条件付自己組織化マップによる階層的分類手法およびその状況解析への応用

• Material Type: 博士論文
• Title: 条件付自己組織化マップによる階層的分類手法およびその状況解析への応用
• Author/Editor: Jaiprakash Narain Dwivedi
• Author Heading: Jaiprakash Narain Dwivedi
• Publication Date: 2017-03-24
• Publication Date (W3CDTF): 2017-03-24
• Alternative Title: Hierarchical Classification Using Conditional Self-Organizing Map and Its Application to Situation Analysis
• Degree Grantor: 九州工業大学
• Date Granted: 2017-03-24
• Date Granted (W3CDTF): 2017-03-24
• Dissertation Number: 甲第289号
• Degree Type: 博士（工学）
• Conferring No. (Dissertation): 甲生工第289号
• Text Language Code: eng
• Subject Heading: Hierarchical Classification; Self-Organizing Map; Situation Analysis; Conditional Self-Organizing Map; Autonomous Vehicle
• Target Audience: 一般
• Note (General): 九州工業大学博士学位論文 学位記番号:生工博甲第289号 学位授与年月日:平成29年3月24日; 平成28年度
• DOI: 10.18997/00006323; https://doi.org/10.18997/00006323
• Persistent ID (NDL): info:ndljp/pid/10952439; https://dl.ndl.go.jp/pid/10952439
• Collection: 障害者向け資料
• Collection (Materials For Handicapped People:1): テキストデータ
• Collection (particular): 国立国会図書館デジタルコレクション > デジタル化資料 > 博士論文; https://dl.ndl.go.jp/collections/A00014
• Acquisition Basis: 博士論文（自動収集）
• Date Accepted (W3CDTF): 2017-09-02T15:28:24+09:00
• Format (IMT): PDF; application/pdf
• Access Restrictions: 国立国会図書館内限定公開
• Service for the Digitized Contents Transmission Service: 図書館・個人送信対象外
• Availability of remote photoduplication service: 可
• Periodical Title (URI): https://doi.org/10.18997/00006323; http://hdl.handle.net/10228/00006323
• Data Provider (Database): 国立国会図書館 : 国立国会図書館デジタルコレクション; https://dl.ndl.go.jp

• Summary, etc.: Classification has been a challenging tasks for a long time. The Classification of road-vehicle situations plays an important role towars the preparation of limited number of set of situations from unpredictable infinite possibility of situation. This limited number of set of situation preparation is needed for the prediction of root cause of collision during autonomous driving. To fulfill this requirement, the hierarchical classification of the different kind of shapes of the road and objects（i.e. vehicles） corresponding to them has been achieved. For the first step of the situation classification, the data considered are the different shapes of the road. These different shapes of the road are cross junction （i.e. four way road） or T-junction （i.e. three way road） or straight road. In this case, the variation among shapes have been observed using U matrix. The most suitable dot distribution representation is used throughout for the representation of the different shapes of the road. The classification method used for this first step of classification is TFSOM×SOM （i.e. Topology Free Self-Organizing Map by Self Organizing Map） algorithm. A vehicle trajectory has been drawn on these different shapes of the road after the first step of situation classification. For an example, Suppose a road starts from point A to point B and between point A and point B, a combination of all the three different shapes of the road discussed above is available. From point A to point B, the position of vehicles has been tracked along with the coordinates of the road. As the vehicle starts travelling from point A, it has to pass through the straight road, cross shapes of the road and T-junction shape of the road till the end of travel i.e. up to point B. To obtain all the different shapes of the road as a simulation result, we used the concept the standard deviation. The standard deviation concept was used as a penalty term because without using this penalty term, we were not able to get the required three different shapes of the road. These three different shapes of the road have been represented using three different colors. For the second step of the situation classification, Two types of data are the coordinates of the position of vehicles and road. In this time, only cross shapes of the road are being considered as the coordinates of shape of the road. This is because that the possibility of number of objects is more than that of T-junction and straight road. In other word, the complexity of the shapes of the road is the potential to define the level of risk of danger. The most suitable dot distribution representation is used throughout for the representation of the cross shapes of the road data. In this hierarchical classification, the first kind of data are road data i.e. cross shape road data. These road data are being classified using TFSOM×SOM algorithm as said above. The second kind of data are the object data i.e. vehicle data being used for classification with proposed method Conditional Self Organizing Map i.e. CSOM algorithm. The quantization error is the average distance between the input data vector and its best matching unit. The simulation result of the quantization error of both CSOM algorithm and SOM algorithm has been compared in order to justify the relevance of the proposed CSOM algorithm over SOM algorithm. The quantization error of CSOM algorithm is less than that of SOM algorithm. Also for same quantization error CSOM method needs almost one third number of learning data in comparison to that of SOM algorithm. The comparison of simulation results of the proposed hierarchical CSOM al-gorithm with the non-hierarchical and hierarchical SOM method has been done in order to justify that the use of this proposed CSOM algorithm is more suitable than those SOM methods.
• DOI: 10.18997/00006323; https://doi.org/10.18997/00006323
• Format (IMT): application/pdf
• Source: sei_k_289.pdf （fulltext）; https://kyutech.repo.nii.ac.jp/record/5111/files/sei_k_289.pdf
• Access Restrictions: インターネット公開
• Data Provider (Database): 国立情報学研究所 : 学術機関リポジトリデータベース（IRDB）（機関リポジトリ）; https://irdb.nii.ac.jp
• Original Data Provider (Database): 九州工業大学 : キューテイカー

Table of Contents

2017-10-02 再収集

2023-03-04 再収集

2023-08-05 再収集

2023-12-06 再収集