Research on Optical Network Port State Objects Detection Algorithm Based on Deep Learning

摘要

The passivity of optical network makes its port usage state impossible to be accurately mastered by remote monitoring system. This study applies a deep learning algorithm to port state detection. First, the algorithm was selected. Second, based on the fixed height-width ratio of the port, k-means cluster analysis was used to determine the height-width ratio for the dimensions of the candidate frame. Finally, the data set was expanded by data enhancement. The experimental results showed that the accuracy of port detection network is as high as 87%. Additionally, it can be applied to other port-intensive devices, providing a certain robustness.