Molecularly imprinted polymer (MIP-QDs) with fluorescence quenching ability toward propazine was synthesized for propazine detection. b(Blue)-MIP-QDs were prepared using ZnCdS/ZnS QDs via reverse micro-emulsion, whereas r(red)-MIP-QDs were synthesized using CdSe/ZnS QDs. By utilizing graphene quantum dots (GQDs) as a stable fluorescence intensity reference, the wide-gamut fluorescence test paper was constructed on the basis of mixing b-MIP-QDs, r-MIP-QDs, and GQDs under the optimal ratio. When analyzing spiked propazine in fish and seawater samples using a test paper, satisfactory recoveries of 104.0 %–114.6 % and 92.0 %–96.4 % were obtained, with corresponding limits of detection of 5.0 μg/kg and 1.0 μg/L, respectively. The RGB extractor was utilized to extract the actual fluorescence color and construct a dataset consisting of R, G, and B values, as well as concentration data from 400 samples. The SVR model of Python 3.9.7 was used to obtain and analyze the concentration and feature data. After optimization, the constructed model achieved a correlation coefficient of 0.98 and an RMSE of only 1.81, indicating high prediction accuracy and excellent generalization ability that meet quenching prediction requirements. As an intelligent and rapid detection method, this model holds significant practical significance.

中文翻译：

---

机器学习辅助的广色域荧光视觉试纸用于测定鱼类和海水样品中的丙嗪

合成了对丙嗪具有荧光猝灭能力的分子印迹聚合物（MIP-QD），用于丙嗪检测。 b（蓝色）-MIP-QD 使用 ZnCdS/ZnS QD 通过反相微乳液制备，而 r（红色）-MIP-QD 使用 CdSe/ZnS QD 合成。以石墨烯量子点（GQD）作为稳定的荧光强度参比，将b-MIP-QD、r-MIP-QD和GQD按最佳配比混合，构建了广色域荧光试纸。使用试纸分析鱼和海水样品中加标丙嗪时，回收率令人满意，分别为 104.0%~114.6% 和 92.0%~96.4%，相应的检出限分别为 5.0 μg/kg 和 1.0 μg/L。 RGB 提取器用于提取实际荧光颜色并构建由 R、G 和 B 值以及 400 个样品的浓度数据组成的数据集。使用Python 3.9.7的SVR模型获取并分析浓度和特征数据。优化后，构建的模型相关系数达到0.98，RMSE仅为1.81，预测精度高，泛化能力强，满足淬火预测要求。该模型作为一种智能、快速的检测方法，具有重要的现实意义。