基于视差信息的多重运动图像轮廓提取仿真  

Simulation of Contour Extraction From Multiple Moving Images Based on Parallax Information

作　　者：张建丰[1] ZHANG Jian-feng(Physical Education College,Zhengzhou University,Zhengzhou Henan 450044,China)

机构地区：[1]郑州大学体育学院,河南郑州450044

出　　处：《计算机仿真》2020年第9期214-217,共4页Computer Simulation

基　　金：河南省教育厅高等学校重点科研项目软科学计划(16A890005);河南省科技厅2016年软科学研究项目(162400410518)。

摘　　要：对运动目标图像进行边缘轮廓特征提取,能够提高运动图像的电子稳像性能,从而准确实现运动目标的检测和识别。但在目标轮廓特征提取过程中存在操作复杂、误差大等问题,所以提出了基于视差信息的多重运动图像轮廓特征提取方法。所提方法在传统Snake模型的基础上,通过不同控制点与中心之间的视差作为主动轮廓模型塑造收缩力和膨胀力,能够有效辅助初始边缘轮廓曲线逐渐向真实边缘轮廓逼近。对初次边缘轮廓进行二次提取,并将其作为模型初始轮廓,能够实现目标凹陷区域图像轮廓的有效收敛。仿真结果表明,所提方法具有较强的适用性以及鲁棒性。The edge contour feature extraction of motion image can improve the performance of electronic image stabilization and realize the detection and recognition of moving object accurately.However,the process of target con⁃tour feature extraction is complex,resulting in large error.Therefore,a method to extract the contour features of mul⁃tiple motion images based on parallax information was proposed.Based on traditional Snake model,the proposed method used the parallax between different control points and center points as the active contour model to mould the contraction force and the expansion force,which effectively assisted the initial edge contour curve to gradually ap⁃proach the real edge contour.Moreover,the initial edge contour was extracted as the initial contour of model,so that effective convergence of image contours in target concave region image.Simulation results show that the proposed method has strong applicability and robustness.

关 键 词：视差信息 多重运动图像 轮廓特征提取 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]