Our sweet little web based Aerial Imagery and UAV Flight Tracker with Ortho-Mosaic/GIS/KMZ Tile Builder. Beautiful Orthophotos!

Hello fellow UAVers,
My UAV partner (Ross) and I have been working on a project that we'd love to get some feedback and additional ideas for. Our intention is to allow free access to this web based tool and also a really cheap subscription model per month (think $4 range) to support hosting costs and the large amount of cpu processing power needed to generate maps. The free users will get all the same features but lower resolution imagery and possibly less accurate georeferenced tiles/tiffs/added watermarks to support the site. No ads.

Our software and backend servers use tons of different open source packages to complete the process of georeferencing and building image mosaics. We are also trying to offer an "easy to use" UI to view each flight and all the associated data. We really hope you enjoy our ideas! We are planning to open the site up to beta users by request in a month or so. Please keep an eye out for more information on that.

We've been having some problems getting our own flight systems up and running, ranging from a bad MUX on an APM 1 (thanks for fixing!), APM2 delays, and code issues with the Tri-Copter platform. This has prevented us from getting amazing imagery on our own at this point. :( Lucky, the awesome guys at Petryx UAV posted a ton of sample imagery which is so cool! Pteryx FTP Sample Imagery

We can't wait to open the site to beta users so you can test your own imagery and flight paths, hopefully at that point we will have our own platform up and running!

So, here we go -- some screenshots and comments:

Upload your flight imagery and Ardulog/GPX/Petryx/CSV log file.
Imagery and Flight Processing Queue
Imagery Dashboard w/ Flight Path, GeoTIFF served via Geoserver, Blended and Georeferenced Mosaics


Same as above, different flight.


Georeferenced View w/ Google Earth API - Shows Tiles, Flight Path


Second Part of Georefenced View - Shows GeoTiff Mosaic Preview and Metadata


Third Part of Georeferenced View - Shows Additional Maps

Blended Mosaic

Our Tricopter platform. Quadcopter and also Skywalker platforms we've been testing!


Quadcopter w/ IR Powershot a490 Camera

We would love to hear some questions and comments!


Thanks for reading and thanks for your support,

DM Team (JP and Ross)

Views: 9425

Comment by Arthur Benemann on February 14, 2012 at 8:38pm

Can you share the way the photos and the APM log are sync?

Comment by Ross on February 14, 2012 at 9:11pm

James...In regards to accuracy we will be able to give an estimate soon. But it depends a lot on where the GPS Sensor is mounted in relation to the camera. On the tri we are flying...the sensor is about 2 inches away from the camera. So that throws things off a little. I am pretty sure the interface lets the user tweak these settings to get more and more accurate.

Thanks for checking it out.

Comment by JP on February 14, 2012 at 9:38pm

@Arthur.benemann Sure, we've got a bunch of php scripts on the backend which process the imagery after it is inserted into the database. The php scripts match the flightlog data/gps location/camera direction/elevation with the timestamp of the images, building kml and other datasets. It also can use the EXIF GPS data of the images but this is often even less accurate vs. APM/flight controller logs. Thx

Comment by JP on February 14, 2012 at 9:45pm

@James -- Ross and I have been talking about a reports function with some information like this. It hasn't been coded yet, but should be pretty straight forward. Also, an embedded option to post your maps/flight on your blog/forum site would be cool. Matching a mosaic against a set of georeferenced control imagery via WMS is what we are working on next, which would yield some good results for a report and contain accuracy information. Right now, the GeoTIFF files that are built are as accurate as the onboard sensors like Ross said. :) Thanks JP

Comment by The Old Flyer on February 14, 2012 at 9:55pm


Your service seems to be exactly what we need.

See my post:  http://diydrones.com/forum/topics/a-successful-apm-aerial-survey-mi...

We have 124 photos taken with a Canon SD1100 from 100 Meters, flown in an APM equipped Kadett.

The mosaic was stitched with PTGui.  After a lot of hand intervention the results were very good.

Georeferencing  was not a requirement for this mission.

Our next attempts at aerial surveying will be with the Canon S100 camera that will give us EXIF GPS data on the photos.  Georeferencing will be a requirement in the future.

Could your service process our GPS EXIF data and APM Log files to produce the position of each photo on a Google map of the area?

Could your service produce the same map from AMP Log files and EXIF data without GPS?

Will your service stitch the photos into Panoramic mosaics?

Is there any way we can help with your development efforts?  

We can supply photos, APM .log, .tlog, .gpx, and .kml or .kmz files from our aerial survey mission. 

Thank you in advance for your efforts.

Comment by JP on February 14, 2012 at 10:12pm

@Irvin Stafford aka The Old Flyer

Hi Irvin, thanks for the link. Awesome work btw! I actually read that a few days ago and sent it to Ross (he is the hardware guy and working on our flight platforms). Sorry I was too busy to respond. We just started building a Skywalker platform and your post is a really good reference. :)

DM can build mosaics from GPS EXIF information but the problem here is most cameras don't have compass heading so it is impossible to get the orientation of the photo. We found a way to hack around this by writing a small php script which will try and calculate the heading of the UAV from GPS waypoint A and GPS waypoint B. 
Additionally, the camera gives us the above sea level elevation which is not ideal. We wrote a script to query the USGS digital elevation service for the GPS waypoint to calculate the above ground elevation and feed this into the database. The APM or Petryx logs seem to be the best option for most accuracy at this point and takes care of most of this.

So the service would produce the same map with or without GPS EXIF but the accuracy would suffer.

The stitching happens two ways, one uses OpenCV and a bunch of control points to automate the building of the mosaic. So, the service can build mosaics on the fly with or without any GPS information using this method.

The second idea uses gdal tools to build the mosaic, which is georeferenced. This is feed into Geoserver automagicly using the rest api and displayed via WMS. Hope this helps, let me know if you have any other questions!

We would love to get some help with testing and development, thank you! There will be a beta soon and beta users get a free account for life. Thx JP 

Comment by MarcS on February 15, 2012 at 1:57am

Ross, you are saying you are calculating the absolute position and orientation of the result by use of GPS? The accuracy will then be limited by GPS and IMU quality, so I guess would be correct to about some meters, at best...  The placement of the gps related to the camera will not have a significant influence, especially in planes, where the movement will make the position even harder to estimate... If you want to have good georeferencing you will probably have to provide some reference points in the pictures.

Comment by Emin Bu on February 15, 2012 at 2:45am

Bravo....what a great project..i am just a beginer in this but i am willing to learn... maybe something from this video helps(but i am sure you already saw it)


Comment by JP on February 15, 2012 at 6:19am

Hi Em Bu- Thanks for the link.

We have definitely looked at Palentier, it is great software and helped inspire this project along with diydrones/arducopter. :)

Comment by Andre S on February 15, 2012 at 7:42am


Interesting tool/concept that is badly needed. I was wondering, being new to aerial photography/image stitching/georeferencing, is there any more detailed background information on your photo processing toolchain available? You mentioned a few things, such as opencv and what the display uses but I'm really interested in the processing part. In the middle and longterm I'm looking for a batch processing solution that ideally would not require any user interaction at all (maybe picking one reference point in a single picture, fixing that picture's orientation or so) and would scale to possibly a few hundred photos. Computing time itself would not be a limiting factor as long as the method is scalable (thinking of Hadoop MapReduce based algorithms). I actually remember I saw a paper on a related topic once but can't quite recall it.

Anyway.. interesting stuff, keep up the nice work. Guess you won't have trouble finding users for this thing here..

Cheers, Andre


You need to be a member of DIY Drones to add comments!

Join DIY Drones

© 2018   Created by Chris Anderson.   Powered by

Badges  |  Report an Issue  |  Terms of Service