Adaptive-image learning and discrimination techniques using classifier systems are presented. The genetic algorithm (GA) is used for a learning strategy in the system. The proposed system learns arbitrary image objects without any prior knowledge of given images and recognizes them. The system also makes up for some general weak points that are present in most learning systems including conventional classifier systems. That is, first, in a learning system, forgetting of knowledge usually occurs if the knowledge is not used for a long time period. The system still maximizes adaptability, but it prevents the system from forgetting useful rules by using the No_Unlearn mode. Second, to improve large-class image classification and learning, a multiple sublength concept has been introduced to genetic algorithms. Third, a triggered GA, which plays an important role in distinguishing two or more similar images by eliminating generalists, is developed.