mm-Wheat: Noncontact Detection of Wheat Moisture Content With Millimeter-Wave Radar

Wireless sensing technology has great potential in detecting grain moisture content. The existing work has proven that noncontact grain moisture detection can be achieved using Wi-Fi and RFID. However, due to the inflexible deployment of Wi-Fi and RFID devices, the usage scenarios of existing solutions will be limited. In this article, we propose mm-Wheat, which is the first system to use lightweight commercial millimeter-wave (mmWave) radar equipment for fine-grained wheat moisture content detection. First, we validate the feasibility of using mmWave radar to detect wheat moisture using the spectrum information and extract the peak zone of the spectrum as the reflection information of the wheat sample. Then, mm-Wheat uses the range-angle map (RAM) to obtain spatial information related to the wheat sample, including angle, distance, and thickness. Next, we design a new dedicated network, termed MFAF-Net, which focuses on learning mmWave radar reflection information obtained from wheat samples. This network not only can capture the fine-grained moisture content characteristics of wheat from the mmWave spectrum but also integrates spatial information to eliminate the interference of wheat container position and changes in wheat pile thickness during the detection process, improving the robustness of the system. We implement a prototype system and evaluate the performance of the mm-Wheat system in a real environment. The experimental results show that the mm-Wheat system has an average detection accuracy of 97.3% for six wheat samples with different moisture contents and has a resolution of 1% for fine-grained moisture content and good robustness.