Abstract: | A subset of colors often needs to be selected to represent a full set. In one such application, a multi‐band color sensor is used to measure reflective color samples, and a matrix transformation method is used to recover the reflectance spectrum of the measured sample. To achieve this, a group of training colors needs to be selected to calculate the transformation matrix. A genetic algorithm (GA) has been developed to optimize the selection of the subset of training colors, and the result is compared with those obtained using random selection or a traditional culling algorithm. In a simulation study, the GA gives better results. |