摘要
为完善我国城市道路交通流参数调查的技术手段,提高数据统计分析的效率与精度,建立一种基于无人机航拍与图像处理的调查与识别方法。以大疆Phantom 4 Pro无人机与数码相机为基础,搭建无人机航拍平台,实现交通流图像的有效采集。在Visual Studio 2017集成编译环境下,依托Open CV开源库,提出车辆识别跟踪与测距算法,通过图像裁剪排除干扰因子,应用cvt Color函数完成灰度化处理,采用中值滤波法实现图像的降噪;引入Canny算子检测车辆边缘,通过Otsu算法进行二值化处理,使用腐蚀操作消除干扰碎片;根据车辆长宽比与面积建立车辆外接矩形,并储存其中心点,提出基于双循环遍历与冒泡排序的车道标定算法,实现车距的提取与车流量、平均速度和车流密度等参数的测算。经检验,该方法调查与识别准确率能够达到95%以上,相比于传统调查方法,具有简单和便捷等优势。
In order to improve the technical means of the investigation of traffic flow parameters of urban roads in China to improve the efficiency and accuracy of data statistics and analysis, an investigation and recognition method based on UAV aerial photography and image processing was built. Based on Dajiang Phantom 4 Pro UAV and digital camera, a UAV aerial platform was built, which realized the effective acquisition of traffic flow image. In the integrated compilation environment of Visual Studio 2017, relying on the open source library of OpenCV, a vehicle recognition, tracking and ranging algorithm was proposed. The algorithm eliminated interference factors through image cutting, invoked the function of cvtColor to finish grayscale processing, adopted median filter to realize image denoising, introduced the Canny operator to detect the edge of vehicle, realized binarization through the Otsu algorithm and eliminated interference fragments using corrosion operation. According to the vehicle aspect ratio and area, the vehicle external rectangles were established and the center points were stored. An algorithm of lane calibration was proposed based on double loop traversal and bubble sorting method, such that the vehicle distance could be extracted and the parameters of vehicle flow, average speed and vehicle density could be calculated. Through examination, the accuracy of this method can reach over 95%, which is more simple and convenient in comparing with the traditional method.
作者
李士尧
鲁元博
裴玉龙
LI Shiyao;LU Yuanbo;PEI Yulong(School of Traffic and Transportation,Northeast Forestry University,Harbin 150040)
出处
《森林工程》
2018年第5期72-77,共6页
Forest Engineering
基金
国家自然科学基金面上项目(51278158)
国家级大学生创新训练项目(201710225086)
