Influence of shadow removal on image classification in riverine environments

Area of Science:

• Environmental Remote Sensing
• Image Analysis
• Geospatial Science

Background:

• Shadows in remote-sensor imagery introduce significant errors in riverine environment classification.
• Accurate mapping of riverine landscapes is essential for ecological and hydrological studies.
• Existing shadow-removal techniques may not fully address the complexities of riverine features.

Purpose of the Study:

• To develop and evaluate a modified shadow-removal algorithm for improved remote-sensing image classification of riverine landscapes.
• To assess the quantitative impact of shadow removal on classification accuracy.
• To enhance the detection of subtle river features often hidden by shadows.

Main Methods:

• Modification of an illumination suppression-based shadow-removal algorithm.
• Incorporation of a user-defined tiling approach to handle spatially complex river features.
• Quantitative evaluation of classification accuracy using aerial photography before and after shadow removal.

Main Results:

• The modified shadow-removal method significantly increased the accuracy of riverine landscape classification.
• The algorithm demonstrated improved detection capabilities for small river channels partially obscured by shadows.
• Shadow removal proved to be a critical preprocessing step for reliable riverine environment analysis.

Conclusions:

• Shadow removal is a vital preprocessing step for accurate remote-sensing classification of riverine environments.
• The proposed modified algorithm effectively mitigates shadow-induced errors, enhancing feature detection.
• This approach offers a valuable tool for researchers and practitioners involved in river landscape monitoring.