以強度投影的適應性視窗法搜尋和追蹤目標物

Abstract

摘 要 本文提出一個有效率的〝Adaptive window method〞來尋找和追蹤目標物。 本文中，由CCD camera 對場景做影像擷取的工作和一部個人電腦做影像分析所組成。隨著電腦的快速發展，電腦使我們能即時地分析並做處理。在處理搜尋或是追蹤目標物時，最常用又簡單的方法就是Correlation-based method，但此法的計算量相當的大，因此本文主要是根據S-I. Chien and S-H. Sung兩人所提出的〝Adaptive window method〞做了修改，目的即為了縮小搜尋的範圍以減少計算量並加快Adaptive window收斂的速度。 我們利用projection-based method代替correlation-based method 來實現Adaptive window，主要的理由是投影是一種統計的量，有利於去除雜訊，降低複雜的背影造成的影響。另外，投影的演算法所需的計算量很小，有利於加速整個追蹤過程。 實驗是以各種人工的真實影像序列來實現本文所提出的演算法，結果顯示此法可適用於多種含複雜背景及雜訊的影像序列，而符合追蹤的目的。

Abstract According to S-I. Chien and S-H. Sung proposed adaptive window method, an efficient algorithm is proposed for the adaptive sizing of a tracking window. In this thesis, the composition of tracking is a CCD camera which collects sequence images and a person computer which analyze image. With the development of a computer quickly, we can analyze and handle instantly in computer. In tracking or following, a common and efficient method is proposed. Correlation-based method is simple, but it must have a large amount of calculation. For the reason, we modify adaptive window method to fit in with tracking. The purpose is in order to reduce calculation and speed up convergence. We propose projection-based method instead of correlation-based method. Projection is a statistical quantity that could reduce the noises. We could achieve benefit of minimizing the influence of complex background and clutters in tracking process. In addition, projection algorithm has less computation complexity that makes the tracking faster. Experimental results using various artificial and real sequences confirm that the proposed algorithm can effectively adjust a tracking window to a moving target and is robust to a complex background and noise.