Dairy farming has been part of agriculture for thousands of years. Today, modern dairy cows are bred specifically to produce large quantities of milk. Dairy cattle are the most efficient of all farm livestock in converting feed protein and energy to food. Body condition scoring is a visual assessment of the amount of fat and muscle covering a cow's bones, regardless of body size. Body condition score (BCS) estimates the mobilization of cattle's energy reserves or the degree of fatness or thinness using a 5-point scale. BCS is an important management tool to maximize milk production and reproductive efficiency while reducing the incidence of cattle's metabolic diseases. It is also an important indicator to determine the amount of dry matter consumption in cows' early lactation. In traditional methods, BCS is evaluated by veterinarians or skilled personnel, which is costly, especially at large industrial farms, and is performed at low speed and over long periods and has low efficiency. Automatic and objective dairy cow body condition scoring can be used as a feed, reproduction, health, and longevity management tool. In this study, the automatic evaluation of BCS by digital imaging method and image processing technique was investigated to replace the traditional manual method. This study aims to establish an automated visual machine system for scoring dairy cows' physical condition using low-cost digital cameras and artificial neural networks. This method is a non-contact and stress-free platform on cows. In this method, images of cows were continuously recorded by a conventional digital camera. The extremities of the cows were identified from the original image and then subtracted from the background. Finally, after pre-processing, the general contour of the end of the cow body was obtained, and 59 different features were extracted. These features were considered as neural network input, and the BCS scores were considered as the network output. The two-stage Key Words: Image Processing, Artificial Neural Network, dir=rtl
