Soaking · Client project
Intelligent Visual Soaking Project
Timed spraying cooled the barn whether or not a cow stood under it. Cameras every 12 metres let the barn spray the square metre a cow is actually standing in — and switch off the moment she leaves.
- 60%less water
- 18.3%more milk per cow
- 8.85MRMB annual benefit
The situation
The project runs tunnel-ventilated barns in a subtropical climate, where heat and humidity hold from March to November. Timed spraying had been the standard answer and it was failing in four ways at once.
- Milk loss. Spraying on a clock could not follow where the cows were. Summer yield on lactating cows fell by 10–18 %.
- Water. Continuous, undifferentiated spraying pushed water use up and with it the volume of manure and wash water the site had to treat.
- Health. Wet bedding raised mastitis and hoof problems, which shows up in the milk cheque long after the spray valve closes.
- Bad sensing. The infrared units then in use read people and vehicles as cows and missed cows standing still at the feed face, so the spraying logic was permanently out of step.
What we built
- Vision instead of infrared. HD cameras were mounted every 12 metres along the headlock line. A deep-learning model separates people, cows and vehicles, tracks each animal and opens only the valve for the zone she is standing in — “zone by zone, spray as she moves”.
- Local control that survives the network. A Micro-environment DB controller runs the barn on its own. Sensors are read every five minutes: spraying and fans start at THI 72 and stop below 68, so the barn never sprays into already-saturated air.
- Working with the tunnel. Spraying and tunnel ventilation were commissioned together, so the air speed the tunnel already produces carries the evaporated water off the cows instead of into the bedding.
- Feed and lane monitoring came along. The same cameras report feed delivery and push-up, which lets the site hold a tighter leftover rate and tune the ration.
What it delivered
| Measure | What changed | Result |
|---|---|---|
| Water | Spraying only over cows | 60 % less water; 145,800 m³ saved across one site from May to October |
| Intake | Cows eat through the heat | Dry matter intake up 15.93 %, from 18.47 kg to 21.41 kg per cow per day |
| Milk | Yield recovered in summer | 18.3 % more milk per cow; 1,255 t more across the year, over 5.77M RMB in additional revenue |
| Cost | Less water, less effluent, less waste | 8.85M RMB net benefit for a single site after equipment investment; feed leftover rate down 30.2 % |
| Quality | Cleaner cows, cleaner milk | Somatic cell count below 200,000 and total bacterial count below 20,000 — both inside EU limits |
We used to spray on a timer, whether there were cows there or not. The water bill was high and the environmental pressure was worse. Now it sprays when it sees a cow, and the recognition is more accurate than the human eye.
Why it holds up
- Precision. Cooling is applied per cow and per zone, not per barn.
- Reliability. Control stays local, so a dropped connection does not stop the barn from cooling itself.
- Economic value. Water, effluent, feed and yield all move in the same direction, which is why the payback is short.
- Simplicity. One interface covers spraying, fans, feed and alarms.
