Deep learning–based intelligent measurement methods and system for CMM

Abstract

Extensive manual intervention and management are typically required when using coordinate measuring machines (CMMs) for inspections in production lines leading to low efficiency. This study presents a deep learning–based intelligent measurement method and system for measuring typical features (including holes, cylinders, balls, steps, and slots) of common components to improve inspection efficiency. This method combines vision sensors and a trigger probe. The You Only Look Once algorithm was employed to learn and achieve intelligent detection of features. An image-matching algorithm based on image inverse perspective transformation was designed, and the ant colony algorithm was implemented to optimize the measurement sequence. Then, an automatic approach for feature measurement path planning was designed. The presented system was tested using CMM, and a component with multiple typical features was measured. Results show that this method and system can be efficaciously implemented for intelligent measurement of typical features.