Surface defect detection of polishing shaft based on parallel computation

JIANG Qing-sheng, LI Yan-biao, JI Shi-ming
（College of Mechanical Engineering, Zhejiang University of Technology, Hangzhou 310023, China）

Abstract: Aiming at the problems of realtime defect detection and defect classification requirements for polishing shaft surface defects in production practice, the surface defect detection system for polishing shafts, defect image acquisition, image features, traditional detection methods, etc. were studied, and the treatment methods of shaft surface defect detection were summarized. Two detection systems based on parallel computing were proposed, i.e. cloud-based Spark multi-machine parallel distributed computer cluster detection system and single-machine multi-core CUDA parallel processing detection system. Through the analysis of Spark architecture multi-machine distributed parallel computer theory, and analysis of the principle of mathematical parallelization combined with CUDA single-machine multi-core parallel operation theory, the experiments were conducted. The results indicate that the singlemachine multi-core parallel processing detection system based on CUDA is suitable for real-time detection, and the multi-machine distributed parallel detection system based on Spark is not suitable for real-time detection systems, but is suitable for big data processing systems.

Key words: image processing; Spark; distributed parallel computing; graphic process unit（GPU）; CUDA