Hi, I uploaded 164 images today to do a simple 2d map of a field. The 4 corners have mapped but the middle is missing and makes it totally unusable. Unfortunately the instability of Dronedeploy and its failure to fully work on a regular basis makes me wary about pulling the trigger and paying for a pro account.
Can someone tell me if I use my phantom 4 pro+ with built in screen, not having the ability to use drone deploy app for planning, I just flew the area and took regular pictures to provide an overlap looking straight down. The 4 corners appear to work ok, just wondering why theres a huge gap in the middle of the map. does this render my drone now useless for mapping? If so then DJI have really spoiled a good thing and taken a backward step.
Like you, I’m new at this and am hoping to add mapping services to my business. I flew a mission today and discovered many blank squares where apparently a photo was not captured.
SD card up to snuff? What are you using? I’m using SanDisk 64GB Extreme - paid a premium so I would be able to capture 4K video. I would think it can handle still space apart by several seconds (flying 13mph today @135’). I was out ~ 1500’ from home and thought there could possibly be connection issues - not sure on this one. I would think that the controller would not be a part of the photo taking process.
What would one do in the real world of business. Would I discover this in my office, then head out and try to fill in the blanks? Refly the mission and cross my fingers?
What altitude (AGL) were you flying? Also what were you imaging? Can you post an image of the failed map?
One of the biggest newbie errors is flying too low. Flying at 135 ft and imaging trees for instance will usually not stitch correctly because the trees do not appear orthagonal across the image. The stitching algorithm cannot then match them up across different images.
Best advise I can give is to start at a high AGL, and then incrementally lower it as necessary
Curious what a typical or suggested altitude would be for construction site mapping, vs. city park mapping, vs. crop mapping? I flew as low as I deemed safe, vs. effective I thought closer would result in a more detailed map:
One may notice a couple missing sections in the top left - another issue I’m trying to figure out. Are missing photos altitude related?
Generally, if you get “holes” in your map it’s because the stitcher can’t find enough detail to work with. This can be caused by many things, but generally the cause comes down to too much uniformity between input images for an area. Think big fields of grass, uniform tree canopy, water, etc.
There are numerous blog posts about ways to mitigate the effect. A good place to start is reading this:
In general, fly higher, increase overlap, alter your capture pattern, and make sure you are getting good in-focus source images. As for a suggested altitude I regularly use the default of ~250ft with no issues. Lower is fine, but you need to compensate for the overlap and sidelap loss. There is also a new app in the App Market called I think “Overlap Optimizer” that can help you in this area.
FlashBuddy, overall that looks very nice. The missing portion along the top left indicates a stitch ‘failure’. There are a couple reasons why this might happen.
One is that there may not be enough duplicate images of this area for the stitcher to work. These engines want 6 - 10 separate images of an area for the stitching to work well. That’s why you’ll often see stitch errors along the edges of your maps.
Secondly, the stitching is done on the image taken (in that instance the trees) not the ground hidden underneath. Therefore the actual drone is imaging trees at maybe 80ft not 135.
I mapped at around 300’ age and some were slightly higher as I had to clear pylons. I have redone the same area today and so far the first map has stitched perfectly well. I flew that at 385’. used Mavic this time instead of the p4p+ so could use DD app.
i’m not really new at this having been using Drone Deploy for nearly 2 years i think.
i started using 64gb cards but then realised very quickly if the data becomes corrupted on a 64gb card then i lose a whole lot of data. I use no larger than 32gb sands extreme cards now with no issues. Usually i’ll use a 16gb card per job.
once the drone sets off the controller isn’t usually the issue. Unless like today mine didn’t RTH where it took off from but was heading for my car roof! Had to quickly shut down DD, open dpi go4 and then cancel rth. interesting. 18 minutes of waiting and 10 seconds of worry.
Not trying to impugn your expertise, sounds like you know what you’re doing. In our experience, when in question we always increase our overlaps n fly higher. Better to lose a little resolution than to have a poorly stitched map.
Also on the RTH, I usually guide it a bit with the controls to land where I want to. It allows that while in DD RTH.
If I get real concerned, or misguide it, I just switch the mode switch to take manual control.
so, having flown the same mission again, it appears its my fault. The drone was getting interference from nearby pylons, not a problem using my magic but the Phantom 4 pro+ just didn’t like it at all. Anyway, I’m happy now, DroneDeploy came thru great for me, the customer is happy. (and has paid the invoice already) and he’s got me a lot more work from the back of it i believe, so I may be upgrading very soon to the pro version. Fingers crossed. and thanks again DD.
Here’s another option to get good resolution and good stitching.
Sometimes I will fly at two altitudes and upload both sets as one map. The higher one supports better alignment and stitching and the lower one supports better resolution. I did this with a P3P at 150 and 300 feet. I uploaded each separately and both together to test variations in resolution. The 150 foot gave .8 in/px resolution and the 300 foot gave 1.5 in/px. Both together gave .9 in/px. If the lower one works, I use it. If not, I lose very little with the combined set. Each subject matter is unique. I’d fly higher over homogeneous (all look the same) fields and forests than construction sites that have more clear reference features to align the images with.