基于自动编码器的塑料制品表面缺陷检测

Surface Defect Detection of Plastic Products Based on Automatic Encoder

基于自动编码器的塑料制品表面缺陷检测方法,使用无监督的方式学习正常图像的隐形特征,通过重建增大正常图像和缺陷图像间的差异,从而检测缺陷。自动编码器方法不需要人工选取特征,同时,不需要使用缺陷图像作为监督信息实现模型学习。以区分正常图像和缺陷图像为目标,对塑料制品表面缺陷进行检测。实验结果表明,自动编码器方法能够有效地检测出多种类型的缺陷,检测准确性较高,具有广泛的应用。

The surface defect detection method of plastic products based on an automatic encoder uses an unsupervised method to learn the invisible features of normal images,and enlarges the difference between normal images and defective images through reconstruction to detect defects.The automatic encoder method does not need to manually select features,and at the same time,it does not need to use defect images as supervision information to realize model learning.To distinguish between normal images and defective images as the goal,the surface defects of plastic products are detected.The experimental results show that the automatic encoder method can effectively detect many types of defects,with high detection accuracy,and has a wide range of applications.