Build a Motion-Triggered Trail Camera With the Raspberry Pi HQ Camera (2026)

Build a motion-triggered trail camera with the Raspberry Pi HQ Camera by pairing a Pi Zero 2W or Pi 4 with a PIR sensor for wake-up, using libcamera or picamera2 to capture stills or clips on trigger, and writing to fast storage sized for how long the camera will run unattended. As of 2026 the whole build fits in a small weatherproof case with a battery pack, runs for weeks between visits, and produces sharper images than any consumer trail cam under $200.

Why build vs buy?

Consumer trail cameras are cheap, fine for a general "did anything walk past" answer, and terrible if you want:

• Better than 8 MP images with configurable optics
• Anything approaching video quality above 720p
• Programmable behavior beyond "PIR triggers, save file"
• Custom uploads, ML classification, or integration with home automation
• Long-term firmware you actually control

The Raspberry Pi HQ Camera is a 12 MP sensor with a CS-mount for interchangeable lenses. Paired with the Raspberry Pi Zero W kit or a Pi 4 board, you get a rig that outclasses commodity trail cams on image quality and gives you full software control. This guide walks the whole build: hardware choices, wiring, capture code, power, and enclosure.

Key takeaways

• HQ Camera + Pi Zero 2W is the low-power sweet spot; Pi 4 if you want ML on-device.
• A PIR sensor wakes the Pi for events; software motion detection is more flexible but hungrier.
• Fast SD or a small SATA SSD determines how long you can shoot before pulling data.
• Battery + solar gets weeks of runtime for a well-tuned build.
• Weatherproofing is boring but the difference between a working camera and a dead one after one storm.

Hardware bill

Core components:

• Compute: Pi Zero 2W (low power) or Pi 4 (on-device ML)
• Camera: Raspberry Pi HQ Camera + a 6mm CS-mount lens (bundled or ~$25 aftermarket)
• Motion: HC-SR501 PIR sensor (a couple dollars)
• Storage: microSD for compact builds; 2.5" SSD for long unattended runs
• Power: 20,000+ mAh USB-C power bank; optional 10W solar panel + charge controller
• Enclosure: IP65 project box with a glass window for the lens
• Odds and ends: real-time-clock module for accurate timestamps if no network

For the storage: a SanDisk SSD Plus 480GB at ~$40 handles months of 1080p video easily; a Samsung 870 EVO 250GB is another good pick with excellent endurance. If the whole rig is USB-powered, a Crucial BX500 1TB via a USB 3 adapter is the go-big option.

Do I need a PIR sensor or can software detect motion?

Both work with different power trade-offs:

• PIR sensor: wakes the Pi only when heat-motion crosses the sensor. Zero CPU load between events. Best for battery/solar builds. Misses cold subjects (some birds, reflected sun on cold days).
• Software motion: keeps the camera running, compares consecutive frames. Catches anything visual. Higher power draw and more storage churn.

For a wildlife camera in a rural area on battery, PIR wins by a landslide — the sensor draws microamps, wakes the Pi only 5–30 times a day, and the battery lasts weeks. For a security camera on mains power, software motion is more flexible.

Wiring the PIR to the Pi

Wire the HC-SR501 output pin to GPIO 17 (or any spare), Vcc to 5 V, GND to GND. Set the two potentiometers on the sensor to short retrigger delay and low sensitivity to start; tune from there.

Software: use gpiod or rpi.gpio to listen for a rising edge on GPIO 17 and trigger a capture. The Adafruit learning system has PIR wiring diagrams and tuning advice that apply directly.

Capture code (libcamera / picamera2)

The Raspberry Pi documentation covers libcamera-still and picamera2 for stills and video. A minimal trail-cam script:

from picamera2 import Picamera2
from gpiod import chip, line_request
from datetime import datetime
import time, os

picam = Picamera2()
config = picam.create_still_configuration(main={"size": (4056, 3040)})
picam.configure(config)

def capture(reason):
 picam.start()
 ts = datetime.now().strftime("%Y%m%d-%H%M%S")
 path = f"/data/captures/{ts}-{reason}.jpg"
 os.makedirs(os.path.dirname(path), exist_ok=True)
 picam.capture_file(path)
 picam.stop()

pir_chip = chip("gpiochip0")
pir = pir_chip.get_line(17)
pir.request(consumer="trailcam", type=line_request.EVENT_RISING_EDGE)

while True:
 if pir.event_wait(sec=30):
 pir.event_read()
 capture("pir")
 time.sleep(2) # simple debounce

The Raspberry Pi HQ Camera product page covers sensor modes and lens compatibility.

Is the HQ camera or a global-shutter sensor better for motion?

The HQ Camera uses a rolling shutter, which can skew fast-moving subjects (running deer, birds mid-wing). A global-shutter sensor freezes motion perfectly at high shutter speeds but has lower resolution and higher cost. For general wildlife where you want good detail on paused subjects, the HQ Camera wins. For chasing high-speed action, a global-shutter build is better.

Storage and how long you can shoot

At 12 MP JPEG (~4 MB per shot):

• 32 GB SD ~ 8,000 captures
• 128 GB SD ~ 32,000 captures
• 480 GB SSD ~ 120,000 captures
• 1 TB SSD ~ 250,000 captures

At 1080p H.264 30fps, ~10 GB per hour:

• 128 GB SD ~ 12 hours of continuous video
• 480 GB SSD ~ 48 hours of continuous video
• 1 TB SSD ~ 100 hours of continuous video

A PIR-triggered camera in the wild rarely fires more than 500 events a day even in busy areas, so a 128 GB SD is plenty for a month. For high-density security or event-heavy urban wildlife, an SSD like the Crucial BX500 1TB via a low-power USB 3 adapter gets you effectively unlimited runtime between visits.

Power budget

Pi Zero 2W idle at ~120 mA, PIR at ~50 μA, camera at ~250 mA during a 2-second capture. Rough math for a build averaging 20 captures/day:

• Idle: 120 mA × 24 h = 2,880 mAh
• Captures: 250 mA × 40 s / day = 2.7 mAh (negligible)
• Total: ~3 Ah per day

A 20,000 mAh USB-C battery bank runs ~6–7 days. Add a small 10 W solar panel with a charge controller and you get indefinite runtime in most latitudes March–October.

Deep power savings come from suspending the Pi between triggers using crontab + a scheduled RTC wake, but the added complexity is only worth it if you're targeting weeks of runtime on a battery-only build.

Enclosure and weatherproofing

The dull, essential part:

• IP65 project box, ~$15–25 on eBay
• Cabochon glass insert for the lens (or a thin acrylic sheet)
• Silicone O-rings on every cable pass-through
• Gore-Tex vent to equalize pressure without letting water in
• Camo tape or matte paint for concealment

Cheap trail-cam enclosures rely on a single rubber cable gland, which fails after a season. Spend an extra hour sealing every hole. Test in a shower before deployment.

Networking (optional)

If you're within Wi-Fi range: the Pi Zero 2W's onboard Wi-Fi handles low-bitrate uploads well. Set up a nightly rsync to a home NAS, or push to S3-compatible object storage. For remote sites, a 4G/LTE hat pairs cleanly.

If you don't need real-time uploads: skip networking entirely. Pull the SD/SSD monthly, offload, redeploy. Simpler, more reliable, longer battery life.

On-device motion classification

The Pi 4 with a small YOLO or MobileNet model can classify captures on-device before storage — you keep only images where the model identifies a person, deer, bird, etc. This cuts storage churn hugely for cameras in busy areas. On the Zero 2W, classification is too slow for the interactive loop; classify off-device after retrieval.

Common pitfalls

• Retrigger too fast. Default PIR delay of ~5 seconds fires the camera during its own settling. Set to 15+ seconds.
• Sensor sun-heated. Direct sunlight on the PIR causes ghost triggers. Shade the sensor housing.
• SD card wear. Cheap SD cards die under 24/7 write load. Use high-endurance cards or an SSD.
• Timestamps drifting. Without a network or RTC, the Pi drifts. Add a $2 DS3231 RTC module.
• Uncalibrated exposure. libcamera's auto-exposure works, but in mixed lighting it needs a --awb custom and manual gain floor.

Bottom line

A Raspberry Pi HQ Camera trail-cam build is a weekend project that outperforms most consumer trail cams on image quality and gives you full control over triggering, storage, and processing. Start with a Zero 2W + PIR + 128 GB SD and a small solar top-up, iterate from there.

Related guides

• Raspberry Pi OS Linux 6.18 LTS: What Changes — the OS underneath
• Self-host a 24/7 FM Radio Station on Pi Zero 2W — another Zero 2W 24/7 project
• Best SSD for Raspberry Pi 5 Boot Drive — storage picks for Pi projects

Sources

• Raspberry Pi HQ Camera — sensor specs and lens options
• Adafruit Learning System — PIR wiring and Pi camera tutorials
• Raspberry Pi Documentation — libcamera and picamera2 API reference

Extended: real-world deployment lessons

After running Pi HQ Camera trail cams across a season, the biggest lessons:

• Test the enclosure in a rainstorm before deploying. Not a shower. A real rainstorm. Water gets in through vents, cable glands, and the lens window in ways showers don't reproduce.
• Cheap PIR sensors have a linear range narrower than the datasheet. Real-world detection at 10 m is unreliable on the HC-SR501. If you need >5 m, use two sensors or upgrade.
• Lens fogging happens. A small silica pack inside the enclosure and an occasional visit to swap it fixes most fogging.
• Ants find everything. Bug guards on all vents. Learned the hard way.

Timelapse variant

The same rig runs as a full-frame timelapse camera by dropping the PIR and swapping to a scheduled capture. Every 5 minutes, capture a still, write to storage, sleep. Six months of deployment produces ~50,000 photos and one very good stop-motion video of a construction site or garden.

Adding cellular for remote uploads

For remote sites without Wi-Fi:

• SIM7600 4G/LTE hat over USB or UART
• Waveshare Iridium module for polar / marine use
• LoRaWAN for tiny metadata uploads (no images)

A cellular upload costs power. A 250 KB image over LTE draws several seconds of ~500 mA. Budget for it: 20 uploads a day are fine on the same battery/solar setup that runs the PIR-triggered stills.

ML classification pipeline

For the Pi 4 path with on-device classification:

1. Capture into a ring buffer
2. Run a MobileNet-SSD or YOLO-nano over the frame
3. If the classifier finds a person/deer/bird, save the frame with metadata
4. Otherwise discard

Frameworks: TensorFlow Lite runs well; ONNX Runtime is another option. Detection latency on a Pi 4 is ~200 ms per 640×640 frame — fine for stills, hard for streaming.

Storage rotation

For long-unattended cameras, add a size-based cleanup script:

find /data/captures -type f -mtime +14 -delete

Or move older captures to lower-cost cold storage on a nightly sync. Losing everything to a full disk is the most common preventable failure.