Comparison of Remote Sensing Image Processing Techniques to Identify Tornado Damage Areas from Landsat TM Data

Area of Science:

• Geosciences
• Remote Sensing
• Environmental Monitoring

Background:

• Remote sensing is effective for large-scale damage surveys after hazardous events.
• Timely and accurate damage assessment is crucial for disaster response and recovery.

Purpose of the Study:

• To compare the accuracy of common image processing techniques for detecting tornado damage tracks.
• To evaluate supervised, unsupervised, and object-oriented classification methods using Landsat TM data.

Main Methods:

• Direct change detection using pre- and post-tornado Landsat TM images.
• Generation of principal component composite images and image difference bands.
• Application of supervised, unsupervised, and object-oriented classification with a nearest neighbor classifier.

Main Results:

• The object-oriented approach demonstrated the highest accuracy in detecting tornado damage.
• Principal Component Analysis (PCA) and Image Differencing showed comparable results.
• Object-oriented classification achieved 15-20% higher accuracy than other methods, while selected PCs improved accuracy by 5-10%.

Conclusions:

• Object-oriented classification is the most accurate technique for tornado damage detection from Landsat TM data.
• Remote sensing, particularly with advanced classification methods, is a valuable tool for post-disaster analysis.