论文部分内容阅读
栅格算法作为星图识别算法中的代表,有着识别速度快、识别率高、导航数据库容量小的优点,应用较广,但其在构造恒星特征模式时存在需要选取临近星进行图像旋转的问题。为克服这一缺陷,提出了一种基于旋转不变特征的星图识别算法。该算法利用模式识别类星图识别算法思想,为每颗恒星建立一种与旋转无关的特征模式,采用改进的Hausdorff距离进行匹配识别。仿真实验结果表明,基于旋转不变特征的星图识别算法对星点位置噪声与缺失星鲁棒性均优于栅格算法,具有良好的识别效果,但其识别速度与存储容量仍需进一步优化。
As a representative of satellite image recognition algorithm, raster algorithm has many advantages, such as high recognition speed, high recognition rate and small capacity of navigation database. However, it has some problems that it is necessary to choose neighboring stars for image rotation when constructing star feature mode . In order to overcome this shortcoming, a star-based recognition algorithm based on rotation invariant features is proposed. The algorithm utilizes the idea of pattern recognition quasar image recognition algorithm to establish a rotation-independent feature pattern for each star, and uses improved Hausdorff distance for matching recognition. Simulation results show that the star recognition algorithm based on rotation invariant feature is superior to the grid algorithm in the noise of star point and the missing star robustness, and has a good recognition effect, but its recognition speed and storage capacity still need to be further optimized .