基于DSP的疲劳驾驶实时检测算法实现

Implementation of the Real-time Driver fatigue Detection Algorithm based on DSP

针对基于DSP的疲劳驾驶实时检测系统,利用Adaboost算法训练人脸与人眼分类器,在分析了基于Hough找圆法与灰度投影法的人眼状态分析算法各自的优缺点后,提出了一种新的基于区域灰度特征的人眼状态分析算法,该算法不需要精确几何模型,利用基于区域特征的灰度均值,具有很强的鲁棒性。将疲劳驾驶检测算法移植到DSP中后,检测算法的帧速率达到18帧/秒,满足了实时检测的性能要求。

In terms of the real - time driver fatigue detection system based on DSP, the paper proposes a new human eye state analysis algorithm based on the regional gray feature after analyzing the advantages and disadvantages of the human eye state analysis algorithm based on the Hough transform method for circle finding or the Gray scale projection method by using the classifier of human face and eye training with the Adaboost algorithm. The new algorithm does not need an accurate geometric module and it has highly robustness by using the gray scale based on statistical average. When the DSP is used in the driver fatigue detection algorithm, the frame rate of the detection algorithm reaches to 18 frames/see. The algorithm can meet the needs of real - time detection.