摘要
为了简化三维物体的识别过程,提高三维物体识别的识别率,该文利用Multi-scale autoconvolution、Trace变换、Zernike矩3种仿射不变性特征,对飞机、汽车、人等三维物体进行视点空间划分,用尽可能少的不等间隔的三维物体的二维投影图像来表达三维物体,并以此为依据进行三维物体识别。在此基础上提出一种针对不同类型物体的仿射不变性特征提取策略,并建立一个实现三维物体任意姿态识别的软件系统平台,应用Princeton形状标准库中的部分模型对该平台进行测试。结果表明,该方法能够取得较好的识别效果,识别率在90%以上。
The viewpoint space of 3-D objects, such as aircraft, automobiles, and humans is partitioned using three affine invariant features based on a multi-scale autoconvolution, a trace transform, and the Zernike moment to express the 3-D objects within minimum two-dimensional projective images. The images are then nonuniformly projected for object recognition to simplify the 3-D recognition process and increase the recognition rate of 3-D objects. The paper presents an affine invariant feature extraction strategy for different 3-D objects that can recognize arbitrary postures of the 3-D object. Tests with models from the Princeton shape benchmark show that the good recognition effect can be achieved by the proposed method and the recognition rate is great than 90%.
出处
《清华大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2009年第1期53-56,共4页
Journal of Tsinghua University(Science and Technology)
基金
国家自然科学基金资助项目(60502013)
国家"八六三"高技术项目(2006AA01Z115)
关键词
模式识别
三维识别
视点空间划分
仿射不变性特征
pattern recognition
3-D object recognition
viewpointspace partition
affine invariant feature