<p>In the in-situ observation of seabed three-dimensional deformation, three-axis acceleration sensors are widely used, but there is a lack of research on the data processing method of acceleration integral displacement, and it is of great significance to improve the accuracy of acceleration integral displacement for the observation of seabed three-dimensional deformation. In this paper, a novel data processing method of three-axis acceleration integral displacement is established. Compared with other acceleration processing methods, the innovation of our approach lies in integrating the Power Spectral Density (PSD) method into the integration process. This method enhances accuracy by mitigating noise and adjusting the optimal cut-off frequency for better signal extraction. Moreover, we combined a three-axis displacement platform with three-axis acceleration sensors to conduct laboratory experiments to obtain the acceleration signal and the measured displacement data, and the data are processed using our proposed method, time-domain integration, and frequency-domain integration methods. Subsequently, the results of the three data processing methods are compared. The experimental data processing results show that the method established in this paper yields smaller errors and more accurate results in the process of acceleration integral displacement. Compared with the other two methods, our proposed method reduces the overall error by 38–59%, which can provide a better data processing method for the integration of three-axis acceleration data in the in-situ observation of seabed three-dimensional deformation.</p>