针对一系列含有圆和直线特征的航天器, 提出了基于嵌入式算法来实现非合作目标的姿态测量。该算法在现场可编程门阵列(FPGA)中完成边缘点的检测和边缘图像的细化; 在数字信号处理器(DSP)中通过基于弧段的快速椭圆检测方法检测目标上的椭圆特征; 同时用快速高精度的二次Hough直线检测方法检测目标上的直线特征。最后, 针对实际在轨任务进行了地面模拟实验, 基于椭圆特征和直线特征计算了目标飞行器和一个CCD相机的相对姿态。结果显示: 该嵌入式方法所需资源较少, 更新率优于3 Hz; 其测量精度随着测量距离的变大而变差, 同一距离3个方向的测量精度不同, 近距离(小于4 m)、深度方向的测量精度优于100 mm, 其他方向最高优于60 mm, 角度精度优于1°, 满足嵌入式系统对资源、在轨更新率和精度的要求。

For the spacecrafts with circle and line features, an algorithm for embedded vision applications was proposed for the relative pose measurement between non-cooperative targets. The edge points were detected by a Canny edge detector and the image thinning was performed in a Field Programming Gate Array(FPGA). Then, elliptic characters on the target were detected by a novel ellipse detection algorithm based on edge contour in a Digital Signal Processor(DSP), and line features on the target were detected by a real-time and high accuracy method base on Hough Transform(HT). Finally, a ground simulation experiment was performed for a practical mission on orbit. and the relative attitude between target spacecraft and CCD camera was calculated by the obtained ellipse and lines. Experimental results show that the proposed method have been implemented in the embedded system and the update rate is more than 3 Hz. The measurement accuracy becomes poorer with the increasing of measuring distance and that in three directions is different even in the same measure distance. When the measuring distance is less than 4 m, the measurement accuracy in the depth direction is higher than 100 mm, other directions is higher than 60 mm and the angle accuracy is higher than 1°. The obtained results meet the requirement of the embedded system for the system resource, update rate and the detection accuracy.