生鲜肉在存储过程中不可避免地会发生品质劣变，其中氨气是重要的劣变挥发物。然而，在冷链存储环境中温湿度的波动会影响氨气检测精度。为克服这一难题，本研究提出了一种基于温湿度的反向传播神经网络补偿模型(based on temperature-humidity back propagation neural network compensation model，THBP)，实现微波传感器氨气检测过程中温湿度补偿。首先，通过分析传感器辐射增益与氨气浓度之间的相关性，并结合无线功率传输模型，构建了无线微波氨气传感模型。其次，通过实时检测生鲜肉劣变过程中释放的氨气，对该传感系统进行了验证。试验结果表明，在低温高湿的冷藏环境下，结合皮尔逊相关性分析的THBP可显著抑制环境温湿度波动对传感性能的影响。经补偿后，射频信号零点频率漂移减少了14 MHz，误差浓度控制在0.06 ppm以内，检测精度提高了31.11%。本研究为动态冷链环境中实现生鲜肉品质劣变的无损检测提供了可靠的理论框架和实践方法。

The quality of fresh meat inevitably deteriorates during refrigerated storage, and ammonia is a critical volatile marker of spoilage. Nevertheless, temperature and humidity fluctuations within the cold chain environment can decrease the reliability of ammonia detection. To overcome this limitation and increase the sensing precision, in this work, a microwave ammonia sensor with temperature and humidity compensation is proposed on the basis of a backpropagation (BP) neural network. By analyzing the correlation between the radiation gain of the sensor and the ammonia concentration and integrating a wireless power transmission model, a new wireless microwave ammonia sensing model was established. The sensing system was experimentally validated through real-time monitoring of ammonia released during the spoilage of refrigerated meat. The results indicate that BP neural network-based temperature and humidity (THBP) compensation with Pearson correlation analysis reduced the radio frequency signal zero-point frequency fluctuation by 14 MHz, limited the absolute error to 0.06 parts per million (ppm), and increased the detection accuracy by 31.11%. This work provides a reliable theoretical framework and practical approach for high-sensitivity, non-destructive monitoring of the quality of fresh meat in dynamic cold-chain.