The first image is the original image which includes both point-like sources and streak-like sources. We transform the gray values in this figure to their log values for better visualization. The second image is the result obtained by our framework. Red boxes represent the streak-like astronomical images and green boxes represent point-like astronomical images. The third image is the result of detection by the classic framework. Because the classic framework cannot classify targets into different types, we assume all the astronomical targets detected by the classic framework can be correctly classified. As we can see in the third image, the classic framework does not perform well for blended sources, in addition, the streak-like target will be detected multiple times by the the classic framework. Therefore, when comparing the detection performance of our framework and that of the classic framework, targets that are detected multiple times are only calculated once.