experiences with precision livestock farming in...

02/11/2016

1

Daniel Berckmans

M3‐BIORES Catholic University Leuven ‐ BelgiumMeasure, Model & Manage Bio Responses

Joint USAHA/AAVLD Plenary Session:Challenges, High Tech Solutions, and Success Strategies in Animal

Agriculture.

17 October 2016Greensboro, USA

Experiences with Precision Livestock Farming in Europe

• Challenges for Livestock Production

• What is PLF?

• Basic principles of PLF

• EU‐PLF Project

• Business Models

• Conclusions

D. Berckmans M3‐BIORES KU Leuven Belgium October 2016 2

Overview

02/11/2016

2

M3‐BIORES team

33

306 A‐Publications 17 products 15 patents389 Conference papers 2 spin‐off companies

Challenges for livestock production

4D. Berckmans M3‐BIORES KU Leuven Belgium October 2016

02/11/2016

3

Worldwide Meat Consumption per capita

Source: FAO (2010)


Worldwide demand for animal products will increase with 70

% by 2050 (FAO, 2016)?

Already over 60 billion animals slaughtered every year

Health: Relationship between animal health and healthy food

Animal welfare (e.g. EU)

Environmental Issues

Social importance

Economic importance including Valorisation of knowledge

Challenges for livestock production


02/11/2016

4

Problem of monitoring animals Livestock farming in the past …

Farmer had the time to use audio‐visual scoring


Numbers of animals

Number of farmers


02/11/2016

5

Resulting in

High number of animals per farm

Less available time per individual animal

More welfare and other problems


3 approaches in Europe with focus on the animal


02/11/2016

6

First approach: Iceberg Indicator (1)


Second approach: Welfare Quality (2)


02/11/2016

7

Europe has invested in a methodology to quantify Animal Welfare

“Welfare Quality”

Procedure: Experts do audio‐visual scoring by visiting farms and looking to (behaviour) of animal.


Tomorrow…Automated Systems

D. Berckmans M3‐BIORES KU Leuven Belgium October 2016

Technology can help to quantitatively measure behaviour, health and performance of animals.

14

02/11/2016

8

Welfare Scoring/Monitoring

15

Welfare Qualityonce a year

TimeTime

Iceberg

indicators

slaughterhouse

PLF-Continuous animal basedmanagement during growth

period


What is Precision Livestock Farming (PLF)?

02/11/2016

9

Precision Livestock Farming

17

Management of livestock by continuous automated real‐time monitoring of health, welfare of livestock, production/reproduction and environmental impact.


Advantages of PLF technology

‐ Objective measurements

‐ Fully automated

‐ Continuous measurements (24h/day, 7 days/wk)

‐ Behavioural responses of animals

‐ Less visits to the animals: biosecurity

‐ Cheaper than human experts


02/11/2016

10

Internationalisation


4 Smart Sensors Workshops

2000: Silsoe2002: Bremen2004: Leuven2006: Gargnano

19

EC-PLFCIGRASABEEurAGENGISAHEAAPICAEW…...

7 European Conferences of PLFSept 2017 Nantes France 8th EC-PLF

EU Projects:-Bright Animal-ALL-Smart-Pigs-BioBusiness-EU-PLF

The basic principles of PLF

02/11/2016

11

21

A living organism is a CITD system


A living organism:Complex


02/11/2016

12

Living organisms are ...


...individually different


02/11/2016

13

Complex Individual

Heartbeat (bpm)

Time (s)

A living organism:


Complex Individual

A living organism:

Individual Response variable

Av. population

N

δ populationδ Individual

Time


02/11/2016

14

Identical Individually different


A living organism:

Complex Individual Time‐Varying

A living organism:

TIME (HOURS)

0 1 2 3 4 5

HE

AT

PR

OD

UC

TIO

N (

W/K

G)

11

12

13

14

15

16

17

MEASUREDMODELLED (1ST ORDER)MODELLED (2ND ORDER)

5 days old

TIME (HOURS)

0 1 2 3 4 5

HE

AT

PR

OD

UC

TIO

N (

W/K

G)

7

8

9

10

11

12

13

MEASUREDMODELLED (1ST ORDER)

30 days oldExample: Heat production of broiler chickens


02/11/2016

15

Complex Individual Time‐Varying Dynamic

A living organism:

1. Measure

2. Model

3. Manage & Monitoring

In an on‐line way

Living organism = CITD ‐ system

Complex Individual Time‐Varying Dynamic

M3‐BIORES


EU‐PLF PROJECT ‘Bright Farm by Precision Livestock Farming

Animal Health Consultancy

Prof. Jos Metz Prof. Noel Devisch Prof. Leo den Hartog Dr. Dieter Schillinger Mr. I. Blanco‐Traba

EU‐PLF Advisory Board

EU‐PLF Farmers

EU‐PLF Partners

02/11/2016

16

Selected farms

Pig farms 10 France, Spain, N‐Ireland, the Netherlands, Italy, Hungary

Poultry farms 5 Spain, UK, the Netherlands, Italy

Cow farms 10 Denmark, the Netherlands,Sweden, Germany, France, Israel, Ireland


Data collection(External Data Storage)

Farm PCRemoteobservation

Internet

Climate controller Feed

controller

Network


02/11/2016

17

Installation of PLF systems poultry farms


Installation of PLF systems pig farms


02/11/2016

18

Data Collection, Assessments, Labelling – Pigs

35

Subject Comment

Audio collection 8.870.000 files, i.e. 30.800 days

Image collection >180 fattening cycles (raw data not stored)

Sound Data Labelled 8.500 minutes of labelled data, >24.000 labels

Assessments 8 assessors trainedBody wounds / tail biting / coughing / sneezing129 farm visits

Data Collection, Assessments, Labelling – Poultry

Number of fattening periods: 90

Number of measuring days: 5.475

Image collection: > 120 Terabytes

Sound collection: 4.906.000 files (5 min)

Welfare assessments: 130


Monitoring of drinking behaviour in pigs (i.c.w. Ughent, Fancom BV)


• Monitoring water usage as indicator for health status

• Estimate hourly water use in a pig pen by analysing hourly duration of drink nipple visits

02/11/2016

19

Model‐based monitoring of water use

Images

Detection of visits

Duration of visits

Transfer function

modelling

Water flow measured

CompareWater use estimated

from image

Water use from water

meters


Model‐based detection of visits

Threshold

Drink nipple

∆ T = duration of the visit


Play

02/11/2016

20

Results


Hourly water use can be estimated with an accuracy of 92% or 200 ml over 13 days

Example “early warning water intake”Swine flu

Fast treatment prevents• High use of antibiotics• Lower feed and water intake

• Secondary bacterialinfections

Animal care takerlogbook

24 hoursWater usage

Time saving

Analysis of drinking behaviour


02/11/2016

21


Infection Monitoring by On-line Pig Sound Analysis

i.c.w. University of Milan, SoundTalks NV, Fancom BV


42

On‐line cough recognition algorithm in pig stables

Health monitoring by on‐line sound analysis:

02/11/2016

22


44

Fixed installationMobile device

Detection of respiratory problems


02/11/2016

23

45

On-line cough recognition algorithm in pig stables


PCM: Results

Animals treated

Pigs ill again

Animals treated

Animals again ill

Pigs ill upon entering


02/11/2016

24


MAIN FUTURE APPLICATIONS REDUCING THE USE OF ANTIBIOTICS

Climate controller

V(t)

T

Q(t)

Antibiotics

sound

Therapeuticdecision

infection

Sound analysis

micro

47

Real time monitoring of problems in a broiler house

i.c.w. Fancom BV

02/11/2016

25

eYeNamic monitor tool

49

Farm manager

Camera network

Farm networkMonitoring software

Imagepre-processing


Vision‐based Early Warning System for Broiler Houses

• Solution?

• Farmers can use automatic tools to continuously monitor the welfare and health of their broilers


02/11/2016

26

Experiment’s ground plan


51

• Detecting malfunctioning in broiler houses

• Produce alarms in real‐time when malfunctioning happens (in feeder or drinker lines, light, climate control, etc.)


02/11/2016

27

Birds and housing

• Experiment rounds: 42 days

• Initial broiler weight: weight of 40±5 grams

• Broiler type: ROSS 308 broilers

• House capacity: 28000 broilers

• Climate control: Fancom FUP1EA2


Farmer logbook and manual video observation as references


02/11/2016

28

Measured vs. modelledanimal distribution

26/10 30/10 03/11 07/11 11/11 15/11 19/11 23/11 27/11 01/12 05/12Date (dd/mm)

Predicted distribution

Measured distribution

Distribution (%)

Prediction window: 1 light period = 5 hours


Event detection

Dis

trib

utio

n (%

)

Date(dd/mm)

Measured valuesSmoothed values within 25% rangeSmoothed values out of 25% rangePredicted values

Normal situation Problem in feeding lines

Feeder line Defect Feeder line


02/11/2016

29

Detected events in the validation experiment over 42 days

Vaccination

Clim

ate control

system

problems

Unloading

Water supply problems

Farmer’s inspection

Light problems

Feeding system problems

Electricity failure

Conclusion: Events in a broiler house could be detected using top-view image analysis with an accuracy of 95.24 %


57

Use of a RTLS to analyse cow activities

cubicles

trough

alleys

Description of the normal time budget of each cow Cow more active (e.g. walking) than normal alarm: oestrus?Cow less active (e.g. resting) than normal alarm: disease?

Antennas

tag

idling

walking

eating resting

CowView


i2

i2 iveissier, 8/12/2015

02/11/2016

30

Descriptive approach based on single basic activities

59

0

5

10

15

20

‐10 ‐9 ‐8 ‐7 ‐6 ‐5 ‐4 ‐3 ‐2 ‐1 0 1 2 3 4 5 6 7 8 9 10

h per day

Days from mastitis treatment

resting

eating

Cows spend less time resting and eating at the onset of mastitis. However, due to large variations the difference is not significant.


Modelling approach based on single activities

60

Start Ketosis event

Alert deviation

End Ketosis event

Back to normal

Just before the problem , cows’ resting time deviates from normal


02/11/2016

31

Some other examples

62

Aggression monitor: Umil, TIHO, Fancom BVCow lameness monitor: i.c.w. Volcani, DeLaval, Wur

Weight estimation: Fancom BV, Agrifirm

Play

Play

Play

Feed intake by sound

Play

02/11/2016

32

More Results

Chickens

Exp.Minutes

Number of peckings per experiment

feed uptake per

experiment (g)

Feed loss per

experiment(g)

Feed intake per

experiment (g)

Feed Intake Per Pecking

(g)

Feed loss per

experiment(%)

1 13 1193 28,63 0,325 28,31 0,0241,14

1

212 759 18,98 0,198 18,78 0,025

1,04

3 10,3 895 24,17 0,222 23,94 0,027 0,92

1 15 1250 32,50 0,236 32,26 0,026 0,73

22 13,5 1283 30,79 0,365 30,43 0,024 1,19

3 15 1460 35,04 0,348 34,69 0,024 0,99

1 7,04 651 16,28 0,168 16,11 0,025 1,03

32 4,35 468 12,17 0,111 12,06 0,026 0,91

3 7,26 533 13,33 0,124 13,20 0,025 0,93

1 6,54 583 13,99 0,145 13,85 0,024 1,04

122 7,43 654 16,35 0,165 16,19 0,025 1,01

3 6,65 573 15,47 0,155 15,32 0,027 1,00

Total‐Average

300,10 25285 633,26 6,22 627,04 0,025 0,98


Business Models

02/11/2016

33

65

Value for farmer:

Euros

Welfare

Health

ConsumerLabour and Time

Social Recognition

Environmental Impact


66

PLF is a tool that helps farmers and stakeholders

Consumer

Veterinarians

Citizens

Researchers

Governments

Press

Companies


02/11/2016

34

67

Feed Farmer Slaughter Retailer

Con‐sumer

PLF Service Provider(funding, set‐up, service, data)

Breeding Vets. OthersTech. Prov.

The PLF Business Model?Cost of PLF investment & operation shared along the value

creation chain by payment for access to data pool


General Conclusions

68

• PLF systems work in real farms, PLF technology will go.

• PLF offers fully automated continuous real time detailed monitoring and management of animals.

• PLF brings the farmer to the individual animals that need his/her attention, active management tool.

• PLF is a tool that helps farmers and stakeholders.

• PLF will allow the animals to drive the system.

• Efficient implementation of PLF needs collaboration betweenresearchers, farmers, industry and stakeholders!


02/11/2016

35

Acknowledgments and Disclaimer

This project has received funding from the European Union’s Seventh Framework Programme for research, technological development and demonstration under grant agreement n° 311825

The views expressed in this presentation are the sole responsibility of the author(s) and do not necessarily reflect the views of the European Commission.


Thank you

For more information you can check our website: http://www.m3‐biores.com

Contact: [email protected]

70

experiences with precision livestock farming in...

Documents