昆虫翅膀图像特征的亚像素级提取方法  

Extraction of subpixel insect wing features in images

作　　者：王颖[1] 张广军[1] 陈大志[1] 

机构地区：[1]北京航空航天大学仪器科学与光电工程学院,北京100083

出　　处：《北京航空航天大学学报》2005年第8期839-842,共4页Journal of Beijing University of Aeronautics and Astronautics

基　　金：国家杰出青年科学基金资助项目 (5 0 12 5 5 18)

摘　　要：昆虫翅脉、翅脉交点及翅膀边缘是昆虫翅膀运动变形三维重构的主要依据.通过分析这三种特征的图像特点,对于翅脉中心,采用Hessian矩阵求出灰度极值点方向,并用二次曲线拟合该方向的灰度变化,通过确定该曲线极值点位置来提取亚象素翅脉中心;然后通过求灰度图像的梯度,在其梯度变化方向求极值点的方法提取翅膀边缘特征;此外根据翅脉方向变化的连续性提出了邻域角度约束方法确定翅脉交叉点,从而最终提取到亚像素级的翅膀边缘、翅脉及交叉点等特征.实际应用证明该方法是较为有效的.Wing vein, junction of veins and wing edges are the fundamental image features for wings 3D reconstruction to obtain insect deformations motion parameters. To determine precise position of the wing features in images, gray properties of wing features were analyzed first. For wing vein features, the gray extreme direction was determined by Hessian matrix. The gray variable was fitted by conic curve along the direction of the gray extreme, so the subpixel location of the vein center was obtained by searching the extremum of the curve. For wing edges, image gray gradient was calculated, and then subpixel step edge of wings was determined by searching the gradient extremum along the gradient change direction. Angle constraint in neighbor was used to determine junctions of wing veins based on the change continuity of vein direction. All wing features were determined at subpixel locations. The experiment results show that the extracting method about wing features is effective.

关 键 词：特征提取 图像处理 昆虫翅膀 边缘 角度约束 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]