Completion of Marine Wireless Sensor Monitoring Data Based on Tensor Mode-n Rank and Tucker Operator

Abstract

The marine monitoring system is one of the frontier technologies actively developed by major countries in the world today. The current ocean monitoring system mainly relies on technologies such as positioning, control, and wireless sensors. The buoy equipped with a variety of wireless sensors continuously collects data on the ocean. However, due to natural environmental influences and malicious tampering by the enemy, the data directly obtained by the wireless sensor buoys may contain large errors. Therefore, we construct the obtained original data into a tensor model, and at the same time replace the data with larger errors into null data “0”. Based on the tensor mode-n rank, we use the alternating direction method of multipliers (ADMM) framework, combined with the tensor Tucker decomposition, and introduce the tensor Tucker singular value operator to it. The overall data can be completed from the existing original data with missing values, so as to optimize the original ocean monitoring data. The method is compared with the data completion based on tensor Tucker decomposition and the linear regression prediction based on principal component analysis. Numerical experiments are given to confirm the superiority of the proposed method.