Exercise 1. Data Preprocessing
LiDAR point cloud data inputs to tools found in the Terrain module must be preprocessed to remove outliers, classify ground points, and remove errors and non-target data points through manual editing. It is important to remove outliers when generating digital elevation products so that the data can accurately represent the areas surveyed. LiDAR360’s ground point classification routines provide fundamental inputs to the Terrain module’s digital elevation model (DEM) and contour line production tools. LiDAR360 also provides the users with both automated and interactive point cloud classification tools that can be used to edit classified results with ease and efficiency.
Outliers are a common type of noise found in unprocessed LiDAR datasets. High-level outliers are usually created when high-flying objects (such as birds or aircraft) pass through a laser scanner’s field of view during the process of data collection. Low-level outliers are returns with extremely low attitudes and are the result of a multipath effect impacting the amount of time it takes for the laser pulse to return to the laser scanner after its first reflection. The Remove Outliers tool in LiDAR360 aims to remove these types of noise and improve the quality of data products derived from the processed point clouds.
1 In LiDAR360, add downloaded sample data into current project. Then go to Data Management > Point Cloud Tools > Remove Outliers. In the tool interface below, accept the default parameters and click OK. For descriptions of the parameters, please refer to the User Guide.
2 Add the resulting TerrainSampleData_Remove Outliers.LiData file into display and turn off the original data from Layers to inspect the results.
If outliers are still present in the results, try rerunning the Remove Outliers tool on the outputs and use a smaller Multiples of std deviation value. Otherwise, manually classify the points to a noise class and exclude them from later processes.
Multiples of std deviation (default value is ‘5’): The factor multiplied by the standard deviation to calculate the maximum tolerance distance for determining outliers. The smaller this value is, the more points will be considered as outliers and deleted from the output point cloud.
Please refer to Tutorial: LiDAR360 Classification – Exercise 2: Interactive Classification for instructions of manual classification in LiDAR360.
Automatic Ground Points Classification
Once low- and high-level outliers have removed from the sample point cloud users can move on to classify ground points before creating DEM and contour line products. Note that users can generate DSM outputs without first classifying ground points.
3 Go to Classify > Classify Ground Points, select TerrainSampleData_Remove Outliers.LiData file as the input. Check UnClassified in From Class and 2-Ground in To Class. To accommodate the building conditions in the sample data, set Max Building Size to 20 m, Max Terrain Angle to 88 degrees, Iteration Angle to 8 degrees, and Iteration Distance to 1.4 m. Leave the other parameters to default and click OK.
- Max Building Size: The maximum length of the building edge that exists in the point cloud scan. The maximum building size can be measured by using the Length Measurement tool in the menu bar.
- Max Terrain Angle: The maximum slope of the terrain shown in the point cloud. This parameter is used to determine whether the points near the ground points belong to the ground or not.
- Iteration Angle: The allowable range of angles between unclassified points and ground points. For undulating terrains, increase this value to accommodate rapid changes in topography over short spatial distances.
- Iteration Distance: This distance threshold limits the amount of space that can exist between the FROM CLASS points and their nearest facet in the triangle mesh created during the running of Classify Ground Points. When the topography is highly undulating, it should be set to a larger value. In addition, the iteration distance should adjust with the iteration angle.
- Reduce Iteration Angle When Edge Length < (m, default value is “2”, Optional): When the triangle length of the point to be classified corresponding to the length of the triangle is less than the threshold, the densification of triangulation network is stopped. This value can prevent the locally generated ground point from being too dense. When this value increase, the ground points will be sparse, and vice versa.
- Only Key Points (Optional): Extract key points of terrain model on the basis of ground point filtering. This function can preserve the key points on the terrain and sparse the points on the flat area. For the specific usage, please see Classify Model Key Points.
- For descriptions of the parameters, please refer to the User Guide.
Ground Points Classification Refinement
In order to produce terrain models with the highest accuracy, it is often the case that manual point cloud classification and editing is needed after the Classify Ground Points tool being run in LiDAR360. Classification refinement can be achieved through Classify Ground by Selected and/or Classify by Interactive Editing in LiDAR360.
Classify Ground by Selected
Due to the complex and varied terrain of point cloud data, it is often difficult to achieve good classification results using a single set of parameters when using Classify Ground Points, especially in areas where a mixture of terrain types (mountains, plains, etc.) can be found.
Classify Ground by Selected tools allows the user to select a specific region of the point cloud, and then run classification using specific algorithm and/or specific parameter settings in the selected area to accommodate its terrain conditions.
Three different algorithms are provided in LiDAR360 for classifying ground points:
- TIN Filter: This filtering method uses an improved progressive TIN densification filtering algorithm described in Zhao et al.,2016. The TIN Filter is the recommended method for classifying ground points in LiDAR360 and the algorithm helps to power its automatic Classify Ground Points tool. This method is the least sensitive to terrain variations, and should have a stable performance under most situations. Please refer to User Guide for detailed introduction of the algorithm: Classify Ground Points.
- Slope Filter: This filtering method extracts terrain based on changes in point cloud slope. This method has the highest efficiency when processing point clouds of smooth topography. Please refer to User Guide for detailed introduction to the algorithm: Slope Filter.
- Conicoid Filter: Ground points are classified by fitting quadric surfaces. This method is suitable for undulating terrain, but not very steep areas. Please refer to User Guide for detailed introduction of the algorithm: Conicoid Filter.
In the following steps, the TIN Filter method is used in a region where automatic Classify Ground Point tool did not extract ground points correctly. Parameters will be adjusted when running TIN Filter to accommodate variation in the local terrain rather than that found in the entire input point cloud.
4 Go to Display by Classification and turn off the UnClassified class by unchecking the box for that Description in the Display column.
5 Rotate the point cloud in the 3D viewer and notice that ground points are missing in the region delimited by the red box shown in the figure below.
6 Go to Classify > Classify Ground by Selected to activate the toolbar.
7 Use the Polygon Selection tool to select the region shown in the image below. Selected points are highlighted in purple.
Because only the Ground class is checked in Display by Classification, only ground points are displayed in the viewer. However, points of all classes (i.e. From Class: UnClassified & Ground) in the polygon regions will participate the following ground point classifying operations.
8 Click on TIN Filter . In the pop-up dialog window, set Max Building Size to 5 (meters). This region was misclassified in the previous step of this tutorial because the large flat terraced areas of the terrain were mistaken by the classifier to be buildings. To fix this error, we can decrease the Max Building Size parameter and reclassify the region by clicking OK to rerun the TIN Filter tool.
9 Click Exit to exit Classify Ground by Selected toolbar.
Classify by Interactive Editing
It is extremely difficult to achieve 100% accuracy with any ground point classification algorithm, therefore classification methods that rely on human-computer interaction are required to meet high accuracy demands. Manual inspection and reclassification operations can be easily performed in LiDAR360’s Profile window.
Please refer to the Tutorial: LiDAR360 Classification – Exercise: Interactive Classification for instructions on interactive classification operations.