Locating Large Native Trees in a Densely Forested Area

Introduction

A client approached Air Inspect Australia with a request to identify and count all the large native trees within a specified area. Due to the density of the forest, traditional methods of tree identification were deemed insufficient. To address this, we utilised advanced point cloud processing and analysis techniques. The client provided a KML file outlining the area of interest and inquired specifically about trees exceeding 20 meters in height.

Methodology

Point Cloud Analysis

We employed drones with LiDAR sensors for data collection to generate a highly dense RGB point cloud in order to conduct analysis to accurately map and classify the trees within the specified area.

1. Aerial Point Cloud - RGB

To begin the analysis, we generated a high-resolution point cloud dataset (300+ points per square metre) representing the area of interest. This dataset was captured using remote sensing technology and processed to display RGB (red, green, blue). The image to the right, reflecting the natural colours of the landscape. This visual representation aids in distinguishing between different types of vegetation and other surface features.

> Top-down View (RGB): The 2D top-down representation provides a top-down view, allowing us to observe the spatial distribution and density of the trees.

> 3D-View (RGB): The 3D view offers a more comprehensive understanding of the vertical structure of the forest, essential for observing and assessing tree heights and canopy coverage.

2. Classified Point Cloud

Following the initial visualisation, the point cloud data was classified based on ASPRS standard LiDAR point classes to differentiate between various vegetation heights.
Point classification separates vegetation from ground points, this is an important step when processing forest data.
The second image below displays the classified point cloud data, the different shades of green depict – low, medium and high vegetation. The colour brown indicating ground. See image to the right.

Feature Extraction for Tree Identification

Once the point cloud data was classified, the next step was to conduct a feature extraction by using specialised software to identify the types of vegetation classified previously.

Once features are extracted (separated based on height) we can now easily identify features such as tree height, volume of vegetation, surface area, etc.

The image below with red circles highlighting trees above 20 meters.

Results

The analysis revealed the presence of several large native trees within the specified area. The post-classification feature extraction successfully identified and highlighted trees exceeding 20 meters in height. The results were validated and cross-referenced with the classified point cloud to ensure accuracy.

This information is crucial for the client’s requirements, providing an accurate count and spatial distribution of significant trees in the dense forested area.

Conclusion

By employing advanced point cloud analysis and feature extraction techniques, we successfully identified and counted the large native trees in the client’s specified area. The RGB point cloud provided a detailed visualisation, while the classified point cloud allowed for a highly accurate tree classification. This case study demonstrates the efficiency of integrating point cloud data with sophisticated GIS based queries to tackle complex forestry analysis tasks. The methodology adopted ensured a comprehensive and accurate assessment, providing the client with reliable data for further environmental planning and management.

Locating Large Native Trees in a Densely Forested Area