摘要
精确地消除活动阴影对运动目标的影响是智能视频监控的核心任务之一,针对当前运动阴影检测中采用的纹理信息过于粗糙、阈值选取需要人工干涉等问题,通过对NCC(归一化互相关)纹理算法进行改进,并结合亮度和归一化颜色特性,提出一种自适应的运动阴影检测方法。以混合高斯模型得到的前景像素为基础,通过阴影在亮度和归一化颜色的特性筛选出候选的阴影区域,结合改进的纹理算法进一步缩小阴影区域范围,最后利用空间后处理得到真实阴影。实验结果表明,该算法在有效降低噪声干扰的情况下能够较好区分局部纹理不明显的运动目标和阴影。
It is one of the core goals to accurately eliminate the effect of shadow on video moving target in an intelligent video monitoring. Aimed at the problem of the current moving cast shadow detection using texture information is excessively rough and requires manual intervention to selection the threshold, an adaptive elimination algorithm based on NCC （Normalized Cross-correlation） merging intensity and normalized color is proposed in this pa- per. Based on the foreground pixels which is obtained with the Gaussian mixture model, the brief steps to reach the goal are as follows： the candidate shadow region is selected through the feature of shadow on intensity and normalized color, then the shadow region is narrowed with the improved algorithm above and finally the real shadow is obtained by spatial analysis. The results show that the algorithm can effectively reduce the noise and then better dis- tinguish the moving objects and shadows which possess of unconspicuous local texture.
出处
《电视技术》
北大核心
2014年第7期178-181,共4页
Video Engineering
基金
国家自然科学基金项目(61162020)
宁夏自然科学基金项目(NZ1138)
关键词
运动阴影检测
归一化颜色特性
纹理
智能视频监控
moving cast shadow detection
normalized color feature
texture
intelligent video monitoring
