WA-YOLO: An explosive material detection algorithm for blasting sites based on YOLOv8.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Safety Engineering

Background:

• Pyrotechnic detection is crucial for blasting safety but challenged by complex environments and irregular object appearances.
• Existing methods struggle with varying object scales and irregular postures of detonator wires.

Purpose of the Study:

• To develop an advanced object detection algorithm for improved pyrotechnic detection in blasting safety.
• To enhance the accuracy and robustness of detecting small and irregularly shaped targets like detonator wires.

Main Methods:

• Proposed WA-YOLO algorithm integrating wavelet-separable convolution (WSDConv) and a multi-scale parallel attention mechanism.
• Incorporated Wise-IoU loss function to improve bounding box precision for irregular shapes.
• Modified Cross Stage Partial (CSP) structure within the neck network for multi-scale object detection.

Main Results:

• Achieved a 12.6% increase in average precision on a custom blasting dataset, with an 8.3% improvement in detonator detection.
• Demonstrated improved performance on the VOC2012 dataset, showing a 1.3% recall and 1.6% average precision increase.
• The WA-YOLO model exhibits strong generalization across different datasets and complex scenarios.

Conclusions:

• The WA-YOLO algorithm offers an effective solution for pyrotechnic and detonator detection in challenging blasting environments.
• The integration of wavelet convolution and attention mechanisms enhances feature extraction and multi-scale detection capabilities.
• The model's robustness and generalization performance make it suitable for real-world safety applications.