小波矩和HMM的三维CAD模型归类与检索算法
|
摘要
为了在工程应用中检索已有的三维CAD模型,以便重用相应零件的设计信息,节省设计和加工成本,提出一种基于小波矩和仿射不变矩特征融合的隐马尔科夫模型(HMM)三维CAD模型归类与检索算法。对三维模型图进行归一化处理,并分别提取归一化图像的小波矩特征值和仿射不变矩特征值;通过K-W检验算法选择出鲁棒性好、稳定性高的组合不变矩特征并进行编码;构造五类三维模型的样本集,将上述特征值作为HMM的输入观测值,通过修正的添加比例因子的多观测序列Baum-Welch(B-W)算法进行模型的训练与识别。将本文算法与其他三种算法进行实验对比,结果表明,本文所提出的算法具有较好的识别率和检索效率,具有一定的实用价值。
In engineering applications,retrieval of existing 3D CAD models and reuse of their design information can reduce the time and cost of design and product. A classification and retrieval algorithm for three-dimensional CAD model with hidden Markov models(HMM) based on wavelet moment and affine invariant moment is proposed in this study. Firstly,the 3D model image is normalized,and the wavelet moment eigenvalues and affine invariant moment eigenvalues of the normalized image are extracted. Then the combined invariant moment features with high robustness and stability are selected by the K-W test algorithm and encoded. Finally constructing five type samples of three-dimensional models and using the above eigenvalues as input observations of HMM. The modified BaumWelch(B-W) algorithm with scaling and multi-observation sequences is used to train and identify the model. The proposed algorithm is compared with other three algorithms. The results show that the proposed algorithm has better recognition rate and retrieval efficiency,and more practical engineering value.
In engineering applications,retrieval of existing 3D CAD models and reuse of their design information can reduce the time and cost of design and product. A classification and retrieval algorithm for three-dimensional CAD model with hidden Markov models(HMM) based on wavelet moment and affine invariant moment is proposed in this study. Firstly,the 3D model image is normalized,and the wavelet moment eigenvalues and affine invariant moment eigenvalues of the normalized image are extracted. Then the combined invariant moment features with high robustness and stability are selected by the K-W test algorithm and encoded. Finally constructing five type samples of three-dimensional models and using the above eigenvalues as input observations of HMM. The modified BaumWelch(B-W) algorithm with scaling and multi-observation sequences is used to train and identify the model. The proposed algorithm is compared with other three algorithms. The results show that the proposed algorithm has better recognition rate and retrieval efficiency,and more practical engineering value.
引文
[1] Hu M K. Visual pattern recognition by moment invariants[J].IRE Transactions on Information Theory,1962,8(2):179-187
[2] Shen D G,Ip H H S. Discriminative wavelet shape descriptors for recognition of 2-D patterns[J]. Pattern Recognition,1999,32(2):151-165
[3]梅雪,林锦国.基于图像边缘小波矩和支持向量机的目标识别[J].计算机工程与科学,2006,28(7):60-61,69Mei X,Lin J G. Target recognition based on wavelet moment and support vector machines[J]. Computer Engineering&Science,2006,28(7):60-61,69(in Chinese)
[4]崔建江,宋星月,陈国坤,等.基于几何形状和小波矩的静脉特征提取与匹配[J].东北大学学报(自然科学版),2009,30(9):1236-1240Cui J J,Song X Y,Chen G K,et al. Feature extraction and matching of vein based on geometrical shape and wavelet moment[J]. Journal of Northeastern University(Natural Science), 2009,30(9):1236-1240(in Chinese)
[5]于剑,程乾生.模糊聚类方法中的最佳聚类数的搜索范围[J].中国科学:E辑,2002,32(2):274-280Yu J,Cheng Q S. The upper bound of the optimal number of clusters in fuzzy clustering[J]. Science in China:Science E,2002,32(2):274-280(in Chinese)
[6]冯宗翰,吴小俊.基于迹比准则与+L-R方法的特征选择算法[J].计算机工程,2011,37(17):136-139Feng Z H,Wu X J. Feature selection algorithm based on trace ratio criterion and+L-R method[J]. Computer Engineering,2011,37(17):136-139(in Chinese)
[7]李伟红,陈伟民,杨利平,等.基于不同Margin的人脸特征选择及识别方法[J].电子与信息学报,2007,29(7):1744-1748Li W H,Chen W M,Yang L P,et al. Face feature selection and recognition based on different types of margin[J]. Journal of Electronics&Information Technology,2007,29(7):1744-1748(in Chinese)
[8]徐旭东,周源华.基于小波矩不变量的模式识别方法[J].红外与毫米波学报,2000,19(3):215-218Xu X D, Zhou Y H. Pattern recognition based on wavelet moment invariants[J]. Journal of Infrared and Millimeter Waves,2000,19(3):215-218(in Chinese)
[9]盛立东.模式识别导论[M].北京:北京邮电大学出版社,2010Sheng L D. Introduction to pattern recognition[M].Beijing:Beijing University of Posts and Telecommunications Press,2010(in Chinese)
[10] Liu Y Y,Zhao F,Shao W H,et al. An hidden markov model based complex walking pattern recognition algorithm[C]//Proceedings of the 2016 4th International Conference on Ubiquitous Positioning,Indoor Navigation and Location Based Services.Shanghai,China:IEEE,2016:223-229
[11] Parcheta Z,Martínez-Hinarejos C D. Sign language gesture recognition using HMM[C]//Proceedings of the 8th Iberian Conference on Pattern Recognition and Image Analysis. Faro,Portugal:Springer,2017:419-426
[12]李晓兵,孙晓丽,夏良正.基于小波矩特征的小波神经网络目标识别[J].东南大学学报:自然科学版,2006,30(S1):90-93Li X B,Sun X L,Xia L Z. Wavelet neural network in automatic target recognition based on wavelet moment[J].Journal of Southeast University:Natural Science Edition,2006,30(S1):90-93(in Chinese)
[13] Ye F,Yi N,Wang Y. EM algorithm for training highorder hidden Markov model with multiple observation sequences[J]. Journal of Information&Computational Science,2011,8(10):1761-1777
[14] ESB. Engineering shape benchmark[EB/OL]. 2007-06-01. http://purdue.edu/shapelab
[15] Wang L W,Zhang Y,Feng J F. On the Euclidean distance of images[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2005,27(8):1334-1339
[2] Shen D G,Ip H H S. Discriminative wavelet shape descriptors for recognition of 2-D patterns[J]. Pattern Recognition,1999,32(2):151-165
[3]梅雪,林锦国.基于图像边缘小波矩和支持向量机的目标识别[J].计算机工程与科学,2006,28(7):60-61,69Mei X,Lin J G. Target recognition based on wavelet moment and support vector machines[J]. Computer Engineering&Science,2006,28(7):60-61,69(in Chinese)
[4]崔建江,宋星月,陈国坤,等.基于几何形状和小波矩的静脉特征提取与匹配[J].东北大学学报(自然科学版),2009,30(9):1236-1240Cui J J,Song X Y,Chen G K,et al. Feature extraction and matching of vein based on geometrical shape and wavelet moment[J]. Journal of Northeastern University(Natural Science), 2009,30(9):1236-1240(in Chinese)
[5]于剑,程乾生.模糊聚类方法中的最佳聚类数的搜索范围[J].中国科学:E辑,2002,32(2):274-280Yu J,Cheng Q S. The upper bound of the optimal number of clusters in fuzzy clustering[J]. Science in China:Science E,2002,32(2):274-280(in Chinese)
[6]冯宗翰,吴小俊.基于迹比准则与+L-R方法的特征选择算法[J].计算机工程,2011,37(17):136-139Feng Z H,Wu X J. Feature selection algorithm based on trace ratio criterion and+L-R method[J]. Computer Engineering,2011,37(17):136-139(in Chinese)
[7]李伟红,陈伟民,杨利平,等.基于不同Margin的人脸特征选择及识别方法[J].电子与信息学报,2007,29(7):1744-1748Li W H,Chen W M,Yang L P,et al. Face feature selection and recognition based on different types of margin[J]. Journal of Electronics&Information Technology,2007,29(7):1744-1748(in Chinese)
[8]徐旭东,周源华.基于小波矩不变量的模式识别方法[J].红外与毫米波学报,2000,19(3):215-218Xu X D, Zhou Y H. Pattern recognition based on wavelet moment invariants[J]. Journal of Infrared and Millimeter Waves,2000,19(3):215-218(in Chinese)
[9]盛立东.模式识别导论[M].北京:北京邮电大学出版社,2010Sheng L D. Introduction to pattern recognition[M].Beijing:Beijing University of Posts and Telecommunications Press,2010(in Chinese)
[10] Liu Y Y,Zhao F,Shao W H,et al. An hidden markov model based complex walking pattern recognition algorithm[C]//Proceedings of the 2016 4th International Conference on Ubiquitous Positioning,Indoor Navigation and Location Based Services.Shanghai,China:IEEE,2016:223-229
[11] Parcheta Z,Martínez-Hinarejos C D. Sign language gesture recognition using HMM[C]//Proceedings of the 8th Iberian Conference on Pattern Recognition and Image Analysis. Faro,Portugal:Springer,2017:419-426
[12]李晓兵,孙晓丽,夏良正.基于小波矩特征的小波神经网络目标识别[J].东南大学学报:自然科学版,2006,30(S1):90-93Li X B,Sun X L,Xia L Z. Wavelet neural network in automatic target recognition based on wavelet moment[J].Journal of Southeast University:Natural Science Edition,2006,30(S1):90-93(in Chinese)
[13] Ye F,Yi N,Wang Y. EM algorithm for training highorder hidden Markov model with multiple observation sequences[J]. Journal of Information&Computational Science,2011,8(10):1761-1777
[14] ESB. Engineering shape benchmark[EB/OL]. 2007-06-01. http://purdue.edu/shapelab
[15] Wang L W,Zhang Y,Feng J F. On the Euclidean distance of images[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2005,27(8):1334-1339