| Wisdom Connects Everything, Enable Digitalization |
AI Egg Counter Camera for Poultry Farms | Automatic Egg Counting & Sorting
AI Egg Counting Camera for Layer Farms and Egg Grading Lines
Counting eggs on a moving belt is a task humans do badly at scale. Our AI visual counting camera detects, tracks and counts every egg in real time, flags broken and dirty eggs, and supports automatic sorting — with all data streamed to a web dashboard.
How It Works — Four Steps
- Detect — a vision model locates each egg in every frame and outputs a confidence-scored bounding box.
- Track — each egg receives a persistent ID and is followed across frames, so overlapping eggs are never double-counted.
- Count — when an egg's trajectory crosses a user-defined counting line, the counter increments; in/out directions are counted separately.
- Act — classification results drive sorting output and trigger a blockage alarm when eggs pile up on the belt.
System Components
| Component | Function |
|---|---|
| AI vision camera | Image capture and on-device inference above the egg belt |
| Detection & tracking engine | Per-egg detection with persistent ID tracking |
| Counting zone configuration | On-screen line/area setup, bidirectional counting |
| Classification module | Normal egg vs. broken/dirty egg (bad egg) |
| Web management console | Live monitoring, video/image parameters, detection & tracking settings, system logs |
| Data dashboard | Hourly curves, week-over-week comparison, 5-minute distribution |
| Alarm & I/O | Blockage alarm output, sorting mechanism linkage |
Feature Matrix
| Capability | Replaces | On-site behaviour |
|---|---|---|
| Egg belt counting | Manual counting staff | Simultaneous multi-egg counting with live totals |
| Broken & dirty egg detection | Visual inspection | Abnormal eggs flagged in a separate class |
| Automatic egg grading | Manual sorting | Qualified vs. unqualified eggs separated |
| Blockage alarm | Line patrol | Alerts when eggs accumulate, reducing breakage |
Deployment Process
Site survey → camera position and counting zone design → model calibration with on-site samples → parallel validation against manual counts → handover with remote parameter tuning.
Measured on Site
| Metric | Result |
|---|---|
| Eggs counted per day (single farm) | 230,000+ |
| Manual intervention during counting | None |
| Data granularity | Hourly curves, 5-minute distribution |
| Comparison views | Today vs. yesterday, this week vs. last week |
FAQ
Does it double-count overlapping eggs?
No. Counting is based on persistent ID tracking, not single-frame detection.
Can it separate broken eggs from dirty eggs?
Abnormal eggs are classified as a distinct category and marked separately on screen.
Does lighting affect accuracy?
Image parameters are adjustable; sites with strong lighting variation should combine supplementary lighting during calibration.
【261112-XF】