本文選題：變化檢測 + Harris角點　； 參考：《浙江大學》2017年碩士論文

【摘要】：圖像技術(shù)在工程應用中使用廣泛,隨著使用范圍的增加,原有的技術(shù)手段與相應算法已不能有效解決新遇到的問題,不斷發(fā)展圖像技術(shù)并用于實際工程中為人類生活帶來各種便利,這是圖像時代的發(fā)展趨勢。論文主要研究基于圖像的變化檢測方法以及相關技術(shù),并結(jié)合實際工程中的相關應用展開論述。論文研究了圖像配準算法和灰度校正算法,這兩種算法可以減少光照變化等情況帶來的干擾。論文詳細介紹了幾種常用的圖像配準算法,分析了這些算法各自的不足。論文主要研究了基于Harris角點的配準算法,著重研究角點的篩選方法,分析了現(xiàn)有的兩種篩選方法的不足,并設計了一種基于矩陣特征值比值的篩選方法。論文研究了基于圖像的變化檢測算法。詳細介紹了基于圖像灰度差值的變化檢測算法和基于LBP直方圖互相關系數(shù)的變化檢測算法并結(jié)合實驗分析其性能。論文還設計了一種對光照變化不敏感的特征,同時利用該特征對圖像進行變化檢測。利用該特征對Haar-like特征改進,設計了基于改進的Haar-like特征和隨機森林的變化檢測方法。最后,論文研究了基于馬爾科夫隨機場并將其應用于圖像的變化檢測。論文介紹了基于馬爾科夫隨機場的圖像分割方法,分析了圖像分割與圖像變化檢測問題的相似之處。本文設計了一種基于馬爾科夫隨機場的融合方法,該方法可以將多種變化檢測算法的檢測結(jié)果圖像融合,融合后的最終結(jié)果比單一的檢測結(jié)果更好。論文結(jié)合上述研究內(nèi)容設計了一個基于圖像的變化檢測系統(tǒng),并給出實驗結(jié)果。
[Abstract]:Image technology is widely used in engineering applications. With the increase of the scope of application, the original technical means and corresponding algorithms can not effectively solve the new problems encountered. It is the development trend of image age to develop image technology and bring all kinds of convenience for human life in practical engineering. This paper mainly studies the image-based change detection methods and related technologies, and discusses the related applications in practical engineering. In this paper, image registration algorithm and gray level correction algorithm are studied, which can reduce the interference caused by illumination change. Several common image registration algorithms are introduced in detail, and their shortcomings are analyzed. This paper mainly studies the registration algorithm based on Harris corner, focuses on the corner selection method, analyzes the shortcomings of two existing screening methods, and designs a selection method based on matrix eigenvalue ratio. The change detection algorithm based on image is studied in this paper. The change detection algorithm based on image gray difference and the change detection algorithm based on the correlation number of LBP histogram are introduced in detail. The performance of the algorithm is analyzed in combination with experiments. A feature which is insensitive to light change is designed and used to detect the change of image. Using this feature to improve the Haar-like feature, an improved Haar-like feature and a change detection method based on random forest are designed. Finally, based on Markov random field, we apply it to image change detection. This paper introduces the method of image segmentation based on Markov random field and analyzes the similarity between image segmentation and image change detection. In this paper, a new fusion method based on Markov random field is designed. This method can fuse the image of various change detection algorithms. The final result of fusion is better than that of single detection. In this paper, a change detection system based on image is designed, and the experimental results are given.
【學位授予單位】：浙江大學
【學位級別】：碩士
【學位授予年份】：2017
【分類號】：TP391.41

【參考文獻】

相關期刊論文 前2條

1 臧克;孫永華;李京;閆志壯;宮輝力;李小娟;趙文吉;;微型無人機遙感系統(tǒng)在汶川地震中的應用[J];自然災害學報;2010年03期

2 張引,潘云鶴;基于模擬退火的最大似然聚類圖像分割算法[J];軟件學報;2001年02期

，

本文編號：1799184