PiPole-Cam Fly's Again Out Bush
In mid-November 2013, a good friend and I travelled South to record GPS coordinates and measure tree height of a patch of Xanthorrhoea grass trees that I had stumbled upon back in April 2013.
The night before, I had upgraded PiPole-Cam with the latest version of Picamera - 0.7. Although I only tested my script with a maximum of 100 photos in each run, it all worked well, and there where no dark photos.
Together, we recorded 252 grass tree positions and their heights. Here is a plot of the grass tree locations that we have recorded so far:
Yes, we did a bit of an intentional clockwise circuit, but didn't mean to skip the central part. I am guestimating that this might be about one third of the patch size, so we have a ways to go in completing this project. Here is a plot of the same data, but histogramming the tree height by colour:
So, enough about Xanthorrhoea grass trees, what about the PiPole-Cam. Well, it was a bit of a hindrance, as we where really there to measure the grass trees, so I left it dangling from two different branches:
But it's perhaps easier to see the photos taken from the two locations, if I remove the background imagery, grass tree GPS data and colour the first position red:
Yes, in the first position (red diamonds), the script cut out after 29 photos x 20 seconds =580 seconds. This was the second time it had cut out, then I realised that I should either run the script via 'screen', or just put it into the background by appending a ' &' to the end of my CLI script. You can see that it took the requested 100 photos on the second run. This ran for a bit over 100 photos x 20 seconds = 2000 seconds = 33.3 minutes.
On the second run though, you can see that the reported photo locations where scattered about the place, even though apart from one wild reading, most where recorded with a horizontal error of under 20m. In fact, removing the one error of 142m, the average GPS error rate was 13.3m and was as low as 9.5m:
As I have previously seen, if a GPS unit is blocked from seeing the full 360 deg. view of the sky, it can still report a good error rate, but the data can be thrown off. On the second site (cyan diamonds, I did have the PiPole-Cam hanging from a large grass tree, so that would have been blocking its view of the Southern sky. Perhaps that's why a lot of the locations are being reported to the North and north West?
In the previous post, I mentioned that perhaps the RPi might be over heating and that might be causing the fade-to-black photos. This time, I had the RPi record the GPS coordinates and CPU temperature to a logfile. It was nice to see that the RPi CPU never got above 55 degrees Celsius. Most of that was ok untill the last 10-13 photographs at the end where it rose to 55 deg. C!. Was this due to the battery voltage running too low? Here is a graph of the CPU temperature taken at 20 second intervals:
Unfortunately, this was the second visit running on the same set of batteries. I had brought along a spare set, knowing that these might give up at some point during the visit, however I had left them back at the car. Interestingly, the RPi was still running, and I could cold-reset it and it would power back up, but I couldn't get a wireless connection to it. Perhaps there wasn't enough battery power/voltage to run the Wi-Pi wireless unit as well as the RPi? Back at home, I measured the battery voltage to be 3.62V each. Before I set out, they where 3.85V. A freshly charged battery is: 4.22V I do have an Adafruit I2C based current and voltage that works well with a RPi (yet to write a blog post on this handy device), but I won't be integrating it on the PiPole-Cam. I will be using them on my future Remote Weather station project.
To finish up, here is a sample photo of a small ~ 1m high (we measure to the top of the dead leaves) Xanthorrhoea grass tree: