This paper presents an automated, intent-driven method for generating multimodal simulated images of space non-cooperative targets to facilitate intent recognition. Existing simulators (e.g., SPIN, SISPO) are often limited by simplistic light source and material models, a narrow range of output data modalities, and crucially, a lack of modeling for the target's dynamic intents. To address these limitations, we developed an automated workflow within Blender that facilitates switching between advanced real-time and offline rendering engines. This workflow, which follows an ‘input → intent-driven trajectory generation → image rendering → output’ pipeline, incorporates high-fidelity 3D models, realistic space illumination, physically-based rendering material properties, and sensor effects. The workflow can generate multi-channel, timestamped image data with corresponding intent labels, based on specified dynamic models of intent. We detail the method's implementation and, by analyzing results obtained with varied simulation parameters, demonstrate its capability to produce time-series image sequences that encapsulate key dynamic features indicative of underlying intent. This approach provides high-quality, data-driven support for the training and evaluation of intent recognition algorithms.