Adaptive Visual Camouflage System

偽裝技術在近代軍事發展中佔了相當重要的一部分，除了視覺上的偽裝外另外還包涵了聽覺、電磁波、熱能以及物體形態上的偽裝。目前大部分的研究都致力於開發出各種反偵測的材料，靠著調整或抑制物體所發出的輻射，覆蓋這些材料的物體可以有效的降低被偵測到的機率。而關於視覺方面卻通常僅有提供一些固定的偽裝圖案，或者是對照周遭環境的平均顏色而改變物體的顏色。為了加強視覺上的偽裝，我們之前開發了一個智能化的動態偽裝系統，它可以因應背景的變化隨時更新偽裝圖案，但是必須事先得到觀察者和背景深度的資訊。在此篇論文中我們設計了兩個方法，其中第一個方法沿用了舊系統架構(穿透式偽裝) ，為了解決估計觀察者深度的問題我們定義了有效偽裝的距離，並且在此段距離中尋找一個最合適的偽裝圖案。之前必須根據觀察者不同的位置找出對應的偽裝圖案，現在在不同深度下只需要同一個偽裝圖案所以也不再需要偵查觀察者的深度。第二個方法主要適用於戶外大型偽裝物，在缺乏觀察者資訊的情況下，因為無法取得相對背景所以只能找出背景中最主要的紋理來當作偽裝圖案。此方法中使用了背景圖中彩度跟亮度的資訊建立了兩個濾波器，可以用來保存背景中包涵主要紋理的區域以及過濾不需要的其它物體。

關鍵字

偽裝；適應性；視覺

並列摘要

Camouflage is an important issue of modern military. Besides visual spectrum, it is also considered in acoustic, the electromagnetic spectrum, thermal and shape. To counteracting various detection systems, most researches are devoted to develop materials that suppress or adjust object’s radiation. Visual concealments are usually some kinds of fixed patterns or average color of the surroundings. Our previous work had constructed an active visual camouflage system, which can change the camouflage patterns as changes occur in the background. But we encountered great difficulties in estimating the foreground and the background depth. In this thesis, we solve the problem by defining the effective range of camouflage and finding the most appropriate patch for camouflage within the range. Information of observer’s depth is then no longer indispensable. We also propose a method, which extracts the dominant texture of the background for camouflage. Two masks utilizing information of chroma and intensity of background image are involved to preserve regions belonging to dominant texture and to filter unwanted object in the image. This method is useful when it lacks observer’s information.

並列關鍵字

camouflage ； adaptive ； visual

參考文獻

[1] Rajesh Nambia,“Modern Camouflage Techniques”.

[6] Inami M., Kawakami N., and Tachi S,“Optical camouflage using retro-reflective projection technology”, The Second IEEE and ACM International Symposium on Mixed and Augmented Reality ,Page(s): 348 - 349, 2003.

[8] Hwann-Tzong Chen and Tyng-Luh Liu, “Finding Familiar Objects and their Depth from a Single Image”, IEEE International Conference on Image Processing, Volume: 6 , Page(s): VI - 389 - VI – 392, 2007.

[9] J Lu, J Dorsey,and H Rushmeier, “Dominant Texture and Diffusion Distance Manifolds”, Computer Graphics Forum,Volume 28, Number 2, 2009.

[11] ZALESNY A., FERRARI V., CAENEN G., and GOOL L. V., “Composite texture synthesis” Int. J. Comput. Vision 62, 1-2 (2005), 161–176.

國際替代計量

Adaptive Visual Camouflage System

全文下載

主題瀏覽