【摘要】 为提高车辆运行效率、提升道路交通管理科技信息化水平,提出一种融合深度感知特征的车辆跟踪态势显示。首先在深度感知成像原理基础上构建车辆信息提取模型,根据SIFT算法的局部性、特殊性、多量性和高效性等优势,并通过特征点测量、边缘点剔出与描述,完成车辆匹配,同时将获得的均值模板在态势显示中做特征匹配,利用遗传算法聚类完成初始编码、种群初始化等一系列操作,获取车辆运动规律,建立滤波方程,根据该方程实现对车辆不同方向的跟踪;最后以可视化技术为载体,将车道空间占比率、车辆区域面积、以及跟踪车辆的密集程度等信息完成可视化实现态势显示。仿真结果证明,方法对车辆跟踪的准确性高、态势显示速度快,达到了智慧化与精细化的车辆跟踪要求,实现了实时态势显示的目的。更多还原

【Abstract】 In order to improve the operating efficiency of vehicle and the informatization level of traffic management, a method of vehicle tracking situation display based on depth perception features was put forward. Firstly, the model of vehicle information extraction was built by the imaging principle of depth perception. According to the advantages of SIFT algorithm, such as locality, particularity, multiplicity and efficiency, the vehicle matching was completed through feature point measurement, edge point extraction and description. Meanwhile, the mean-value template was used to match the features in the situation display. And then, genetic algorithm clustering was used to complete the initial coding and population initialization, and thus to obtain the vehicle running law. Based on the filtering equation, the vehicle could be tracked in different directions. Finally, the visualization technology was used as the carrier, so that the information such as the ratio of lane space, the area of vehicle, and the density of tracking vehicles were visualized to achieve the situation display. Simulation shows that the proposed method has high accuracy for vehicle tracking and fast speed of situation display. Meanwhile, this method achieves the intelligent and refined vehicle tracking and the real-time situation display.更多还原