Non-allelic recombination interactive genetic algorithms for colour imagery reproduction in colour scheming

Abstract

Using image extracted colours as the colour source for product colour matching is a common design strategy. The process of colour scheme design involves both source colour combination optimisation and continuous variable optimisation. As a commonly used optimising tool, interactive genetic algorithm (IGA) recombines the colours between alleles, that is, the same area in different colour schemes. Two types of invalid schemes are prone to arise during optimisation. The first is the colour duplication in a colour scheme's areas, which are supposed to be painted differently. The second is the unexpected new colours generated through interpolation between alleles. This article proposes a non-allelic IGA (NA-IGA) method, which allows colour recombination among non-alleles, to avoid invalid schemes and improve algorithm efficiency. The crossover, evaluation, and selection operators are redesigned. NA-IGA tries to maintain the features of extracted colours in the colour schemes while taking user preference into consideration. Thus, interaction methods for colour matching optimisation are designed to extend the evaluation mode of traditional IGA, including colour shuffling, continuous fine-tuning, and so forth. An application test is conducted in the rail vehicle's painting design. The comparison experiment with traditional IGA shows that NA-IGA significantly reduces the generation of invalid schemes and significantly improves efficiency and effectiveness of the optimisation.