基于先验形状信息和水平集方法的车辆检测  

Vehicle detection based on shape priors and level set

作　　者：赵璐[1] 于慧敏[1] 

机构地区：[1]浙江大学信息与电子工程学系,浙江杭州310027

出　　处：《浙江大学学报（工学版）》2010年第1期124-130,共7页Journal of Zhejiang University：Engineering Science

基　　金：国家自然科学基金资助项目(60872069);航天基金资助项目

摘　　要：针对交通视频检测应用,提出一种基于先验形状信息和主动轮廓模型的运动车辆检测方法.算法首先利用颜色信息和边缘信息检测并去除车辆阴影,提取车辆的初始轮廓;为了改善车辆轮廓的提取精度,在进一步的车辆分割中引入车辆形状的先验知识,用水平集符号距离图像的隐含表示建立车辆的先验形状模型,并以先验的车辆形状模型作为约束构造出主动轮廓能量函数;将第一步获得的车辆轮廓作为车辆分割演化曲线的初始轮廓,采用变分法求解能量函数的极小值,利用形状配准和水平集方法演化车辆的分割曲线,得到准确的运动车辆轮廓.将该方法应用于实际采集的交通视频,获得了很好的测试结果.A moving vehicle detection method based on the prior shape knowledge and active contour method was proposed for the application of traffic video detection.First,the shadow was eliminated for obtaining the initial contour of the vehicle using the color and edge information.Then,an implicit shape model with level set signed distant image was built to improve the accuracy of vehicle contour pick-up,and an active contour energy function with the restriction of the existing shape priors was constructed.The obtained vehicle contour was set to initial contour of the vehicle segmentation evolvement contour,and then the minimal value of the energy function could be found by variational method.The precise contour of the vehicle was obtained by applying the shape alignment and level set method to evolving the initial contour.Experimental results of real traffic sequences proved the good effectiveness of the proposed method.

关 键 词：车辆检测 水平集 先验形状 形状配准 阴影检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]