Quadratic Detection Water Depth Estimation Algorithm Based on M-estimation

To accomplish outlier detection and data quality control for MBES (Multi-beam echo sounding), a quadratic detection water depth method based on M-estimation has been proposed. The single Ping data has been processed. In order to improve the tolerance, the preliminary results of M-estimation are used to identify suspicious areas. The sliding window for M-estimation and the historical window for suspicious areas are used to conduct a quadratic investigation on the data. We estimate the position marked as abnormal by M to be the last point in the quadratic inspection window and process a number of depth measurements up to that point. Since what we examine are the left-hand historical data and the window length has been increased, the proportion of the valid input samples can be increased correspondingly. The suspicious measurement values processed by the algorithm are marked and automatically eliminated at the next input. A hypothesis model was established using a Bayesian model, and multi-depth estimation was established by examining the continuity of different results. The depth with the greatest number of models was used as the estimation depth, while the remaining depth models served as post-processing reference values. This algorithm provides the results of the depth estimation of the corresponding sounding values of each beam, and the detection of outliers can be performed by selecting various thresholds based on the estimation results. The experimental results demonstrate that abnormal pulse outliers and cluster outliers can be detected, and the details can be well reserved. The proposed algorithm has superior estimation performance and improved tolerance.