2. Progressive Morphological Filter (PMF) Cluster Analysis

Ground Classification in Wetlands Using Progressive Morphological Filter (PMF)

The Progressive Morphological Filter (PMF) is based on the method described in Zhang et al. (2003) with some technical modifications. It is a widely used algorithm for separating ground and non-ground points in LiDAR datasets, and it can be adapted effectively for wetland environments. PMF operates by applying a series of morphological operations—specifically, opening filters with progressively increasing window sizes—to identify and remove elevated features such as vegetation and structures.

In wetland settings, PMF is particularly useful because:

• Flat terrain: Its elevation-based filtering is well-suited to low-relief landscapes.

• Vegetation clutter: PMF can suppress low-lying vegetation by tuning slope and elevation thresholds.

• Sparse ground returns: It can still detect ground points by gradually expanding the search window and refining elevation constraints.

The algorithm works by:

  1. Applying a morphological opening to the point cloud using a small window.

  2. Increasing the window size iteratively to capture larger non-ground features.

  3. Comparing the elevation difference between the original surface and the filtered result.

  4. Classifying points as ground if the elevation difference is below a user-defined threshold.

For wetland applications, careful tuning of parameters like initial window size, maximum window size, slope threshold, and elevation threshold is essential to avoid misclassifying vegetation or water surfaces as ground.

Summary

##     Analysis Complete!
##     Total rasters processed: 16
##     Valid rasters plotted: 16
##     Global value range: -1.00 to 102.00

Display Final Output

The legend shows the value scale for the raster data. Here’s what it means:

Color gradient: Pink (low values) → Cyan → Blue → Black (high values) Numbers on the legend: The actual data values in each raster cell For your PMF (Cloth Simulation Filter) cluster data:

-1.00: Minimum value (possibly representing ground points or the lowest elevation/classification) 312.00: Maximum value (possibly representing the highest elevation or different feature classification) Values in between: Represent different heights, classifications, or cluster IDs depending on what your CSF processing assigned What the colors tell you:

Pink areas: Lowest values (close to -1.00) - likely ground level or lowest features Cyan/Light blue areas: Medium-low values Dark blue areas: Medium-high values Black areas: Highest values (close to 102.00) - likely elevated features or specific cluster types The legend essentially acts as a color-to-value decoder, allowing you to interpret what each color in the raster represents numerically. When all plots share the same global legend (range = -1.00 to 102.00), you can directly compare values across different rasters because the same color means the same value in all subplots.