頁籤選單縮合
題名 | 基於類神經網路整合學習兩類攝影資訊之非校正型影像伺服系統=A Neural-Network-Based Co-operative Learning for Un-calibrated Visual Servoing System |
---|---|
作者 | 彭守道; 許毅然; 蔡明宏; Peng, Shou-tao; Sheu, Yih-ran; Tsai, Ming-hung; |
期刊 | 南臺科技大學學報 |
出版日期 | 20061200 |
卷期 | 31 民95.12 |
頁次 | 頁155-163 |
分類號 | 448.94 |
語文 | chi |
關鍵詞 | 非校正式互助學習系統; 自我組織映射網路; 徑向基底函數神經網路; Cooperative system; Self-organization maps; Radial basis function neural network; |
中文摘要 | 本文旨在發展一套整合固定式與移動式兩類攝影系統的互助式學習系統。此系統將監督與非監督式的類神經網路進行整合,並以互助學習的方式,建立出一類整合型的影像伺服系統;其中,移動式的影像伺服資訊,是由非監督式的自我組織映射進行建立;而固定式的影像伺服資訊,則是以監督式的徑向基底函數神經網路進行學習。此整合系統除了能以固定式影像伺服資訊引導攝影機進行監視/拍攝外,也能藉由主動式視覺法的修正方案,漸進地調整兩類神經網路中的學習參數。透過此互助學習方式,固定式攝影系統將可逐漸提升影像伺服資訊的正確性。實驗結果顯示,經過充分的學習次數後,系統便可藉由固定式攝影系統的引導, 加上移動式攝影系統的修正,達到攝影鏡頭快速定位的特性。 |
英文摘要 | In this paper, we present a visual servoing system that combines and integrates the static-camera system and moving-camera system for the cooperative surveillance monitoring. This proposed system employs neural-network-based un-calibrated leaning algorithms for the integration. In the development, it uses neural network of Radial basis function (RBFNN) and that of Self-Organization maps (SOM), respectively, to construct the hand-eye maps for the two camera systems. Information acquired from the static camera is applied by the RBFNN to aim the moving camera at a detected object, and the SOM, with the information from the moving camera, consecutively drives the camera again to locate the image of the object at nearby the center of the filed of view. The resulting SOM information is used to adjust the parameters of both RBFNN and SOM. Experimental results validate this proposed strategy. |
本系統之摘要資訊系依該期刊論文摘要之資訊為主。