Aircraft Detection for Remote Sensing Image Based on Bidirectional and Dense Feature Fusion.

Area of Science:

• Computer Vision
• Remote Sensing
• Artificial Intelligence

Background:

• Aircraft detection in remote sensing images is crucial for military applications.
• Existing object detection methods struggle with low accuracy and high missed detection rates for aircraft.
• Sophisticated environments in remote sensing images present challenges for target identification.

Purpose of the Study:

• To develop an advanced object detection method for aircraft in remote sensing imagery.
• To address the limitations of current methods, specifically low detection accuracy and high missed detection rates.
• To improve the identification of aircraft targets in complex remote sensing environments.

Main Methods:

• A novel object detection method based on bidirectional and dense feature fusion is proposed.
• The method enhances the YOLOv3 detection framework by incorporating a feature fusion module.
• This module enriches feature maps by integrating shallow and deep features for better detail representation.

Main Results:

• Experimental results demonstrate significant improvements in detection accuracy and reduction in missed detections.
• The proposed method effectively addresses the challenges posed by complex remote sensing environments.
• A notable increase of 1.57% in Average Precision (AP) for aircraft detection was achieved compared to the standard YOLOv3.

Conclusions:

• The bidirectional and dense feature fusion method offers a superior solution for aircraft detection in remote sensing.
• The enhanced YOLOv3 framework with feature fusion proves effective in improving detection performance.
• This research contributes to more reliable aircraft surveillance and monitoring using remote sensing technology.