摘要
本文提出了一种基于超分辨率图像重建的质心细分定位的新方法。在图像识别与匹配中,往往需要用到物理、数字等的特征提取方法,当给定的图像分辨率低时,就会使得所提取出的特征产生不可忽略的误差。为了解决这一问题,本文以实拍星图分辨率低的局限性为例,并结合传统的质心提取方法得到观测星图中任意两颗星的角距,验证新方法降低误差的有效性。实验结果表明,在同等系统误差条件下,相对于原始星图求得的星角距,基于超分辨率重建后的星图所得到的观测星的角距值更接近于真实角距,精度提高了29.56%,即新方法提取到的特征更加精确。
A novel method is put forward in subdividing locating of star image in this paper, which is based on super resolution image reconstruction that can generate high resolution image with more information contained by using one or more low resolution images. When the resolution of given image is low, it makes the features extracted errors not be ignored. In order to solve this problem, taking the limitations of real star map with low resolution for instance, we use the traditional method of centroid extraction to obtain any two stars＇ angular distance in star map, verifying the effectiveness of the new method in reducing errors. By comparing the angular distances with real angular distances in star database, the results show that angular distance achieved from super resolution reconstruction star map is closer to real value, with the precision 29.56% improved than that from original star map.
出处
《电子设计工程》
2015年第9期127-130,134,共5页
Electronic Design Engineering
关键词
稀疏表示
超分辨率
字典学习
星图识别
质心细分
sparse representation
super resolution
dictionary learning
star Identification
subdivided Locating
