缩放
计算机科学
稳健性(进化)
工作流程
计算机视觉
粒子群优化
人工智能
合成
计算机图形学(图像)
高斯分布
覆盖
工程制图
图像处理
面子(社会学概念)
运动估计
相似性(几何)
作者
Lan Lan,Dewen Cheng,Yongtian Wang,Tong Yang
出处
期刊:Applied optics-OT
[Optica Publishing Group]
日期:2025-11-26
卷期号:64 (35): 10622-10622
被引量:3
摘要
This paper presents an automated framework for early stage design of mechanically compensated zoom systems. The method integrates three stages: (1) thin-lens modeling, in which Gaussian modeling with particle swarm optimization generates feasible four-group thin-lens layouts; (2) thin-lens refinement, which enforces practical optical and mechanical constraints including F-number, half field-of-view (FOV), minimum spacing, and smooth zoom trajectories; and (3) real-lens substitution and quick optimization, where each thin-lens group is matched, scaled, and replaced using a database-driven strategy based on focal power, F-number, and ray-angle similarity with principal-plane alignment. A large-scale robustness evaluation across varied zoom ratios, half-FOV angles, and length-scaling factors shows 76% end-to-end success. Representative 3×, 20×, and 50× examples further demonstrate that the workflow reliably produces compact and manufacturable zoom configurations with smooth motion and balanced imaging performance. Overall, the framework provides higher efficiency, stable numerical behavior, and greater structural diversity than manual design, offering a practical foundation for data-driven zoom-lens development and seamless integration into optical-design environments.
科研通智能强力驱动
Strongly Powered by AbleSci AI