基于图像纹理分析的动态车辆识别方法研究

Study on Dynamic Vehicle Identification Method Based on Image Texture Analysis

在智能交通系统中,动态图像识别技术是系统应用的基础核心技术之一。以应用于交通监控、智能驾驶系统等场景的HSV空间动态车辆识别为基础,研究并论证提出了新的检测识别方法,实现对运动车辆的检测识别、目标追踪、驾驶辅助等功能。研究问题的难点是,如何从复杂的背景中分割运动物体,是检测方法能否有效的至关重要的一步,在研究了目前存在的各种方法之后,提出了一种新的基于阴影检测的HSV空间自适应背景模型的车辆追踪检测算法,算法基于HSV空间图像处理,采用最大类间方差法获取相邻帧二值化阈值,利用纹理信息进一步确定动态图像以及确认图像范围。通过截取由监控系统获取的视频信息,并对其进行图像处理检测车辆移动轨迹。从监控视频信息中获取两帧不同时刻的图像信息,在HSV空间进行相邻帧检测。由于阈值的选择将直接影响判断精度,本研究将固定阈值法进行了改进,该阈值是通过统计模型对整幅图像上灰度值进行计算,并通过最大类间方差法确定阈值。最后经过实际视频图像验证,仿真试验流程清晰,试验结果达到预期设想。

In intelligent Transport Systems,dynamic image recognition technology is one of the basic core technologies of system application. Dynamic vehicle identification（ In HSV space） is applied to the scenes such as traffic monitoring,intelligent driving system. On this basis,a new detection identification method is studied,proved and puts forward to realize the functions of moving vehicle detection and recognition,target tracking,driver assistance and so on. The difficulty of the study is how to segment moving objects from the complex background,which is a crucial step for an effective detection method. After studying the various existing methods,the vehicle tracking detection algorithm for a new HSV space self-adaptive background model based on shadow detection is proposed. Based on HSV space image processing,the algorithm uses maximum between-class variance to obtain the binarized threshold of adjacent frames,and then determine the dynamic image and confirm the image scope by using texture information. The moving trajectory of vehicle is detected by intercepting the video information obtained by the monitoring system and processing the image.The image information of 2 frames at different time is obtained from surveillance video information,and the adjacent frame detection in HSV space is conducted. Because the selection of threshold value will directly affect the judgment accuracy,the fixed threshold method is improved,in which the threshold value is calculated through calculating the gray levels on the whole image,and determined by maximum between-class variance. Finally,the actual video image verification shows that the flowchart of simulation test is clear,andthe experimental result meets expectation.