Quantitative analysis and identification of unknown shaped defects have always been difficult and challenging in the quality control of micro pipes. A series of algorithms for defect detection and feature recognition is presented in this study. A lightweight convolution neural network (LCNN) is introduced to realize defect discrimination. A shallow segmentation network is employed to cooperate with LCNN to obtain pixel-wise crack detection, and a feature recognition algorithm for quantitative measurement is presented. The experimental results show that the proposed algorithms can achieve defect detection with an accuracy of 98.5%, segmentation with mean intersection over union of 0.834, and latency of only 0.2 s. It can be used for online feature recognition and defect detection of the inner surface of a hole.
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
Cited By
You do not have subscription access to this journal. Cited by links are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.
You do not have subscription access to this journal. Figure files are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.
You do not have subscription access to this journal. Article tables are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.
You do not have subscription access to this journal. Equations are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.