Hi Forum - first of all, posting in Aerial photography topic as thought it was the most appropriate, please someone correct me if this should be in a different area.
I'd like to select the best most appropriate airframe to perform camera tracking missions to film activities like snowboarding, skiing, mountain biking.
What I have found so far
It seems that their is very little information on the subject,
I found this page on the legacy manual which is no longer updated.
Then, I looked in the updated wiki to find a page on camera stabilization, which is a start, but doesn't include information on camera tracking - now, if this is all you have to do to enable camera tracking and gimbal setup, that's great, however I'm sure there would be more examples of this if it was this easy...
There is a link on the camera stabilization page which should give information on camera tracking, however from reading, it only looks like it is to do with antenna tracking, not camera tracking.
Some more googling revealed various posts on the subject, with little or no information on how it can be achieved...
This next link deserves a bit more of an explanation, It seems a camera tracking feature has been developed, but the instructions state that the camera can only point to the right, and only 45 degrees to the left. If this is correct, how would the gimbal be able to focus on a point 90 degrees to the left. does the plane simply fly right hand circuits of the target?
More Recently in Jan 2013, Tristan tried to do something similar tracking a snowboarder with a quad
more promisingly, it looks like a system has been developed that tracks a gps point, a camera gimbal on a fun cub - however I have found little information on how this is achieved in this example.
I have experience with the Arduplane setup and have been doing basic mapping with a pair of X5's and Canon cameras pointing straight down. I have relatively good success with this. But am ready to move to a new challenge!
I'd like to build a setup that is capable of taking video of a certain point on the ground (fixed waypoint and moving groundstation). I started searching around for this type of information and have found next to nothing in the way of examples of this functionality in use.
I am guessing that the 'point camera here' function of the APM planner manipulates a servo controlled gimbal to point the camera at a certain chosen point on the ground.
My question is this, what kind of airframe would be best for this, fixed wing or multirotor? I know I can achieve 20 minutes of flight time with either, if built well.
What to do next
I could try to fit a roll gimbal to the underneath of a fun cub to test the mechanics of a gimbal, as I have a fun cub in the box I could power with an APM 2.5.
or......Get some sort of multi-rotor.
Due to the huge lack of any evidence to support that this works - I am not sure whether or not to pursue the challenge! however, if anyone has achieved this - then please let me know!
So many questions! :)
For tracking you need a gimbal with tilt/pan/roll. Others can already be discarded. The gimbal also must have GPS support, because it wouldn't be able to figure out where you're pointing at. Even then, it'd probably assume a point at 0 altitude on the geoid (WGS84), instead of actual geometry, otherwise it needs access to DTM information.
There are two approaches; one that integrates with the flight controller and others that use a standalone IMU yet is connected to a GPS through the AP for example. The difference is that a standalone IMU on the gimbal allows you to not worry about the FC it's integrated with. The other requires a specific brand/project for the AP as well.
Multirotors by far are a lot easier to do this work with. MR's can temporarily hold still in the air allowing the subject time to get moving. On a plane you need to synchronize when the person should start moving and you're a lot more limited in choosing how you want to direct the shot. As the plane is always moving, physics dictate it for you.
Gimbals on planes make a bit more sense for surveillance at higher altitudes, as this reduces the travel that a gimbal needs to perform to keep a subject in view. This is at the cost of resolution and definition however, since the subject is further away. You can do it with planes, but as the previous paragraph shows, more accurate planning is necessary either in terms of synchronization or flight planning.
Align has a gimbal (G800) where you basically manipulate the rotation axises to look at a subject. The software than calculates what you're looking at when you're moving to that point and maintains the rotations 'correctly'. It does need to make some assumptions on the current altitude of the vehicle to figure out angles.
The other complexity is thus how you communicate with a vehicle what you're looking at. You can have the vehicle calculate what you're looking at a la align, but you can also apply a ground station approach where you upload coordinates.
If you're trying to do aerial filming work, for all the reasons above, I think it makes a lot more sense to consider a helping hand framing the shot for you. You'll get a lot more flexibility, because once you're piloting the craft you're not likely to have a lot of time figuring out how to instruct the craft what the look at. A second person has all the time to do this and you communicate all the complexity with people instead of clumsy computers.
Hi Gerard, thanks for the information. it is really useful.
From reading, it seems like a multi-rotor is going to be the preferred solution. Although i do like the look of the align G800, that may be out of the price bracket for a little while!
I was thinking along the lines of sending co ordinates from the ground to the plane, A-to tel the plane/quad where to fly over, and B, to tell the gimbal where to point at.
After doing some more research around multi-rotors, certainly the ones in my price bracket, it seems like the top speeds are only around 20mph, which is no where near quick enough for filming activities like snowboarding and skiing.
Which only adds to the dilemma!
I am not massively fussed with framing the shot from any specific angle, If an aircraft camera gimbal could accurately track a fast moving subject, I am sure any video acquired would be adequate for my requirements.
Thanks for the help - it looks like i'll need to wait for gimbal control to catch up before continuing with this idea!
Well, if it's for fast moving action, I really doubt you'll get anywhere through an autopilot ground station. It just takes too long to figure out where the craft is, your subject is and then command the gimbal to "look" at that.
Although the G800 looks really cool, it's also mounted on a heavy helicopter that you only get 8-12 minutes flight time out of. The momentum of those blades is so high that if you fall on top of anyone, you're certainly going to kill that person. I was present with a pilot flying that thing and it's a bit scary to be honest.
Multirotors can go 60mph, like the TBS Discovery or the QAV540G. In those cases you're going to use more powerful motors (in the range of 1000-1100kV), which allows you to expend more energy to go faster forward yet maintain enough lift. Skiing and that stuff is in the mountains and the same potential energy converted into kinetic energy for skiers can be applied to multirotors to gain forward speed. 20mph is certainly on the low end of the spectrum regarding top speed. the TBS Discovery and the QAV540G are more towards the 50mph range, depending on air density, weight and other factors. Both of those have a really nice gimbal setup and allow you to see through the flight camera as well as the gimbal to frame the shot, yet not at the same time.
I don't think automatic gimbal tracking through video is quite there yet. first of all, a number of more professional tracking solutions utilize infrared video to do the actual tracking. The other is that with analog received video you may have snow or discarded frames which would confuse the tracking algorithm. I'm thinking you're better off maintaining a constant relative position with the action to allow the gimbal to do its work or get another guy to tweak the angles now and then to frame the shot.
Hmmm... couple things.
Forget about the G800 gimbal. It's the last of the old dinosaur servo driven gimbals, and performance is not very good. They couldn't even film a decent promotional video for it.
The helicopter it is attached to is capable of much, much longer flight times than 8-12 minutes. I've gone 24 minutes with a similar machine, and it can go longer still with more batteries. The 8-12 minute thing is typical, but only because most people don't know how to set them up properly. But yes, those large helis... not the type of thing you would go flying all over the place wherever you wanted. Flying one at a location like a public ski hill would be pretty reckless IMO.
Some multirotors are capable of flying at higher speeds, but they are specialized machines, not really capable of carrying a camera, at least not at 60 mph. Maybe I've missed something here, but I've never seen it. 40 mph, is more realistic. And I've not really seen many examples of that even. Lots of hovering.
If anybody has some examples of FPV multirotor flights at 40+mph, I'd love to see them.
Is that 24 mins with the gimbal attached? That gimbal is quite a heavy thing. This setup used 12S batteries with carefully aligned blades. Would be interesting to know how this can be doubled on such a standard setup. We never fully depleted batteries, but 8mins usually got us back to nominal voltage.
There's some builds here where they claim up to 60mph, some even 66. I'm not confirming nor denying it's true. This one is in Oz or something, somewhat arid air so less density so less thrust:
When going downhill and hitting forward I'd say you'd be able to get more than 40mph out of it. Not all quads are equipped with these stronger 1000kV motors and 10" props, which makes a lot of difference. Burns the batteries quicker, but if it's needed...
Hi Richard and Gerard,
The ground station solution is not such a bad idea, and prooved to work for me.
I've been working on camera tracking for the last two months to do exactly the same thing as you, except that in my case it's to be used with surfers. I'm using video traking based on OpenCv wich is a grate open source library for video applications. It's power is also it's inconveninent beacause it needs some computational power. Two options are therefore possible : the first one is to get video signal to a computer on the ground through fpv gear. Then process video traking and commands on the ground station, and send back commands to your copter. To send back those gimbal's commands to the drone, I have used an arduino that acts like a dubby controller through trainer port and controls two channels of transmiter.
The second one is to equiped you copter with more computational power onboard. That's why i'm now using the little but yet very powerfull linux computeur Udoo. It is basically an arduino due embed on a quad core microcontroller. It's maybe a little overkilling, but it allows me to have a totaly autonomus copter with vision based control.
In both cases, you could find some realy good tutorials on using OpenCV (like this ones from Kyle) and I can share with you some of my c++ codes.
So far I'have not had any problem with latency when using the first solution, and video traking proved to be realy accurate. For your application it will be too, because the more the object you are traking contrasts with the background, the more traking is accurate and nowboarding are generally wearing colored things...
Maybe my approch of the problem can help you...
PS : sorry for my "French" english...
No, that's true, no camera attached. But I'm pretty sure I could do it.
I've seen quads do 30 m/s. Just not with a camera attached. That video... is hard to see what's going on there. Is he going downwind?
Hi Antoine, Cheers for the information, this all looks very interesting, i think i'll start reading. That could be a little tricky considering my coding skills! but i'll definitely give it a go...
Have you successfully filmed any surfers yet? I guess you'll be using a multi-rotor of some sort?
From what has been discussed above, it looks like the QAV line of multirotors will perform well, and should be able to keep up with fast paced activities, however it looks like they are US based, so postage etc is a bit of an issue to the UK.
I'll do some reading and report back on this thread with my progress...
My first hurdle is to choose, acquire and successfully set-up a multi-rotor. I expect I am going to get a lot of differing opinions about this question, but i'll give it a go anyway...
What multi-rotors are people using as a first/intermediate build?
I'm thinking of a 3DR or a DJI, I've noticed that there are multi-rotors that have two props on each arm, is this for the extra lift, or to provide redundancy? I have seem multiple videos of motors/escs failing, therefore, is redundancy a good idea? does 2 props on an arm even provide redundancy if one of the motors fails? how does the Autopilot know to add more power to that motor?
Here are the platforms i was thinking of (all powered by an APM 2.5)
These quads are readily available in the Uk - however most don't advertise the top speed, probably because of the multiple different configurations...
Id ideally like to carry a gopro, and possibly a 250g point and shoot camera alternatively. I know this isnt the forum for discussing aerial platforms, but any suggestions on what multi-rotor to get would be great!
That's good news - I am currently looking around at the different multi-rotors - and have a sort of shortlist below... further down in this thread...