欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
分析了传统MeanShift方法中，特征模型表述缺乏像素点空间位置信息的不足，提出了一种改进特征模型表述方法以解决这个问题。首先对图像模板区域进行分块处理，分别计算每个子块内像素点的空间分布和颜色分布情况，然后以此构建联合特征空间。并且，采用同样的方式计算当前帧候选目标区域的特征模型表述。然后，采用Bhatta-charyya距离的负对数形式，进行特征相似性度量，以获得目标在当前帧的新位置。实验结果表明，提出的跟踪算法是可行的，且具有更好的鲁棒性和跟踪精度，在目标发生旋转和尺度变化及背景区域有相似颜色干扰情况下，取得了更准确的跟踪结果。
英文摘要：
Abstract：After analyzing the shortcoming that lack of pixels＇ spatial information in feature space representation of classical Mean Shift,we proposed an novel method to cope with it. At first, segment the model region of the image into several squares,and compute the spatial distribution and color distribution of every pixel in each square. Then, the joint feature space is constituted. And do the same for candidate target region of the current frame. After that, a new similarity measure method is adopted to search the new location of target. The results show that, the improved method is feasible, and can track the object more robust, accurate and quickly. While there contains rotation, scale change and similar color in the background, the results is more accurate.
同期刊论文项目
同项目期刊论文
Analysis of harmonic oscillation synchronization for the single-mass nonlinear system under harmonic
Variable structure synchronization control for the single-mass nonlinear sharp resonance vibration m
期刊信息
位置：