Exercise 3. Tree Segmentation

In order to extract individual tree attributes such as locations, heights, canopy diameters and so on, clusters of points representing trees in the point cloud need to be segmented into individual trees.

LiDAR360 provides three methods for tree segmentation and we will introduce you to each of them in this exercise:

  • CHM Segmentation.
  • Point Cloud Segmentation
  • Layer Stacking Segmentation.

For this exercise, you can either use the output from Exercise 1 as your input, or use downloaded sample data ALSData_Remove Outliers_Normalize by DEM.LiData and ALSData_Remove Outliers.LiData. Pay extra attention to which input data is used when following the exercise instructions.

CHM Segmentation

1 Generate DEM

1.1 Load the ALSData_Remove Outliers.LiData point cloud (after removing outliers and prior to normalization) into the software.

1.2 Go to Terrain > DEM. Leave the parameters to default. Click OK.

LiDAR360 ALS Forest
LiDAR360 ALS Forest

2 Generate DSM.

2.1 Go to Terrain > DSM. Leave the parameters to default. Click OK.

LiDAR360 ALS Forest
LiDAR360 ALS Forest

To generate DSM based on spike-free TIN, select TIN for Interpolation Method, and select Spike-Free TIN method. Spike-free TIN can be used to generate Pit-Free CHM, which can improve CHM segmentation results in certain scenarios.

3 Generate CHM.

3.1 Go to Terrain > CHM, set Input DSM and Input DEM data to match the following, and click OK.

LiDAR360 ALS Forest
LiDAR360 ALS Forest

4 CHM Segmentation.

4.1 Go to ALS Forest > Segmentation > CHM Segmentation, select the CHM file as your input, set Sigma to 1, and accept default settings and click OK.

LiDAR360 ALS Forest

4.2 Once the segmentation completes, the software will prompt you to add the results into display. Set your parameters to match the screenshot below and click Apply. The CSV results will be added in to show tree locations, heights, crown diameters, and crown areas.

LiDAR360 ALS Forest
LiDAR360 ALS Forest
LiDAR360 ALS Forest

In addition to the CSV file, CHM Segmentation also creates a shapefile of tree boundaries. Right-click at Vector under Layers in Project panel > Import Data > select the shapefile ALSData_Remove Outliers_DSM_CHM_CHM Segmentation.shp to open.

LiDAR360 ALS Forest

4.3 Though attributes of individual trees are available in the CSV file, the LiDAR point cloud is not segmented into individual trees yet. Go to ALS Forest > Segmentation > Point Cloud Segementation from Seed Points. Select the ALSData_Remove Outliers_Normalize by DEM.LiData dataset as Point Cloud File, and ALSData_Remove Outliers_DSM_CHM_CHM Segmentation.CSV as Seed File.

LiDAR360 ALS Forest

Once segmentation completes, the display should switch to Display by TreeID. If not, click on the toolbar to apply the effect.

LiDAR360 ALS Forest

The CSV seed file can also be generated using the Generate Seed Points from CHM tool under Segmentation toolset in ALS Forest model, which uses the same algorithm as CHM Segmentation.

Point Cloud Segmentation

Point Cloud Segmentation can directly segment LiDAR point cloud, which can reduce the influence of under-canopy information loss in the CHM segmentation method. Individual tree information, including tree location, tree height, crown diameter, crown area and crown volume can be obtained from the segmentation results.

1 Add the ALSData_Remove Outliers_Normalize by DEM.LiData.LiData point cloud to LiDAR360.

2 Go to ALS Forest > Segmentation > Point Cloud Segmentation, input your normalized point cloud data, accept the default settings and click OK.

Please refer to our User Guide to understand the parameter settings of the tool: Point Cloud Segmentation.

LiDAR360 ALS Forest

Once segmentation completes, the display should switch to Display by TreeID. If not, click on on the toolbar to apply the effect.

The segmentation process also outputs a CSV file that contains tree attributes, including tree IDs, XY coordinates, heights, DBH, crown diameters, crown areas, and crown volumes.

LiDAR360 ALS Forest
LiDAR360 ALS Forest
LiDAR360 ALS Forest

Layer Stacking

LiDAR360 also provides option to use Layer Stacking algorithm to segment trees. Two major steps are involved in Layer Stacking segmentation: generate seeds from Layer Stacking, and segment from seed points.

1 Generate Seeds from Layer Stacking

1.1 Go to ALS Forest > Segmentation > Generate Seeds from Layer Stacking. Select the ALSData_Remove Outliers_Normalize by DEM.LiData dataset as input file. Accept the default settings and click OK.

LiDAR360 ALS Forest

A CSV file of seed points consisted of 4 columns –TreeID, TreeLocationX, TreeLocationY, and TreeLocationZ –will be created.

LiDAR360 ALS Forest

All the seed points generated from layer stacking share the same TreeLocationZ value, which is the maximum Z value of the point cloud. Tree heights will be calculated later with Point Cloud Segmentation from Seed Points.

2 Point Cloud Segmentation from Seed Points

2.1 Go to ALS Forest > Segmentation > Point Cloud Segmentation from Seed Points. Click on the blanks under Point Cloud File and Seed File to select the corresponding inputs accordingly. Click OK.

LiDAR360 ALS Forest

Load in the segmentation results, and then use the Display by TreeID tool to update your display.

The segmentation process also outputs a CSV file that contains tree attributes, including tree IDs, XY coordinates, heights, DBH, crown diameters, crown areas, and crown volumes.

LiDAR360 ALS Forest
LiDAR360 ALS Forest

Segmentation Results Inspection and Editing

ALS Editor allows the user to closely examine tree segmentation results, and make edits at the same time by adding to and deleting from seed points. The user can then rerun segmentation on the point cloud data using modified seed points to achieve higher segmentation accuracy.

1.Add Results

1.1 Launch the editor from ALS Forest > ALS Editor. Go to Editor > Start Edit, and then select the normalized point cloud.

1.2 On the ALS Editor toolbar, click on Open Seed Point File and then choose the CSV result of segmentation. As shown in the screenshot below, specify the appropriate column headers to Tree ID, X, Y, Z, and Crown Diameter respectively, and set the last two columns to Ignore. Set Skip Lines to 1, and then click Apply.

LiDAR360 ALS Forest

Tree IDs will be displayed by default. If Tree IDs are blocking the point cloud in display, you can turn them off by going to the Seed Setting tool and unchecking Show Seed ID. You can also set Seed Size to 1 to increase visibility.

LiDAR360 ALS Forest
LiDAR360 ALS Forest

2.Examine Results

2.1 Use the Filter Trees tool to filter point cloud data based on tree IDs, heights, and crown areas. Trees falling within the specified attribute range will become highlighted. Set the Tree Crown Areas range to 10 – 18.358 m², and find the point cloud within this attribute range highlighted.

LiDAR360 ALS Forest

2.2 Click on Profile tool and start drawing a hexagonal selection area in the main display window. The point cloud in the selection area will show in a profile view window, where you can view the data in 3D.

The tree colored red in the profile view below has a larger-than-actual crown area due to under-segmentation:

LiDAR360 ALS Forest

3.Edit Results

3.1 Use Add Seed Points tool to add a seed point on the treetop.

Seed points can be added in either the main display or the profile viewer. It’s recommended to add seed points on or near treetops in order to ensure segmentation accuracy.

LiDAR360 ALS Forest

3.2 Use Select Seed Points tool to select and highlight incorrect seed points in the main display.

LiDAR360 ALS Forest

3.3 Use Delete Selected Seed Points tool or the Delete key on your keyboard to delete selected incorrect seed points.

LiDAR360 ALS Forest

While working with profile views, you can use the Pan Profile tool to drag and pan your work area in the main display to easily move around in the profile view. This allows you to work with a fixed area size, and save the pain of drawing a new profile each time.

4.Save and Rerun

4.1 When you are done inspecting/editing, click on Save Seed Points File to save the seed points to a CSV file.

4.2 As needed, you may want to rerun segmentation based on your edited seed points. Click on Clear Tree ID tool to clear existing TreeIDs, and then click on Point Cloud Segmentation Based on Seed to rerun segmentation.

LiDAR360 ALS Forest

results matching ""

    No results matching ""