Contactless Tracking And Material Sensing Of Mobile Objects Using COTS RFID Devices

In recent years,Radio Frequency(RF)signal sensing has become a hot topic.From coarse-grained motion tracking to fine-grained breathing sensing,many related tech-nologies have been born.The main advantage of RF signal sensing is that it does not require the target to carry any sensor,which is much more convenient than traditional wearable methods.On the other hand,wireless communication has become the main means of communication for terminal equipment now.If ubiquitous wireless signals can replace specialized sensor equipment,the structure of the wireless sensor network will be more simplified.In order to realize the intelligent sensing scheme based on RF signals,people have explored various schemes using different protocols,including WiFi,Bluetooth,Zigbee,Radio Frequency IDentification(RFID)signals,etc.Among them,RFID-based solu-tions have been widely used as an identification technology.In addition to its low price,RFID has the advantages of concise protocol,moderate signal propagation distance,and convenient target marking[1].Therefore,it is used for many sensing applications such as trajectory tracking,posture recognition,etc.However,most of existing RFID-based sensing systems are based on contact,which means RFID tags need to be attached to the sensing targets.In the field of non-contact sensing,RFID technology still faces severe challenges such as low accuracy,instability in mobile conditions and poor real-time performance,which limits its application in practical application.In response to these problems,in this dissertation,we propose a contactless RFID-based sensing system,which including desktop-level localization systems for static tar-gets.Real-time target tracking system,material and motion sensing system for non-contact moving targets.All the above systems use commercial RFID equipment with-out any hardware modification.After the system is implemented,our experimenters use electric slide rails to simulate the mechanical movement environment and move ob-jects by hands to simulate the real movement environment for experimental evaluation.Comprehensive experiments show that the localization error of the static target is less than 1cm,and the tracking accuracy of the dynamic target is as high as 1.3cm,which is 200%higher than the current trajectory tracking technology.For the four common material types,the accuracy of material identification is higher than 95%,even if it is interfered by surrounding people.In the action recognition subsystem,the experimenter simulates the shopping behavior in the supermarket.The system can recognize the four actions of picking up,putting back,putting in the shopping cart and putting it from the shopping cart,and its recognition accuracy is all above 80%.Our main contributions are as follows:1)We propose a contactless sensing system based on commercial RFID devices,which can track the localization of a target without any sensor in real time,identify the material of a moving target,and perform contactless movement recognition.2)We use the RFID signal reflection model and propose a difference-based method to distinguish moving targets from other nearby static objects and eliminate errors caused by hardware noise.3)We designed a unique feature extraction scheme based on the signal reflection model of the moving target in the material recognition module.In addition,tag arrays are used to replace expensive wireless signal transceivers,and the information collected from multiple tags is fully combined to improve tracking and recognition accuracy and enhance the robustness of the system.