Detection of Small Aerial Object Using Random Projection Feature With Region Clustering

Area of Science:

• Computer Vision
• Remote Sensing
• Artificial Intelligence

Background:

• Existing moving object detection methods struggle with small aerial objects against dynamic backgrounds.
• Accurate detection of small aerial objects is crucial for applications like remote sensing and early warning systems.

Purpose of the Study:

• To propose a novel method for accurate small aerial object detection.
• To address the limitations of current techniques in distinguishing small objects from complex backgrounds.

Main Methods:

• Block segmentation to reduce frame redundancy.
• Random Projection Feature (RPF) for block characterization.
• Moving direction estimation to filter dominant motion.
• Variable Search Region Clustering (VSRC) with color features for pixelwise target extraction.

Main Results:

• The proposed method demonstrates superior performance in detecting small aerial objects.
• Effective handling of dynamic backgrounds and scale variations was observed.
• Outperforms state-of-the-art methods in maintaining object integrity.

Conclusions:

• The novel method provides a robust solution for small aerial object detection.
• The approach enhances accuracy in challenging scenarios with dynamic backgrounds and scale variations.