中文摘要：

高功率激光装置中靶的定位精度直接影响激光打靶的成功率。提出一种基于立体建模的进化式微小靶标位姿检测方法,该方法抛弃已有的图像处理算法对单幅图像进行特征提取的过程,通过建立立体模型将多视角图像中的目标相关联,实现相互制约的立体匹配。在模型优化方面,模拟人眼的＂扫视＂与＂注视＂功能改进遗传算法,使模型优化能同时兼顾精度与速度,为实现快速、高精度的靶自动检测与定位提供技术基础。实验搭建了3-CCD视觉检测平台并通过视觉反馈控制对标靶进行定位调整,所得定位精度小于5μm,角度误差小于0.05°。实验结果满足激光打靶要求,可应用于实际工程中。

英文摘要：

In high power laser device,target positioning accuracy affects the success of target hitting directly. This paper proposes an evolutionary pose measurement method for target positioning based on solid modeling; discarding the single image characteristic extraction using existing image processing methods,this method uses a solid model to relate the targets imaged with multi-cameras. The mutual restricted matching is achieved and the measurement error can be reduced. The adaptive genetic algorithm is used for model optimization in order to improve the accuracy and speed,aiming at providing key technology for fast target-laser selfvadjustment. A 3-CCD vision-based target positioning experimental system was constructed,which utilizes visual feedback control strategy to adjust the target pose. The precision of the target positioning is less than 5μm and the orientation error is less than 0. 05°,which satisfies the system precision requirement and can be used in practical engineering applications.