Rychagov S.
🚗 LPR · ANPR · Vehicle Tracking

AI Video Analytics for Vehicles and License Plate Recognition

Entry/exit control, license plate recognition (CRNN + CTC), vehicle monitoring, white/black lists, checkpoint and parking integration.

Key facts

License plate recognition

CRNN (CNN + BiLSTM + CTC) trained on Russian plates with Cyrillic (АВЕКМНОРСТУХ). Multi-frame stabilization, plate accuracy 95%+.

Vehicle detection & classification

YOLOv26: cars, trucks, buses, motorcycles. ONNX/TensorRT inference for realtime edge deployment.

Entry/exit tracking

Automated line-crossing detection with plate number, vehicle type, direction, and timestamp.

White/black lists

Instant plate check against access databases. Automatic barrier, REST API webhook, or MQTT alert on violation.

FAQ

What cameras and protocols are supported?

RTSP/IP cameras (Hikvision, Dahua, ONVIF), USB cameras, video files (MP4, AVI), YouTube streams.

How accurate is license plate recognition?

CRNN + CTC stabilizes results across multiple frames, supports Cyrillic, logs to CSV. Plate accuracy 95%+ on a 4,891-image validation set.

Can it be trained for other plate formats?

Yes, CRNN + CTC architecture is universal for OCR sequences. Replace the alphabet and fine-tune on new data.

How many streams does one GPU handle?

Up to 16 RTSP streams at 15 FPS on a single GPU (TensorRT). Horizontal scaling by adding nodes.

Limitations and when it may not fit

OCR quality depends on camera resolution, shooting angle, lighting, and plate condition.

Dirty, damaged, or non-standard plates reduce accuracy. Manual verification available via dashboard.

Realtime processing requires an NVIDIA GPU (CUDA). CPU is limited to batch file processing.

Vehicle control — manual logging, missed cars, no analytics

At parking lots, checkpoints, and warehouse territories, entry/exit is tracked manually or with simple sensors without plate recognition. Non-standard plates, operator errors, and absent fleet tracking lead to losses and uncontrolled access.

Blacklists checked by eye, visit logs on paper, reports compiled over days. YOLOv26 + CRNN automate the entire pipeline: from detection to realtime alerts.

How it works

1

Install cameras at checkpoints

RTSP/IP cameras (Hikvision, Dahua, ONVIF) at entry/exit points. Setup in minutes.

2

Configure lists and rules

Whitelist for authorized vehicles, blacklist for unwanted, automatic barrier, webhook notifications.

3

YOLOv26 + CRNN recognize vehicles and plates

YOLOv26 detects vehicles, CRNN (CNN + BiLSTM + CTC) reads plates with Cyrillic. TensorRT accelerates inference.

4

Get analytics and integrations

Entry/exit journal, fleet tracking, plate statistics, REST API for barriers, 1C and CRM integration.

Architecture Pipeline

Camera (RTSP / ONVIF)
    ↓
YOLOv26 Detection + Tracking (ONNX / TensorRT)
    ↓
CRNN OCR (CNN + BiLSTM + CTC)
    ↓
Event Engine (entry/exit, whitelist/blacklist, speed)
    ↓
Dashboard / REST API / Webhooks / MQTT / Telegram

Integrations

RTSP / IP cameras

Hikvision, Dahua, ONVIF-compatible. Connect any manufacturer.

Barriers & PACS

Automatic opening by whitelist. Integration via REST API or relay.

1C and CRM

Export journals, plate database sync, subscription reports.

Telegram / Webhooks

Realtime notifications on blacklist violations, entry/exit, and anomalies.

REST API

Full access to events, recognized plates, and analytics for custom integrations.

MQTT

Lightweight protocol for IoT integrations and edge devices.

ONNX / TensorRT

Inference optimization for GPU acceleration and edge deployment.

CSV / Database

Full logging of recognized plates and events for analytics and audit.

Features

🔢

License plate recognition (LPR)

CRNN + CTC for Russian plates with Cyrillic. Multi-frame stabilization, CSV logging. Plate accuracy 95%+.

🚗

Vehicle classification

YOLOv26: cars, trucks, buses, motorcycles. Confidence score + tracking per object.

🚧

Entry / exit

Automated crossing detection with plate number, vehicle type, direction, and timestamp.

White/black lists

Instant access check. Auto-open barrier for authorized, realtime alert for blacklisted.

📍

Fleet tracking

Track fleet vehicles across the territory, monitor routes, measure on-site time.

📊

Vehicle counting

Count vehicles by direction, parking occupancy statistics, peak hour analytics.

Speed estimation

Estimate vehicle speed on monitored sections. Detect speeding with plate reference.

🅿️

Parking occupancy

Real-time monitoring of occupied/free spaces. Integration with displays and navigation.

Where it's used

Parking systems — plate recognition, automatic barriers, occupancy monitoring, speed estimation.

Residential complexes — access control, resident whitelist, visit logs, 1C integration.

Logistics and warehouses — fleet tracking, entry/exit logging, delivery control, vehicle classification.

Checkpoints and industrial — blacklist/whitelist, PACS integration, MQTT for edge, territory security.

Transport and roads — traffic counting, speed estimation, violation recording, peak load analytics.

Dangerous zones — vehicle access control in restricted areas, realtime perimeter violation alerts.

Interested in practical LPR experience?

We can show you how YOLOv26 + CRNN recognize plates, track entry/exit, and manage whitelist/blacklist on real RTSP streams.

Connect on Telegram