基于DTW的交警指挥手势识别方法

发布时间：2019-02-14 08:07

【摘要】：对于日益成熟的无人驾驶技术,如何快速准确地识别交警的手势成为无人驾驶领域中一个重要的研究内容。提出一种基于DTW算法的交警指挥手势识别方法。使用Kinect传感器获取人体关节点数据,并根据交警手势特点进行预处理,建立训练模板库;深入分析了该模板库具备的两个特征,即类内高内聚性以及类间低耦合性,由此给出交警指挥手势识别的实现步骤和算法。实验结果表明,该方法能自动识别各种交警指挥手势,具有识别准确度较高、实时性较强、稳定性好的特点。
[Abstract]:For the increasingly mature driverless technology, how to quickly and accurately recognize the hand gestures of traffic police becomes an important research content in the field of driverless. A signal recognition method for traffic police command based on DTW algorithm is proposed. The Kinect sensor is used to obtain the data of the human body, and the training template library is established according to the characteristics of the traffic police gesture. In this paper, two characteristics of the template library, that is, high cohesion in class and low coupling between classes, are analyzed, and the steps and algorithms of signal recognition for traffic police command are given. The experimental results show that the method can recognize all kinds of traffic police command gestures automatically, and has the characteristics of high recognition accuracy, high real-time performance and good stability.
【作者单位】：南京财经大学信息工程学院;
【基金】：国家科技部科技支撑项目(BAH29F01) 国家自然科学基金资助项目(60802087)
【分类号】：TP391.41

【相似文献】