State Perception Method for UHF RFID Tag Groups in Conveyor Belt Scenarios

In the application of Radio Frequency Identification (RFID) technology, the density and distribution state of tag groups are critical factors affecting identification performance. This paper proposes a tag group state perception method for conveyor belt moving scenarios. First, an RFID tag group density perception algorithm based on ReliefF feature selection and the K-means clustering algorithm is introduced. This algorithm selects important features using ReliefF and classifies tag densities using the K-means algorithm. Grid search is employed to optimize the K-means clustering parameters, enhancing the model's performance.Second, a tag group distribution perception algorithm based on the XGBoost classifier is presented. This algorithm classifies tag group distributions using the XGBoost classification algorithm, with hyperparameters fine-tuned through grid search and three-fold cross-validation to ensure robustness and high stability.Finally, the outputs of these two algorithms are integrated to determine the overall state of tag groups. A test scenario is constructed to validate the proposed algorithms. Results indicate that the RFID tag group density perception algorithm based on ReliefF and K-means achieves a purity of 96.67%, the tag group distribution perception algorithm based on XGBoost achieves an accuracy of 96.94%, and the overall RFID tag group state perception method achieves an accuracy of 93.89%. These findings demonstrate the effectiveness of the proposed method in conveyor belt scenarios, providing valuable support for optimizing RFID system reading strategies.