Developing a machine vision system for detecting laying hens

O. Geffen, Y. Yitzhaky, N. Barchilon, S. Druyan, I. Halachmi

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

The Israeli laying hens industry is regulated by quota; a farm can produce eggs according to a fixed number of hens. With the new community cages now integrated into the Israeli egg industry, a manual head count of the hens is an impossible task. The aim of this study is to develop a machine vision system that automatically counts the hens, and helps the regulator to control the industry. The hen house that was used is 87 m long stacked on six floors, with 37 community cages set in a row, each cage is 2.4 m long, 0.54 m tall, and 0.74 m depth, housing 18-34 hens. The hen house has a narrow path along the cages. Consequently, a wide-angle camera was applied (HD Action Camera 1080p, wide angle 170 deg' lens) in order to frame the entire cage in a single field of view. The camera was mounted on a steel arm 0.85 m from the cages. The arm was connected to the feeder that moves along the cages. Videos with 30 fps were processed with an AI detection algorithm called Faster R-CNN. A feeding event appeared to be an adequate time to count the hens, as all hens were lined up in front of the cage, visible to the camera, making it possible to count. The detection algorithm was trained to detect hens in cages; it was tested on 4,000 images and got an accuracy of 80%. The algorithm count was compared to human observer count used as ground truth. The accuracy can be improved by further training the algorithm parameters.

Original languageEnglish
Title of host publicationPrecision Livestock Farming 2019 - Papers Presented at the 9th European Conference on Precision Livestock Farming, ECPLF 2019
EditorsBernadette O'Brien, Deirdre Hennessy, Laurence Shalloo
PublisherOrganising Committee of the 9th European Conference on Precision Livestock Farming (ECPLF), Teagasc, Animal and Grassland Research and Innovation Centre
Pages428-433
Number of pages6
ISBN (Electronic)9781841706542
StatePublished - 1 Jan 2019
Event9th European Conference on Precision Livestock Farming, ECPLF 2019 - Cork, Ireland
Duration: 26 Aug 201929 Aug 2019

Publication series

NamePrecision Livestock Farming 2019 - Papers Presented at the 9th European Conference on Precision Livestock Farming, ECPLF 2019

Conference

Conference9th European Conference on Precision Livestock Farming, ECPLF 2019
Country/TerritoryIreland
CityCork
Period26/08/1929/08/19

Keywords

  • Deep learning
  • Faster R-CNN
  • Laying hens
  • Object detection

Fingerprint

Dive into the research topics of 'Developing a machine vision system for detecting laying hens'. Together they form a unique fingerprint.

Cite this