精确的血迹年龄预测具有重大的法医学价值。利用可见反射光谱技术与偏最小二乘法(PLS)相结合分析预测血迹年龄。遗传算法与偏最小二乘法相结合被用来选择有效光谱区间。与全光谱PLS 模型相比较，建立在优化光谱区间的遗传区间偏最小二乘法(GA-iPLS)模型具有更好的预测能力。结果表明GA-iPLS 能合理地选择有效光谱区间，提高预测能力。在考虑取自不同个体血迹特异性的情况下，建立在2.00~48.00 h时间段和48.00~1080.00 h时间段的GA-iPLS 模型的相关系数(Rp)、预测标准误差(RMSEP)和剩余预测偏差(RPD)分别为0.9949/0.9924、1.59 h/43.56 h、10.32/8.42。两个GA-iPLS模型的结合可代替建立在2.00~1080.00 h时间段的GA-iPLS模型精确预测2.00~1080.00 h 时间段的血迹年龄。结果表明可见反射光谱与GA-iPLS 模型在法医学领域可成为一种可靠的精确预测血迹年龄的方法。

Accurate age estimation of bloodstains can provide enormous forensic value. Visible reflectance spectroscopy technique combined with partial least squares (PLS) is applied to accurately estimate the age of bloodstains for forensic purposes. Genetic algorithm (GA) combined with PLS is used to select the most efficient spectral intervals. The genetic algorithm interval partial least squares (GA- iPLS) models built in the optimal intervals, present better predictive capability than full-spectrum PLS model. GA-iPLS can validly select desirable intervals and improve predictive ability. Considering the effect of the specificity of bloodstains on models, the GAiPLS models built in age period from 2.00 h to 48.00 h and in age period from 48.00 h and 1080.00 h are achieved with correlation coefficient (Rp) of 0.9949/0.9924, root-mean-square error of prediction (RMSEP) of 1.59 h/43.56 h , residual predictive deviation (RPD) of 10.32/8.4243 respectively, which can be used to accurately predict the age of bloodstains from 2.00 h to 1080.00 h instead of GA-iPLS model in 2.00~1080.00 h. The results demonstrate that visible reflectance spectroscopy combined with GA-iPLS will be a reliable tool to accurately estimate the age of bloodstains for forensic practical applications.