Research On Pork Freshness Detection Method Based On WiFi Channel State Information

Meat freshness testing is an important measure to ensure food safety,improve food quality and safeguard the lawful consumer interests.Traditional methods for meat freshness testing are time-consuming,inefficient and destructive to the meat itself,while modern methods for meat freshness testing have limitations such as high cost and inconvenient deployment.Therefore,in order to meet the needs of different market inspection scenarios,this paper proposes a pork freshness detection method based on WiFi channel state information to meet the needs of different market inspection scenarios,and proposes an in-depth study on data pre-processing and efficient modelling,and constructs a pork freshness detection model based on broad learning system and deep learning to achieve fast and accurate detection of pork freshness.The main research contents are as follows:(1)In this paper,a CSI data collection platform is built using commercial WiFi equipment to obtain CSI data of pork at different freshness levels.To address the problems of prominent outliers and excessive high frequency noise in the raw data,Hampel filtration,Chebyshev II filter,subcarrier selection and z-score were used for data pre-processing to construct a high-quality CSI feature dataset.The CSI amplitude and phase data extracted from the dataset were also correlated with the freshness of pork to provide a data base for subsequent experiments.(2)In this paper,the freshness classification of pork was carried out.Pork was tested for volatile salt nitrogen using a semi-trace nitrogen determination method at 4°C under refrigeration,and was classified as fresh,sub-fresh and spoiled based on the volatile salt nitrogen content in accordance with the relevant national standards,providing a basis for the setting of freshness identification labels.(3)In order to further improve the speed and accuracy of pork freshness recognition,this paper designs a fast pork freshness detection system based on broad learning system to address the problems that traditional machine learning networks usually have such as insufficient generalization ability and difficulty in handling complex data.The experimental results show that the system has an average detection accuracy of over 90% for the freshness of a single part of pork,and can achieve faster online detection to meet the real-time detection requirements.(4)Aiming at the defect that broad learning system cannot be applied to pork freshness detection of multi-parts,this paper designs an accurate detection system of pork freshness based on CNN-Bi LSTM-AM.Feature extraction of CSI data was performed by convolutional neural network and bidirectional long and short-term memory network,while attention factor was introduced to focus on freshness features of different parts to achieve accurate detection of multi-part pork freshness.The experimental results,show that the system can effectively identify the freshness features of different parts,and the average detection accuracy of multi-part pork freshness is up to 97%.The major point of innovation in this article is the first use of WiFi channel state information technology to build a meat freshness detection device.The application of this method is promising as it is fast,accurate,easy to deploy and non-destructive.