一种地球紫外中心快速提取方法

A Fast Earth Ultraviolet Center Extraction Algorithm

针对地球紫外中心指向高精度提取问题,提出一种基于梯度统计的快速、低存储需求的紫外地平圆盘中心提取算法。首先,考虑到星载计算机与高内存消耗的滤波算法、实时导航需求的冲突,采用结合Sobel边缘算子与局部二值模式(LBP)算子的改进边缘快速提取方法,有效剔除背景噪声并准确提取地球紫外临边特征。然后对非完整临边边缘采用最小二乘拟合得到中心的精确位置。实验结果表明,该方法抗噪声,可实现亚像素级地球中心提取,并显著降低存储需求和计算时间开销。对于1046×746的图像,该算法需要的存储空间仅为1100 k Byte,运算时间在20 ms内,满足自主导航的需求。

A fast earth center extraction algorithm based on gradient statistic is proposed to accurately extract the earth ultraviolet center point. First,a fast edge detector combining Sobel edge operator with LBP（ Local binary pattern） feature descriptor is proposed to balance computer property of on-board computer and high memory consumer of filter algorithm.This proposed edge detector can both filter background noises and extract earth ultraviolet edge features. Then earth center is obtained by using least square fitting forthe non-integrity ring edge. Experiments show the proposed method is noise immunity and provides a sub-pixel level earth center,with a remarkable low requirement for memory capacityand real-time computation. Specially,at memory space of 1100 k Byte,the method provides accurate earth center within 20 ms,thus satisfying the requirement for autonomous navigation.