摘要
提出了一种融合线性特征的局部纹理运动车辆阴影检测方法。首先基于连续帧视频图像信息建立初始背景模型;通过背景差法获取包含阴影的运动目标区域,同时依据该运动区域信息实时更新背景;结合亮度信息,利用改进局部二值模式的纹理算子描述运动区域纹理,并根据海明距离进行粗分类,快速检测出运动区域中的阴影覆盖区;进一步对阴影覆盖区域进行纹理信息的线性特性判断,排除车辆自阴影区域,获取背景阴影,得到真实车辆目标。实验结果表明,该方法提高了阴影和车辆自阴影的检测准确度,且速度快,可满足实时性要求。
How to detect the background shadows of the moving vehicles quickly in order to obtain the real vehicles is a tough problem for vehicle detection in video images . We put forward a method of moving shadow detection based on local texture features integrated with linear characteristics . Firstly , the background model can be made on the basis of successive frame video image information , and a moving target area containing shadows can be obtained via the background subtraction method and the background can be updated according to the moving area information in real time . And then the description of texture can be achieved through its brightness and the improved Local Binary Pattern （ LBP ） texture operator, then conduct the rough classification using Hamming distance to detect the shadows areas in the moving area quickly . Further, the linear characteristics analysis on texture information has been conducted on the shadows areas to preclude the self shadows of vehicles and obtain the background , finally obtaining the real vehicle targets . According to the experiment, this method improves the detection accuracy of shadows and self shadows of the vehicles at high rate of efficiency , meeting the requirement of real time .
出处
《微型机与应用》
2013年第14期44-46,50,共4页
Microcomputer & Its Applications
基金
福建省自然科学基金项目(2012J01263)
关键词
线性特性
LBP纹理
运动车辆
阴影检测
linear characteristics
LBP texture
moving vehicle
shadow detection
