YOLOv8-Logistic Integration for Raptor Tail Feather Microstructure Identification

Feathers are crucial evidence for identifying species in criminal cases, combating poaching, and curbing illegal trade. This study selected SEM images of tail feathers from 10 raptor species as raw materials, constructed a dataset via semi-automated annotation with X-AnyLabeling, and used the YOLOv8 algorithm for feature detection—accurately localizing key microscopic structure. A logistic regression model was developed for species classification using quantified features from YOLOv8, and a visualization platform was built with PyQt5, forming the “YOLOv8 + Logistic” intelligent identification system. Experimental results show an overall classification accuracy of 90.0% on the test set, with most species achieving 100% precision and recall, and an average AUC ≥0.97. This system provides efficient, objective species identification for frontline law enforcement, supporting the fight against raptor-related wildlife crimes.