复杂场景中航天器靶标的快速识别

发布时间：2018-04-16 06:24

  本文选题：航天器靶标 + 目标识别　； 参考：《光学精密工程》2016年04期

【摘要】：基于靶标的几何特征提出了一种快速识别算法来解决目前航天器靶标识别存在的问题。设计了一款带有线段与圆环图案的合作靶标。利用高斯滤波去除图像中的噪声,运用Canny检测算子得到边缘图像,并跟踪得到单像素边缘序列。然后,通过判断非共线四点是否共圆排除大部分不可能构成圆的边缘,利用同一圆周的两段子弧对应相同圆心和半径的几何特征实现圆检测。最后根据靶标圆与线段的几何关系排除干扰,实现靶标的准确识别。实验结果表明,该靶标识别算法对噪声、光照、旋转等不敏感,能够在多种复杂场景中快速、准确地识别靶标,处理时间小于125ms,满足实时位姿测量8帧的需要。目前,该算法已经成功应用于工程样机。
[Abstract]:Based on the geometric features of target, a fast recognition algorithm is proposed to solve the problem of spacecraft target recognition.A cooperative target with line segment and ring pattern is designed.Gao Si filter is used to remove the noise in the image, and the Canny detector is used to get the edge image, and the single pixel edge sequence is obtained by tracking.Then, by judging whether the non-collinear four points exclude most of the edges of the circle that can't be formed, the circle detection is realized by using the geometric features of two sections of arc corresponding to the same center and radius of the same circle.Finally, according to the geometric relationship between the target circle and the line segment, the interference is eliminated and the target recognition is realized.The experimental results show that the target recognition algorithm is insensitive to noise, illumination and rotation, and can recognize the target quickly and accurately in many complex scenes. The processing time is less than 125 Ms, which meets the need of 8 frames of real time pose measurement.At present, the algorithm has been successfully applied to engineering prototype.
【作者单位】： 中国科学院长春光学精密机械与物理研究所;中国科学院大学;白城师范学院;哈尔滨工业大学;
【基金】：国家973重点基础研究发展计划资助项目(No.2013CB733103)
【分类号】：V52;TP391.41
，

本文编号：1757652