一种基于CLM的服务机器人室内功能区分类方法
|
摘要
基于CLM(无码本模型)提出一种规避码本的室内功能区表示与建模方法.首先,在灰度级图像的基础上提取SURF(加速鲁棒特征)描述子;然后,运用空间金字塔方法将图像分成规则区域,在向量空间引入高斯流形,将每个区域用单高斯模型表示,并将其联合构成混合高斯模型以表示整幅图像;最后,将图像的高斯模型与改进的SVM(支持向量机)分类器联合使用,实现室内功能区的分类.在Scene 15数据集上的实验结果表明,本文方法相较于传统的构建码本方式在分类识别精度上提升约20%,同时对方向变化、光照不均匀等情况具有较好的鲁棒性,有效提升了服务机器人对室内功能区的认知能力.
A representation and modeling method of indoor affordance areas based on CLM(codebookless model) is proposed to avoid using codebook. Firstly, multi-scale SURF(speeded-up robust feature) descriptors are extracted on greyscale image. Then, the image is divided into some regular regions using the spatial pyramid method. By introducing Gaussian manifolds into vector space, each region is denoted as a single Gaussian model, and the mixed Gaussian model is combined to represent the whole image. Finally, the Gaussian model and the modified SVM(support vector machine) classifier are utilized to classify the indoor affordance areas. The experimental results on Scene 15 datasets show that the proposed method improves the classification accuracy by about 20% compared with the traditional codebook construction methods, is more robust to direction changes and uneven illumination, and effectively enhances the ability of service robots to cognize indoor affordance areas.
A representation and modeling method of indoor affordance areas based on CLM(codebookless model) is proposed to avoid using codebook. Firstly, multi-scale SURF(speeded-up robust feature) descriptors are extracted on greyscale image. Then, the image is divided into some regular regions using the spatial pyramid method. By introducing Gaussian manifolds into vector space, each region is denoted as a single Gaussian model, and the mixed Gaussian model is combined to represent the whole image. Finally, the Gaussian model and the modified SVM(support vector machine) classifier are utilized to classify the indoor affordance areas. The experimental results on Scene 15 datasets show that the proposed method improves the classification accuracy by about 20% compared with the traditional codebook construction methods, is more robust to direction changes and uneven illumination, and effectively enhances the ability of service robots to cognize indoor affordance areas.
引文
[1]Lazebnik S,Schmid C,Ponce J.Beyond bags of features:Spatial pyramid matching for recognizing natural scene categories[C]//IEEE Computer Society Conference on Computer Vision and Pattern Recognition.Piscataway,USA:IEEE,2006:2169-2178.
[2]Azim T.Fisher kernels match deep models[J].Electronics Letters,2017,53(6):397-399.
[3]Bo C J,Lu H C,Wang D.Weighted generalized nearest neighbor for hyperspectral image classification[J].IEEE Access,2017,5:1496-1509.
[4]Song Y,Li Q,Huang H,et al.Low dimensional representation of Fisher vectors for microscopy image classification[J].IEEE Transactions on Medical Imaging,2017,36(8):1636-1649.
[5]Zhang C J,Xiao X,Pang J B,et al.Beyond visual word ambiguity:Weighted local feature encoding with governing region[J].Journal of Visual Communication and Image Representation,2014,25(6):1387-1398.
[6]Yang Y B,Zhu Q H,Mao X J,et al.Visual feature coding for image classification integrating dictionary structure[J].Pattern Recognition,2015,48(10):3067-3075.
[7]Zhou W G,Yang M,Li H Q,et al.Towards codebook-free:Scalable cascaded hashing for mobile image search[J].IEEE Transactions on Multimedia,2014,16(3):601-611.
[8]Grauman K,Darrell T.The pyramid match kernel:Discriminative classification with sets of image features[C]//IEEE International Conference on Computer Vision.Piscataway,USA:IEEE,2005:1458-1465.
[9]Li F F,Perona P.A Bayesian hierarchical model for learning natural scene categories[C]//IEEE Computer Society Conference on Computer Vision and Pattern Recognition.Piscataway,USA:IEEE,2005:524-531.
[10]Bo L F,Sminchisescu C.Efficient match kernel between sets of features for visual recognition[C]//Advances in Neural Information Processing Systems 22.Red Hook,USA:Curran Associates Inc.,2009:135-143.
[11]Boiman O,Shechtman E,Irani M.In defense of nearestneighbor based image classification[C]//IEEE Computer Society Conference on Computer Vision and Pattern Recognition.Piscataway,USA:IEEE,2008:1992-1999.
[12]Peng Z L,Li Y,Cai Z Q,et al.Deep Boosting:Joint feature selection and analysis dictionary learning in hierarchy[J].Neurocomputing,2016,178(S1):36-45.
[13]Nakayama H,Harada T,Kuniyoshi Y.Global Gaussian approach for scene categorization using information geometry[C]//IEEE Computer Society Conference on Computer Vision and Pattern Recognition.Piscataway,USA:IEEE,2010:2336-2343.
[14]Wang Q L,Li P H,Zhang L,et al.Towards effective codebookless model for image classification[J].Pattern Recognition,2016,59(S1):63-71.
[15]Na Y,Liao M M,Jung C.Super-speed up robust features image geometrical registration algorithm[J].IET Image Processing,2016,10(11):848-864.
[16]Li P H,Wang Q L,Zhang L.A novel earth mover’s distance methodology for image matching with Gaussian mixture models[C]//IEEE International Conference on Computer Vision.Piscataway,USA:IEEE,2013:1689-1696.
[17]Lovric M,Min-Oo M,Ruh E A.Multivariate normal distributions parametrized as a Riemannian symmetric space[J].Journal of Multivariate Analysis,2000,74(1):36-48.
[18]Amari S.Differential geometry of statistical models[M]//Lecture Notes in Statistics,vol.28.Berlin,Germany:SpringerVerlag,1985:11-65.
[19]Arsigny V,Fillard P,Pennec X,et al.Fast and simple calculus on tensors in the Log-Euclidean framework[C]//8th International Conference on Medical Image Computing and ComputerAssisted Intervention.Berlin,Germany:Springer-Verlag,2005:115-122.
[20]Pennec X.Probabilities and statistics on Riemannian manifolds:A geometric approach[R].Nice,France:INRIA,2004.
[21]Stein C.Lectures on the theory of estimation of many parameters[J].Journal of Mathematical Sciences,1986,34(1):1373-1403.
[22]Carreira J,Caseiro R,Batista J.Semantic segmentation with second-order pooling[C]//12th European Conference on Computer Vision.Berlin,Germany:Springer-Verlag,2012:430-443.
[23]Carreira J,Caseiro R,Batista J,et al.Free-form region description with second-order pooling[J],IEEE Transactions on Pattern Analysis and Machine Intelligence,2015,37(6):1177-1189.
[2]Azim T.Fisher kernels match deep models[J].Electronics Letters,2017,53(6):397-399.
[3]Bo C J,Lu H C,Wang D.Weighted generalized nearest neighbor for hyperspectral image classification[J].IEEE Access,2017,5:1496-1509.
[4]Song Y,Li Q,Huang H,et al.Low dimensional representation of Fisher vectors for microscopy image classification[J].IEEE Transactions on Medical Imaging,2017,36(8):1636-1649.
[5]Zhang C J,Xiao X,Pang J B,et al.Beyond visual word ambiguity:Weighted local feature encoding with governing region[J].Journal of Visual Communication and Image Representation,2014,25(6):1387-1398.
[6]Yang Y B,Zhu Q H,Mao X J,et al.Visual feature coding for image classification integrating dictionary structure[J].Pattern Recognition,2015,48(10):3067-3075.
[7]Zhou W G,Yang M,Li H Q,et al.Towards codebook-free:Scalable cascaded hashing for mobile image search[J].IEEE Transactions on Multimedia,2014,16(3):601-611.
[8]Grauman K,Darrell T.The pyramid match kernel:Discriminative classification with sets of image features[C]//IEEE International Conference on Computer Vision.Piscataway,USA:IEEE,2005:1458-1465.
[9]Li F F,Perona P.A Bayesian hierarchical model for learning natural scene categories[C]//IEEE Computer Society Conference on Computer Vision and Pattern Recognition.Piscataway,USA:IEEE,2005:524-531.
[10]Bo L F,Sminchisescu C.Efficient match kernel between sets of features for visual recognition[C]//Advances in Neural Information Processing Systems 22.Red Hook,USA:Curran Associates Inc.,2009:135-143.
[11]Boiman O,Shechtman E,Irani M.In defense of nearestneighbor based image classification[C]//IEEE Computer Society Conference on Computer Vision and Pattern Recognition.Piscataway,USA:IEEE,2008:1992-1999.
[12]Peng Z L,Li Y,Cai Z Q,et al.Deep Boosting:Joint feature selection and analysis dictionary learning in hierarchy[J].Neurocomputing,2016,178(S1):36-45.
[13]Nakayama H,Harada T,Kuniyoshi Y.Global Gaussian approach for scene categorization using information geometry[C]//IEEE Computer Society Conference on Computer Vision and Pattern Recognition.Piscataway,USA:IEEE,2010:2336-2343.
[14]Wang Q L,Li P H,Zhang L,et al.Towards effective codebookless model for image classification[J].Pattern Recognition,2016,59(S1):63-71.
[15]Na Y,Liao M M,Jung C.Super-speed up robust features image geometrical registration algorithm[J].IET Image Processing,2016,10(11):848-864.
[16]Li P H,Wang Q L,Zhang L.A novel earth mover’s distance methodology for image matching with Gaussian mixture models[C]//IEEE International Conference on Computer Vision.Piscataway,USA:IEEE,2013:1689-1696.
[17]Lovric M,Min-Oo M,Ruh E A.Multivariate normal distributions parametrized as a Riemannian symmetric space[J].Journal of Multivariate Analysis,2000,74(1):36-48.
[18]Amari S.Differential geometry of statistical models[M]//Lecture Notes in Statistics,vol.28.Berlin,Germany:SpringerVerlag,1985:11-65.
[19]Arsigny V,Fillard P,Pennec X,et al.Fast and simple calculus on tensors in the Log-Euclidean framework[C]//8th International Conference on Medical Image Computing and ComputerAssisted Intervention.Berlin,Germany:Springer-Verlag,2005:115-122.
[20]Pennec X.Probabilities and statistics on Riemannian manifolds:A geometric approach[R].Nice,France:INRIA,2004.
[21]Stein C.Lectures on the theory of estimation of many parameters[J].Journal of Mathematical Sciences,1986,34(1):1373-1403.
[22]Carreira J,Caseiro R,Batista J.Semantic segmentation with second-order pooling[C]//12th European Conference on Computer Vision.Berlin,Germany:Springer-Verlag,2012:430-443.
[23]Carreira J,Caseiro R,Batista J,et al.Free-form region description with second-order pooling[J],IEEE Transactions on Pattern Analysis and Machine Intelligence,2015,37(6):1177-1189.