A highly automated moving object detection package
A highly automated moving object detection package
复制标题
高度自动化的移动物体检测包
DOI：
10.1111/j.1365-2966.2004.07217.x
复制
发表时间：
2004
影响因子：
4.8
通讯作者：
B. Gladman
中科院分区：
文献类型：
--
作者：
J. Petit;M. Holman;H. Scholl;J. Kavelaars;B. Gladman
随着大型CCD拼接相机的部署及其在大规模巡天中发现太阳系天体的应用，需要一种能够以近乎自动化的方式处理大数据负载的快速检测算法。我们在这里展示一个我们开发的算法。我们的方法通过使用两种独立的检测算法并结合结果，在产生低误检率的同时保持了高效率。这些属性对于减少与搜索这些庞大数据集相关的操作时间至关重要。我们将该算法应用于加拿大-法国-夏威夷望远镜（CFHT） CFH12K相机获得的两个不同的拼接数据集。将每个单独算法的检测效率和误检率与两者的组合进行比较，我们发现我们的方法将误检率降低了几百到一千倍，而将“极限幅度”（检测率降至50%）仅降低了0.1-0.3等。极限幅度与人类操作员眨眼图像相似。我们的完整管道还通过在数据集中植入人工对象来表征整个系统的巨大效率。包的检测部分是公开的。
With the deployment of large CCD mosaic cameras and their use in large-scale surveys to discover Solar system objects, there is a need for fast detection algorithms that can handle large data loads in a nearly automatic way. We present here an algorithm that we have developed. Our approach, by using two independent detection algorithms and combining the results, maintains high efficiency while producing low false-detection rates. These properties are crucial in order to reduce the operator time associated with searching these huge data sets. We have used this algorithm on two different mosaic data sets obtained using the CFH12K camera at the Canada–France–Hawaii Telescope (CFHT). Comparing the detection efficiency and false-detection rate of each individual algorithm with the combination of both, we show that our approach decreases the false detection rate by a factor of a few hundred to a thousand, while decreasing the ‘limiting magnitude’ (where the detection rate drops to 50 per cent) by only 0.1–0.3 mag. The limiting magnitude is similar to that of a human operator blinking the images. Our full pipeline also characterizes the magnitude efficiency of the entire system by implanting artificial objects in the data set. The detection portion of the package is publicly available.
影响因子：
3.2
作者：
V. Mannings;A. Boss;S. Russell
通讯作者：
V. Mannings;A. Boss;S. Russell