虫害损害大量储粮，长期食用虫蛀粒小麦会造成营养不良，甚至诱发疾病，因此对受损粒小麦的检测工作刻不容缓。本文提出一种新的基于碰撞声的方法结合极限学习机与蜻蜓优化算法检测小麦虫蛀粒与发芽粒。从时域频域两方面提取信号特征，包括：时域短时窗口最大幅值与方差、峭度、3阶Rényi熵、功率谱均方根。随后采用极限学习机进行分类，并用蜻蜓算法优化相应参数。实验结果显示，93%的完好粒、95%的虫蛀粒及87%的发芽粒被正确识别，表明了本文所提出算法对受损粒小麦检测的有效性。%Insects will destroy significant amounts of stored grain, and long-term feeding on damaged wheat kernels will result in malnutrition, even induce diseases, therefore, the work of detection of damaged wheat kernels is of great urgency. In this paper, a novel method based on impact acoustics combining extreme learning machine (ELM) with dragonfly algorithm was proposed for detection of insect-damaged wheat kernels (IDK) and sprout-damaged ones. Discriminant features, including the maximum amplitudes and variances in time-domain short-time windows, kurtosis, the third-order Rényi entropies and the mean square roots of power spectrum, were extracted both from the time-domain and frequency-domain. Subsequently, ELM was used for classification with dragonfly algorithm for parameter optimization. The experiment results demonstrated that 93.0%of undamaged wheat kernels, 95.0% of insect-damaged wheat kernels and 87.0% of sprout-damaged ones were correctly detected, which indicated the effectiveness of the proposed algorithm for detection of damaged wheat kernels.
展开▼
