Addressing the issues of low efficiency and imprecision in current multi-line LiDAR point cloud fusion methods, this paper presents an innovative algorithm based on the Inertial Measurement Unit (IMU) and an enhanced Iterative Closest Point (ICP) technique. First, voxel filtering is used to downsample the point cloud, then the discrete points in the LiDAR point cloud are eliminated by the Statistical-Outlier-Removal filter, and the IMU information is introduced to complete the distortion correction of the point cloud; Further, the improved sampling consistency initial alignment (SAC-IA) algorithm and the improved ICP algorithm are applied for the initial and accurate registration of the feature point clouds of the current and historical frames. The algorithm is tested in two different scenes, in the outdoor experimental environment, compared with the traditional Normal Distributions Transform (NDT), Fast Point Feature Histograms (FPFH) and NDT-ICP algorithms, the root mean square errors of the registration of this algorithm are 87.60%, 59.25% and 87.88%, respectively. In the indoor environment, the root mean square errors are 74.69%, 37.90% and 81.32%, respectively. These results demonstrate the superior accuracy of point cloud registration offered by our algorithm, affirming its capability to achieve high-precision fusion between point clouds of two frames.