Real-time recognition of sows in video: A supervised approach

Forskningsoutput: TidskriftsbidragArtikelVetenskapligPeer review

Sammanfattning

This paper proposes a supervised classification approach for the real-time pattern recognition of sows in an animal
supervision system (asup). Our approach offers the possibility of the foreground subtraction in an asup’s image processing
module where there is lack of statistical information regarding the background. A set of 7 farrowing sessions of sows,
during day and night, have been captured (approximately 7 days/sow), which is used for this study. The frames of these
recordings have been grabbed with a time shift of 20 second. A collection of 215 frames of 7 different sows with the
same lighting condition have been marked and used as the training set. Based on small neighborhoods around a point,
a number of image local features are defined, and their separability and performance metrics are compared. For the
classification task, a feed-forward neural network (NN) is studied and a realistic configuration in terms of an acceptable
level of accuracy and computation time is chosen. The results show that the dense neighborhood feature (d.3x3) is the
smallest local set of features with an acceptable level of separability, while it has no negative effect on the complexity of
NN. The results also confirm that a significant amount of the desired pattern is accurately detected, even in situations
where a portion of the body of a sow is covered by the crate’s elements. The performance of the proposed feature set
coupled with our chosen configuration reached the rate of 8.5 fps. The true positive rate (TPR) of the classifier is 84.6%,
while the false negative rate (FNR) is only about 3%. A comparison between linear logistic regression and NN shows
the highly non-linear nature of our proposed set of features.
Originalspråkengelska
TidskriftInformation Processing in Agriculture
Volym1
Utgåva1
Sidor (från-till)73-81
ISSN2214-3173
DOI
StatusPublicerad - jul 2014
MoE-publikationstypA1 Tidskriftsartikel-refererad

Vetenskapsgrenar

  • 113 Data- och informationsvetenskap

Citera det här

@article{1e112851962d4b7f89288b0ac0c3dda0,
title = "Real-time recognition of sows in video: A supervised approach",
abstract = "This paper proposes a supervised classification approach for the real-time pattern recognition of sows in an animalsupervision system (asup). Our approach offers the possibility of the foreground subtraction in an asup’s image processingmodule where there is lack of statistical information regarding the background. A set of 7 farrowing sessions of sows,during day and night, have been captured (approximately 7 days/sow), which is used for this study. The frames of theserecordings have been grabbed with a time shift of 20 second. A collection of 215 frames of 7 different sows with thesame lighting condition have been marked and used as the training set. Based on small neighborhoods around a point,a number of image local features are defined, and their separability and performance metrics are compared. For theclassification task, a feed-forward neural network (NN) is studied and a realistic configuration in terms of an acceptablelevel of accuracy and computation time is chosen. The results show that the dense neighborhood feature (d.3x3) is thesmallest local set of features with an acceptable level of separability, while it has no negative effect on the complexity ofNN. The results also confirm that a significant amount of the desired pattern is accurately detected, even in situationswhere a portion of the body of a sow is covered by the crate’s elements. The performance of the proposed feature setcoupled with our chosen configuration reached the rate of 8.5 fps. The true positive rate (TPR) of the classifier is 84.6{\%},while the false negative rate (FNR) is only about 3{\%}. A comparison between linear logistic regression and NN showsthe highly non-linear nature of our proposed set of features.",
keywords = "113 Computer and information sciences",
author = "Ehsan Khoramshahi and Juha Hietaoja and Anna Valros and Jinhyeon Yun and Matti Pastell",
year = "2014",
month = "7",
doi = "10.1016/j.inpa.2014.07.002",
language = "English",
volume = "1",
pages = "73--81",
journal = "Information Processing in Agriculture",
issn = "2214-3173",
publisher = "Elsevier Scientific Publ. Co",
number = "1",

}

Real-time recognition of sows in video : A supervised approach. / Khoramshahi, Ehsan; Hietaoja, Juha; Valros, Anna; Yun, Jinhyeon; Pastell, Matti.

I: Information Processing in Agriculture, Vol. 1, Nr. 1, 07.2014, s. 73-81.

Forskningsoutput: TidskriftsbidragArtikelVetenskapligPeer review

TY - JOUR

T1 - Real-time recognition of sows in video

T2 - A supervised approach

AU - Khoramshahi, Ehsan

AU - Hietaoja, Juha

AU - Valros, Anna

AU - Yun, Jinhyeon

AU - Pastell, Matti

PY - 2014/7

Y1 - 2014/7

N2 - This paper proposes a supervised classification approach for the real-time pattern recognition of sows in an animalsupervision system (asup). Our approach offers the possibility of the foreground subtraction in an asup’s image processingmodule where there is lack of statistical information regarding the background. A set of 7 farrowing sessions of sows,during day and night, have been captured (approximately 7 days/sow), which is used for this study. The frames of theserecordings have been grabbed with a time shift of 20 second. A collection of 215 frames of 7 different sows with thesame lighting condition have been marked and used as the training set. Based on small neighborhoods around a point,a number of image local features are defined, and their separability and performance metrics are compared. For theclassification task, a feed-forward neural network (NN) is studied and a realistic configuration in terms of an acceptablelevel of accuracy and computation time is chosen. The results show that the dense neighborhood feature (d.3x3) is thesmallest local set of features with an acceptable level of separability, while it has no negative effect on the complexity ofNN. The results also confirm that a significant amount of the desired pattern is accurately detected, even in situationswhere a portion of the body of a sow is covered by the crate’s elements. The performance of the proposed feature setcoupled with our chosen configuration reached the rate of 8.5 fps. The true positive rate (TPR) of the classifier is 84.6%,while the false negative rate (FNR) is only about 3%. A comparison between linear logistic regression and NN showsthe highly non-linear nature of our proposed set of features.

AB - This paper proposes a supervised classification approach for the real-time pattern recognition of sows in an animalsupervision system (asup). Our approach offers the possibility of the foreground subtraction in an asup’s image processingmodule where there is lack of statistical information regarding the background. A set of 7 farrowing sessions of sows,during day and night, have been captured (approximately 7 days/sow), which is used for this study. The frames of theserecordings have been grabbed with a time shift of 20 second. A collection of 215 frames of 7 different sows with thesame lighting condition have been marked and used as the training set. Based on small neighborhoods around a point,a number of image local features are defined, and their separability and performance metrics are compared. For theclassification task, a feed-forward neural network (NN) is studied and a realistic configuration in terms of an acceptablelevel of accuracy and computation time is chosen. The results show that the dense neighborhood feature (d.3x3) is thesmallest local set of features with an acceptable level of separability, while it has no negative effect on the complexity ofNN. The results also confirm that a significant amount of the desired pattern is accurately detected, even in situationswhere a portion of the body of a sow is covered by the crate’s elements. The performance of the proposed feature setcoupled with our chosen configuration reached the rate of 8.5 fps. The true positive rate (TPR) of the classifier is 84.6%,while the false negative rate (FNR) is only about 3%. A comparison between linear logistic regression and NN showsthe highly non-linear nature of our proposed set of features.

KW - 113 Computer and information sciences

U2 - 10.1016/j.inpa.2014.07.002

DO - 10.1016/j.inpa.2014.07.002

M3 - Article

VL - 1

SP - 73

EP - 81

JO - Information Processing in Agriculture

JF - Information Processing in Agriculture

SN - 2214-3173

IS - 1

ER -