Automated Honey Bee Subspecies Identification Using Advanced Wing Venation Analysis and Adaptive Hierarchical...

Accurate honey bee subspecies identification is vital for biodiversity conservation and pollination resilience, yet current methods face critical limitations. Classical morphometric techniques, reliant on manual wing vein measurements, suffer from subjectivity and poor scalability across hybrid populations, while deep learning approaches demand extensive labeled datasets and exhibit limited interpretability in noisy field conditions. Crucially, existing methods fail to reconcile scalability with the ability to analyze phenotypic gradients in hybrid specimens. To bridge these gaps, we propose a novel computational framework integrating adaptive image processing with topology-aware clustering to enable scalable, label-free subspecies discrimination. Our pipeline begins with robust image preprocessing using nonlocal means (NLM) denoising and contrast-limited adaptive histogram equalization (CLAHE)-specifically optimized to enhance low-contrast venation patterns in field-collected images-followed by adaptive thresholding and morphological processing to isolate wing veins. Subsequent Zhang-Suen skeletonization extracts graph-based feature maps that uniquely encode both local vein geometry and global network topology, addressing the limitations of traditional morphometrics. Unlike supervised methods, our adaptive hierarchical clustering (AHC) algorithm dynamically infers subspecies clusters without predefined labels, enabling robust discrimination of hybrid populations and phenotypic intermediates. Evaluated on 26,481 wing images, the method achieves a silhouette score of 0.72 and improves classification accuracy by 26.1% over traditional morphometric techniques, demonstrating superior performance in noisy and hybridized datasets. This work resolves a key challenge in apid taxonomy by combining the interpretability of manual methods with the scalability of automation, providing conservationists with a practical tool for ecological monitoring.