扬声器纸盆缺陷的机器视觉检测方法研究

针对目前扬声器纸盆外观缺陷检测主要依靠人工,其工作效率低、易出现误检的现象,提出一种基于机器视觉的检测技术。通过对检测系统的组成和软件算法的设计进行研究,从两个不同的角度对目标缺陷区域进行特征提取,并由此提出两种不同的缺陷识别方法。实验结果表明,机器视觉检测技术能够较好地适用于扬声器纸盆外观缺陷检测,同时,采用基于BP神经网络的识别方式,其正确识别率可达94.8%,符合工业检测要求,具有较高的推广应用价值。

Research in Method to Detect the Defects of Loudspeaker Cone by Machine Vision

Aiming atthe low working efficiency, error prone of the appearance defect detection of the loudspeaker cone resulting from relying on manual work,proposing a detection recognition technology based on machine vision. Through the research of the composition of the detection system and the design of the software algorithm, the feature extraction of the target defect area is made from two different angles, and two different defect identification methods are proposed. The experimental results show that the machine vision detection technology can be well applied to the defects identificationof loudspeaker cone, meanwhile, after adopting the identification method based on BP neural network, it’s correct recognition rate can up to 94.8%, which meets the requirements of industrial inspection and has high value of popularization and application.