基于多目标优化方法,针对几个常用于评价多光谱相机性能的参数峰值信噪比(PNSR)、拟合优度系数(GFC)、CIE色差评价系数(CIEDE2000)和均方误差(MSE),及其对应的最大值或最小值,按照最优排序集分的方法,得到了各评价参数均可以得到较好取值的最优滤色器。和前期研究相比,待选滤色片数量由45增加为1035,同时成像场景的反射率样本数量由24增加为1269。结果表明,待选滤色片数量和反射率向量空间增加后,最优滤色器的性能更好;同时描述最优滤色器的光谱透过率向量矩阵自身的三个参数即滤色器向量矩阵的条件数(Cond)、峰值均匀度指数(UF)和相邻滤色片交叠度指数(OLP)的数值也更加稳定。实验结果可为通过滤色器向量矩阵定量选取宽带多光谱滤色片提供参考。

Based on the multi-objective optimization method, an optimal filter with good values for each evaluation parameter is obtained. This is achieved according to the cumulative scoring method of optimal ordering for several parameters often used to evaluate the performance of multi-spectral cameras such as peak signal to noise ratio (PNSR), goodness of fit coefficient (GFC), CIE color difference index(CIEDE2000) and mean square error (MSE) and their corresponding maximum or minimum values. Compared with the previous study, the number of filters to be selected increases from 45 to 1035; however, the number of reflectivity samples of the imaging scene increases from 24 to 1269. The results show that the performance of the optimal filter is better when the number of filters and the reflectivity vector space increase. At the same time, the values of the three parameters of the spectral transmittance vector matrix of the optimal filter, namely the condition number (Cond), the peak uniformity index (UF), and the adjacent filter overlap index (OLP), are also more stable. The experimental results can provide a reference for the quantitative selection of wideband multispectral filters via the vector matrix of the filter.