关键词: 无人机, 遥感图像, 低空经济, 深度学习, 铁路外部环境, 未知风险目标检测

Abstract: Common hazards as well as unknown risks (including mudslides, rockfalls, animal intrusion, etc.) in the external environment of railroad seriously threaten the safe operation of railroads, requiring frequent time-consuming and laborious inspections by inspectors, but the scope of inspections is still very limited. At present, the low-altitude economy has become the country's new quality productivity representatives, and the UAV has innate inspection advantages of high altitude, long distance, and small influence by the terrain and railroad maintenance skylight. Aiming at the challenge of sparse unknown risk samples with random uncertainty in the external environment of the railroad, this paper utilizes UAVs for remote sensing image acquisition along the line and proposes an unknown risk detection framework based on Faster R-CNN. First, a novel objectness and multi-classification decoupling training strategy is designed and integrated in the unknown risk detection framework, which significantly improves the generalized object detection performance and avoids misclassifying unknown risk objects as background. Second, the virtual feature synthesis method of VOS has been improved, and similarity-based feature space sampling has been designed to obtain a generalized unknown risk object feature representation by performing multivariate Gaussian distribution parameter estimation and resampling based on the construction of instance-level object feature space. Subsequently, an energy-based uncertainty measurement is utilized to measure the uncertainty of the instance-level features, and losses are calculated accordingly to induce the network to optimize the decision boundaries for common and unknown risk categories. Finally, quantitative and qualitative experimental analysis are conducted on the collected external environment of railroad dataset, open-source drone dataset, and generalization test data. The proposed method achieved 95.7% mAP50 in common hazard identification, while achieving 98% and 80.8% Recall50 in the test set and generalization test data, respectively. The experimental results show that the proposed method has high detection ability for unknown risk objects while ensuring high recognition rate of common hazard categories.

Key words: UAVs, Remote Sensing Imagery, Low-altitude Economy, Deep Learning, External Environment of Railroad, Un-known Risk Object Detection