轮廓随机序列模型及其分级检测技术  

Multi-step detection of moving object based on stochastic sequence model

作　　者：任明艺[1] 李在铭[1] 

机构地区：[1]电子科技大学通信与信息工程学院,成都610054

出　　处：《计算机应用研究》2008年第1期292-293,299,共3页Application Research of Computers

基　　金：国家"863"计划资助项目(2004AA823120);国家自然科学基金资助项目(10376005)

摘　　要：研究了运动目标图像随机轮廓模型,它包含四特征模型和三参数非平稳随机序列描述,进而拟订了轮廓检测定理;然后建立了轮廓分级检测系统,根据轮廓分割了目标图像。系统包含二阶时差分变换、全域自学习的高信噪比轮廓点二元聚类检测;中信噪比轮廓点自学习的局域检测;在时空域基于封闭和Markov关联准则的低信噪比轮廓点检测。实验仿真给出了良好的结果。Stochastic contour model of moving objects, which including four-characteristic model and three-parameter non-stationary stochastic sequence description was studied. Then the contour detection theorem was proposed. A multi-step contour detection system was established and the moving object was segmented by it, The detection system included binary clustering detection of high-SNR contour points based on second-order temporal deference transform and universal self-learning, medium- SNR points detection by local self-learning and Iow-SNR points detection based on closed rule and Markov correlative rule. The experiment emulation shows a good result.

关 键 词：运动目标 轮廓随机序列模型 图像分割 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]