Machine learning analysis of wing venation patterns accurately identifies Sarcophagidae, Calliphoridae and Muscidae fly species

• DOI: 10.1111/mve.12682
• Publication Date: 2023-07-21T10:05:30Z
• Authors: Min Hao Ling; Tania Ivorra; Chong Chin Heo; April Hari Wardhana; Martin Jonathan R...; Siew Hwa Tan; Zulqarnain Mohamed; Tsung Fei Khang
• Keywords: Forensic Entomology; Lucilia

In medical, veterinary and forensic entomology, the ease affordability of image data acquisition have resulted in whole-image analysis becoming an invaluable approach for species identification. Krawtchouk moment invariants are a classical mathematical transformation that can extract local features from image, thus allowing subtle species-specific biological variations to be accentuated subsequent analyses. We extracted invariant binarised wing images 759 male fly specimens Calliphoridae, Sarcophagidae Muscidae families (13 variant). Subsequently, we trained Generalized, Unbiased, Interaction Detection Estimation random forests classifier using linear discriminants derived these inferred identity test samples. Fivefold cross-validation results show 98.56 ± 0.38% (standard error) mean identification accuracy at family level 91.04 1.33% level. The F1-score 0.89 0.02 reflects good balance precision recall properties model. present study consolidates findings previous small pilot studies usefulness venation patterns inferring identities. Thus, stage is set development mature analytic ecosystem routine computer image-based importance.