[Object-oriented segmentation and classification of forest gap based on QuickBird remote sensing image.]

Area of Science:

• Forestry
• Remote Sensing
• Geospatial Analysis

Background:

• Traditional forest gap extraction methods are insufficient for regional-scale analysis.
• High spatial resolution remote sensing imagery offers a viable alternative for large-area forest gap assessment.

Purpose of the Study:

• To develop and evaluate an object-oriented classification approach for regional forest gap extraction using QuickBird imagery.
• To determine the optimal segmentation scale and assess the classification accuracy for identifying forest gaps.

Main Methods:

• Object-oriented image segmentation of QuickBird data at multiple scales (10-100).
• Evaluation of segmentation results using intersection area ratios (RAor and RAos).
• Support Vector Machine (SVM) classification using spectral characteristics at the optimal scale.

Main Results:

• The optimal segmentation scale was determined to be 40, balancing segmentation accuracy.
• Overall classification accuracy reached 88% (Kappa=0.82) at the optimal scale.
• The object-oriented approach demonstrated significant accuracy improvements over varying scales.

Conclusions:

• Combining high-resolution remote sensing with object-oriented classification provides an efficient and accurate method for regional forest gap extraction.
• This approach can effectively replace traditional field investigations and manual interpretation for large-scale forest monitoring.