SS-OPDet: A Semi-Supervised Open-Set Detection Framework for Dead Pine Wood Detection.

Area of Science:

• Forestry and Environmental Science
• Computer Vision and Machine Learning
• Plant Pathology

Background:

• Pine wilt disease is a global threat to pine forests, requiring effective detection of affected trees.
• Traditional supervised methods struggle with unknown objects, leading to false positives and high annotation costs.
• Accurate detection of dead pine wood is essential for disease control and forest management.

Purpose of the Study:

• To develop an efficient and accurate semi-supervised open-set detection framework for identifying dead pine wood.
• To overcome limitations of traditional methods in handling unknown interfering objects and reduce annotation burden.
• To improve the detection of pine wilt disease for better forest monitoring.

Main Methods:

• Proposed SS-OPDet, a semi-supervised open-set detection framework utilizing labeled and unlabeled data.
• Integrated a Weighted Multi-scale Feature Fusion module for enhanced feature representation.
• Implemented a Dynamic Confidence Pseudo-Label Generation strategy to reduce noise and leverage unlabeled data.

Main Results:

• SS-OPDet achieved 84.73% average precision (APK) and 94.48% recall (RK) on 7733 UAV images.
• Demonstrated low Absolute Open-Set Error (AOSE) of 271 and Wilderness Impact (WI) of 0.0917%.
• Cross-region validation confirmed the framework's robustness and generalization capabilities.

Conclusions:

• The SS-OPDet framework offers a cost-effective and accurate solution for detecting dead pine wood.
• This method significantly improves timely detection of pine wilt disease.
• Provides substantial benefits for forest monitoring and management strategies.