Faisal, ShahOoi, Melanie Po-LeenKuang, Ye ChowThawdar, YaminnAbeysekera, Sanush K.Reutemann, PeterFletcher, DaleFleming, RachelMcClelland, Heather2026-10-282026-10-282025Faisal, S., Ooi, M. P. -L., Kuang, Y. C., Thawdar, Y., Abeysekera, S., Reutemann, P., Fletcher, D., Fleming, R., & McClelland, H. (2025). Identifying biological fluids using hyperspectral near-infrared imaging and machine learning for forensics analysis. IEEE Instrumentation and Measurement Technology Conference Proceedings. https://doi.org/10.1109/I2MTC62753.2025.11078980979-8-3315-0500-42642-2069https://hdl.handle.net/10289/18725Near-infrared hyperspectral imaging captures subtle spatial and spectral characteristics of biological fluids on fabrics for forensics analysis. It is a non-destructive approach that could potentially be used for onsite crime scene investigations. However, current data processing methods lack the accuracy and sensitivity required for real-world applications. This paper utilised a 900-1700nm hyperspectral camera to capture blood, semen, and urine fluids that were systematically applied onto a patterned rayon fabric. Multi-Later Perception (MLP) and 1D Convolutional Neural Network (1D-CNN) classifiers were then trained to identify them. A dual evaluation approach, combining machine learning model evaluation metrics (F1-score and accuracy) with uncertainty quantification (Entropy-based and Modal Predictionbased) was then used to assess the performance of the classifiers. The results indicated that the MLP slightly outperformed the 1DCNN for fluid-specific classification tasks, achieving slightly better accuracy and F1-score and lower uncertainty in predictions.enThis is an accepted version of a paper presented at the 2025 International Instrumentation and Measurement Technology Conference (I2MTC). © 2025 IEEE.computational imagingdeep learningforensicshyperspectral imagingmachine learninguncertaintyConference Contribution10.1109/I2MTC62753.2025.110789802642-20774605 Data Management and Data Science46 Information and Computing Sciences4611 Machine Learning