Research On Adaptive Weighting Algorithm Based On Internet Of Things

Information fusion is an attribute characteristic of biology,which belongs to the root of animals and plants to perceive the changes of surrounding environment and behavior,and is a common phenomenon in ecosystem.The human eyes,ears,nose,skin and limbs are used as sensors to predict and judge facts that are or are about to happen using information currently collected on environmental conditions,sounds,odors,and prior information available.Because of the environmental noise and equipment failure,the information obtained by a single sensor is limited and incomplete,which leads to the unreliable and unreliable sensor data collected.Therefore,in the automation industry,data fusion should coordinate the use of multi-sensors to collect target information,expand the scope of single sensor data collection,so as to carry out multi-level,all-directional,multi-angle information processing,parameter fusion estimation to obtain target status and feature prediction of a multi-level information automatic processing.Based on the theory of knowledge and the practical problems of application system in the research of multi-sensor information fusion,this paper explores deeply.Firstly,the concept of multi-sensor information fusion is defined and the related theory and technology are summarized.The application algorithm of multi-sensor information fusion is introduced in detail,and the data fusion level is divided into three layers,namely,data layer,feature layer and decision layer.The adaptive weighting algorithm on the data layer is studied.For the problems of noise,low precision and poor anti-interference,it is proposed to eliminate the existence of abnormal data,use the redundancy of data,increase confidence,find the super data with the highest trust,fill the excluded abnormal data and reduce ambiguity.The focus of the paper is as follows:Due to environmental interference,system noise,equipment failure,when the sensor carries out the original data sample collection,there will be data error.By using the distribution map method,the data samples are arranged in order,the results of analysis and processing are compared,the suspicious data or abnormal data are excluded,and the credible data set is formed;Using confidence matrix and relation matrix to calculate the best data set and the highest supported sensor data,that is,the generation of hyperdata;finally,using hyperdata to replace suspicious data or abnormal data,in the state with the smallest total variance error,the target parameter estimation of each sensor allocation weight evaluation factor is carried out.The optimized fusion results can not only reduce the amount of redundant data,but also improve the accuracy and robustness of parameter estimation.