Comparison of Different Types of Lens for Defect Detection of Inner Thread Based on Deep Learning

Due to limitations of operation space, the inspection of internal threads in mechanical products is problematic not only for skilled operators but also for the design of automated inspection systems. With the introduction of advanced vision lenses, different inspection systems are constructed, allowing detection and defect classification in internal threads. This paper introduces two methods to acquire images of the internal surface in M8 Nuts. By applying Convolutional neural networks (CNNs) to identify defects in the captured images, the characteristics of the two measurement systems are compared, providing general evaluations for the construction of vision systems using the YOLO algorithm for defect detection.