Sea ice is a significant factor in ship navigation, and accurately identifying its thickness is crucial for timely decision-making to ensure ship structural integrity and personnel safety. However, current methods for sea ice thickness measurement, such as shipborne imaging devices and synthetic aperture radar altimeters, have limitations such as high costs, difficulties in data acquisition, and poor timeliness. In this article, we present an approach to detect sea ice thickness using a combination of sensors and machine learning techniques. Vibration signals from ship-ice collisions are collected using acceleration vibration sensors, undergo data preprocessing and time-domain alignment, and the true value of sea ice thickness is obtained through field measurements and a sea ice image information identification system. Time-frequency domain features are extracted, and feature quantity optimization is performed, followed by machine learning regression algorithms to predict sea ice thickness. Experimental results show that excellent regression results can be obtained using a single node for thickness prediction, with an R2 value of 0.908 and an RMSE of 1.374 cm. Our findings demonstrate that analyzing vibration signals generated by ship-ice collisions during vessel navigation can effectively identify sea ice thickness. This approach provides a practical and cost-effective alternative for sea ice thickness measurement and identification in the field.