Glass panel surface defect detection method based on small sample learning

Abstract

The invention discloses a glass panel surface defect detection method based on small sample learning. The method comprises the following steps: collecting a small number of defect glass panel images and marking bounding boxes and defect categories; carrying out preprocessing and quantity expansion on the small number of glass panel defect images; constructing a network for identifying and positioning the defect images of the glass panel; and performing defect detection on the glass panel image by using the trained defect detection model, and outputting a defect frame and a defect type. According to the method, data enhancement, transfer learning and L2 regularization are used for relieving a small sample problem, random jitter is carried out on a labeling box to increase diversity of a frame so as to improve model robustness, a global ROI extraction layer is used for introducing background information for candidate region features, and the weight of each sample in loss is adaptively changed so as to improve model performance. The method is suitable for a surface defect detection task with only a small number of glass panel images, and the detection precision is high.

Description

technical field [0001] The invention relates to the technical field of glass panel detection, in particular to a method for detecting surface defects of glass panels based on small sample learning. Background technique [0002] Glass panels are in great demand in industries such as computers, communications, and consumer electronics. With the continuous growth of market demand, the requirements for the quality of glass panels are also getting higher and higher. Traditional manual inspection methods require a large number of well-trained workers, often consume a lot of manpower, are inefficient, and due to individual subjectivity, standards may vary greatly. In addition, due to the special optical properties of glass, long-term inspection work is harmful to The workers' eyes have some damage. With the development of optical technology and computer technology, many automatic optical inspection solutions have been proposed for surface defect inspection tasks. This non-contact...

AI Technical Summary

Problems solved by technology

[0007] In order to solve the above problems, the present invention proposes a glass panel surface defect detection method based on small sample learning, which realizes efficient and accurate detection of glass panel defects when there are only a small number of sample images of glass panel defects, the labels are not accurate enough, and there are many small targets. Panel surface defects, the specific steps are as follows:

Images