PTCDet: advanced UAV imagery target detection.

Area of Science:

• Computer Vision
• Machine Learning
• Remote Sensing

Background:

• Object detection in drone aerial images faces challenges with small objects and complex backgrounds.
• Existing models often struggle with accuracy and robustness in these scenarios.

Purpose of the Study:

• To propose an improved object detection model, Perception and Target Capture Detector (PTCDet), for enhanced accuracy and robustness in drone aerial imagery.
• To address the specific challenges of small object detection and complex backgrounds.

Main Methods:

• Developed the Multiple Feature Extraction Attention (MFEA) module for multidimensional feature map augmentation to improve small object detection.
• Introduced the Weighted Perceptive Field Augmentation (WPFA) module to enhance contextual awareness and feature representation.
• Utilized an Enhanced Scale Fusion Detection (ESFD) module based on multiscale feature fusion to improve detection by generating larger scale feature maps.
• Implemented the Inner Focaler IoU loss (INFL) function to accelerate bounding box regression and improve generalization.

Main Results:

• PTCDet demonstrated superior performance compared to other detection algorithms on three public datasets.
• On the VisDrone dataset, PTCDet achieved improvements of 6.21% in map@0.5 and 4.21% in map@0.5:0.95 compared to the YOLOv8 baseline.
• The model showed excellent performance in handling complex backgrounds and detecting small objects.

Conclusions:

• PTCDet provides an effective and robust solution for object detection in drone aerial images.
• The proposed modules (MFEA, WPFA, ESFD) and INFL function significantly contribute to improved detection accuracy and generalization.
• PTCDet offers a promising advancement for aerial surveillance and analysis applications.