I have attempted mobile object tracking at ranges of 2 to 20 m using OpenCV and RTK GPS with mixed results. Has anybody tried to use a ultra-wide band transceiver for this purpose? If so, what was your setup and how was the tracking performance for your application?
1 Like
Hi Richard,
Could you please share some more details about your setup?
What kind of environment are you operating in?
What algorithms are you running?
I haven’t gotten around to using any UWB in projects yet but have seen plenty of examples in application, but they still require additional sensors in combination for good results
Hi Liam, my previous projects using CV and GPS for object tracking were to track and video a fellow skier or cyclist whilst both the camera base and target person were moving independently in an outdoor environment. The results were okay most of the time, however reliability was not “fit for purpose”. I notice that a follow me suitcase using UWB is being marketed and wondered if this technology is more reliable for slower object tracking, such as a follow me golf cart. There are tutorials to set up a UWB module on the web, but I have not seen UWB modules being used for object tracking. My algorithm for CV was based mostly on OpenCV MOSSE. For GPS I modified UBlox’s setup from their F9P Application Notes and used proportional motor control in python, but the weakness was signal strength, i.e. loss of RTK fix. I’d now like to try UWB tracking if there is evidence that it is suitable for these sorts of applications.
1 Like
Hi Richard,
Interesting project!
Without knowing more about the project I would advise that if the recording station and the point of interest could communicate and also include an IMU (atleast an accelerometer) on each.
UWB would be a good alternative to GPS in regards to getting initial angles if the POI has been lost - but likely only up to a range of 10s of m (I haven’t actually used UWB, so unsure of practical range).
What method are you using for RTK?
How are you integrating the CV and GPS measurements?
What kind of camera are you using and what range are you expecting the system to work at?
1 Like
It seems that some UWB modules have a range of up to a few hundred meters which is more than I require. For tracking with RTK GPS in my project a few years ago I use the Ublox ZED F9P module in moving base mode, connected by a pair of Xbee’s, by parsing the RELPOSNED message.
My CV and GPS tracking projects were separate and not integrated due to size/weight limitations of being helmet mounted.
I used a separate camera for CV tracking and a GoPro for video recording at a range of 2 to 20m.
Another newer technology for object tracking is Bluetooth Angle of Arrival, but like UWB, there does not appear to be any current use in an outdoor/dynamic environment.
2 Likes
Hi Richard,
I would be sceptical with the hundred or so m range especially in an environment with moving objects getting in the way of signals.
To get a good angle measurement you’ll need to spread out the base stations receivers quite a bit.
It might be worth posting on a control systems forum or Reddit group.
If you could share a bit more about the software stack you are using for the CV we might be able to help more.
What performance are you after as well?
Yes, it appears the that the two UWB base stations (on a helmet or golf cart) may be too close for effective triangulation. Nevertheless, sensors on commercially available follow me golf carts and suitcases appear to be using UWB tracking.
The main problem with my OpenCV helmet mounted object tracking (at about 90fps on a ski slope) was the challenging environment, i.e. the MOSSE tracker code worked fine most of the time but if the camera person moved their head too rapidly then the target skier would be lost. Then both skiers would need to stop, and the target skier manually redesignated (as there are often other skiers in the vicinity), which is too disruptive. In theory, if the tracking method uses an effective tag/beacon (such as GPS or UWB) then tracking has the potential to be more reliable.
I guess I will just have to trial UWB tracking myself to see if it’s suitable for my applications. I suspect it may work for a “follow me” golf trolley as it is big and slow, but it won’t be suitable for helmet-based camera tracking.
2 Likes
Hey @Richard83832,
What an awesome project!
To increase the distance between the two base stations, I have had luck with triangulation projects in the past by just throwing it on top of a long pole. Probably not a super useful idea for a helmet mount but I could see this working out for the golf cart.
As for the helmet-mounted station, is the camera mounted directly to the helmet or do you have some kind of stabilisation implemented? Gimbal rigs seem to be constantly dropping in price, maybe you could implement something like that into the helmet rig?
