一种快速的目标骨架提取算法  

Fast target skeleton extraction algorithm

作　　者：王松伟[1] 李言俊[1] 张科[1] 王征[1] 

机构地区：[1]西北工业大学航天学院,陕西西安710072

出　　处：《红外与激光工程》2009年第4期731-736,共6页Infrared and Laser Engineering

基　　金：航天支撑基金资助项目(N7CH0009);航空支撑科技基金资助项目(07c53007);航空科学基金资助项目(20070153005)

摘　　要：提出了一种已知骨架点邻域骨架点的判别准则,首先确定邻域骨架点的大致位置,根据图像离散化的特性在大致位置附近进行骨架点的精确判断,据此可以直接判别初始骨架点邻域的骨架点,以此生成目标的整个骨架结构。同时利用曲线演化与最大圆弦长的方法进行骨架的多尺度控制。算例研究表明:该算法保证了目标骨架的连通性与单像素性,而对骨架的多尺度控制能较好地表示图像中的视觉重要成分,且算法的运算复杂度是骨架点数级的,大大提高了运算速度,为将目标骨架应用于实时目标识别提供了支持。An adjacent skeleton point discriminant criterion with known skeleton point was proposed According to the criterion, the adjacent skeleton point of an initial skeleton point could be discriminated by the following two-stage method. Firstly, the rough positions of skeleton points could be determined. Then, the accurate position for skeleton point in the neighbor of the initial point could be judged by utilizing the discrete feature of digital image. In this way, the whole skeleton structure of target could be acquired. Meanwhile, curve evolution and the maximum disk methods were used to fulfill the multi-scale control for the skeleton. Experimental results show that the connectivity and 1-pixel thinning of the skeleton are guaranteed; the visual important part can be represented to meet human vision by the multi-scale control of skeleton; the complexity of the proposed algorithm is proportional to the number of the skeleton points that make it possible for the applications of skeleton based methods in real-time target recognition

关 键 词：骨架提取 边界 视觉重要成分 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]