Automated tree detection from 3D lidar images using image processing and machine learning.

Area of Science:

• Computer Vision
• Remote Sensing
• Machine Learning

Background:

• Accurate tree extraction from 3D data is challenging due to the presence of non-tree objects.
• Automated methods are needed to improve efficiency and accuracy in forestry and environmental monitoring.

Purpose of the Study:

• To develop and validate an automated method for tree detection and extraction from 3D lidar images.
• To achieve high accuracy in identifying trees amidst other environmental objects.

Main Methods:

• A novel approach combining 3D image processing and machine learning techniques was employed.
• Objects in 3D lidar data were segmented, projected onto 2D images, and classified using a one-class support vector machine (SVM).

Main Results:

• The proposed method successfully extracted trees from 3D images with 95% accuracy.
• The classification step effectively distinguished trees from non-tree objects.

Conclusions:

• The integrated 3D image processing and machine learning approach provides an accurate and efficient solution for automated tree extraction from lidar data.
• This method holds potential for applications in large-scale forest inventory and analysis.