中文摘要：

RFID数据采集的不可靠性降低了RFID应用中数据的准确性，并进一步对复合事件的检测产生影响。目前以RFID读数为粒度的清洗方法只能在一定程度上降低原始采集错误的发生频度，而复合事件检测过程又很少对其进行处理。为解决上述问题，将RFID数据从数据层抽象到逻辑语义层作为处理的粒度，提出了复合事件相互之间的约束规则，进行误检处理。通过挖掘已知复合事件之间的相关性对后续发生的事件进行误检判断，考虑了具体应用的逻辑语义，保证了RFID数据的可靠性。

英文摘要：

The unreliability of RFID data collection reduces the accuracy of data in RFID applications, further impacts on composite event detection. The existing algorithms which take primitive RFID readings as granularity can only reduce the occurrence frequency of collection mistake, and composite event detection hardly deal with those mistakes. In this paper, data are transformed from data level to logic area level as the interpolation granularity, the proposed constraint rules between composite events, and processing false detection. The reliability of the RFID data can be ensured by mining the correlation between the known composite events, judging the false detection of follow-up events, and considering the logical semantics of the specific application.