The Arena Fragmentation Test(AFT) is one of the important tasks for designing a high-explosive warhead. Inorder to measure the statistics of a warhead in the test, fragments of a warhead that penetrate steel plates aredetected by using complex and expensive measuring equipment. In this paper, instead of using specific hardware tomeasure the statistics of a warhead, we propose to use an image processing based object detection algorithm todetect fragments in AFT. To this end, we use a hard-thresholding method with a brightness feature and apply amorphology filter to remove noise components. We also propose a simple yet effective temporal filtering methodto detect only the first penetrating fragments. We show that the performance of the proposed method iscomparable to that of a hardware system under the same experimental conditions. Furthermore, the proposedmethod can produce better results in terms of finding exact positions of fragments.