项目以无线传感器网络在土遗址健康监测中的应用为背景，针对土遗址WSN监测数据的质量问题，以及土遗址WSN监测数据异常模式识别中的个体差异和病害差异带来的挑战，致力于研究并提出面向土遗址预防性保护的异常模式识别方法，探索土遗址赋存环境和生存状态之间的科学规律，以期为预防性保护提供科学决策依据。项目通过引入指数平滑预测模型和投票策略，对土遗址WSN时序监测数据进行有效清洗，为后续的异常模式识别奠定了基础；通过引入危险理论，借助危险信号浓度的量化算法，对土遗址健康度进行科学量化，并以此为依据设置特征权重，构造基于健康度—支持向量机的异常模式识别方法，有效解决面向个体的土遗址WSN监测数据异常模式识别问题。借助于证据权重（WoE）理论，有效描述土遗址多因素耦合作用，并以此为依据设计权重计算公式，构造基于WoE—支持向量机的异常模式识别方法，有效解决面向病害的土遗址异常模式识别问题。

On the background of application of health monitoring for earthen sites with wireless sensor network（WSN），the project tries to seek appropriate and efficient abnormal pattern recognition method, aims at exploring scientific laws between the environment and survival state，which is expected to explore decision-making basis for preventive protection of earthen sites. In allusion to the challenge of individual differences and diseases differences, and the data quality issues, multiple, and unbalance time series monitoring data of earthen sites, the project is trying to introduce exponential smoothing forecasting model and voting strategy to clean the monitoring data to lay foundations for the subsequent abnormal pattern recognition. By using danger theory and the danger signal quantization algorithm, based on the health degree of the earthen site, feature weights can be set. Then an abnormal pattern recognition method is designed for the individual differences oriented abnormal pattern recognition based on WSN monitoring data. On the basis of rate of health risk deterioration to describe the coupled action of multiple factors, weights can be set, Then an abnormal pattern recognition method is designed for the diseases differences oriented abnormal pattern recognition based on WSN monitoring data.