A Maize Kernel Loss Monitoring System for Combine Harvesters Based on Band-Optimized Discrete Wavelet Transform

Precise distinguishing of maize blends and the evaluation of kernel losses enhances the accurate measurement of harvest loss. To address the low accuracy and poor anti-interference ability of traditional maize kernel detection methods under complex conditions, this paper proposes a multi-channel kernel impact detection algorithm based on discrete wavelet transform (DWT). The algorithm extracts feature band energies of kernel impacts through DWT multi-resolution analysis and counts kernels based on the duration of the energy signal. Therefore, weak signals are able to be effectively detected, thus correcting the missed errors that traditional monitoring systems produce for weak kernel signals. The monitoring system’s efficacy was assessed across various operational conditions. Test findings reveal that within the operating ranges of kernel flow rate of 20–40 kernels/s, sensor mounting angle of 30–60°, and mounting height of 300–500 mm, the system’s average detection accuracy reaches 94.4% and maintains good stability under different conditions. Compared with traditional detection systems, the system designed in this research exhibits superior sensitivity to weak kernel signals and higher monitoring accuracy. Finally, it was verified via practical field experiments that the designed sensor basically achieved the expected performance, and the recognition accuracy of the kernels in the mixture reaches 94%.