Really dumb question...probably. Licensed Surveyor here willing to do what it takes to QA/QC our LiDAR point clouds with OSSDA ground?ÿproofing shots/reporting and GCPs as we are already on site picking?ÿup underground utilities and usually boundary work for design purposes. Because of the groundwork we already have several GCPs set as well that we occupy, check, and/or backsight for the ground data. We have a UAS with a base that we set up on an established GCP and running RTK observation along with the drone's IMU to produce the point cloud.
If the more experienced professionals of each application would share...How many GCPs should we set up on a project? I am told and can see how it would be much?ÿless than a photogrammetric process would require. The equipment configuration is also much different than a manned aerial collection as we have the base unit running?ÿconcurrently on a GCP much more like an RTK setup.
We have a camera mounted as well so we are getting colored point clouds and are producing?ÿan orthophoto. Because of this, I feel the need to go ahead and set out targets in access of the LiDAR need. That way we would have the availability for photogrammetry as a QA/QC, data backup, or just another deliverable to market as a client option. All that being said I'd like to know my?ÿbare minimum for each application and deliverable.
Any comments, advice, or links to resources that are not a vendor's?ÿpromotional material 🙂 are welcome. Thank you!
It would help if we knew what SUA, LIDAR and camera units you were using - also the level of accuracy your customers are looking for and the typical area you are covering on each project.
You might, as standard, consider reprocessing your RTK data as PPK, just to be sure that no gremlins sneak into the RTK results.
Here's a link to What I use for my projects. It may be way overkill for how you're using LiDAR in comparison to my project work expectations.
?ÿ
https://pubs.usgs.gov/tm/11b4/pdf/tm11-B4.pdf
?ÿ
Good Luck.
Sorry to let this thread die for a while there....
Our drone is a DJI M600 Pro, our LiDAR sensor is Phoenix Systems Alpha 32, and the?ÿcamera is a Sony A6000.
Our most restrictive work is also?ÿour bread and butter. Design surveys for transportation design that vary from a single subdivision?ÿroad to?ÿa longer highway project. Our tolerances?ÿwould be 0.1', which is my need to get as tight as it can get. It's just that I do not know what is possible. We also do a fair amount of H&H floodplain surveys that can get to large acreages. All of our projects have a lot of ground components- boundary, leveled benchmarks, utils, ground PIDs, and other ground supplemented data as needed. We are already checking very tight with our UAS, better with it mounted on a vehicle but I'd like to verify with some guys that have been out there doing this for a while.
Thanks!
As it happens we have just run some analysis this week on a number of old test flights over an area of road about 1km. long. The camera positions were processed as PPK (just reprocess your RTK values) and minimal tolerance set on the resulting camera coordinates. Including three GCP's in each of the flights significantly improved the residuals by up to around 50%. Typically we were left with nearly all the data within 3cm. (The test site has around 70 check points - some on the road and some on surrounding land).
The A6000 normally produces very good results, flying at around 300ft. and 80% overlaps should get what you need. For narrow surveys you need to keep the GCP well out to each side, probably having to fly several times the width you need but with RTK then the spacing along the route could probably be around 500 to 1000m. (in pairs opposite each other if you can get the access).
On the width concern with long route surveys. Is it giving the software more common points to help during the PPK or the conventional trig problems with weak triangles more like a baseline?ÿvs a network?
I've read on case studies that on a square site, then 5 GCP should be your minimum. 4 around the corner with one in the middle like a playing card 5. Going to a 7, 9, or more doesn't add much more accuracy to the scan. Would you say this fits your experience with your test flights?
Our experience over several years is that very thin strips are at risk of getting a cross tilt - they simply are not wide enough to get an inherent stability. It's basically the common survey problem of using a very short baseline (across the line).
For the flying the scenarios are slightly different for rotary or fixed wing - if you are flying across the survey strip the fixed wing turns effectively extend the lines but you need to allow for the aircraft to come back onto line at the start of the turn - for rotary work the flight plan lines need to be extended beyond the site.
If we are flying along the strip we normally add at least two additional lines outside those you would theoretically set in the flight plan.
Using PPK or RTK as control is equivalent of "mob rule" - generally the huge amount of data will produce a good average for the result, but occasionally it will go awry and produce anomalies. The introduction of a few GCP should limit or eliminate any problems. (Even fixed wing LIDAR benefits from the occasional GCP).
Without PPK or RTK we would always aim for 9 points on any site: for larger sites the aim would be a maximum of around 250 metres, with odd ones missed out if we could get quick access. With PPK/RTK then 500 to 1000 metres should be fine. On our test site the 3 GCP took the accuracy down close to the realistic camera limits. If you can put extra ones in while passing and you have a few minutes then they won't harm - all adds to peace of mind.
A few weeks ago we helped somebody else out on a coastal survey - some 5km. of beach and sea wall. They knew best, cut the flying tight with only 3GCPs. I've been back there this week (hence the delay in replying to you) to run a GPS string along the whole length of the sea fall path in order to sort out their problems. A little extra flying and ground control can save weeks of data hassle. There's no such thing as a productive short cut - it's either a thought through working practice or a bodge.
The 5km. was done in three flights. You can do this on linear surveys where the pilot can walk the route by simply walking along ahead of the aircraft until it overtakes and gets 500 metres ahead of you or runs out of battery power. You then take manual control, bring it back to where you have got to and land, move forward, change batteries and then continue - directing the aircraft to a waypoint a couple BEFORE you terminated the previous flight. (This only works if you are flying ACROSS the route, obviously). Do allow adequate width - they didn't since they knew best.
That flight was planned as a single operation, so they could break off and rejoin the flight plan at any convenient position. The downside of this method is that you need to set the safety circle to cover the whole site - in this case 6km. My personal preference would have been to put in a little more work and create 3 overlapping flight plans, each taking in 2.5km. of beach. Each of those would then have needed a safety circle of 2km., much more manageable.
?ÿ
Chris,
Thanks for the general intervals on GCPs. It's a helpful frame of reference.
The other thing we experienced is a lot of noise on our LiDAR when our flight paths are not with/straight against the headwinds. So a route survey more than likely might be zig-zaged?ÿto where a walking pilot would be able to keep up easily. It's the price we pay for living on the plains. We have a lot of no flight days because of wind speeds.
What sort of wind speed can you fly in - the M600 is a pretty hefty beast.?ÿ We are flying a fixed wing - Quest 200 - which is quite a stable platform. We have had it up in winds of 35kts. but realistically we would try not to fly in steady windspeeds of more than 20kts. Our normal flying speed would be 35kts. at 350 ft. so in strong winds we need to boost to 50kts which rather shortens the flight duration (normal duration around 50 minutes).
We cutoff?ÿours at much lower wind speed at the recommendation of our?ÿguys doing the processing. We are doing roadway design work so we have to be as tight as the drone can possibly get. Our cutoff is 10 mph (13knts) just because above that we've seen a lot more noise. It's usually not a problem to plan the best day in any week because we are a smaller firm and the guys flying/processing are also doing other work. I think it could be said that being on the plains, 10 mph may be the average for the day but it can double in straight line wind gusts frequently.
Once we have a heavier aerial workload or are doing hydrology surveys I think we'd try under 20 mph (23 knts) if we had to and watch our?ÿresiduals. I honestly am not sure of our normal speed but we fly at 150ft. Thanks for the frame of reference, seems like we are always flying blind...lol
We cutoff?ÿours at much lower wind speed at the recommendation of our?ÿguys doing the processing. We are doing roadway design work so we have to be as tight as the drone can possibly get. Our cutoff is 10 mph (13knts) just because above that we've seen a lot more noise. It's usually not a problem to plan the best day in any week because we are a smaller firm and the guys flying/processing are also doing other work. I think it could be said that being on the plains, 10 mph may be the average for the day but it can double in straight line wind gusts frequently.
Once we have a heavier aerial workload or are doing hydrology surveys I think we'd try under 20 mph (23 knts) if we had to and watch our?ÿresiduals. I honestly am not sure of our normal speed but we fly at 150ft. Thanks for the frame of reference, seems like we are always flying blind...lol
Factual correction - you applied your conversion the wrong way round, 23mph is 20 kts.
With an A6000 on board you should be able to fly at 200ft+ and still get 2cm. ground resolution. There would be less ground effect on the wind at that height so the aircraft should be a bit steadier, although as long as you get a good coverage any tilted shots will actually improve the end result rather than hinder it (gives some intersections at a better angle). If you are getting noise, is the shutter speed too low or have you set the camera to auto-focus. We find it is better to set the camera focus to flying height before take-off, so it doesn't try and change focus when it should be taking a photo. I assume you are flying a grid pattern which goes crosswind, rather than flying legs up and down the survey alignment. As I said before look for around 85% overlap (so you should get around a dozen photos of each point) and map a wider strip than you need (so any quality drop-off is likely to fall outside the important area).
