| With the development of information technology,many scholars have applied WIFI wireless LAN technology in the field of personnel intrusion detection.The outstanding advantage of this technology is that the inspected personnel do not need to carry wireless transceiver equipment,only rely on the existing indoor WIFI environment,can realize the detection and identification of personnel intrusion activities.Compared with Received Signal Strength Indication(RSSI),Channel Status Information(CSI)can depict the characteristics of the channel environment in the process of multi-path propagation of wireless signals more finely,thus achieving better performance.However,the existing CSI-based intrusion detection methods have the following problems:on the one hand,in the complex indoor environment with occlusion,its detection performance has declined significantly;on the other hand,previous research has remained at the level of discretization of collected data,and no solution of real-time detection system with higher practical value has been proposed.In view of the above two challenges,this thesis first investigates the current main intrusion detection technology.An intrusion detection method based on CSI is improved by combining the passive intrusion scenario and the processing scheme of CSI.Based on this method,a real-time intrusion detection system in indoor environment is designed and implemented.The main work and innovation of this thesis are as follows:Referring to the existing passive personnel intrusion detection technology based on CSI,intrusion detection is divided into two main processes:data processing and detection.Referring to the current processing scheme for CSI,the original processing method for CSI is improved.Principal Component Analysis(PCA)is introduced to optimize the feature data extraction method in intrusion detection process.Standard deviation and first-order absolute difference mean of time vector after dimensionality reduction are proposed as the feature parameters to measure the fluctuation of data.Support Vector Machine(SVM)is used to complete classification and discrimination.Furthermore,combined with the multi-link voting strategy to improve the overall detection effect of the method,and ultimately achieve the detection and discrimination of personnel intrusion behavior.The experimental results show that in the complex indoor scene with occlusion,the average accuracy of the detection method is improved by about 2%,up to 96%compared with the existing methods.A real-time intrusion detection system based on CSI in indoor environment is designed and implemented.The system adopts modular design,which divides the whole system into real-time data acquisition module,data processing module,feature extraction and classification discrimination module,and interactive display module.This paper focuses on the overall operation process of the system and the technical implementation details of each module.The detection results in typical indoor scenarios show that the detection system can guarantee more than 80%detection accuracy for intrusions with different moving rates(the numerical range is about 0.5-2 m/s)when the transceiver is 5 m apart.When the distance between the transceiver and the receiver is in the range of 2m-7m,the detection accuracy of the detection system can reach more than 85%for intrusions occurring at normal walking speed(about 1-1.5m/s),which meets the expected results. |
