Pi Spycam: Prototype Deployment

Good Morning from my Robotics Lab! This is Shadow_8472, and today I am trying to solve an animal misbehavior problem around the house, and we don’t know who’s done doing it. And I’m literally dusting off an old project to help. Let’s get started!

Blinkie Pie is a Raspberry Pi 3B+ with a case themed after a PacMan ghost. I modified the files to include a Pi camera module looking out its left eye, but got bogged down with computer vision trying to automate critter punishment. Well, in the present day, one or both of our cats keeps doing something in the same area of the house, and a video feed served over HTTP (to access via browser) could let us monitor things on a second screen. After a brief survey of open source projects, Motion looks to be exactly what I think I need. [1]

DietPi looks like a good base OS. The article I found says it’s based on Debian Buster [2], but I’m only after a short-term project. I checked on Pkgs.org and Motion is listed as in the Debian Buster ARM repositories. [3]

Modern DietPi is based on Debian Bookworm as I found out when I downloaded it. Motion was present, but setup was annoying. I’ll spare the blow by blow, but the terminal program serving as its installer doesn’t think like me. Sometimes it ran quite slow (Pi 3B+ being old?). Credit to the install script where credit is due, but it got hung up trying to update, so I had to drop out and unsnag it manually.

Motion was obtuse to get working. While pounding around for a solution, I found a curated list of software calling the project motionEye. It didn’t find the camera module until after a reboot or two. My only confirmation was terminal activity in response to motion in front of the camera. And working from the command line, I can’t check on Motion’s web interface. A few hours of diagnostic research later, I used curl over SSH and found data on Blinkie Pie’s localhost:8081, but not over network_ip:8081 – which is a wise default configuration with the base OS lacking a firewall, but annoying for my low security use case of monitoring cats. I overrode this setting with a config option at ~/.motion/motion.conf.

# Restrict stream connections to the localhost.
stream_localhost off

I now had images around once per second, but they were coming through upside-down because of how I mounted my camera module in its case. I Modified a similar setting to allow the webUI, but it was very limited. I’d spent a while Thursday night trying to solve this flip issue from the config file, but I only ended up jamming terms from documentation in ways that didn’t work. I circled back around on Friday afternoon and used a search function to find settings for both rotation and flip, but later realized the system only flipped vertically (probably because I used two lines).

With high to mediocre hopes, I deployed Blinkie Pie to watch over Sabbath. The image came out dark, so I found an old desk lamp I had stashed in a closet and left a few other lights on. The good news is that it was more stable than during tentative testing. The only difference I can think of I that I was only SSH’ing over Wi-Fi once instead of for two links.

To access the recordings, my first instinct was to use SCP, a program to transfer files over the SSH protocol. DietPi operating system does not include SCP by default. Instead, I logged in with the FISH protocol, which doesn’t require anything special on the other end besides SSH. Bonus: I could use it with Dolphin (KDE’s file browser). Unfortunately, the default motion detection settings mostly caught human family. Our little, white dog starred in a couple automated recordings, but my black Labrador was seen in one clip being ordered to lay down in the observation zone without any footage from when he got up and left. At no point did I see a cat who wasn’t being carried.

Takeaway

This project didn’t catch anycreature in the act, but I’m still satisfied with my progress this week, and I intend to follow it up later this month where I tweak the configuration to be more sensitive to smaller creatures.

Final Question

Have you any suggestions on tracking cats with Motion or a similar technology?

Works Cited

[1] C. Schroder, “How to Operate Linux Spycams With Motion,” linux.com,July 10, 2014. [Online]. Available: https://www.linux.com/training-tutorials/how-operate-linux-spycams-motion/. [Accessed Mar. 11, 2024].

[2] C. Cawley, “The 8 Best Lightweight Operating Systems for Raspberry Pi,” makeuseof.com, Nov. 7, 2021. [Online]. Available: https://www.makeuseof.com/tag/lightweight-operating-systems-raspberry-pi/. [Accessed Mar. 11, 2024].

Available:

[3]pkgs.org, [Online]. Available: https://pkgs.org/search/?q=motion. [Accessed Mar. 11, 2024].