基于小麦碰撞声音的霉变区域识别仿真分析

Mildew Region Recognition Simulation Analysis Based on the Wheat Collision Sound

研究小麦霉变区域的识别优化问题,由于小麦霉变粒、受损颗粒与完好粒的物理构造不同,碰撞声信号与正常小麦信号一定存在差异。差异特征各项属性很容易受到外部噪声的干扰发生改变,造成传统的模式识别方法很难准确识别。提出了一种加入干扰分类去除因子的小麦碰撞差异特征霉变识别方法。在小麦碰撞过程中,确定分类决策因子,并且将小麦碰撞声音的残差平方和SSE引入核函数,设计一个小麦碰撞约束超平面,确保碰撞中的系数特征点中的正类输入以及负类输入,分别处于该约束超平面的两端,实现对霉变小麦碰撞特征的干扰消除。实验表明,改进算法实现了小麦霉变区域的特征识别的优化。

Research the recognition optimization of the wheat mildew region.The paper proposed a mildew identification method which is based on difference characteristics of wheat collision and removes interference factors.The classification decision factors were determined in the process of wheat collision,the sum of squared residuals SSE of the wheat collision sound was introduced into a kernel function,and a wheat collision constraint hyperplane was designed to realize the interference elimination of mildew wheat collision characteristics.The simulation results show that the algorithm realizes the optimization of wheat feature recognition.