Outdoor Cameras and Live View: Entrance, Driveway, Garden, but Not a Fake Alarm
Outdoor Cameras and Live View: Entrance, Driveway, Garden, but Not a Fake Alarm
A camera shows a picture; an alarm detects and reacts to an intruder. Those are two different jobs, often confused. This lesson covers configuring outdoor cameras well in HA, and draws a clear line around where their role ends.
Budget around 85 minutes. Example entities: camera.driveway, camera.entrance, binary_sensor.driveway_pir, switch.nvr_recording.
A camera is not an alarm system
A camera's image-based motion detection catches a pixel change, not an intruder. Wind moving a branch, a passing car's shadow, or a lighting shift all generate the same event as a person would. The real alarm system from Module 18 relies on dedicated sensors and zone logic: there, a camera is a visual add-on, not the core of detection.
Choosing an Outdoor Camera: What to Look For
For an outdoor install, what matters is the enclosure's protection rating (IP66 minimum if there's no protective overhang), an operating temperature range realistic for your winters, and how it integrates with HA: cameras with an RTSP stream available locally, with no forced manufacturer cloud, give you the most control and the lowest latency. Popular in the HA community are models with local RTSP plus ONVIF integration, then handled by a detection engine like Frigate if you want advanced image analysis (recognizing a person, a vehicle, a package) instead of simple pixel-based motion detection.
Recording Events, Not a Continuous Stream
Continuous 24/7 recording from every camera fills a drive fast and makes it harder to find the moment that matters. A more sensible pattern: short clips triggered by an event from a dedicated PIR sensor, not by image-based motion detection alone, with a few seconds of pre-event buffer if your recording system supports it.
automation:
- alias: "Driveway: record a clip on PIR motion"
id: driveway_record_clip_on_pir_motion
mode: single
triggers:
- trigger: state
entity_id: binary_sensor.driveway_pir
to: "on"
conditions:
- condition: state
entity_id: input_boolean.outdoor_automation_enabled
state: "on"
actions:
- action: camera.record
target:
entity_id: camera.driveway
data:
filename: "/media/recordings/driveway_{{ now().strftime('%Y%m%d_%H%M%S') }}.mp4"
duration: 20
- action: notify.mobile_app_your_phone
data:
title: "Motion on the driveway"
message: "Motion detected. Recording saved, check it when you get a chance."
data:
tag: "driveway-motion"
importance: lowThe notification's priority is deliberately low (importance: low): driveway motion is everyday noise, not a critical event. Contrast this with the alerts from Module 18, where a real intrusion into an interior zone gets high priority: distinguishing priorities is exactly what separates a useful system from one you tune out after a week of spam.
An Outdoor Camera Dashboard
type: grid
columns: 2
cards:
- type: picture-entity
entity: camera.driveway
camera_view: live
name: "Driveway"
- type: picture-entity
entity: camera.entrance
camera_view: live
name: "Entrance"
- type: picture-entity
entity: camera.garden
camera_view: live
name: "Garden"
- type: picture-entity
entity: camera.intercom
camera_view: live
name: "Pedestrian gate / intercom"camera_view: live shows the live stream instead of a static thumbnail: check network load with several cameras displayed at once, especially over remote access away from the home network from Module 7.
Privacy: Neighbors and Public Space
Don't aim cameras at a neighbor's property or a public sidewalk
Set the field of view to cover your own property. Many modern cameras and NVR systems (Frigate included) let you define privacy masks blocking out specific parts of the frame: use that feature if the camera's physical placement doesn't fully eliminate someone else's space from the shot.
Frigate: A Local NVR with Object Detection, and Which Cameras Fit It
Frigate is a complete, local NVR (Network Video Recorder) with real-time object detection, running entirely on your own hardware, with no video sent to the cloud. It recognizes specific event categories, like a detected person, a car in the driveway, or a dog in the garden, and passes those events straight into Home Assistant as ready-made entities and automation triggers.
| Element | Recommendation |
|---|---|
| AI detection accelerator | Hailo8/Hailo8L, Google Coral (no longer recommended for new builds), OpenVINO on Intel Arc, Nvidia GPU, Rockchip devices with RKNN |
| Cameras | Reolink, Dahua, Hikvision, and Amcrest, in that order of popularity in the Frigate community |
| A specific model worth considering | Reolink RLC-823A: a 4MP camera with a 2.8-12mm varifocal lens, solid night vision, standard RTSP with no workarounds needed |
Frigate integrates directly with the Home Assistant media browser, giving you low-latency camera entities plus sensors and switches that feed further automations and notifications, covered in Module 22.
Recording at Night: Infrared vs. Color Night (Starlight)
| Night technology | Characteristics |
|---|---|
| Infrared (IR, black-and-white image) | works in complete darkness, black-and-white image, cheaper cameras |
| Color night (starlight/full-color) | needs at least a little ambient light (moonlight, a lamp), keeps color, easier to make out details like clothing color |
| A motion-triggered LED spotlight | some cameras have a built-in spotlight that fires on motion detection, adding a deterrent |
For object recognition through Frigate, color night usually gives better detection results than a pure infrared image, because object-recognition models are trained mostly on color photos.
Storing Footage: A Local NVR vs. the Cloud
Frigate, covered in this lesson's main content, stores footage locally, meaning no monthly subscription, but it needs its own drive with adequate capacity and a thought-out retention policy (similar to the recorder database's purge_keep_days, covered in the maintenance module). Cloud options (built into some Reolink cameras, or Ring, for instance) are more convenient to set up, but come with recurring fees and store your video outside your control, which for many people taking this course, valuing locality and privacy, is a strong argument for the local option.
Dedicated Lighting to Support Nighttime Monitoring
Cameras running in color-night mode, covered in this lesson's main content, need at least a little ambient light to preserve color. Discreet, low-power steady lighting (not a blinding floodlight, just a gentle fill light on the monitored area) can work alongside the lighting automation from Lesson 4, switching on at a low brightness level precisely during the hours when nighttime monitoring matters most.
Integrating Cameras with License Plate Recognition
Similar to the gate automation from Lesson 5, a camera aimed at the driveway can serve two purposes at once: general monitoring and license plate recognition for vehicles entering the property. Frigate supports this through additional text-recognition plugins, though it requires good image quality and the right camera angle, different from what plain vehicle-presence detection needs.
Backup Power for Cameras During an Outage
Video monitoring matters most precisely during unusual situations, which sometimes include a power outage. PoE cameras can share the same UPS that backs up the network switch and the Home Assistant server itself (tying back to the maintenance module), keeping recording going through a brief power outage, when the risk of a break-in can genuinely run higher than usual.
Monitoring Ethics: Telling Guests and Workers About Cameras
Beyond the question of neighbor privacy covered in the main content of this lesson, it's worth remembering the ethical, and often legal, obligation to disclose camera monitoring to people who regularly visit the property: a gardener, a renovation crew, a babysitter. A simple, visible sign noting the presence of cameras near the entrance, while it may feel like a formality, is good practice regardless of local legal requirements.
How to Test This Lesson
1. Check every camera's frame for neighbor privacy and public space.
2. Walk through the PIR's field of view and confirm the recording actually saves.
3. Check disk usage after a week: is event-based recording actually saving space?
Common Mistakes
Image-based motion detection treated as an "alarm": a flood of false alerts from wind and shadows.
Continuous recording with no retention limit: a full drive within weeks.
A camera aimed at a sidewalk or a neighbor's property: real legal and neighborly risk.
Practical Task
☐ Set up event-based recording triggered by a PIR sensor, not by image-based detection.
☐ Build a dashboard with a live view of every outdoor camera.
☐ Review and adjust each camera's frame for neighbor privacy.
Key Takeaways
A camera is video, an alarm is detection and response: don't confuse the roles.
Event-based recording triggered by a dedicated sensor, not by the camera alone.
Check neighbor privacy in every single camera's frame.
What's Next
Next lesson: Outdoor Dashboard: Garden, Gate, Pedestrian Gate, Irrigation, Weather, Tank, and Diagnostics. We tie everything in this module into one coherent view.
