基于时空预处理DS证据的同质传感器数据融合

Homogeneous Sensor Data Fusion Based on Spatio-temporal Preprocessing DS Evidence

针对多个传感器数据融合时,融合结果受异常和噪声影响,并且容易丢失局部环境特征的问题,提出一种基于时空预处理的DS证据方法。首先,通过设计的空间和时间一致性指标剔除可疑异常数据,并采用区域熵捕获特征数据;然后,根据特征位置和传感器空间关系定义约束条件,计算数据可信度;最后,以可信度为权值组合所有数据。以综合管廊内的甲烷浓度传感器作为数据源进行实验,数据融合结果能够准确反映甲烷浓度正常和泄漏情况,融合算法具有噪声过滤和特征保持性能。

In view of the problem that the fusion results of multiple sensors are affected by anomalies and noises and are prone to lose local environmental characteristics,a DS evidence method based on spatio-temporal preprocessing was proposed.Firstly,the designed spatial and temporal consistency indexes were used to eliminate the suspicious abnormal data,and the regional entropy was used to capture the characteristic data.Then,according to the feature location and spatial relationship of sensor,the data reliability was calculated.Finally,all data was combined with a weight of credibility.The methane concentration sensor in the integrated pipe gallery was used as the data source for the experiment.The data source can accurately reflect the methane concentration and leakage,and the fusion algorithm has the performance of noise filtering and feature retention.