天文图像序列中弱目标的实时检测算法

Real time detection algorithm of weak targets in astronomical images

针对天文图像中运动弱小目标的检测问题,在分析天文CCD图像特点的基础上,根据待检测目标运动状态的不同,提出:1)在检测动目标时,对基于图像对称差分运算方法进行了改进,改进后的方法性能优于图像差分法,且硬件实现容易。该方法以连续三帧序列图像为一组处理对象,在进行绝对差运算之前,对图像进行对比度增强及均值滤波;2)使用形态学滤波的方法实现单帧静止多目标的检测,该方法采用top-hat算子完成背景的估计与目标的检测。为了实时实现所提出的动目标及静止目标的检测算法,设计了DSP+FPGA硬件架构方案,并进行了外场实验。实验的结果表明,检测算法在硬件加速的情况下可以实时有效地检测到SNR≈2的弱小目标,并可以同时实时保存原始图像数据。

In view of problem of detecting weak targets in astronomical images, based on the analysis of the characteristics of the astronomical CCD images and the different moving states, two improved methods are proposed： 1） The symmetrical difference operation algorithm is improved for detecting the moving targets. The performance of the improved algorithm is better than that of image difference method and it is easier to realize with hardware. In this algorithm the consecutive three flame images are treated as one group. Enhance image contrast and filter the images before absolute difference operation. 2） Morphological filter algorithm is adopted to detect static multi-targets in single image. The top-hat operator is applied to estimate the background and detect the targets. In order to realize the detection algorithms mentioned above in real time, the hardware device was developed with DSP and FPGA. The fieldwork experiment result shows that the weak targets of SNR≈2 can be effectively detected and the massive original images can be saved.