Processing point cloud data requires tremendous amounts of time and effort. However, this whole workflow can be done with ease with the use of ScanX’s machine learning algorithms. In this guide, you will learn about how to create your very first ScanX project with the help of a very powerful feature: Automatic Classification.
What is Automatic Classification?
A raw point cloud data is just a collection of sets of points in a 3D space with various attributes such as colour, brightness, and coordinates. This alone will be hard to use in actual applications since the points are not classified according to their groupings such as material type or object name.
Traditionally, it took a lot of time to classify such points. A person needs to manually select each cluster of point cloud and identify the correct classes for them. Fortunately, with the help of Machine Learning algorithms, ScanX can do this automatically.
The main automated classification objects in ScanX are ground, vegetation, and buildings. However, there are also options to classify Roof Edges, Shrubs, Gutters, and Overhang Trees. This makes the whole process easier and more efficient.
Creating a Project
In your ScanX Dashboard, select Create a new project on either the left-side of the User Interface or the bottom right side. Once this has been selected, you will be routed to a new page.
Once finished, you will need to move on to a second step which involves determining accuracy for automatic building and tree classification in terms of cm or mm.
There is a need to select which type of scanner has been used for the scan such as TLS, UAV LiDAR, UAV Photogrammetry, Vehicle Mobile, or Backpack LiDAR. Also, area types such as Forestry, Town, Mining, Urban, or Wilderness should be selected.
Lastly, there are boxes to be ticked regarding the type of device used or selection of grounds input only.
The next step involves the classification of noise options. Noise processing profile is to be selected in terms of mm, cm, or 0.5 m. furthermore, you are given a choice to transform coordinates, classify noise points, or remove them. Once finished, you will be prompted to upload the dataset. In this case, we will just use a demo data set.
Viewing the Project + More Classifications
Once the project is opened, you will see that objects such as vegetation, ground, noise, building, and water are colour-coded. Here is a view of one.
You can select or deselect certain objects to make them visible or not. In this case, I unselected all the ground parts.
For a deeper classification of the ground, you can go to the process button on the side and click on ground extraction.
A target point cloud needs to be selected, and parameters are to be defined. Once done, the Execute button can be selected.
There are other deep object classifications such as smaller buildings, shrubs, roof gutters, edge of the roof, and overhang trees! These can be accessed in this part so don’t forget to explore them.