Publication: Early detection of Newcastle disease virus - Infected broiler chickens using machine learning based on optical chromaticity and morphological features
| dc.contributor.author | Mohd Anif Akhmal Abu Bakar | |
| dc.date.accessioned | 2025-05-19T02:25:38Z | |
| dc.date.available | 2025-05-19T02:25:38Z | |
| dc.date.issued | 2025-05-15 | |
| dc.description.abstract | The Newcastle disease (ND) outbreak in chickens has raised significant concerns since the Newcastle disease virus (NDV) is highly contagious in chickens and poses a catastrophic impact on populations which can lead to economic losses. Conventional methods of detecting infected chickens rely on manual observation and laboratory tests for confirmation. This method often suffers from delayed detection, leading to major disease outbreaks. Therefore, this research aims to detect NDV-infected chickens in the early stages of infection using image processing techniques and various machine learning classifier models (ML). The research is divided into two main phases. Phase 1 focuses on developing an innovative technique to detect chickens infected with various types of diseases caused by bacteria or viruses based on the optical chromaticity of the chicken comb. This phase analyses the distinctive features of the chromaticity of the chicken comb represented by parameters x, y, and z using the CIE XYZ color space. Chromaticity features were extracted to classify healthy and infected chickens using ML. Phase 2 involved an experimental chicken trial that was carried out for 50 days using specific pathogen free broiler chickens. Surveillance cameras with online connections were set up inside the chicken cages to capture the chicken images without human intervention, simulating actual farm conditions. This phase focuses specifically on detecting NDV-infected chickens based on the combination of the chromaticity and morphological features. The morphological features of the standing chicken were represented by parameters such as circularity, convexity, eccentricity, roundness, and belongation. Morphological and chromaticity features were extracted at 36 hours post infection (after NDV infection) with a 12-hour time step onwards, indicating the severity of the symptoms. This approach enables the development of ML and neural networks to detect infected chickens with respect to the severity of symptoms. Based on the chromaticity analysis in the first phase, the color of the infected chicken's comb transitions from red to green and yellow to blue. The development of various ML using the chromaticity features showed that Logistic Regression (LR), Support Vector Machine (SVM) with Linear and Polynomial kernels performed the best with 95% accuracy. For the well-structured experimental chicken trial in Phase 2, it was established that within 36 hours of NDV-infection, LR and SVM models with Polynomial kernels predict and detect infected chickens at an accuracy of 78.33% and achieved 82.39% after feature optimization. After 72 hours of infection, the best detection is achieved by SVM with Linear kernel model, achieving an accuracy rate of 91.67%. Overall, Phase 1 of this research has established that the chromaticity of chicken comb features can be utilized to detect infected chickens. The study carried out in Phase 2 is the first report on detecting NDV-infected chicken as early as 36 hours post-infection using image processing and ML with an accuracy rate of >75% based on the combination of chicken's comb chromaticity and morphological features. These significant findings contribute to advancing modern agricultural automation technology for early disease detection, ultimately enhancing food security | |
| dc.identifier.uri | https://irepository.uniten.edu.my/handle/123456789/37996 | |
| dc.language.iso | en | |
| dc.title | Early detection of Newcastle disease virus - Infected broiler chickens using machine learning based on optical chromaticity and morphological features | |
| dc.type | Resource Types::text::Thesis | |
| dspace.entity.type | Publication | |
| oaire.citation.endPage | 137 | |
| oaire.citation.startPage | 1 | |
| oairecerif.author.affiliation | #PLACEHOLDER_PARENT_METADATA_VALUE# |