Intelligent generation of product bionic image forms via multimodal emotion-weighted quantification

To address key limitations in traditional bionic image design, including the lack of quantitative analysis of user preferences and the poor controllability of AI-driven picture generation, which often leads to homogeneous outputs with limited emotional resonance, this study proposes a data-driven bionic image generation method empowered by StyleGAN (DL-BBI). The proposed approach unfolds in three stages. First, a Multimodal Emotion Cognition Experiment was conducted by integrating visual-linguistic-expression data with emotional cognitive saliency to extract biological image features. Second, an Affective Weighted Quantification Method was proposed to process users' multimodal subjective and objective emotional preference data, thereby constructing an image stimulus graphic aligned with user preferences. Finally, by applying a StyleGAN latent space linear interpolation technique, the controllable fusion of product-biological image features was achieved, enabling intelligent generation of bionic product forms. Using a companion robot as an example, the method's feasibility and superiority were validated. Results show that this approach effectively enhances the emotional resonance between the generated design solution and users, with the "Friendly" score of the design solution improving by approximately 22.6% compared to the original solution.