WiPd: Contactless Water-Injected Pork Detection Using Commodity WiFi Devices

Detection of water-injected pork is of great significance for pork safety and food quality monitoring. In this paper, we propose a nondestructive wireless sensing system, termed WiPd, for rapid detection of water-injected pork using commercial WiFi devices. We first verify the feasibility of water-injected pork detection using WiFi channel state information (CSI). We then design the WiPd system consisting of a sensing module, a preprocessing module, and a detection model module. In the sensing module, commercial WiFi devices are used to collect different types of CSI data from normal pork and water-injected pork. In the data preprocessing module, the CSI ratio model is used to eliminate the environment and hardware noise, and then the subcarrier selection and normalization methods are carried out for feature extraction. In the detection module, a double-layer long short-term memory (LSTM) network is designed to detect water-injected pork with an online detection method. Finally, we evaluate the proposed WiPd system with extensive experiments, and the results show that WiPd can achieve an average detection accuracy of more than 98% in line-of-sight (LOS) scenarios.