摘要
轮廓是运动目标检测及识别的重要依据。针对视频序列,采用一种基于改进Snake模型的方法对运动目标进行轮廓提取。首先通过部分MPP算法与外接矩形结合的方法,实现模型初始轮廓线控制点的自动设置;进而改进梯度能量函数,引入梯度矢量流算子,使得Snake轮廓能较好收敛于目标凹陷部位;并根据实际情况调整其他能量函数项的权重,最后通过8邻域贪心算法求得总能量最小化时的目标轮廓线。实验证明,该算法能对视频序列中的目标轮廓实现较可靠的提取。
Contour forms an important basis for motion detection and recognition. According to the video sequence ,this paper adopts a method based on novel snake model to extract the contour of motion object . Firstly ,the model controlling points of initialized contour are obtained by a method which combines MPP algorithm and circumscribed rectangle algorithm ; secondly ,it improves the gra- dient energy function ,and introduces gradient vector operator to enable the model to converge to boundary concavities of the ...
出处
《微计算机信息》
北大核心
2008年第10期293-295,共3页
Control & Automation
基金
国家自然科学基金资助项目(基于几何散射理论提取高距离分辨雷达目标特征方法研究-60272059)
