University of Tehran Press Iranian Journal of Biosystem Engineering 2008-4803 56 2 2025 08 23 Identifying and classifying cow types based on spinal end deviation using machine learning Identifying and classifying cow types based on spinal end deviation using machine learning 51 67 103941 10.22059/ijbse.2025.398902.665603 FA Mohsen Daneshman Vaziri Researcher, Agricultural Engineering Research Department, Isfahan Agricultural and Natural Resources Research and Education Center, AREEO, Isfahan, Iran.. Abdollah Imanmehr Assistant professor, Agricultural Engineering Research Department, Isfahan Agricultural and Natural Resources Research and Education Center, AREEO, Isfahan, Iran. 0000-0002-4556-0904 Mohsen Heidarisoltanabadi Associated professor, Agricultural Engineering Research Department, Isfahan Agricultural and Natural Resources Research and Education Center, AREEO, Isfahan, Iran. 0000-0001-7892-3487 Journal Article 2025 07 19 In dairy farms, machine learning operations can be used to identify and classify cow types based on body condition scoring (BCS) using features extracted from images. In particular, machine learning algorithms can analyze the curvature of the spine, often by identifying key points and fitting a line or curve, to distinguish between different breeds of cattle and assess their condition. In this study, machine learning models that have been frequently used in computer science in recent years, including SVM, KNN, and CNN, were used in conjunction with a pre-trained deep learning network Resnet50 to enhance the success of the architectures. In each of the algorithms, image features were extracted, registered, and merged to identify the type of cows, and finally, the pre-trained CNN algorithm based on deep learning was able to correctly identify the type of cow with the highest accuracy (93 percent). Therefore, by combining this processing system with the imaging mechanism, it is possible to identify and classify cows based on various states and physical characteristics in cattle environments in a shorter, simpler, and more user-friendly time. This approach eliminates the need for manual extraction of livestock features, reduces the use of human resources, and achieves improved recognition accuracy. In dairy farms, machine learning operations can be used to identify and classify cow types based on body condition scoring (BCS) using features extracted from images. In particular, machine learning algorithms can analyze the curvature of the spine, often by identifying key points and fitting a line or curve, to distinguish between different breeds of cattle and assess their condition. In this study, machine learning models that have been frequently used in computer science in recent years, including SVM, KNN, and CNN, were used in conjunction with a pre-trained deep learning network Resnet50 to enhance the success of the architectures. In each of the algorithms, image features were extracted, registered, and merged to identify the type of cows, and finally, the pre-trained CNN algorithm based on deep learning was able to correctly identify the type of cow with the highest accuracy (93 percent). Therefore, by combining this processing system with the imaging mechanism, it is possible to identify and classify cows based on various states and physical characteristics in cattle environments in a shorter, simpler, and more user-friendly time. This approach eliminates the need for manual extraction of livestock features, reduces the use of human resources, and achieves improved recognition accuracy. https://ijbse.ut.ac.ir/article_103941_d8fbe9b5c978b18e835381ef3f6c412b.pdf