基于面部识别的管制员疲劳风险评估技术研究

发布时间：2018-11-20 21:36

【摘要】：据统计,空中交通管理中13%的运行差错也直接与管制员的疲劳相关;其中,管制工作中约18%的失误与管制员的疲劳有关。因此,有效的实时监测管制员的疲劳状态对保证空管运行安全,提高空管安全水平有非常重大的意义。本论文的主要任务是通过对面部识别关键技术的研究,设计一个基于OpenCV平台的管制员疲劳监测系统,实时监测管制员的疲劳状态。通过摄像头实时获取管制员图像,将所获取的图像信号传送给PC进行处理:1、采用Adaboost算法、级联分类器以及ASM模板匹配算法实现对人脸的快速定位;2、在检测到人脸的基础上,首先选用基于统计学习的人眼检测算法初步确定人眼睛和嘴巴矩形区域,其次采取基于ASM模板匹配的人眼检测算法进一步精细确定人眼和嘴部区域。3、用局部二值化技术和直方图均衡化来增强图强对比度,将检测的人眼和嘴巴图像变为黑白图像,并识别眼睛和嘴巴状态;根据PERCLOS值(眼睛闭合时间占某单位时间的百分比)、平均闭眼时长、打哈欠频率(被测对象某单位时间内总计打哈欠的次数)综合分析计算,判断管制员的疲劳等级;当管制员处于严重疲劳时系统报警。经反复实验证明,管制员疲劳监测系统各模块设计合理、功能完备,尤其对人脸识别、人眼定位、人眼和嘴巴状态识别、疲劳等级判断等基本准确,成功率高达85%,能满足实地监测的需求,系统存储的PERCLOS值等数据对管制员岗位的疲劳风险管理也有一定的参考价值。
[Abstract]:According to statistics, 13% of the operational errors in air traffic management are also directly related to the fatigue of the controller, and about 18% of the errors in the control work are related to the fatigue of the controller. Therefore, effective real-time monitoring of air traffic controllers' fatigue is of great significance to ensure the safety of air traffic control and improve the safety level of air traffic control. The main task of this thesis is to design a controller fatigue monitoring system based on OpenCV platform through the research of the key technology of facial recognition, and monitor the fatigue state of controller in real time. The controller image is acquired by the camera in real time, and the image signal is transmitted to PC for processing. 1. Using Adaboost algorithm, concatenated classifier and ASM template matching algorithm to realize the fast localization of face; 2. On the basis of detecting human face, we first select the human eye detection algorithm based on statistical learning to determine the rectangular region of human eyes and mouth. Secondly, the human eye detection algorithm based on ASM template matching is used to determine the area of human eye and mouth. 3. The local binarization technique and histogram equalization are used to enhance the strong contrast of the image, and the detected human eye and mouth image is turned into black and white image. And recognize the state of the eyes and mouth; According to the PERCLOS value (the percentage of eye closure time to a certain unit time), the average time of closing eyes and the frequency of yawning (the total number of yawning times in a given unit time of measured object), the fatigue grade of the controller was judged. System alarm when controller is seriously tired. After repeated experiments, it is proved that the controller fatigue monitoring system has reasonable design and complete function, especially for human face recognition, human eye location, human eye and mouth state recognition, fatigue grade judgment, etc., and the success rate is as high as 85%. The data stored in the system, such as PERCLOS value, can meet the needs of field monitoring. It also has some reference value for the fatigue risk management of controller post.
【学位授予单位】：中国民航大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：V355.1

【相似文献】