【摘要】 针对装备建模中采用传统基于草图的3D建模技术生成模型质量较差的问题，提出新的3D模型智能化生成和改进方法。从手绘草图中提取形状和视角特征，采用基于有向距离场的深度神经网络，根据2D草图形状特征生成3D模型，基于2D-3D对齐策略，在预测的手绘草图视角下，使3D模型的可微分渲染轮廓接近真实的手绘草图轮廓。提出支持手绘交互式的3D模型设计改进方法，可根据手绘轮廓线对3D模型进行调整。试验结果表明：采用所提方法生成的3D模型形状特征更接近手绘草图，可应对手绘草图视角的不确定性，降低人工改进成本，为装备生成式概念设计提供技术支持。更多还原

【Abstract】 In order to solve the issue of poor quality of 3D model generated by traditional sketch-based 3D modeling technology, a novel intelligent 3D model generation and optimization method was proposed. The shape and perspective features were extracted from hand-drawn sketches. Deep neural network based on the signed distance field was used to generate a 3D model based on the 2D sketch shape features. Based on the 2D-3D alignment strategy, the differential rasterization of 3D model was made under the predicted sketch perspective. An interactive hand-drawn 3D model design improvement method was proposed, which could adjust 3D silhouettes according to hand-drawn contour lines. The experimental results showed that the 3D model generated by the proposed method could get closer to hand-drawn sketch and cope with the uncertainty of the perspective of hand-drawn sketches, and reduce the cost of manual optimization. The research results could provide technical support for the generative conceptual equipment design.更多还原