摘要
三维物体的快速准确识别是研究的热点.根据局部特征变换的特点,提出了不变角度轮廓线的识别方法.算法通过点云矢量特征对物体进行局部分割,利用欧式距离、测地距离以及角度三个变量,建立其不变角度轮廓特征描述,进一步提取不变矩特征,构建特征向量数据库集.被识别物体的特征描述和数据库中特征进行夹角余弦匹配,可完成物体的识别.通过识别实验以及识别算法性能分析,结果表明算法具有较高的识别率和识别效率,可以用于复杂点云物体识别.
The fast and accurate identification of three-dimensional objects is a hot research topic. Ac- cording to the characteristics of local feature transform, one new method called the invariant angle con- tour line was proposed. Local segmentation of the objects was realized by point cloud vector feature, and then using Euclidean distance, measuring the distance and angle of the three variables, the invariant point outline feature description was established, after further extracting invariant moments feature, the feature vector database would be constructed. Through the cosine matching between feature description of the object and database, the object recognition can be accomplished by comparing the threshold value. The recognition experiment results show that the algorithm has a high recognition rate and recognition efficiency, and can be used for complex point cloud objects recognition.
出处
《四川大学学报（自然科学版）》
CAS
CSCD
北大核心
2017年第4期759-763,共5页
Journal of Sichuan University(Natural Science Edition)
基金
国家自然科学基金民航联合基金(U1633127)
民航局科技项目(20150215)
四川省科技项目基金(2015JY0188)
关键词
三维物体
识别
不变角度轮廓线
不变矩
3D object
Recognition
Invariant angle contour
Invariant moments
