Build a Driveway and Gate Camera with License Plate Recognition Using Raspberry Pi and Frigate

A camera pointed at a driveway or gate is only really useful if it can tell you something specific happened — not just that "motion was detected," but that a particular vehicle arrived. Automated license plate recognition (ALPR) used to require expensive dedicated hardware; today a Raspberry Pi, a reasonably sharp camera, and the plate-recognition plugin ecosystem built around Frigate NVR can do a genuinely good job of reading plates on vehicles entering your own property and logging them, triggering an automation, or alerting you the moment an unrecognized vehicle shows up. This project builds a dedicated Pi-based ALPR node that feeds into Frigate and Home Assistant, covering camera placement, the recognition pipeline, and the privacy and legal considerations that come with running plate recognition on your own property.

What This Project Is (and Isn't)

This is a self-hosted, local-network home security tool for your own driveway, gate, or parking area — recognizing vehicles you own or regularly expect, and flagging ones you don't. It is not a substitute for, and shouldn't be treated as, a law-enforcement-grade system, and you should not point it at public roads or use it to track vehicles that aren't entering your own property. Plate recognition accuracy on a hobby setup is good but not perfect — expect occasional misreads on plates at odd angles, in heavy rain, or on unusual state plate formats — so treat the output as a strong hint for automation and search, not a legal record.

Hardware and Camera Placement

A Raspberry Pi 5 with the Hailo AI Kit (or a Pi 4 paired with a Coral USB accelerator) gives you enough inference throughput to run object detection and plate recognition without maxing out the CPU, especially if you're also running Frigate for general NVR duties on the same box. Camera placement matters more than camera resolution for plate reads: mount so the camera looks nearly straight down the direction of travel rather than at a steep side angle, keep the plate area filling a reasonable portion of the frame at the point vehicles typically stop or slow (a gate, a specific driveway choke point), and add IR illumination or use a camera with strong low-light performance if the read zone isn't well lit at night — plates are retroreflective and respond well to IR at the right angle, but poorly to a camera relying purely on ambient light after dark.

Software Stack

Frigate handles the object detection and clip recording, using its vehicle detection to crop and pass the plate region to a recognition engine — a common, well-documented approach on the self-hosted side pairs Frigate with a plate-recognition add-on (running OpenALPR-style OCR or a more modern CRNN-based recognizer in Docker) that watches Frigate's event stream and returns plate text on car/vehicle detections. Feed the recognized plate strings into Home Assistant via MQTT, where you can maintain a simple "known plates" list and trigger different automations for recognized versus unrecognized vehicles — unlock a gate automatically for a known plate, or just push a notification with a snapshot for an unknown one. Keep the whole pipeline local: there's no reason to send driveway footage or plate data to a cloud service for this kind of project, and running it all on your own network is both more private and removes any ongoing subscription cost.

ComponentRoleNotes Raspberry Pi 5 + Hailo AI Kit (or Pi 4 + Coral)Runs Frigate and handles inferenceHailo kit gives meaningfully more headroom for multiple camera streams PoE IP camera (2MP-5MP)Captures the read zoneHigher resolution helps less than good angle and lighting Frigate NVRObject detection, recording, event triggersRuns in Docker alongside the plate-recognition add-on Plate recognition add-onOCRs the cropped plate region from vehicle detectionsWatches Frigate's MQTT event stream Home AssistantAutomation, known-plate list, notificationsTies recognized/unrecognized plates to real actions

Tuning for Accuracy

Most read failures on a hobby setup trace back to motion blur, poor angle, or insufficient contrast rather than the recognition model itself — a vehicle moving quickly through the frame at a low shutter speed will smear the plate characters beyond what any OCR can recover. If your camera supports manual shutter speed control, favor a faster shutter even at the cost of some added gain/noise in low light, since a slightly noisier but sharp frame reads far better than a smooth but blurred one. Set Frigate's detection zone tightly around the actual read area rather than the whole driveway, which both improves accuracy (less irrelevant motion competing for processing) and reduces false triggers from passing pedestrians or animals.

Privacy and Legal Considerations

Laws around recording vehicles and license plates vary by state and country, and some jurisdictions have specific rules about ALPR data retention or require signage if you're recording a shared driveway or semi-public access point. Keep the camera's field of view confined to your own property, set a reasonable data retention period in Frigate rather than keeping plate logs indefinitely, and if the driveway or gate is shared with neighbors or is visible from a public right-of-way, check your local regulations before deploying — this is exactly the kind of home project where "it works" and "it's fine to run this way" aren't automatically the same thing.

Wrapping Up

Pairing Frigate with a plate-recognition add-on turns a driveway camera from a generic motion sensor into something that can distinguish "the mail truck again" from "a car that's never been here before," entirely on hardware you own and a network you control. The build itself is mostly software integration once the camera is well placed — the real engineering work is in camera angle, lighting, and shutter tuning, not in the recognition model itself.

Related Guides

Loading_