Standalone camera that obscures subject
July 8, 2021 7:48 AM   Subscribe

I need a camera that I can set up and let run for a few days that will record a subject and obscure their identity.

This will be for a study for a new product. The subject will generally be in one spot for hours, and we just need to know when they've moved. We don't need to see any details of the person, and we think that subjects would be more comfortable knowing that their face would be obscured.

I've looked at things like the Kinect camera, and similar things from Orbbec and Intel, but they all appear to require a connection to a computer, and they also seem more intended towards developers than end users. Something like a trail camera or security camera might work, but those are intended to identify the subject, not hide the subject, so I haven't found anything that fits the bill (or maybe I'm using the wrong search terms).

• The camera needs to work on its own. It can't require a connection to a computer.
• The camera should have an SD card slot.
• It should be able to run continuously until the card is full.
• It can be AC or battery powered
• It doesn't need to record high resolution video, and ideally, we'd be able to set the frame rate very low to allow greater recording time.
• It doesn't really matter how it obscures the subject, so maybe something like a thermal camera would work.
• It needs to time stamp the videos so we can see the time that the subject moved.

Any ideas?
posted by jonathanhughes to Technology (14 answers total) 1 user marked this as a favorite
 
You could use a trail cam with a physical filter over the lens - enough that you can make out the location of the subject, but can't make out details.
posted by zamboni at 8:21 AM on July 8, 2021 [2 favorites]


I could easily *make* something like this, say with a raspberry pi and it's camera peripheral, but I don't know of anything off the shelf, it's a bit niche to be a product. The thermal camera idea isn't bad, look at FLIR cameras maybe.
posted by RustyBrooks at 8:26 AM on July 8, 2021 [3 favorites]


Do you have control over the lighting in the space? Do you have control over the color of the floor/walls/etc? Do you have control over the clothes the people wear?

If yes on some or all of those questions you could probably just HORRIBLY over or under expose the video and have obscured people while still being able to track motion.

There are also tools that just sense motion which might work for your purposes. This thing attached to a Raspberry Pi could easily track motion and record timestamps.
posted by gregr at 8:33 AM on July 8, 2021 [1 favorite]


Best answer: This paper uses frosted glass (with tape to adjust light transmission) to blur footage for a parking capacity detection system.
posted by zamboni at 8:34 AM on July 8, 2021 [6 favorites]


Some clarifications:
It's sort of implied, but not explicitly stated: You only need the retroactive ability to determine if the person moved, and do not need any sort of real-time notification?
What level of accuracy are you looking for, or how much movement do you need to detect? Do you want to know if the person blinks, turns their head, moves a finger, moves their arm, crosses their legs while sitting in a chair, shifts their weight while standing, rolls over in bed, or moves more than 1/5/10 feet away from where they started? Is it acceptable to miss detecting that they moved but returned to the same position before the next image was taken?
Do you need the camera to flag the motion, or is it acceptable to require a human to review the recording and manually identify it?
What physical constraints are there on the position of the camera relative to the environment and to the subject?
Are the lighting conditions consistent throughout the recording period, and between one recording period and the next?
What skill level will the operators of the camera have? Does it need to be a one-button start, or would they be able to asses the image quality and fine-tune before each recording period?

One option would be to take a standard digital camera with manual focus, intentionally set the image to be out of focus, and set it up to take a time-lapse. You could also add one or more lens filters. Depending on the model of camera, that may be more-or-less complex, and you may not be able to hit the tradeoff between anonymizing the face but detecting the movement with enough precision, but a lot of people have a camera sitting around, so you might be able to try it out before committing. I'll note, also, that CHDK is an alternate firmware that adds additional functionality to a lot of Canon cameras, so you may be able to get time-lapse and manual focus out of a camera that doesn't support it out of the box, with a bit of legwork.
posted by yuwtze at 8:48 AM on July 8, 2021


Or on re-reading the question, I missed the "few days" portion. I was assuming that the intent was that the subject should be motionless for the entire period. It seems like the intent is more to let the subject move naturally, but you want to know where they moved within a particular space?
posted by yuwtze at 8:56 AM on July 8, 2021


If you're at all comfortable with wiring, you probably don't need a camera at all, just a PIR sensor, a 9v battery, and a simple digit counter. You could probably knock out 4-5 of them for the price of a camera, and arrange them around your subject so that the count increases whenever a movement is captured.
posted by aspersioncast at 9:27 AM on July 8, 2021


Would an out-of-focus image work? If you can manually set the focus on the camera you can adjust it so that the person is just a vague blob.
posted by It's Never Lurgi at 9:49 AM on July 8, 2021 [1 favorite]


Response by poster: “You only need the retroactive ability to determine if the person moved, and do not need any sort of real-time notification?”

Correct

“What level of accuracy are you looking for, or how much movement do you need to detect?”

The subject will be in a bed, so we need to see if they get out of bed, but ideally, we’d be able to tell if they roll over or make similar larger movements.

“Is it acceptable to miss detecting that they moved but returned to the same position before the next image was taken?”

No. But we figure that a frame rate of one per second would suffice.


“Do you need the camera to flag the motion, or is it acceptable to require a human to review the recording and manually identify it?”

The footage would be reviewed after the fact, based on the time stamp.


“What physical constraints are there on the position of the camera relative to the environment and to the subject?”

It would probably need to be mounted to a wall

“Are the lighting conditions consistent throughout the recording period, and between one recording period and the next?”

No. The lights could be one or off (and it could be day or night), so some level of low light capability would be very helpful.


“What skill level will the operators of the camera have? Does it need to be a one-button start, or would they be able to asses the image quality and fine-tune before each recording period?”

Ideally, it would require minimal skill. Turning it on and letting it run would be the best.


And out-of-focus image might help, and the suggestion above to use frosted glass or plastic may be the simplest option. The raspberry pi ideas could definitely be an option, but I’m hoping for something simple and unobtrusive that also doesn’t look like it was put together from scraps.

thanks!
posted by jonathanhughes at 10:33 AM on July 8, 2021


Searching for "motion detector camera" brings up options from about $12 up. Some of the ones at about $35 look credible. Read carefully. Some require a computer connection, some that don't may require a computer for setup or to download the pictures. Some are meant for outdoor use and to turn on lights.

Is plug-in power OK or is battery power a must?
posted by SemiSalt at 12:40 PM on July 8, 2021 [1 favorite]


I think the out of focus footage or frosted glass cover ideas are likely cheapest to implement, but an infrared camera like the ones used for thermal imaging would give you footage that is great for identifying movement but terrible for identifying people and would work great in low light.
posted by Aleyn at 12:57 PM on July 8, 2021 [1 favorite]


I forgot to add photographers have used Vaseline on lenses to get blurred edges for a long time. With a little experiment, it should work to hide identity.
posted by SemiSalt at 1:09 PM on July 8, 2021


Best answer: Blurring:
Cut a small piece of kitchen cling wrap and use a rubber band to secure it over the lens. Smear some petroleum jelly on the plastic. Alternatively, cut a piece from a plastic bag that isn’t perfectly transparent (like the bags in a supermarket produce section) and you may not need to use any petroleum jelly at all. Layering a couple pieces might give you best results. You’ll need to experiment a bit to find the right amount of blurring.

The Camera:
I wouldn’t rely on a trail/wildlife camera to detect all the motion that might be relevant for your purposes. They can miss a lot of stuff.

A camera with continuous video recording (CVR) capabilities with timestamp enabled might be a better option. When you review the footage you can just play back at high speed or scroll through it quickly until you spot any relevant motion. Most of these will store video in the cloud, but there are a few that have SD card options. Some of them, especially ones that are labeled 'smart' will only record motion events on the SD card, so make sure you're getting full CVR on the SD card. There are a couple of cameras on this page that fit the bill.

If you are at all open to the cloud option, a company called arlo has cameras, wall mounts, and AC power adapters that could work very well for you.
posted by theory at 4:35 PM on July 8, 2021


There are raspberry pi projects but I also ran across an iphone app geared towards anonymized news recording that might work.
posted by radio other at 7:11 PM on July 8, 2021


« Older Help me take amateur photos of arteries in my...   |   How do I entertain a teen around Crown Heights? Newer »

You are not logged in, either login or create an account to post comments