To improve the efficiency and detail performance of the Poisson surface reconstruction algorithm’s reconstruction results, a point cloud search method based on a hybrid tree balances the conflict between time complexity and space complexity of octree and binary tree technology. In the point cloud search stage, the density evaluation and filtering of the point cloud were used by introducing multiple energy terms, and adaptive point cloud up sampling was used for the sparse part of the point cloud to ensure the details and accuracy of the reconstructed model. The results show that the speed of the hybrid tree reconstruction algorithm increased by 33% and 15% on average compared with the Poisson surface reconstruction algorithm and the screened Poisson surface reconstruction algorithm. In addition, the details of the reconstructed model can be better maintained to obtain the minimum error. This study provides a reference for the surface reconstruction of point clouds.