CAJ | 학술논문

为解决传统智能交通系统中由于阴影检测算法的精确性导致错误提取交通参数的问题，提出一种基于YCbCr颜色空间和局部傅里叶变换的阴影检测方法，通过分析阴影区域和背景区域在YCbCr颜色空间的不同特性后，综合考虑YCb-Cr颜色空间的亮度和色度信息确定候选阴影区域，利用局部傅里叶变换系数各偶数阶矩的纹理差异及加权交叉熵作为纹理相似度的度量方法获得阴影区域和背景区域的优化分割。实验表明，该方法阴影检测的有效率在85.3%以上，并且能很好的满足实时场景要求，为后续交通参数的提取等工作奠定了良好的基础。
위해결전통지능교통계통중유우음영검측산법적정학성도치착오제취교통삼수적문제，제출일충기우YCbCr안색공간화국부부리협변환적음영검측방법，통과분석음영구역화배경구역재YCbCr안색공간적불동특성후，종합고필YCb-Cr안색공간적량도화색도신식학정후선음영구역，이용국부부리협변환계수각우수계구적문리차이급가권교차적작위문리상사도적도량방법획득음영구역화배경구역적우화분할。실험표명，해방법음영검측적유효솔재85.3%이상，병차능흔호적만족실시장경요구，위후속교통삼수적제취등공작전정료량호적기출。
In order to solve the difficulties of extracting the wrong traffic parameters that caused by the accuracy of shadow de-tecting methods in traditional intelligence traffic system,a method of shadow detection and removal is proposed based on YCbCr color space and Local Fourier transform ( LFT ). Firstly, analyse the intensity and chromacity characteristics of the YCbCr color space to obtain the rough regions of the shadows, Secondly, use the texture discrimination performance of LFT coefficients even moments and weighted cross-entropy method for moving shadow removal exactly in video sequences. Experiments on different scenes suggest that effective detection rate of proposed method is over 85.3 percent, and can satisfy the requirement of real-time processing, and set a good foundation for the extracting of traffic parameters.