學術產出-Theses
Article View/Open
Publication Export
-
題名 基於三元編碼之三維特徵描述子建構與模型比對
Local Ternary Descriptor for 3D Model Matching and Retrieval作者 王崇銘
Wang, Chong Ming貢獻者 廖文宏
Liao, Wen Hung
王崇銘
Wang, Chong Ming關鍵詞 積木風格模型
三元化區域特徵描述子
三維模型比對
Block-style model
Local ternary descriptor
Model comparison日期 2017 上傳時間 3-Jan-2018 16:20:19 (UTC+8) 摘要 Qmodel Creator是一款製作三維模型的軟體,其模型特色為積木風格,顧名思義,是以積木或是立方體所構成的模型。然而,目前三維模型檢索系統多數以關鍵字搜尋,缺點為需要大量時間對三維模型進行關鍵字標記。本論文提出三維特徵描述方法(3D Local Ternary Descriptor,3DLTD),嘗試基於內容本身進行三維模型檢索。此方法由二維影像的三元特徵描述延伸而來。首先,利用PCA找出主軸分佈,藉此篩選非相關的模型。接著,依照每個模型的bounding box分割兩次區塊,第一次分割的單位稱為cube,第二次分割稱為voxel,再根據與邊界的距離賦予voxel對應的權重,距離邊界越近權重越大,同時隨機從voxel樣本對關係中取64組樣本對進行三元化編碼。最後把編碼完的字串經由bipartite weighted matching做匹配。相較於3DBRIEF方法,本架構可以改善效率的問題,使用三元化編碼機制加快其運算速度,並以voxel為單位加入權重機制,以區別加權voxel位置之間造成的影響程度。 實驗結果顯示本研究方法對於積木風格模型的樣本集,相對於二元化編碼,採用三元化編碼不僅可以提升描述力和抗噪力,同時因為區域的劃分,降低特徵描述子維度,節省模型比對的時間和空間,也能維持整體模型比對之效能。
Qmodel Creator is a software for designing block (lego) style 3D models using intuitive drawing methods. The produced models are composed of cubes, which are conveniently encoded using voxel-based representation. In order to provide the search function for 3D models, keywords or tags have to be inserted manually, which is time-consuming and not cost-effective.In this thesis, we proposed a 3D feature descriptor defined as 3D local ternary descriptor (3DLTD) to support content-based search and retrieval for models generated using Qmodel Creator. This descriptor is extended from a class of 2D features known as local binary descriptors (LBD) for image matching. To begin with, principal component analysis (PCA) is employed to align model orientation to filter out irrelevant samples. After the alignment, we first partition the bounding box of each model into cubes and then divide cubes into voxels. Weights are assigned to each voxel according to its distance to the boundary. We randomly select 64 pairs of voxels in each cube and generate a ternary code based on the relationship between each pair of voxels. Finally, weighted bipartite matching is employed to compute the similarity between two models. Compared with 3DBRIEF, another method extended from LBD, our proposed framework is more robust and efficient. The inclusion of weights differentiates the contribution of different voxels and have effectively enhanced the performance of 3DLTD. Experimental results indicate that 3DLTD is suitable for comparing and searching voxel-based 3D models. Thanks to two-level partition, feature representation and distance computation are greatly simplified. Ternary encoding also promotes `describility` and noise immunity while maintaining efficiency in 3D model search and retrieval.參考文獻 [1] HU, Chen-Chi, et al. Intuitive 3D cubic style modeling system. In: SIGGRAPH Asia 2015 Posters. ACM, 2015. p. 27.[2] LOWE, David G. Object recognition from local scale-invariant features. In: Computer vision, 1999. The proceedings of the seventh IEEE international conference on. IEEE, 1999. p. 1150-1157.[3] BAY, Herbert; TUYTELAARS, Tinne; VAN GOOL, Luc. SURF: Speeded up robust features. In: Computer Vision–ECCV 2006. Springer Berlin Heidelberg, 2006. p. 404-417.[4] TOLA, Engin; LEPETIT, Vincent; FUA, Pascal. Daisy: An efficient dense descriptor applied to wide-baseline stereo. IEEE transactions on pattern analysis and machine intelligence, 2010, 32.5: 815-830.[5] CALONDER, Michael, et al. BRIEF: Binary robust independent elementary features. In: Computer Vision–ECCV 2010. Springer Berlin Heidelberg, 2010. p. 778-792.[6] LEUTENEGGER, Stefan; CHLI, Margarita; SIEGWART, Roland Yves. BRISK: Binary robust invariant scalable keypoints. In: Computer Vision (ICCV), 2011 IEEE International Conference on. IEEE, 2011. p.2548-2555.[7] RUBLEE, Ethan, et al. ORB: an efficient alternative to SIFT or SURF. In:Computer Vision (ICCV), 2011 IEEE International Conference on. IEEE, 2011. p. 2564-2571.[8] OJALA, Timo; PIETIKAINEN, Matti; MAENPAA, Topi. Multiresolution gray-scale and rotation invariant texture classification with local binary patterns. IEEE Transactions on pattern analysis and machine intelligence, 2002, 24.7: 971-987..[9] TANGELDER, Johan WH; VELTKAMP, Remco C. A survey of content based 3D shape retrieval methods. Multimedia tools and applications, 2008, 39.3: 441-471.[10] ALAHI, Alexandre; ORTIZ, Raphael; VANDERGHEYNST, Pierre. Freak: Fast retina keypoint. In: Computer vision and pattern recognition (CVPR), 2012 IEEE conference on. Ieee, 2012. p. 510-517.[11] WANG, Zhenhua; FAN, Bin; WU, Fuchao. Local intensity order pattern for feature description. In: Computer Vision (ICCV), 2011 IEEE International Conference on. IEEE, 2011. p. 603-610.[12] TVERSKY, Amos. Features of similarity. Psychological review, 1977, 84.4: 327.[13] SUNDAR, Hari, et al. Skeleton based shape matching and retrieval. In: Shape Modeling International, 2003. IEEE, 2003. p. 130-139.[14] VRANIC, Dejan V.; SAUPE, Dietmar. 3D model retrieval. 2004. PhD Thesis. University of Leipzig.[15] PAQUET, Eric, et al. Description of shape information for 2-D and 3-D objects. Signal processing: Image communication, 2000, 16.1: 103-122.[16] VRANIC, Dejan V.; SAUPE, Dietmar; RICHTER, Jörg. Tools for 3D-object retrieval: Karhunen-Loeve transform and spherical harmonics. In: Multimedia Signal Processing, 2001 IEEE Fourth Workshop on. IEEE, 2001. p. 293-298.[17] HORN, Berthold Klaus Paul. Extended gaussian images. Proceedings of the IEEE, 1984, 72.12: 1671-1686.[18] KAZHDAN, Michael; FUNKHOUSER, Thomas; RUSINKIEWICZ, Szymon. Rotation invariant spherical harmonic representation of 3 d shape descriptors. In: Symposium on geometry processing. 2003. p. 156-164.[19] SHEN, Yu-Te, et al. 3D model search engine based on lightfield descriptors. In: Proc. eurographics. 2003.[20] MATSUDA, Takahiro; FURUYA, Takahiko; OHBUCHI, Ryutarou. Lightweight binary voxel shape features for 3D data matching and retrieval. In: Multimedia Big Data (BigMM), 2015 IEEE International Conference on. IEEE, 2015. p. 100-107.[21] VRANIĆ, Dejan V. An improvement of rotation invariant 3D-shape based on functions on concentric spheres. In: Image Processing, 2003. ICIP 2003. Proceedings. 2003 International Conference on. IEEE, 2003. p. III-757-60 vol. 2.[22] ZHANG, Lisha, et al. Survey on 3D shape descriptors. FundaÃgao para a Cincia ea Tecnologia, Lisboa, Portugal, Tech. Rep. Technical Report, DecorAR (FCT POSC/EIA/59938/2004), 2007, 3.[23] CHEN, Ding‐Yun, et al. On visual similarity based 3D model retrieval. In: Computer graphics forum. Blackwell Publishing, Inc, 2003. p. 223-232.[24] LIAO, Wen-Hung. Region description using extended local ternary patterns. In: Pattern Recognition (ICPR), 2010 20th International Conference on. IEEE, 2010. p. 1003-1006.[25] TAN, Xiaoyang; TRIGGS, Bill. Enhanced local texture feature sets for face recognition under difficult lighting conditions. IEEE transactions on image processing, 2010, 19.6: 1635-1650. 描述 碩士
國立政治大學
資訊科學學系
103753015資料來源 http://thesis.lib.nccu.edu.tw/record/#G0103753015 資料類型 thesis dc.contributor.advisor 廖文宏 zh_TW dc.contributor.advisor Liao, Wen Hung en_US dc.contributor.author (Authors) 王崇銘 zh_TW dc.contributor.author (Authors) Wang, Chong Ming en_US dc.creator (作者) 王崇銘 zh_TW dc.creator (作者) Wang, Chong Ming en_US dc.date (日期) 2017 en_US dc.date.accessioned 3-Jan-2018 16:20:19 (UTC+8) - dc.date.available 3-Jan-2018 16:20:19 (UTC+8) - dc.date.issued (上傳時間) 3-Jan-2018 16:20:19 (UTC+8) - dc.identifier (Other Identifiers) G0103753015 en_US dc.identifier.uri (URI) http://nccur.lib.nccu.edu.tw/handle/140.119/115462 - dc.description (描述) 碩士 zh_TW dc.description (描述) 國立政治大學 zh_TW dc.description (描述) 資訊科學學系 zh_TW dc.description (描述) 103753015 zh_TW dc.description.abstract (摘要) Qmodel Creator是一款製作三維模型的軟體,其模型特色為積木風格,顧名思義,是以積木或是立方體所構成的模型。然而,目前三維模型檢索系統多數以關鍵字搜尋,缺點為需要大量時間對三維模型進行關鍵字標記。本論文提出三維特徵描述方法(3D Local Ternary Descriptor,3DLTD),嘗試基於內容本身進行三維模型檢索。此方法由二維影像的三元特徵描述延伸而來。首先,利用PCA找出主軸分佈,藉此篩選非相關的模型。接著,依照每個模型的bounding box分割兩次區塊,第一次分割的單位稱為cube,第二次分割稱為voxel,再根據與邊界的距離賦予voxel對應的權重,距離邊界越近權重越大,同時隨機從voxel樣本對關係中取64組樣本對進行三元化編碼。最後把編碼完的字串經由bipartite weighted matching做匹配。相較於3DBRIEF方法,本架構可以改善效率的問題,使用三元化編碼機制加快其運算速度,並以voxel為單位加入權重機制,以區別加權voxel位置之間造成的影響程度。 實驗結果顯示本研究方法對於積木風格模型的樣本集,相對於二元化編碼,採用三元化編碼不僅可以提升描述力和抗噪力,同時因為區域的劃分,降低特徵描述子維度,節省模型比對的時間和空間,也能維持整體模型比對之效能。 zh_TW dc.description.abstract (摘要) Qmodel Creator is a software for designing block (lego) style 3D models using intuitive drawing methods. The produced models are composed of cubes, which are conveniently encoded using voxel-based representation. In order to provide the search function for 3D models, keywords or tags have to be inserted manually, which is time-consuming and not cost-effective.In this thesis, we proposed a 3D feature descriptor defined as 3D local ternary descriptor (3DLTD) to support content-based search and retrieval for models generated using Qmodel Creator. This descriptor is extended from a class of 2D features known as local binary descriptors (LBD) for image matching. To begin with, principal component analysis (PCA) is employed to align model orientation to filter out irrelevant samples. After the alignment, we first partition the bounding box of each model into cubes and then divide cubes into voxels. Weights are assigned to each voxel according to its distance to the boundary. We randomly select 64 pairs of voxels in each cube and generate a ternary code based on the relationship between each pair of voxels. Finally, weighted bipartite matching is employed to compute the similarity between two models. Compared with 3DBRIEF, another method extended from LBD, our proposed framework is more robust and efficient. The inclusion of weights differentiates the contribution of different voxels and have effectively enhanced the performance of 3DLTD. Experimental results indicate that 3DLTD is suitable for comparing and searching voxel-based 3D models. Thanks to two-level partition, feature representation and distance computation are greatly simplified. Ternary encoding also promotes `describility` and noise immunity while maintaining efficiency in 3D model search and retrieval. en_US dc.description.tableofcontents 第一章 緒論 11.1 研究背景 11.2 三維模型之建構 21.3 三維模型比對 51.4 研究目的 51.5 主要貢獻 61.6 論文架構 7第二章 相關研究 82.1 二維區域影像特徵 82.1.1 基於區域梯度統計 82.1.2 基於點對關係 92.1.3 基於點群關係 112.2 三維特徵比對流程 112.3 三維特徵編碼 132.3.1 基於統計的特徵 152.3.2 基於延伸擴展的特徵 152.3.3 基於射線的特徵 162.3.4 基於投影的特徵 172.3.5 基於圖(graph)的特徵 182.3.6 基於立體像素(voxel)的特徵擷取 192.4 小結 23第三章 研究方法 253.1 三維特徵描述子之建構 263.1.1 Pixel值與Voxel值關係 283.1.2 描述子建構 – 3DLTD 293.1.3 權重設定 333.2 比對方法建構 343.2.1 快速篩選機制 343.2.2 比對方法 353.2.3 bipartite weighted matching 36第四章 實驗結果與分析 384.1 實驗樣本 384.2 樣本標準差設定 394.3 實驗一 404.4 實驗二 434.5 實驗三 474.6 實驗四 494.7 實驗五 514.8 實驗結果小結 54第五章 結論與未來展望 56參考文獻 57 zh_TW dc.format.extent 2920730 bytes - dc.format.mimetype application/pdf - dc.source.uri (資料來源) http://thesis.lib.nccu.edu.tw/record/#G0103753015 en_US dc.subject (關鍵詞) 積木風格模型 zh_TW dc.subject (關鍵詞) 三元化區域特徵描述子 zh_TW dc.subject (關鍵詞) 三維模型比對 zh_TW dc.subject (關鍵詞) Block-style model en_US dc.subject (關鍵詞) Local ternary descriptor en_US dc.subject (關鍵詞) Model comparison en_US dc.title (題名) 基於三元編碼之三維特徵描述子建構與模型比對 zh_TW dc.title (題名) Local Ternary Descriptor for 3D Model Matching and Retrieval en_US dc.type (資料類型) thesis en_US dc.relation.reference (參考文獻) [1] HU, Chen-Chi, et al. Intuitive 3D cubic style modeling system. In: SIGGRAPH Asia 2015 Posters. ACM, 2015. p. 27.[2] LOWE, David G. Object recognition from local scale-invariant features. In: Computer vision, 1999. The proceedings of the seventh IEEE international conference on. IEEE, 1999. p. 1150-1157.[3] BAY, Herbert; TUYTELAARS, Tinne; VAN GOOL, Luc. SURF: Speeded up robust features. In: Computer Vision–ECCV 2006. Springer Berlin Heidelberg, 2006. p. 404-417.[4] TOLA, Engin; LEPETIT, Vincent; FUA, Pascal. Daisy: An efficient dense descriptor applied to wide-baseline stereo. IEEE transactions on pattern analysis and machine intelligence, 2010, 32.5: 815-830.[5] CALONDER, Michael, et al. BRIEF: Binary robust independent elementary features. In: Computer Vision–ECCV 2010. Springer Berlin Heidelberg, 2010. p. 778-792.[6] LEUTENEGGER, Stefan; CHLI, Margarita; SIEGWART, Roland Yves. BRISK: Binary robust invariant scalable keypoints. In: Computer Vision (ICCV), 2011 IEEE International Conference on. IEEE, 2011. p.2548-2555.[7] RUBLEE, Ethan, et al. ORB: an efficient alternative to SIFT or SURF. In:Computer Vision (ICCV), 2011 IEEE International Conference on. IEEE, 2011. p. 2564-2571.[8] OJALA, Timo; PIETIKAINEN, Matti; MAENPAA, Topi. Multiresolution gray-scale and rotation invariant texture classification with local binary patterns. IEEE Transactions on pattern analysis and machine intelligence, 2002, 24.7: 971-987..[9] TANGELDER, Johan WH; VELTKAMP, Remco C. A survey of content based 3D shape retrieval methods. Multimedia tools and applications, 2008, 39.3: 441-471.[10] ALAHI, Alexandre; ORTIZ, Raphael; VANDERGHEYNST, Pierre. Freak: Fast retina keypoint. In: Computer vision and pattern recognition (CVPR), 2012 IEEE conference on. Ieee, 2012. p. 510-517.[11] WANG, Zhenhua; FAN, Bin; WU, Fuchao. Local intensity order pattern for feature description. In: Computer Vision (ICCV), 2011 IEEE International Conference on. IEEE, 2011. p. 603-610.[12] TVERSKY, Amos. Features of similarity. Psychological review, 1977, 84.4: 327.[13] SUNDAR, Hari, et al. Skeleton based shape matching and retrieval. In: Shape Modeling International, 2003. IEEE, 2003. p. 130-139.[14] VRANIC, Dejan V.; SAUPE, Dietmar. 3D model retrieval. 2004. PhD Thesis. University of Leipzig.[15] PAQUET, Eric, et al. Description of shape information for 2-D and 3-D objects. Signal processing: Image communication, 2000, 16.1: 103-122.[16] VRANIC, Dejan V.; SAUPE, Dietmar; RICHTER, Jörg. Tools for 3D-object retrieval: Karhunen-Loeve transform and spherical harmonics. In: Multimedia Signal Processing, 2001 IEEE Fourth Workshop on. IEEE, 2001. p. 293-298.[17] HORN, Berthold Klaus Paul. Extended gaussian images. Proceedings of the IEEE, 1984, 72.12: 1671-1686.[18] KAZHDAN, Michael; FUNKHOUSER, Thomas; RUSINKIEWICZ, Szymon. Rotation invariant spherical harmonic representation of 3 d shape descriptors. In: Symposium on geometry processing. 2003. p. 156-164.[19] SHEN, Yu-Te, et al. 3D model search engine based on lightfield descriptors. In: Proc. eurographics. 2003.[20] MATSUDA, Takahiro; FURUYA, Takahiko; OHBUCHI, Ryutarou. Lightweight binary voxel shape features for 3D data matching and retrieval. In: Multimedia Big Data (BigMM), 2015 IEEE International Conference on. IEEE, 2015. p. 100-107.[21] VRANIĆ, Dejan V. An improvement of rotation invariant 3D-shape based on functions on concentric spheres. In: Image Processing, 2003. ICIP 2003. Proceedings. 2003 International Conference on. IEEE, 2003. p. III-757-60 vol. 2.[22] ZHANG, Lisha, et al. Survey on 3D shape descriptors. FundaÃgao para a Cincia ea Tecnologia, Lisboa, Portugal, Tech. Rep. Technical Report, DecorAR (FCT POSC/EIA/59938/2004), 2007, 3.[23] CHEN, Ding‐Yun, et al. On visual similarity based 3D model retrieval. In: Computer graphics forum. Blackwell Publishing, Inc, 2003. p. 223-232.[24] LIAO, Wen-Hung. Region description using extended local ternary patterns. In: Pattern Recognition (ICPR), 2010 20th International Conference on. IEEE, 2010. p. 1003-1006.[25] TAN, Xiaoyang; TRIGGS, Bill. Enhanced local texture feature sets for face recognition under difficult lighting conditions. IEEE transactions on image processing, 2010, 19.6: 1635-1650. zh_TW