飞机结构X射线裂纹图像智能评定

飞机结构X 射线图像评定过程存在复杂背景下裂纹分割不准、检出难等问题。基于高效层聚合网络提出一种飞机结构X 射线裂纹图像智能评定模型（ELAN-Seg），将ELAN-Seg 模型和DeepLabv3+模型的射线图像裂纹分割能力进行对比，结合图像处理技术对模型分割的裂纹长度进行评估，利用飞机强度试验及外场维护过程采集的X 射线图像对模型进行验证。结果表明：分割的最小裂纹长度约为3 mm，ELAN-Seg 模型对复杂背景射线图像裂纹分割更加准确，裂纹漏检率小于3.8%，该模型具有工程适用性。

Intelligent evaluation of X-ray crack image of aircraft structure

Aiming at the problems of inaccurate crack segmentation and difficult detection under complex background in the process of aircraft structure X-ray image evaluation, an intelligent evaluation model ELAN-Seg for X-ray crack images of aircraft structures is proposed based on efficient layer aggregation network. The backbone network of the model is stacked by the efficient layer aggregation network. The low-level and high-level features extracted from the backbone network are enhanced by using SPPCSPC spatial pyramid and ASPP spatial pyramid respectively, so as to realize multi-level and multi-scale feature extraction of input features. The attention mechanism is embedded after the enhanced feature extraction, and the low-level and high-level features after the enhanced feature extraction are fused, so as to realize the intelligent segmentation of the crack area of the X-ray image. The model was verified by X-ray images obtained from aircraft strength test and field operation process, and the crack leakage rate was less than 3.8%, indicating that the method has engineering applicability.