Research On Intelligent Sensing Of Humans And Commodities In Warehouse Based On Passive RFID

In recent years,the Internet of Things(IOT)has become an important part of new generation of information technology.One of the typical applications of the Internet of Things is the application in intelligent warehouse systems.The management of warehousing is mainly about automatic management of commodity and staff in warehouse.Radio frequency identification(RFID)technology play a crucial role in both types of the management.RFID is a kind of non-contact automatic identification technology based on electronic coupling or backscattering.Now,RFID has been widely used in remote automatic inventory of goods,product cardinality estimation,personnel access control,positioning & tracking,motion recognition and humancomputer interaction and so on.At present,although the research and application of RFID around the smart warehouse environment have been extensively carried out,There are still some key problems that have not yet been well resolved.On the one hand,as the Deploying of RFID systems in warehouse on a larger scale,it is urgent for the readers to communication with a large number of tags within a limited period of time to collect specific attributes of the commodities,which can supply service for upper applications in real time.On the other hand,with the trend of combining RFID wireless communication technology and sensing technology,the applications of sensing human actions with RFID sensing tag draws broad attention.However,most researches on action awareness based on RFID focus RSSI or phase information from tags,which are susceptible to multipath propagation and ambient electromagnetic noise.Less work try to study high-precision human motion recognition by collecting sensor readings with passive RFID sensing tags worn by the personnel.In order to solve the above problems,we make a research on efficient data acquisition and accurate motion recognition technology in RFID systems.On the one hand,we study how to improve the information exchange between readers and tags in the RFID system,sensing the quantity of goods without collecting tag IDs,which can accelerate upper applications;on the other hand,we study skills to collect RFID sensor readings efficiently and make full use of information about personnel actions,which will provide technical Support machine interaction applications.The specific contributions are summarized as follows:(1)In view of the quantity perception of warehousing commodities,we study the technology for fluxility estimation of large-scale RFID tags: the reader scans tags attached to commodities at different times or places to obtain two unequal frames.Then we estimate the fluxility of commodoties through joint analysis.Since each time slot in the frame has two states,empty and busy(busy can be further divided into singleton and collision),we first make a study on fluxility estimation based on partial slots including empty slots and busy slot.Then we further analyse signal collision of the time slot in the frame and study the fluxility estimation based on all slots to improve estimation precision.(2)As to sensing on humans in warehouse,we study gesture recognition based on RFID sensing tags.We study a three-level filtering scheme for removing noise;Then we study skills of border check to eliminate redundant borders;For the sparse data,we study method of increasing sampling rate by multi-sampling and one transmission scheme,which further improve the quality of data;After extracting features,different machine learning classifiers are applied to train and recognize various hand gestures.(3)After theoretical analysis,we verify the fluxility estimation and hand gesture recognition by experiment.For the fluxility estimation We compare our solution with existing method respectively on accuracy with load-factor,the ratio of two tag sets,the proportion of common tags,the protocol time and running time.As to hand gesture recognition,we define multiple macro and micro evaluations for single gestures and the overall recognition by different classifiers;Then We test the accuracy after increasing sampling rate and the accuracy after extending gestures.In summary,we study two kinds of intelligent sensing schemes for humans and commodities management in warehouse.As for sensing to commodities,we study a fluxility estimation,which will further improve the management of warehouse.As for sensing to humans,we study hand gesture recognition solution based on passive RFID sensing tag which will promote the development of smart warehouse.