Extraction of Cattle Retinal Vascular Patterns with Different Segmentation Methods

In the field of animal husbandry, the process of animal identification and recognition is challenging, time-consuming, and costly. In Türkiye, the ear tagging method is widely used for animal identification. However, this traditional method has many significant disadvantages such as lost tags, the ability to copy and replicate tags, and negative impacts on animal welfare. Therefore, in some countries, biometric identification methods are being developed and used as alternatives to overcome the disadvantages of traditional methods. Retina vessel patterns are a biometric identifier with potential in biometric identification studies. Preprocessing steps and vessel segmentation emerge as crucial steps in image processing-based identification and recognition systems. In this study, conducted in the Kars region of Türkiye, a series of preprocessing steps were applied to retinal images collected from cattle. Fuzzy c-means, k-means, and level-set methods were utilized for vessel segmentation. The segmented vascular structures obtained with these methods were comparatively analyzed. As a result of the comparison, it was observed that all models successfully performed retinal main vessel structure segmentation, fine vessels were successfully identified with fuzzy c-means, and spots in retinal images were detected only by the level-set method. Evaluating the success of these methods in identification, recognition, or disease detection will facilitate the development of successful systems.