This is the fixed camera that watches the plant, not the room: one view, framed tight enough to matter, captured with locked settings on a schedule, every frame carrying its own reference. Point it down a propagation bench and it counts what germinated. Point it at a single high-value plant and it catches the lesion while it is still the size of a pencil dot. Its whole product is a stack of frames you can honestly compare.
The need.
What you want to know: what is my crop actually doing, and is it changing?
Your eye is a fine instrument with a terrible memory. It cannot compare today's green against eleven days ago, and it was not in the room at 6am. A fixed station with locked capture produces the record your memory cannot: growth rate you can see, color drift you can trust, and the early small stuff (a lesion, a colony, an off-color corner of a tray) while it is still cheap to act on.
The solution.
When you are done, one framed view of the crop lands at your endpoint on a schedule, every frame shot with the same locked settings and carrying a Station Card. This build captures and pushes; it does not analyze or alert. The comparisons, the time-lapse, and any plant-vision analysis happen downstream, on frames worth analyzing, and can be re-run years later with better methods because the raw frames were kept.
The config: the parts.
| Part | Job | Cost |
|---|---|---|
| Raspberry Pi Zero 2 W (or any recent Pi) | Runs the capture schedule and pushes each frame onward. | $15–35 |
| HQ Camera + a CS-mount lens | The sensor, behind optics you choose for the framing the job needs. | ~$50 + $20–40 lens |
| Camera Module 3 (budget alternative) | Same discipline, fixed lens, autofocus you will lock once and never touch. | ~$25 |
| microSD card + power supply | Storage for the system; mains or PoE power. This config assumes power exists. | $15–25 |
| A rigid mount | The geometry is the instrument. A clamp arm or bracket that nothing can nudge. | $10–25 |
| The Station Card | Printed identity, scale, and color reference, clipped inside the frame. | A few dollars |
Why the HQ Camera is the pick: it takes a lens you choose, and the lens is usually the real limit on what a camera resolves, not the megapixels. An inexpensive CS-mount lens matched to your framing beats the molded plastic in front of most sealed cameras. The camera hardware guide covers the parts; run your framing through the resolution calculator before you buy anything.
The Station Card here.
The card is not an accessory on this build; it is half the instrument. Clip it at the edge of the frame, flat to the camera, and leave it there. Every frame then carries the station's identity (which bench, which view), a scale bar that makes the millimeters-per-pixel figure checkable per frame instead of taken on faith, and color patches that let any frame be corrected after the fact. When the camera is ever bumped, replaced, or upgraded, the card is what lets the record continue instead of starting over. Print a fresh one when it weathers; the identity carries forward.
The build discipline.
The hardware is an afternoon. The discipline is what makes it an instrument:
- Frame for the question. Decide what the smallest thing you need to see is, then set the distance so enough pixels land on it. Tight on one plant beats wide on a room.
- Lock everything. Exposure, gain, and white balance are set once, by hand, and never left on auto. The Pi's camera stack supports full manual control, which is exactly why it anchors this config. On a Module 3, lock the autofocus at the working distance on day one; focus hunting across a season quietly destroys scale comparability.
- Mount rigid. The fixed geometry is what makes frames comparable for free. If the mount can flex when someone brushes it, it will.
- Shoot on the light schedule. Capture at the same times each day, under the same declared light state. A frame at 10:00 lights-on compares with every other 10:00 lights-on frame; a frame at a random moment compares with nothing.
The software.
Coming
The capture software for this config (the scheduled capture with locked settings, the card check, and the push to your endpoint) is being written and is not downloadable yet. This page is the spec it is being built to. Until it ships, the config is buildable by hand by anyone comfortable with a Pi: the camera stack's own tools support locked manual capture on a schedule today, and the discipline above is the whole recipe.
Where the frames go.
Same principle as every reading on this site, with one honest difference: frames are bigger than numbers. A high-quality still every fifteen minutes runs on the order of 2 GB per camera per month, so the endpoint needs disk, not just an inbox. Keeping the raw frames is the point; they are the part that can be re-analyzed in three years with methods that do not exist yet. See Connectivity for the endpoint choice, and budget the storage before the fleet grows, not after.
Honest limits.
- We have not run this station through a season. The parts, the arithmetic, and the capture discipline come from published specifications and the camera stack's documentation. When our builds have been in a grow room for months, what we learn lands here.
- It sees one view. The fixed frame is the feature and the limit. More views are more stations; the close-up of something suspicious is the phone's job.
- It needs power and network. Off-grid is the Solar Field Camera's territory, and the two do not trade places well.
Frequently asked questions.
What is the best camera setup for monitoring plant health over time?
A fixed camera with locked exposure, white balance, and focus, framed tightly on the plants you care about, shooting scheduled stills with a reference card in the frame. A Raspberry Pi with the HQ Camera and a lens you choose is a strong affordable base because its camera stack allows full manual control, which consumer cameras mostly do not. The lock-everything discipline matters more than the brand.
Is a Raspberry Pi camera good enough to detect plant pests or disease?
Framed correctly, yes, for the larger early signs: at roughly 0.3 mm per pixel a 12 MP sensor puts around 8 to 10 pixels across an aphid or a young powdery mildew colony, which is the edge of reliable detection. Framing sets that number as much as the sensor does, so check your distance and field of view with a resolution calculator before building. Spider mites are below any room-scale framing and remain a close-up job.
Why lock the camera exposure instead of using auto mode?
Because auto exposure and auto white balance re-decide on every frame, so the same plant photographs differently day to day and the changes you care about drown in the camera's adjustments. Locked settings make the camera boring and the record honest: what changes between frames is the crop.
How much storage does a plant time-lapse camera need?
Plan on the order of 2 GB per camera per month for high-quality 12 MP stills every fifteen minutes, so roughly 25 GB per camera per year, and scale from there. Keeping the raw frames rather than just derived numbers is worth the disk: raw frames can be re-analyzed later with better methods.