Research On Multi-sensor Data Fusion Algorithm For Environmental Monitoring In Metrology Laboratory

Measurement and calibration are the technical basis of modern social and economic activities,national defense construction,social development and scientific research,as well as an important legal means to maintain trade fairness,promote market development and protect the rights and interests of consumers.However,the accuracy of measurement testing,calibration and standard traceability is seriously affected by the environment of metrology laboratories.Therefore,the environment monitoring of metrology laboratory plays a pivot role of ensuring the accuracy and reliability of measuring experiments,and has become an essential link in the construction of metrology laboratory.Traditional environment monitoring of measurement laboratory based on singletyped sensors is difficult to perform a comprehensive and reliable monitoring of the environment,especially for large-scale and precision measurement and detection environment.On the basis of in-depth discussion for metrology laboratory of environmental monitoring multiple source sensor data fusion architecture and hierarchical multi-sensor data fusion method,this paper provides a reliable laboratory environment for measurement experiments.The main contributions of this thesis are summarized as follows:(1)A two-level fusion-based multi-sensor data fusion framework is established.Since the data form,sampling rate,data volume and target concerns of sensory data from heterogeneous multi-source sensors in measuring lab environment monitoring system have great difference with high redundancy,a kind of two-stage fusion architecture i.e.,the data level and decision level,is introduced for the multi-source heterogeneous sensory data fusion.It can effectively reduce the difficulty of the design and implementation of data fusion algorithms and improve the fusion performance of multi-source heterogeneous sensory data.(2)An adaptive integrated over-sampling method for the unbalanced data processing based on a variation-optimized Gaussian mixture model is proposed.Aimed at solving the problem of imbalanced distribution of data samples with respect to different environmental states,a kind of variation Bayesian Gaussian mixture model is established for the adaptive oversampling of environmental samples,since the Gaussian mixture model has the merit of fitting any unknown distribution.The proposed method can achieve a relatively balanced environmental monitoring data sets,for the purpose of achieving unbiased state classifier.Experimental results show that the proposed method can effectively improve the classification accuracy of traditional classifiers for the unbalanced data set processing.(3)A data-level fusion algorithm for multi-source heterogeneous sensors based on an improved BP neural network is proposed.To solve the problems of information redundancy and detection error in the data of sensors of the same type,a BP neural network model for the data fusion of sensors of the same type was designed,and a BP weight adjustment strategy based on Grey Wolf algorithm was proposed to adaptively adjust the weight of the neural network model.Simulation results show that the proposed method greatly improves the performance of BP neural network,and the data level fusion effect is improved.(4)A D-S evidence theory-based decision level fusion algorithm is proposed to achieve a comprehensive evaluation result of the experimental environment of metrology laboratory.Since the measuring environment requires a variety of monitoring data and it is difficult to achieve a comprehensive evaluation result of environmental state,the D-S evidence theory-based multi-sensor data fusion algorithm is introduced based on the data level fusion result of BP neural networks for various kinds of sensors,to achieve a comprehensive assessment of environmental condition.Experimental results demonstrated that the proposed method can effectively improve the credibility of the lab environment state comprehensive evaluation.Finally,based on the actual requirements of metrology laboratory,the corresponding experimental simulation system is developed,and the algorithm effect of the proposed data fusion algorithm of the two-level fusion architecture is evaluated.Simulation results show that the proposed two-level fusion architecture can effectively improve the data fusion efficiency of the measurement laboratory monitoring system,providing technical support for accurate laboratory measurement,value transfer,labeling and traceability.