In this paper, we first extract visual elements from the pattern, color, and copy features of brand products using saliency detection methods. Based on the multi-instance collaborative learning mechanism, combined with the semantic correlation relationship between samples, user connection, social distance and other social factors, the correlation between data and brand entities is reasoned to accurately obtain brand entity information. Finally, the social brand entity information is used to optimize the brand value communication strategy of the information flow, including the innovation of intelligent advertisement and scenario-based visual communication, as well as the interactive intelligent advertisement of the social information flow. The practical application found that the brand relevance accuracy detection rate is up to more than 0.93, and the NDCG of visual communication element evaluation of different brands is between 0.93.9 and 0.98.6. The best relevance coefficient and communication effect is achieved when the three elements of pattern, color and copy are combined, with the relevance coefficient reaching 0.95 and the conversion rate being 2.9%. In the digital era, the innovation of visual communication art provides more inspirations and strategic support for brand communication, helping brands to better establish association with consumers in the digital era, and realizing the enhancement of brand image and maximization of communication effect.