【摘要】 针对现有的目标检测算法检测表面亮度低的小尺度星系时效果不理想的问题，该文提出了一种基于掩码机制与目标交叉认证的低表面亮度的小尺度星系检测方法。首先，针对天文图像设计了一个基于目标总数变化率的阈值确定方法来获取阈值；其次，设计了基于掩码机制的目标去除方法和基于自适应半径的点源区域获取方法，结合图像分割和点源检测算法生成非检测目标掩码，与原图进行逐点相乘去除图中体积较大、亮度较高的非检测目标，得到亮度微弱、体积较小的候选者；然后，利用图像分割技术获取候选体分割区域，计算区域面积和质心坐标定位候选者；最后，通过目标交叉认证的方法将候选者与星表中真实记录的星体进行坐标差值计算获取星系目标。实验与分析表明，在SDSS(Sloan Digital Sky Survey)天文数据集上该方法对低表面亮度的小尺度目标检测率可达约94.90%,星系的识别率可达到约89.21%,都高于经典的目标检测算法。更多还原

【Abstract】 The existing object detection algorithms are not effective in detecting small-scale galaxies with low surface brightness. To solve this problem, a low-brightness small-scale galaxy detection method based on mask mechanism and target cross-identification is proposed. Firstly, a method based on the change rate of total number of objects is designed to obtain the threshold for astronomical images. Secondly, a objects removal method based on mask mechanism and a point-like sources regions location method based on adaptive radius are designed. At the same time, the above methods combine image segmentation and source detection to generate a non-detection objects mask, and non-detection objects, such as large objects, small point sources and noise, are removed by multiplies original image point-by-point with mask, and candidates are obtained with low surface brightness and small scale. Then, the segmentation region of the candidate is obtained by image segmentation technology, and the region area and centroid coordinates are calculated to locate the candidate. Finally, the coordinate difference between the candidates and the stars recorded in the star-catalog is calculated by cross-identification method to obtain the galaxies. Experiments and analysis show that on the SDSS(Sloan Digital Sky Survey) astronomical datasets, the detection rate of the proposed method for small-scale objects with low surface brightness can reach about 94.90%,the recognition rate of galaxies can reach about 89.21%,which is higher than that of the classical object detection algorithm.更多还原