The work aims to explore the cultural connotation, improve the innovation and diversity of traditional color innovation design, and solve the problems of lack of cultural heritage and low design efficiency in traditional color innovation design. Firstly, the explicit traditional color library and the implicit cultural image database were constructed by web crawler, clustering algorithm and natural language processing, and the traditional color culture dataset was constructed by combining the semantic difference method and ResNet18. Then, C-WGAN was used to construct a model to generate traditional color innovation design schemes, which could creatively produce product color design schemes that conformed to the color matching rules and were rich in cultural imagery. Finally, with the Wuqiang New Year painting as the research object of traditional color culture, the tourist sightseeing bus was used as the design application object to carry out the innovative design of traditional color, and a design system was built to verify the effectiveness and applicability of the method proposed in the paper. A traditional color innovation design method based on deep learning was proposed, which could quickly design product color design schemes that conformed to cultural imagery, and realize the integration of traditional colors, cultural heritage and modern products. This paper verifies the application potential of deep learning technology in traditional color innovation design, provides a feasible and reliable idea for traditional color innovation design, and promotes the innovation and inheritance of traditional color in modern product design.