小空间密集环境下基于深度学习的RFID信号补全与抗干扰方法

RFID Signal Completion and Anti-Interference Method Based on Deep Learning in Small Space Dense Environment

摘要: RFID技术作为智慧仓储的核心技术之一，凭借其非接触式识别、读取范围广、存储信息量大等优势，在物品追踪和管理方面发挥着重要作用。然而，在小空间、多类别密集存储的环境中，RFID面临信号碰撞、错读漏读、环境适应性不足等挑战。针对上述问题，提出小空间密集环境下基于深度学习的RFID信号补全与抗干扰方法。首先，针对小空间密集环境下RFID标签密集分布、信号重叠问题，在信号读取时采用多维度信号获取方法，优化标签附着介质，根据环境变化实时进行射频优化；其次，应用强耦合松弛化分解技术和盲源分离算法对读取到的RFID信号进行建模和分离，提高信号质量和可读性；最后，对于残缺数据，基于MDO多目标优化和深度学习的多尺度联动数据融合补全方法，训练卷积神经网络学习数据特点，实现不完整RFID信号的数据补全。实验结果表明，所提方法能够有效解决小空间密集存储环境下RFID信号碰撞导致的数据漏读错读问题，将识别准确率提升至99%，显著提高了RFID定位感应的精确度和可靠性。

Abstract: RFID technology, as one of the core technologies of smart warehousing, plays an important role in item tracking and management with its advantages of non-contact identification, wide reading range, and large amount of stored information. However, in environments with small space and dense storage of multiple categories, RFID faces challenges such as signal collision, misreading, and insufficient environmental adaptability. To address the above issues, a deep learning based RFID signal completion and anti-interference method is proposed for small space dense environments. Firstly, to address the issues of dense distribution and signal overlap of RFID tags in small space dense environments, a multidimensional signal acquisition method is adopted during signal reading to optimize the tag attachment medium and perform real-time RF optimization based on environmental changes; secondly, strong coupling relaxation decomposition techniques and blind source separation algorithms are applied to model and separate the read RFID signals, improving signal quality and readability; finally, for incomplete data, a multi-scale linkage data fusion completion method based on MDO multi-objective optimization and deep learning is used to train convolutional neural networks to learn data characteristics and achieve data completion of incomplete RFID signals. The experimental results show that the proposed method can effectively solve the problem of data leakage and misreading caused by RFID signal collision in small space dense storage environments, improve the recognition accuracy to 99%, and significantly improve the accuracy and reliability of RFID positioning and sensing.