Abstract: In response to the existing clustering methods that have wasted CPU and GPU resources on nodes during 3D modeling of tilt images and the inefficiency of 3D modeling,this paper proposes a cluster 3D modeling method based on compatibility. Firstly,by dividing the 3D reconstruction task based on two-step submission,the execution efficiency of the cluster is improved while the modeling effect remains unchanged. Then,we propose the compatibility task node allocation method for the waste of cluster node resources,and complete the implementation of the 3D modeling cluster method based on the compatibility task allocation. Finally,the experimental test of large-scale image 3D reconstruction is conducted to verify the superiority of the compatibility theory on the optimization of the allocation method under the experimental environment of big data,which proves that the algorithm of this paper has significantly improved the efficiency of 3D modeling.

Key words: 3D reconstruction, real-world modeling, cluster systems, compatibility algorithms, tilt image