一种高准确率的交通监控视频车辆检测算法  

A High Accuracy and Low Compute Cost Approach in Traffic Video Vehicles Classification

作　　者：李乐平[1] 高杨[2] 

机构地区：[1]内江职业技术学院,四川内江641000 [2]南方医科大学,广州510515

出　　处：《计算机测量与控制》2015年第3期852-854,857,共4页Computer Measurement &Control

基　　金：四川省科技厅支撑计划项目(2013GZ0030)

摘　　要：针对高速视频监控的车辆识别易受几何变换影响且准确度不高的问题,提出了高效的阴影去除和车辆分类两种视频处理算法来提高车辆识别性能;利用高斯分布像素匹配实现车辆形状的提取;利用阴影区域前景和背景的对比度参数进行有效的阴影去除;利用Hu不变矩表征车辆,从而克服了几何变换的影响;使用7D欧氏特征空间最小距离算法将车辆与指定车辆模型进行比较并分类;对比试验结果证明,本算法具有较高的阴影检测性能和极佳的车辆分类性能,明显优于其他同类型算法。For the problem that highway video traffic surveillance is easily influenced by Geometric transformation and has low accuracy, proposed two approaches which contain shadow elimination and vehicles classification to improve the vehicles detect performance. Realize vehicles shape abstract with Gaussian distribution and pixel match; for the shadow elimination with the parameter contrast compare to realize the shadow elimination efficiently; use the Hu moments to describe the vehicle and overcome the influence of geometric transformation; classify the vehicles by compare with the typical vehicle model with 7-D Euclidean feature space distance. The simulating results prove that the proposed approach has better performance of shadow detective and vehicles classification than the same category approaches.

关 键 词：高速公路 阴影去除 车辆分类 Hu不变矩阵 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]