SUPPLEMENTARY MATERIAL AVAILABLE ONLINE Indian Journal of Animal Sciences 93 (4): 401–405, April 2023/Article https://doi.org/10.56093/ijans.v93i04.131010 Principal component analysis in pig breeds identification SANKET DAN 1* , SATYENDRA NATH MANDAL 1 , PRITAM GHOSH 1 , SUBHRANIL MUSTAFI 1 and SANTANU BANIK 2 Kalyani Government Engineering College, Kalyani, Nadia, West Bengal 741 235 India Received: 2 December 2022; Accepted: 24 March 2023 ABSTRACT Maintaining the purity of pig breeds is an essential task for their economic value. The traditional breed identifcation methods through coat colour are prone to error due to huge intra-breed variation. This paper uses principal component Analysis (PCA) to classify the pig breeds using their images. Individual images of fve diferent pure breeds were captured from organized farms in India under both controlled and uncontrolled environments. Three diferent image sets were created, containing images in the controlled, uncontrolled, and mixed environment image sets. With 80:20 training to testing datasets, 93% accuracy was found in the proposed method of principal component analysis. Finally, two performance-based comparative analyses of our method were done with PCA-based methods and other renowned techniques used for animal breed identifcation, wherein our PCA method outperformed others in both comparative scenarios. Keywords: Breed identifcation, Confusion matrix, Euclidean distance, Image space, Principal components Animal Husbandry and livestock sector leads a major role in the socio-economic growth of the Country (Neethirajan and Kemp 2021). According to the World Development report 2008, 70% of the rural economic growth depends on the livestock sector (Pica et al. 2008). Animal breed detection is one the major tools for boosting the fastest growing livestock sector. Genotype based marking procedures like whole-genome sequencing, microsatellite markers etc. have been used for recognition of animal breed. The phenotypic features such as muzzle print, body shape, coat colours and pattern have been used as the recognition trait(s) of individual animals and also the individual breed (Lahiri et al. 2011, Andrew et al. 2016, Kumar et al. 2018). Based on the phenotypic features, some research works were done for identifcation of animal breed like dog using Convolution neural network, Artifcial neural network, Deep learning (Hsu 2015, Ráduly et al. 2018, Borwarnginn et al. 2019, Mandal et al. 2020, Fuad et al. 2021). Principal Component Analysis has been widely used for face detection, animal recognition, image compression etc. In this paper, pig breeds were recognized from their images using Principal Component Analysis. The individual pig from fve pig breeds were captured from organized farms located at diferent places in India. Total 1000 images were captured to build the pig breed dataset where 500 images were captured in a controlled environment and the rest of the images are from uncontrolled environment. In both cases, the side profle of each pig was captured. The captured images were divided 80:20 ratio as training and test sets. The principal component analysis was applied on training sets and the training templates (Central mean, image space and project train image set) were developed. In test phase, the test template (projected test image) was developed for each pig image using central mean and training image space from training phase. The Euclidean distances were calculated among projected test image and each column of projected training image set. The minimum value (Minimum Euclidean distance) was calculated and corresponding column was selected in project training image set. The breed of test image was predicted as same breed at selected column in projected training set. Finally average precision, recall F1-score and accuracy were computed and a comparison was made with accuracies with other breeds from other established method. MATERIALS AND METHODS In this paper, the principal component analysis was applied on captured pig images and breeds of images were predicted. The details of methodology is divided into some steps as shown in Fig. 1. The steps were operated sequentially where frst block was pig image capturing and fnal step was pig breed prediction. Pig image capturing and image grouping: The individual pig was captured in controlled and uncontrolled environments using both mobile phones and DSLR camera. Among the registered and exotic pure breed pigs available in India, fve breeds named Duroc, Ghungroo, Yorkshire, Present address: 1 Kalyani Government Engineering College, Kalyani, Nadia, West Bengal. 2 ICAR-National Research Centre on Pig, Rani, Guwahati, Assam. * Corresponding author email- sanketdan@gmail.com 81