A KL Divergence-Based Test Strip Classification Algorithm with Attention-Contrastive Learning Network

In recent years, Biochemical test strips have become widely used in daily life for applications such as urine testing, pool water quality checks, and pregnancy tests due to their convenience, speed, and low cost. However, the classification accuracy of these test strips is often affected by subjective factors like psychological influences, color blindness, and age-related vision problems. In this paper, we propose KLCANet (KL-ContraAttention Network), a KL divergence-based classification algorithm utilizing an attention-contrastive learning network to improve the accuracy and robustness of test strip color block classification. We conduct experiments on a dataset of 1,657 pool water quality test strips that we collected, and the results show that KLCANet achieves superior accuracy compared to existing algorithms, offering better performance and computational efficiency.