Study on Vision Measurement for Levitation Gap of Magnetic Levitation Ball Based on Convolutional Neural Network
Jie Kong, Yongzhi Jing, Chenhao Zhang, Jianhua Hao, Qian Cheng, Qianwen Gong · 2019
With the development of deep learning, the Convolutional Neural Network (CNN) is widely used in object classification and pattern recognition. It has enabled computer to achieve better performance than humans in specialized computer vision tasks. This paper takes the magnetic levitation ball system as the research object. Aiming at the shortcomings of traditional methods for levitation gap measurement, a new method is proposed by combining machine vision and CNN image processing technology. The convolution neural network algorithm is used to build the gap measurement model, and the training set is used to train the model. The experimental results show that using convolution neural network image processing technology to realize the levitation gap measurement of magnetic levitation ball system has high distance measurement accuracy and good performance. The proposed CNN model provides correct gap data with the maximum error of 0.16mm for full scale and the average error of 0.07mm for full scale in the test set.