为了提高在全天球自主工作模式下星图识别的成功率和鲁棒性, 本文结合kNN算法和有向图理论的思想, 构造出具有强约束的有序星点集模式, 提出基于有序星点集的全天球自主星图识别算法。该方法首先利用k近邻算法的思想, 以导航星点为中心, 对位于其一定范围内的导航星进行了分类划分。然后基于有向图理论, 以距离中心点星最近的导航星为基准, 按照顺时针顺序对分类的导航星进行排序, 构造出具有强约束特性的有序星点集作为星图识别的特征。实验结果表明: 在存在星点位置误差和伪星点的情况下, 本文提出的基于有序星点集全天球自主星图识别算法具有很强的抗噪声能力、抗伪星点干扰能力和鲁棒性。在星点质心位置达到3像素时, 基于有序星点集星图识别算法成功率仍然可以达到99.8%, 比三角形识别算法和栅格识别算法的识别成功率高16%以上; 在存在3颗伪星点的情况下, 基于有序星点集星图识别算法成功率为98.4%, 比三角形识别算法和栅格识别算法高10%以上。

In order to improve the success rate and robustness of all-sky automation star identification under star position error and pseudo stars, a novel star identification is proposed, which utilizes ordered star point set as star pattern. First, stars in certain range are classified by kNN algorithm as the center is a specified star, then based on the graph theory, with the nearest navigation star as reference, the classified stars are ordered in a sequence by the clockwise order. Hence, a star pattern under strong constraints is established. The numerical results imply that the proposed star identification algorithm is robustness to star position noise and false stars. Experimental results indicate that under the condition that star position error is 3 pixels, the success rate of the proposed algorithm is 99.8%, which is 16% more higher than the triangle identification algorithm and grid identification algorithm. In the case that there are 3 pseudo stars in the star image, the success rate of the proposed algorithm is 98.4%, and it is 10% more higher than the triangle identification algorithm and grid identification algorithm. It is obviously that the proposed star identification own very high success rate and strong robustness in harsh environment.