is a bit slower than the Autel system. DJI uploads the mission to the aircraft prior to launch. Of course, this is assuming we’ve already achieved an approval from DJI to fly in the restricted airspace, on top of the FAA blanket approval. If we don’t, we may find (and have found) ourselves unable to fly once on-site, due to glitches or slow response from DJI.
Yuneec is the slowest to deploy, given six props that must be detached for transport. Powering the ST16 Controller, attaching props, and waiting for GPS lock often requires up to five minutes. The mission planning tool (DataPilot) is significantly more robust than DJI’s GSPro, third party Litchi or other planning apps, and is far more robust than Autel Explorer’s mission planner. DataPilot also essentially ensures the mission will fly correctly, as it auto-sets the camera angle for different types of flight, reducing the margin for pilot error. The Yuneec H520 is superior in high winds, holding accurate position in reasonably high winds nearing 30mph.
All three aircraft turn out very usable models. All aircraft capture very usable, high-quality images. All of the aircraft are, within reason, accurate to ground points prior to being tied to GCP.
We were surprised to find we prefer the Autel EVO and are now completing this project after having acquired an Autel EVO II Pro with a 1” camera and 6K video.
Why?
Foremost, the Autel EVO family offered the most accurate positioning compared to the other aircraft in the many, many missions flown over this site. With dozens of comparison datasets, the Autel also offered the fastest deployment, and ability to fly well in high winds when necessary. The cost of the Autel EVO and EVO II Pro make this an exceptionally accessible tool and entirely reliable. That the Autel EVO requires no authorization from an overseas company, particularly in areas where we already have authorizations from the FAA, is significant to us, and the image quality is superior to either of the other aircraft.
We also greatly appreciate the small size of the aircraft, as it takes little space in our work truck, and our clients appreciate that we’re not invasive when working residential areas for them. The aircraft isn’t nearly as noisy as other aircraft, resulting in fewer people paying attention to the UA on the jobsite. The bright orange color, coupled with our FoxFury D3060 light kit (used even in daylight) assists in being able to see the aircraft quite easily, even when up against a white sky or dark building background.
We also of course, appreciate the speed in deployment. With safety checks, LAANC authorizations, planning a mission, and powering on remote and aircraft, the Autel EVO is deployable in under two minutes. When flying in G airspace, from case to airborne can be accomplished in under 30 seconds.
Battery life on the EVO 1 is substantial at 25 minutes, while our newly acquired EVO II Pro offers 40 minutes of flight time with incredible images to feed into Pix4D or other post-flight analytics software.
Of greatest importance, the EVO provides the most accurate XYZ location in-flight compared to the other aircraft. For those not using GPS systems such as the DTResearch 301 that we’re using on this project, accuracy is critical, and being able to ensure clean capture with accurate metadata is the key to successful mapping for input to Autocad applications.
