This system is an end-to-end Smart Canteen & Kitchen Food Safety Management Platform that integrates IoT (Internet of Things) devices and AI Computer Vision algorithms. Rather than just simple video recognition, the platform integrates deeply into every phase of daily canteen operations. It establishes a closed-loop food safety workflow covering food procurement, warehouse inventory, nutritional meal planning, face-based morning health checks, cooking process supervision, food sample retention, and automated alert response.
- Supplier Admission & AI Evaluation: Full lifecycle supplier credentials vetting and cancellation workflows. The AI automatically grades suppliers based on historical delivery rates, fulfillment quality, and price stability to assist in identifying and replacing high-risk suppliers.
- Smart Purchasing Decisions: Automatically compares historical purchase prices, real-time inventory, and live supplier quotes to recommend optimal purchase quantities and target suppliers.
- Smart Temperature & Humidity Monitoring: Integrates IoT sensors inside cold storage and dry warehouses to detect environmental anomalies.
- Intelligent Stock Pick Suggestions (FEFO Rule): Generates automated warehousing and dispatch advice using "First Expired, First Out" and batch priority logic.
- AI Demand Forecasting: Combines historical consumption patterns, recipe plans, and dining headcounts to predict future ingredient requirements using time-series and machine learning models.
- AI Ingredient Loss Analysis: Automatically flags high-loss ingredients, timeframes, and root causes (e.g., spoilage, improper storage) to minimize food waste.
- AI Menu Recommendations: Automatically designs daily/weekly/monthly meal schedules balanced for nutrition, cost budgets, customer preferences, and specific health profiles (e.g., diabetes, hypertension).
- AI Dietary Profiles & Evaluation: Conducts nutritional dietary analysis for target groups (e.g., children, elderly, patients), calculates macro distributions (protein, fats, carbs, calories), and offers optimization tips.
- Cooking Parameters Preset: Links recipes to strict safety parameters (e.g., core cooking temperature for meat must exceed 70°C).
- Automatic Retention Logs: Instantly generates retention task logs linked to daily cooking records for complete traceability.
- AI Quality Assessment: Uses image recognition to grade cooked dishes based on color, texture, and doneness, giving automated feedback.
- 48-Hour Expiration Warnings: Automatically flags and reminds staff to discard food samples after 48 hours of cold storage.
- Face & Identity Verification: Allows kitchen staff to perform check-ins via facial recognition terminals.
- Compliance Vetting: Validates body temperature, checks health certificate validity, and uses computer vision to detect hand wounds or skin infections, blocking non-compliant staff from starting their shifts.
- Real-Time Video Feeds: Connects to multiple IP camera streams with support for recording and playback.
- AI Behavior Violations: Automatically detects kitchen non-compliance, such as missing chef hats/masks/gloves, smoking, mobile phone usage, raw/cooked cross-contamination, unwashed hands, unattended stoves, or pest/rodent presence. Captures video clips and raises instant alerts.
- Staff Behavioral & Efficiency Reports: Scores staff hygiene and operational compliance to generate training and efficiency recommendations.
- Multi-Device Pairing: Connects cameras, food testing kits, gas monitors, and temperature sensors.
- Real-Time Cooking Curves: Links temperature sensors to cooking burners. Clicking "Start Cooking" triggers real-time temperature collection (every 30 seconds) to plot cooking curves for safety audits.
[Web Admin Portal] (Vue 3 + TS + Element Plus) --> Administrative Staff (Procurement, Inventory, Alerts)
[Kitchen KDS Terminal] (Vue 3 + Vite) --> Kitchen Staff (Cooking Tasks, Temperature Curves)
[Mobile App / Mini-Program] (Uni-app) --> Handheld Operations (Inbound/Outbound, Mobile Health Checks)
|
v
[API Gateway (Nginx / Nacos)]
|
v
[Spring Cloud Microservices Cluster]
(Auth-service, Device-service, Procurement-service, Warehouse-service, etc.)
|
+-------------+-------------+
| |
v v
[Database Layer (MySQL + Redis)] [Edge AI Inference Service] (YOLOv8 + RTSP Video Stream)
- Backend: Spring Boot / Spring Cloud (Nacos Registry & Configuration, Sentinel) + MyBatis-Plus + MySQL + Redis.
- Web Frontend: Vue 3 + Vite + TypeScript + Pinia + Element Plus.
- Kitchen Terminal: Specialized Kitchen Display System (KDS) layout with real-time SSE/WebSocket data updates.
- AI Inference: YOLOv8-based model execution on RTSP video feeds.
To keep the git repository lightweight, raw dataset JSON files under doc/ have been compressed. You must run the restore script first after cloning the project:
./scripts/dev/restore-datasets.shThe platform requires MySQL, Redis, and Nacos. Spin them up using docker-compose:
docker-compose up -dExecute the unified startup script in the root directory:
./scripts/dev/start-all.sh --takeoverDefault local access URLs:
- Web Admin Portal:
http://localhost:5175 - Default Credentials:
admin/admin
Navigate to your camera recognition algorithms directory:
cd path/to/camera-recognition
source .venv/bin/activate
python review_app.py --config config.yaml --host 127.0.0.1 --port 18081Verify that the http://127.0.0.1:18081/annotated.mjpg stream is active to view AI bounding boxes in the Admin Portal under "Video Monitoring".
For database schemas, API specs, and technical details, refer to: