民机复合材料目视检查中的眼动与疲劳检测

疲劳是导致民机复合材料结构目视检查差错的重要诱因，疲劳检测对于减少人的差错，保障飞行安全具有重要意义。对基于眼动行为的疲劳度量与检测方法进行了研究，建立了民机复合材料结构目视检查实验场景，利用Tobii眼动仪提取了正常状态和疲劳状态下的目视检查眼动数据，分析了瞳孔直径、平均注视时间、平均注视频率、平均眼跳时间、平均眼跳频率、注视热点与轨迹和扫视速度等眼动行为与疲劳的关系，进而提取了能表征疲劳的瞳孔直径、平均注视时间和扫视速度3种眼动指标，以该指标构建特征向量，利用支持向量机（SVM）方法构建了目视检查疲劳检测模型。研究发现疲劳状态下的目视检查平均注视时间更长，扫视速度更慢、瞳孔直径减小，右瞳孔减小程度更大，核函数为径向基函数和高斯函数的SVM方法对疲劳的检测效果好。研究结果表明，利用SVM方法训练由瞳孔直径、平均注视时间和扫视速度构成的眼动特征向量能有效检测目视检查中的疲劳状态。

Eye movement and fatigue detection in visual inspection of civil aircraft composite materials

Fatigue is an important cause of human error in visual inspection of civil aircraft composite material components. The measurement of fatigue has great implications on reduction of human error and flight safety. To measure and detect fatigue, an eye movement behavior based method is proposed. The experimental scene of composite material visual inspection is established. Tobii eye tracking is used to extract the eye movement data in the experiment under normal working condition and fatigue condition. The relationships between fatigue and means of pupil diameter, fixation time, fixation frequency, saccadic time, saccadic frequency, fixation heat map and saccadic velocity were analyzed. Then, three kinds of eye movement indexes, namely, pupil diameter, saccadic velocity and average fixation time, which could represent fatigue, are extracted to construct feature vectors for Support Vector Machine (SVM) method to build the fatigue detection model. It is found that the average fixation time is longer, and the saccadic velocity and the pupil diameter are smaller under fatigue in visual inspection, especially the pupil of the right eye. The SVM method with kernel function as radial basis function and gaussian function has a good effect on fatigue detection. The experimental results show that the SVM method using the eye movement feature vector can effectively detect the fatigue state in visual inspection.