The complete process of point cloud data processing includes steps such as point cloud data preprocessing, entity model establishment, and data result output.

The 'point cloud' result obtained by scanning is a set of points containing the three-dimensional coordinates of the points and the color attributes, and the architectural surveying needs to obtain a two-dimensional line drawing composed of 'line' and 'face'. How to transform a 3D point cloud into a 2D line drawing is the key to the processing and application of the point cloud. There are several ways to convert a three-dimensional point cloud into a two-dimensional line drawing: one is to directly measure the relevant data of the building on the point cloud model, and draw a picture according to the traditional ancient building mapping method; the second is to intercept a part of the point cloud to form a point. The cloud is “sliced”, and then the “slice” is imported into the relevant software to form a two-dimensional line drawing after fitting; the third is to make an orthographic projection “point cloud” image relative to the surface of a certain area of the building. After the projection image matches the photo color information, it is imported into other processing software and then 'draw' into a two-dimensional line drawing. In addition, continuous orthophoto maps are required for building groups, and architectural attachments for special shapes can be represented by contour maps.

In general, it is difficult for a 3D scanner to scan a single direction to obtain complete point cloud data of a scanned target. It is reflected that a scanned entity information is usually completed by several scans, but each scanned image is in the position of the scanner. The zero-point local coordinate system, that is, the coordinate system of the point cloud data obtained each time it is scanned is independent and uncorrelated. But in fact, each point cloud array data is part of the scanning scene, then it is necessary to convert these point cloud array data into the same coordinate system, so the point cloud data obtained is stitched and matched.

In the splicing process of point cloud data or the operation of 3D data in the processing software, a series of three-dimensional transformations such as translation, rotation and scaling are bound. In order to express the scan results in the form of an image, viewing any part of the point cloud at any angle is essentially the transformation and processing of the three-dimensional graphics. In general, the premise for splicing two scanned images is that there should be coincident portions in the two scanned images, that is, a part of the target object should be scanned in the two scans before and after, roughly overlapping The part should account for 20~30% of the whole image. If the proportion of the overlapping part is too small, it is difficult to ensure the precision of the splicing. If the proportion is too large, the scanning times and the splicing workload will be increased.

Preprocessing the 3D point cloud data generally requires reprocessing the acquired raw data, checking the integrity of the data and the consistency of the data, standardizing the data format, and performing point cloud filtering. Due to the complicated working environment of the scanner in the field, especially when working on the construction site, the movement of the construction machine, the movement of the personnel, the obstruction of trees, the building blockage, the construction dust and the uneven reflection characteristics of the scanning target itself will cause scanning. Obtaining unstable points and noise points of point cloud data,

The existence of these points is not expected in the scan results. These point cloud data should be removed in the post-processing. This process is called point cloud filtering. Point cloud filtering is an important process of data preprocessing. The data results have a significant impact.

Generally speaking, if the initial point cloud data coincides with the operation of the same name, only a small amount of point cloud search calculation work is needed to complete the precise splicing of the point cloud data. Otherwise, a large amount of calculation work is required to complete, and the number of calculations is superimposed. It also depends on the set error reference. According to the needs of data usage, it is also possible to perform merging of the spliced data, denoising, deleting, and reducing the density of the point cloud.