Abstract: To investigate the effects of expert strategies and environmental conditions in demonstration data collection on the performance of embodied intelligence models, this paper proposes a demonstration data collection strategy based on dual Vision in Action (ViA) segmented sampling. Illumination, background, and tabletop conditions are controllably configured to construct comparative datasets. Based on a dual-arm collaborative robotic platform, an industrial raw-material grasping and placing task with clutter interference is conducted. Grasp success rate, placement success rate, and task execution position deviation are recorded to evaluate model performance in the validation stage. The experimental results show that, compared with data collected using the dual ViA segmented sampling strategy under conventional environmental conditions, data collected by combining the dual ViA segmented sampling strategy with illumination, background, and tabletop condition control increase the model grasp success rate by 52 percentage points and the placement success rate by 61 percentage points, and reduce the task execution position deviation by 70% in the validation stage. These results indicate that dual ViA segmented sampling combined with controllable environmental condition collection can improve the task success rate of the model in complex scenarios and reduce placement position deviation. Expert strategy organization and environmental condition control can improve demonstration data quality and enhance model validation performance in complex manipulation scenarios.