Large spatial steel structures have numerous joints, which are susceptible to construction errors and often encounter challenges related to low measurement efficiency during construction. In order to quickly and accurately measure the construction error of structural spherical joints, this paper proposes an intelligent measurement method based on 3D scanning. Firstly, the structural point cloud model is established using 3D scanning technology. Secondly, the study examines the impact of various noise ratios on the accuracy of spherical fitting using the random sampling consistency method. It is found that when the noise ratio for the sphere exceeds 80%, the random sampling consistency method fails to achieve accurate spherical fitting. Based on this, the spherical point cloud is clustered by combining the genetic algorithm. The noise ratio of the clustered point cloud is controlled, and then the clustered point cloud is fitted spherically using the random sampling consistency method. This approach achieves accurate cluster fitting of two spherical nodes in a partitioned point cloud with a noise ratio of more than 95%. Finally, the effectiveness of the proposed method is verified through an engineering example in Haikou.Results show that the method can successfully achieve batch automatic fitting of ball nodes in the overall point cloud model of the structure. It also allows for the acquisition of actual ball node coordinates and assessment of construction errors by comparing them with theoretical values. These findings hold significant implications for improving the efficiency of construction surveys for large-scale spatial steel structure ball joints.