As a brilliant treasure of human civilization, the non-legacy of the Shandong Yellow River Basin has always been an important carrier for people’s pursuit of beauty and unique expression. The rapid development of science and technology has resulted in the introduction of artificial intelligence and digital media technology, which have brought new vitality to cultural and creative visual communication design. The article introduces deep learning technology to analyze and extract non-legacy elements of the Yellow River Basin in Shandong, and deep learning technology mainly includes image recognition and image style migration technology. After studying the image recognition algorithm based on the migration of the generative adversarial model, the article proposes an image style migration model that integrates multi-scale discriminators, applies the decoupling training strategy so that the encoder is trained only when the adversarial loss is the largest, and finally sums up the visual communication design strategy optimized based on deep learning. Through the survey and study of 10 users’ satisfaction with the cultural and creative products designed by using the method of this paper, the total score rates of the four evaluation indexes, namely, content design, presentation design, element design and structural design, are 0.82, 0.83, 0.84 and 0.85, respectively, which the users highly praise. As a result, the visual communication design method proposed in this paper is effective for non-heritage cultural creations in the Shandong Yellow River Basin.