Pre-processing of point cloud Due to the obstruction and obscuration of the scanning target by external environment factors during the scanning process of the 3D laser scanner, such as the occlusion of moving vehicles, pedestrian trees, and the reflection characteristics of the entity itself, it is necessary to Line filtering eliminates unstable points and error points contained in point cloud data. In practice, you need to choose the appropriate filtering algorithm to automate this process.

Point cloud registration uses control point registration to register the point cloud to the control network coordinate system; the point cloud missing from the target uses the common area to find the same name point for pairwise registration, when the point name of the same name cannot be found, Use the manual registration method. The latter two methods are two-two registration. In order to convert all point clouds to a unified control network coordinate system and the control point registration method to get the point cloud, the two stations must be assigned one of the two stations. Go to the point cloud under the control network coordinate system.

When the data collected by the point cloud splicing field is imported into the software, it will be automatically spliced ??according to the coordinate points. However, due to the error of the human operation and the corner frame, some point cloud joints are not ideal. At this time, manual splicing is required, and some coordinates are not compensated. The splicing of the sweeping surface also needs to be handled manually. When manually splicing, the point cloud should be properly compressed, and the feature points of protrusion, sharp corner and different planes should be selected to reduce the operation error. If the scanning is performed at 1 cm laser interval, the error after splicing is preferably less than 3 mm.

Establishing a three-dimensional model After the building is digitized into a large number of discrete spatial point cloud data, a three-dimensional model of the building is constructed on this basis.

Due to the discreteness of the point cloud itself, the point cloud's vulnerability repair will cause certain defects in the model. It is necessary to repair and adjust it in the polygon stage to obtain an accurate physical digital model. Due to the complex and diverse shape of the building, it is difficult to fully automate the repair of the current grid. There are two main methods for the repair of the 3D laser scanner point cloud data: when the hole appears in the plane area, such as a hole in a window or a wall, linear interpolation can be used to fill the hole data; when the hole appears in the non-planar area For example, a vulnerability occurs on a cylinder, and a quadric interpolation method can be adopted.