Should motion detection use images or still frames?

Question

Some Context

For a project I'm working on, I'm creating a security camera that detects motion. When the camera has detected motion, it checks again to see if that motion was caused by a human. This avoids the camera being triggered by animals, cars etc.

Now there are two main ways I thought I could do this.

1. The original way I was going to do this was capture 5 frames. I would then process these images to check for motion by averaging the pixels to create a single image, and comparing that to the image from the last 5 frames. If a certain threshold of pixels had change, motion would be detected.

2. I could do this using a stream and continuously checking the live image for motion. This seems better for the next phase, which is to identify if that motion is caused by a human.

I'm going to be using this project on this Picamera from Amazon. I have a Raspberry Pi Model 3B+

Based on this hardware, which is the most viable option. Theoretically, I can code both but would doing it in real time put too much pressure on a Raspberry Pi? I don't want to overwork it, but I do want to get the most out of it.

I did ask about this question on Meta: Help narrowing down a broad question but it didn't receive any replies

Answer 1

one answer no one would like to hear when asking a question is probably "it depends", but in your case: it really depends.

As you have a concrete use case, you could start with a simple project: By setting up motion (https://motion-project.github.io/motion_config.html) and playing around with the configuration options for a motion, e.g., how long should a video be recorded, when pixel changes over a certain threshold are detected. You can then setup an alert informing you whenever a new motion is detected. This way you can get a feeling for what is missing in your current approach and what suffices your needs. If you do not like to completely start from scratch - there are a lot of blog posts on surveillance with raspi and motion, where you can decide for the approach which best suits your needs.

A few more notes/examples on 'it depends':

• Sometimes the first idea is always 'lets use a camera for solving problem XYZ', but after discussing in detail it turns out that another sensor would do a better job
• Think about the priorities, is it real-time you most care about, e.g., because you want to open the door before your visitors are ringing the door bell? Is it the detection of human and/or the separation between human vs. animal? Is it the motion and with the motion you are also interested in direction, acceleration, ...?
• ...

btw. the title confuses me: the phrase "still frame" in image processing refers to image