A timely post,
Have been looking at the GPS issue from a different angle this week - improve the hardware in the camera for better accuracy even in harsh environments removing the need for any post capture additional work.
Watch this demonstration of one of the latest chips.
Jump to 3:45 for the start of the demo.
@JimGayes360 great find with that video.
I’ve heard a lot of people running Street View cameras on top of these cars using these chips (where you have the advantage of powering the system using the car).
Seems like a fun weekend project to build a complete receiver: https://learn.sparkfun.com/tutorials/gps-rtk2-hookup-guide/all
Were you thinking of buying one of these chips or a full receiver kit @JimGayes360 for GSV work?
I’m still not sure of how I’d power the system (safely and consistently) when on the move. Any advice would be welcomed
Just doing some initial research on how it might be possible to improve GSV captures for better accuracy and been reading up on what might be involved.
As you know canopy cover, bridges and tunnels are the enemy for accurate GPS location.
For GSV video mode captures it’s not possible to do anything in post I don’t think? (Software is not my specialty) So to my way of thinking it would be easier to make highly accurate captures in the first place.
With a high quality GNSS solution attached to the camera you would also be able to capture individual timelapse captured panos as well as video mode.
I’d seen the Sparkfun board and a few similar products but I’d not come across the setup link that you shared, so will read with great interest later.
Getting it to work with the camera maybe a hurdle and needs further investigation.
Your question about powering the unit I think can be answered quite easily - Power bank battery.
Reading this part of the article -
A 3.3V regulator is provided to regulate the 5V USB down to 3.3V the module requires. External 5V can be applied or a direct feed of 3.3V can be provided. Note that if you’re provide the board with 3.3V directly it should be a clean supply with minimal noise (less than 50mV VPP ripple is ideal for precision locating).
The 3.3V regulator is capable of sourcing 600mA from a 5V input and the USB C connection is capable of sourcing 2A.
I have a 20,000Mah battery on my backpack already to power my camera and phone on long captures.
Powering this module alone - 20,000/600 = 33.3 hours, so shouldn’t be an issue.
There is a brilliant person - SviMik who has connected the f9p to a Kandao Qoocam 8k using the audio input. Its a pretty sophisticated setup but you likely understand hardware better than I.
I shared the link to this discussion with him so hope he may chime in.
All the best.
Since people are asking, I’m posting a bit more details:
I use two GPS modules on my car: ublox F9P and ublox LEA-6R. The first one is a RTK receiver with 10cm accuracy, and the second one is a dead-reckoning device for underground navigation (and whenever RTK receiver fails due to signal loss - driving under a bridge, thick tree canopy, etc).
For F9P I used this board and antenna for my setup:
I also connected ESP32 as a WiFi adapter and then I share internet from my phone so it would be able to connect to a base station. It is also possible to use a Bluetooth adapter for that purpose, although it will make it dependent on a running app in your phone, which is slightly less reliable. A 3G module can be used too, but it will require an additional SIM card.
There is another great receiver specifically for cars - u-blox LEA-6R. It works even with intermittent GPS signal (you can go into a tunnel or underground parking and it will still works), it doesn’t need a base station nor Internet connection, bust must be connected to your car’s ABS sensor, so it would be able to sense your car’s movements when GPS isn’t available. Its precision is 2m and it’s a bit cheaper than F9P, but again it’s strictly a car solution, while F9P can be used on any kind of vehicle or even when walking.
The LEA-6R board I use was made by a Russian enthusiast from OSM forum, you can read more about it here (in Russian, but there are screenshots and photos too):
This device is able to back up the navigation for 15 minutes with no GPS signal, still providing positioning with 2m accuracy.
An example of underground parking track:
I was really pushing it to the limits here, but even with that many turns and a bit longer than 15 minutes driving underground it produced a track that can be used to create a map of that underground facility.
In order to merge the signals from two GPS receivers into one I use Arduino Mega board, switching to whichever receiver has a better solution at this moment. It goes then to my phone via bluetooth and to my laptop via USB, both recording for redundancy.
To synchronize external GPS with the camera I use QooCam audio input, feeding it both timepulse and NMEA signals. Then I use my own program to extract frames from the video file and attach GPS tags to them.
So my workflow would be to connect the power and audio cables to QooCam, put it onto my car roof, start the video recording and go driving. Once done - I go home, move the files to my PC, stitch the video with QooCam Studio (it’s painfully slow, 2-4fps is a typical speed, so can be left overnight as well), then I run my program to convert the video into geotagged jpegs, and then run mapillary_tools to upload these jpegs. With a single PC I’m able to post-process about 1 hour of video per day, which gives me 50 km mapped distance or 10k images uploaded on an average scenario.
F9P with ESP32:
A test of accuracy where I took the location from a produced jpeg file
and put the coordinates on top of aerial image (red dot).
The camera might need some calibration though, mine was 1 frame (1/30 second) behind the audio track, but that offset is constant, so need to find only once for your camera model.
impressive setup! Is it all hobby for you? Where do you use it for?
I have read that the QooCam has problems in consitancy when used longer (blurred focus among other things). You record for an hour, what is your experience with the camera?
This particular setup is for OpenStreetMap mapping, which is my hobby. My main occupation is programming, though I do hardware development occasionally too.
The main disappointment with QooCam is the stitching software, which is really slow. Other than that I do not have any immediate complaints. I hadn’t had a chance to compare, but from reviews its quality is compared to GoPro Max (though GoPro doesn’t have an audio input which I found useful). Anything better would be one digit more expensive, or would be limited to take pictures only, while I record video because it gives me much more flexibility for post-processing. And since QooCam photo and video resolution is exactly the same, I don’t feel any trade-off when recording videos for extracting pictures from it later. I’m not saying QooCam is a perfect camera, but I can’t suggest anything better for the same budget.
I have also made a remote for QooCam using Arduino, so I would be able to start/stop recording and monitor the free space without needing a second phone just for that (because my phone is already busy sharing internet over WiFi and providing me navigation). Here’s what the remote looks looks like: https://youtu.be/7t6Pf2q4g2Y
Have you tried Hugin? I know that one isn’t fast also (but I love the quality… it’s much, much better than the stitching software for my YI-360, and also I can use it via the commandline… which helps me a lot in processing/automation)
I’ve never used Hugin, but I’ve heard it needs a template for each camera model, and since my camera was too new there wasn’t a template available (maybe it is now?), and I didn’t bother make it myself. The other argument was that I’ve also seen some comparisons with other programs, and QooCam Studio does better with color aberration. Alas, can’t remember the details, but I found the examples, with QooCam Studio on the right:
After that I decided to not waste my time experimenting if the outcome is likely to be worse anyway.
Yikes… could you post an unedited image here?
PS: might be a good idea to start a QooCam topic? We’re getting a bit off topic here
Not sure, but I guess this is the one:
And he is using MistikaVR, so it wasn’t Hugin.
A bit too late to start a new topic with several posts already here. Maybe a moderator can split the thread?