Abstract:
The efficiency of the system could be greatly improved by using the Radio Frequency Identification (RFID) batch tag identification in the quality traceability of agricultural products. In this study, a tracing system of agricultural products based on RFID technology was designed. This system consisted of a tracking subsystem that was from the planting side to the consumption side and a tracking subsystem that was from the consumption side to the planting side. This system designed the information processing process of agricultural product planting, storage information processing, transportation information processing, distribution and distribution information processing, and terminal consumption information processing. This system also designed a three-level Internet of things architecture. Compared with the traditional agricultural product traceability system based on bar code, the one based on RFID technology was more flexible and accurate and had a better application prospect. In the agricultural product traceability system, especially in the courses of delivery and storage, batch identification of an entire truck of product tags needed to be conducted inbound and outbound. For frequently used forklift transport, entire tags attached to products in a forklift also needed to be identified simultaneously so that automatic rapid processing of transport, loading, and allocation of warehouse location could be accomplished. Batch tag identification was the most obvious advantage of RFID, but it would produce tag signal interference with each other and result in tag collision. Therefore, the anti-collision method was the key technology to the agricultural product traceability system. In this study, a model of agricultural product traceability system based on radio frequency identification technology was established, and the implementation of the storage link was presented. In the process of traceability, batch identification of an entire truck or forklift of the agricultural product greatly improved the management efficiency and shortened the circulation time of the product. However, the tag collision problem was inevitable due to the non-cooperation mechanism among tags. Therefore, the tag anti-collision method was a key technology of the agricultural product traceability system. A tag anti-collision method was excogitated and a frame slotted Aloha tag anti-collision method that adaptively and dynamically adjusted frame length was proposed aiming at traceability application environments of agricultural products. In this method, the initial tag population size was estimated when the number of successful slots, idle slots, and collision slots was acquired using statistical calculation only after the first frame identification so that the length of the following frame could be optimized and the optimal identification could be actualized, significantly improving the overall throughput performance. Based on the traceability system designed and the tag anti-collision method proposed, this study used Monte Carlo simulation to compare the throughput performance of the proposed method with the Vogt method and Q algorithm. The global throughput was an important index to measure the performance of RFID batch identification, and it was defined as the ratio of the time occupied by the transmission information to the total time consumed by the whole identification cycle. Taking the tracking and traceability of tea as an example, the throughput performance and delay performance of the anti-collision method for batch identification tags were simulated. Besides, the effect of initial frame length on throughput performance was simulated. The simulation results of RFID batch tag identification in agricultural product quality traceability showed that the global throughput was between 20% and 32%. It also showed that the proposed anti-collision algorithm applied to the designed agricultural product traceability system significantly improved the identification efficiency, and the overall throughput performance was improved by more than 30% compared with the Q algorithm.