Research On RFID-based Theft Detection Technique

In recent years,the total loss of large-scale shopping malls and warehouses due to theft has reached up to 100 billion dollars.And the existing theft detection systems have some limitations.The traditional video surveillance method is greatly affected by the light condition.And the current RFID-based theft detection systems cannot effectively handle some false alarm cases caused by non-human factors.For example,the detection range of the RFID system is not clear,so the RFID system could identify a stationary tag in the shopping malls.The existing RFID-based theft detection system did not consider this situation.In addition,merchants have different degrees of interest in different value products.However,there is no effective RFID system that can provide different reading sampling probabilities for products with different values.This paper jointly leverages the RFID and Kinect computer vision technique to propose a multi-source data fusion monitoring system.Specifically,the RFID system is used to collect the phase time series of the product tag.At the same time,the Kinect system is used to collect the position time series of the joint position of the person.The conversion formula convertes position time serise into the phase time series of the joint point of the human.In the case of error-free alarm and false alarm,the dynamic time warping(DTW)algorithm and the hypothesis testing model are used to complete the matching tracking of the tag trajectory.At the same time,for tags with different values,this paper proposes the multi-array tag sampling(MTS)method to allow the reader to select the tags with different sampling probabilities.And the sampling probability is determined by the binary strings embedded in the tag memory.In summary,the RFID-based theft detection system is combined with the Kinect system to realize tag tracking and differential probability sampling.The multi-array tag sampling method is applied to realize different sampling probabilities for commodities with different values.The theoretical correctness,feasibility analysis,the experimental correctness,and feasibility analysis are proved at the end of the paper.