针对目前常用的人眼检测方法误判率高,计算量大的特点,提出了一种基于多次分割的机器视觉人眼检测方法,进行疲劳驾驶检测研究.首先,将图像由RGB颜色空间转换至HSL空间,设定HSL空间分割阈值得到人脸初分割图像(第一次分割),经过滤波,膨胀、腐蚀等操作后与源图像进行掩膜.然后,按照初分割方法再进行第二次和第三次人眼分割;最后利用统计双眼面积和的方法实现疲劳检测.动态实验中,人眼识别准确率为93％,疲劳检测准确率为90％,表明该方法能较好地解决复杂背景中人限定位问题,准确率较高、速度快,算法移植性强.%In allusion to the disadvantages of the methods we usually use, The multiple segmentation method, which can detect the eyes in effect, is used in the fatigue determination. Firstly, after converting the RGB image to HSL color space, the binary image was gained by image segmentation, dilate and erode. Then, after masking with the source image, the second and third image segmentation is completed by the same method. Finally, fatigue determination iss achieved by the area of the eyes. In the online test, the discrimination of eyes is 93% , and the discrimination of fatigue determination is 90%.
展开▼
