三维目标离面旋转和非线性光照识别研究  

Research of three-dimensional object recognition under out-of-plane rotation and nonlinear illumination

作　　者：鲍毅[1] 冯少彤[1] 聂守平[1] 王亮[1] 季瑾[1] 

机构地区：[1]南京师范大学江苏省光电技术重点实验室,南京210097

出　　处：《仪器仪表学报》2007年第8期1440-1444,共5页Chinese Journal of Scientific Instrument

基　　金：江苏省自然科学基金(BK2006726);江苏省高校自然科学基金(06KJB140062)资助项目

摘　　要：针对三维目标在离面旋转以及非线性光照条件下的识别问题,采集了180幅不同旋转角度的图像作为训练集,利用主分量分析法选取了20个特征向量,显著降低了特征空间的维数。利用这些特征向量对图像进行分解和重构,在保持较高计算精度的前提下减少了计算的复杂性。提出了利用原始图像向量与重构图像向量夹角余弦值来判别目标真假的方法。模拟结果表明,该方法能有效识别不同旋转角度的目标,同时能够消除非线性光照对目标识别的影响。In order to recognize 3D object under out-of-plane rotation and nonlinear illumination, 180 images with different rotation angles were taken as the training set of the images. Using principal component analysis method, twenty feature vectors were selected, which greatly reduces the dimension number of the feature space Image decomposition and reconstruction were carried out based on these feature vectors, and calculation complexity was decreased while higher calculation accuracy was retained. The object recognition rule based on the cosine value of the angle between original image vector and reconstructed image vector is proposed. Computer simulation results show that the proposed method can effectively recognize the object with different rotation angles and eliminate the influence of nonlinear illumination on object recognition.

关 键 词：目标识别 主分量分析 三维目标 特征向量 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]