论文部分内容阅读
为提高驾驶员注视特征辨识的准确性和可靠性,运用多层信息融合决策方法建立驾驶注视行为模式辨识体系。采用BP神经网络的模式分类技术与Dempster-Shafer证据推理融合决策技术相结合的解决方案,克服了复杂驾驶环境,多作业工况条件,以及注视表现特征的多元性对驾驶员注视行为模式辨识的影响,实现对不同驾驶注视行为模式的判别。研究结果表明:以双眼间的横向宽度信息和嘴巴到双眼连线中点之间的纵向距离信息,构建表征驾驶员面部朝向的“T”特征信息,同时结合驾驶员眼睛闭合度以及驾驶员虹膜—巩膜比例与位置信息,三者共同作为驾驶注视特征的表征参量进行融合分析,实现了对驾驶注视行为表现模式状态的辨识和判别。
In order to improve the accuracy and reliability of driver’s gaze recognition, a multi-layer information fusion decision-making method is used to establish the driving gaze behavior pattern recognition system. The solution combining the pattern classification technology of BP neural network with the Dempster-Shafer evidence inference and fusion decision technology overcomes the problems of complex driving environment, multi-working conditions and multi-characteristics of gaze performance. Influence, realize the discrimination of different driving gazing patterns of behavior. The results show that the information of “T” characterizing the driver’s facial orientation is constructed based on the information of the horizontal width between the eyes and the information of the longitudinal distance between the midpoint of the mouth and the line of the two eyes. Combining the driver’s eyes closed degree and driving Iris - sclera ratio and location information, the three together as a driving gaze characteristics of the characteristics of the parameters of the fusion analysis to achieve the behavior of driving gaze pattern recognition and discrimination.