現今的生產線自動化程度仍低，必須以人工進行重複性的取放動作。使用機器人進行少量多樣的產品自動化組裝時，分別遭遇兩大難題：一為難以迅速自主地進行零組件的取放組裝，二為機器人使用設定不易而難以快速針對不同產品的組裝進行設定。這兩大難題造成目前電子零組件取放與組裝仍相當程度仰賴人工裝配，形成自動化生產製造的瓶頸。採用機器人進行隨機夾取動作，重點在於機器人取料時須能夠自動估測容器內零組件的姿態，進而判斷如何夾取。無論針對新進零組件的建模，或者增進姿態估測的準確度，強健特徵點分析均為一重要的核心模組，在本論文中發展一應用於自主物件模型學習技術之強健特徵點分析技術，以利後續產線上之即時自動姿態估測。

The two major problems in the automatic assembling production line by industrial robot are, the first, random bin picking and placing the components in high speed and the second, teaching the robot to assemble product with different product models and components. These difficulties makes the electric component assembling relies on human nowadays and the human operation could be the bottleneck of production. The key point of random bin picking is automatic pose estimation while the components are in the container. After pose estimation, the robot should automatically plan its action and pick up the component. The 3D feature saliency mapping is an important core technique in 3D modeling and poses estimation. We represented a feature saliency analysis method for automatic modeling in this paper. The result assists the offline database building and the real time automatic pose estimation.