Full metadata record
DC Field | Value | Language |
---|---|---|
dc.contributor.author | 莊凱傑 | en_US |
dc.contributor.author | Chuang, Kai-Chieh | en_US |
dc.contributor.author | 胡竹生 | en_US |
dc.contributor.author | Hu, Jwu-Sheng | en_US |
dc.date.accessioned | 2014-12-12T02:41:53Z | - |
dc.date.available | 2014-12-12T02:41:53Z | - |
dc.date.issued | 2013 | en_US |
dc.identifier.uri | http://140.113.39.130/cdrfb3/record/nctu/#GT070160054 | en_US |
dc.identifier.uri | http://hdl.handle.net/11536/74907 | - |
dc.description.abstract | 本論文提出了一個高擬真度模擬與驗證平台。此平台目前適用於工件隨機堆疊取放演算法(random bin picking)的模擬與驗證。此平台能自動產生大量的演算法實驗數據,並讓使用者能夠藉由這些數據,了解其演算法的特性並做出改善。因為此平台擁有逼真的模擬環境,使得模擬得到的結果能與真實環境得到的結果相近。 本篇論文專注於建立此模擬與驗證平台和建立高擬真度環境的方法。首先,此平台能讓使用者指定模型,並透過物理引擎模擬該模型的多物體隨機堆疊情形。接下來,藉由此平台的虛擬深度感測器擷取該場景,並將逼真的擷取資料傳送給姿態估測演算法。經過姿態估測演算法對收到的資料處理後,將演算法結果傳送回此平台。最後,此平台利用存在於模擬世界中的正確資料(ground truth),驗證演算法結果的準確度。藉由不斷的模擬與驗證,此平台能提供大量的準確演算法統計特性。 本篇論文將提供讀者此平台的建立方法,以及如何改進原來既有的模擬平台來達大逼真的模擬結果。並在最後,利用此平台來驗證工業技術研究院的姿態估測演算法。並透過此平台未姿態估測演算法找到最佳的參數。 | zh_TW |
dc.description.abstract | We present a realistic evaluation platform for a random bin picking scenario on a simulator, targeted to evaluate the performance of an algorithm, for instance, pose estimation, segmentation, etc. It’s a fully automated process. First it will simulate objects stacking randomly in a box. Secondly, a highly realistic virtual depth sensor will capture the scene and send the data to the targeted algorithm. Then, our evaluation platform waits for the algorithm result and evaluates the algorithm by comparing the result with ground truth. By running this process continuously, we can obtain a vast amount of valuable statistics data about the performance of the algorithm. In this thesis, we will mention the architecture of our evaluation platform, and focus on how we try to model an active stereo depth sensor by simulating the real sensor’s attributes and noise, making it as realistic as possible. | en_US |
dc.language.iso | en_US | en_US |
dc.subject | 深度感測器模擬 | zh_TW |
dc.subject | 雜訊模擬 | zh_TW |
dc.subject | 隨機堆疊取放演算法 | zh_TW |
dc.subject | 姿態估測演算法 | zh_TW |
dc.subject | 分割演算法 | zh_TW |
dc.subject | 模擬驗證平台 | zh_TW |
dc.subject | Virtual RGB depth sensor | en_US |
dc.subject | Synthetic noise | en_US |
dc.subject | Random bin picking | en_US |
dc.subject | Pose estimation | en_US |
dc.subject | Segmentation | en_US |
dc.subject | Simulation | en_US |
dc.subject | Evaluation | en_US |
dc.title | 高擬真度之工件隨機堆疊取放演算法的模擬與驗證平台 | zh_TW |
dc.title | A High Fidelity Simulator for Random Bin Picking Scenario Evaluation | en_US |
dc.type | Thesis | en_US |
dc.contributor.department | 電控工程研究所 | zh_TW |
Appears in Collections: | Thesis |