Translating conventional wisdom on chicken comb color into automated monitoring of disease-infected chicken using chromaticity-based machine learning models

Mohd Anif A A Bakar; Pin Jern Ker; Shirley G H Tang; Mohd Zafri Baharuddin; Hui Jing Lee; Abdul Rahman Omar

doi:10.3389/fvets.2023.1174700

Translating conventional wisdom on chicken comb color into automated monitoring of disease-infected chicken using chromaticity-based machine learning models

Front Vet Sci. 2023 Jun 21:10:1174700. doi: 10.3389/fvets.2023.1174700. eCollection 2023.

Authors

Mohd Anif A A Bakar¹, Pin Jern Ker¹, Shirley G H Tang², Mohd Zafri Baharuddin¹, Hui Jing Lee³, Abdul Rahman Omar^{4

5}

Affiliations

¹ Department of Electrical and Electronics Engineering, College of Engineering, Institute of Sustainable Energy, Universiti Tenaga Nasional, Kajang, Malaysia.
² Center for Toxicology and Health Risk Studies (CORE), Faculty of Health Sciences, Universiti Kebangsaan Malaysia, Kuala Lumpur, Malaysia.
³ Department of Electrical and Electronics Engineering, College of Engineering, Institute of Power Engineering, Universiti Tenaga Nasional, Kajang, Malaysia.
⁴ Department of Veterinary Pathology and Microbiology, Faculty of Veterinary, Universiti Putra Malaysia, Serdang, Malaysia.
⁵ Institute of Bioscience, Universiti Putra Malaysia, Serdang, Malaysia.

Abstract

Bacteria- or virus-infected chicken is conventionally detected by manual observation and confirmed by a laboratory test, which may lead to late detection, significant economic loss, and threaten human health. This paper reports on the development of an innovative technique to detect bacteria- or virus-infected chickens based on the optical chromaticity of the chicken comb. The chromaticity of the infected and healthy chicken comb was extracted and analyzed with International Commission on Illumination (CIE) XYZ color space. Logistic Regression, Support Vector Machines (SVMs), K-Nearest Neighbors (KNN), and Decision Trees have been developed to detect infected chickens using the chromaticity data. Based on the X and Z chromaticity data from the chromaticity analysis, the color of the infected chicken's comb converged from red to green and yellow to blue. The development of the algorithms shows that Logistic Regression, SVM with Linear and Polynomial kernels performed the best with 95% accuracy, followed by SVM-RBF kernel, and KNN with 93% accuracy, Decision Tree with 90% accuracy, and lastly, SVM-Sigmoidal kernel with 83% accuracy. The iteration of the probability threshold parameter for Logistic Regression models has shown that the model can detect all infected chickens with 100% sensitivity and 95% accuracy at the probability threshold of 0.54. These works have shown that, despite using only the optical chromaticity of the chicken comb as the input data, the developed models (95% accuracy) have performed exceptionally well, compared to other reported results (99.469% accuracy) which utilize more sophisticated input data such as morphological and mobility features. This work has demonstrated a new feature for bacteria- or virus-infected chicken detection and contributes to the development of modern technology in agriculture applications.

Keywords: agriculture; chicken comb; chromaticity; classification model; diseases-infected chicken; energy; image processing; machine learning.