摘要
基于视频序列的移动物体提取在各领域有着积极的研究意义。传统方法从灰度域提取物体特征向量往往无法准确给出物体位置等信息。而在色彩域内,可利用多种特征向量进行有效的图像分割。为了提取可视化图形中的阴影区域,提高车辆识别与跟踪的鲁棒型,通过在RGB色彩空间内利用欧式度量法和HSI色彩空间中的特征向量实现上述目的。并建立了统计模型并通过区域增长的理论利用车辆空间信息避免了近似色车辆带来的误判,提高算法的可靠性。实验结果表明,运动物体的阴影可以取得较好地提取结果。
The study of moving object based on video sequence has its positive significance in all areas. Traditional methods usually fail to tell the object information precisely through feature vectors subtracted in gray area. Various vectors can be used within the color area for effective image segmentation. Euclidean distance is computed for RGB color space and feature vectors are constructed in HSI space to realize the shadow segmentation. In order to im- prove the robustness and reliability of the algorithm, a statistic model is formed, and the theory region growing is introduced to avoid the false decisions caused by the vehicles of similar color to the shadow. The result of experiments show that it can figure out the shadow effectively.
出处
《计算机仿真》
CSCD
2008年第10期254-256,271,共4页
Computer Simulation
关键词
阴影分割
欧氏距离
特征向量
统计模型
Shadow segmentation
Euclidean distance
Feature vectors
Statistic model
