基于NMF的罐装食品真空度声学检测方法

Acoustic Detection Method for Vacuum Degree of Canned Foods Based on NMF

目前频谱峰值法被广泛用于罐装食品真空度检测,其是将罐盖振动所产生声音的谱峰频率是否在合适范围作为检测依据。而实际应用中出现了部分不合格类型产品的谱峰频率也处于设定的合适范围,导致该方法做出误判。为此,将罐装食品真空度检测归结为单分类问题,提出基于非负矩阵分解的单分类算法。该算法仅需从合格品的数据学习得到分类器,用于判断被检品真空度是否合格。设计一个罐装食品真空度检测实验系统,将本文方法、传统的频谱峰值法以及典型的单分类方法进行对比实验。

At present the spectral peak method is widely used to detect the vacuum degree of canned foods. Whether the spectral peak frequency of the sound produced by the vibration of the can lid is in the appropriate range is taken as the detection basis of this method. However, the spectral peak frequency of some unqualified products is also within the set appropriate range appeared in practical application, which leads to the misjudgment of this method. In order to improve the detection performance, this paper firstly proposes to treat the vacuum degree detection of canned food as a one-class classification problem, and presents a one-class classification algorithm based on non-negative matrix factorization. The classifier only needs to be learned from the dataset of qualified products in this algorithm, and then it can be used to judge whether the vacuum of the detected product is qualified or not. We constructed an experimental system for the vacuum degree detection of canned food, and compared the proposed method with the traditional spectral peak method and the typical one-class classification methods.