發(fā)布時間：2018-06-14 13:26

  本文選題：SURF + 向量空間模型�。� 參考：《電子科技大學(xué)》2017年碩士論文

【摘要】：最近幾年來圖像信息識別和處理的技術(shù)越來越受到關(guān)注,在學(xué)術(shù)方面有研究越來越多的成果,在日常生活場景、工業(yè)、醫(yī)學(xué)、航空航天等方面都得到廣泛的應(yīng)用,圖像信息識別技術(shù)尚且還有很多的潛在應(yīng)用場景和研究意義。于此同時,21世紀(jì)以來,包括Apple、Samsung和各大國產(chǎn)手機廠商的智能手機產(chǎn)品的圖像采集性能和計算能力飛速增長和移動智能手機設(shè)備的普及,人們越來越依賴手機這樣的設(shè)備來提供各種服務(wù)。本文分析了在公共建筑,例如商場環(huán)境下,位置表示非常有局限性的情況,并且室內(nèi)收不到GPS信號的情況下,通過使用智能手機實現(xiàn)室內(nèi)定位功能成為一個熱門應(yīng)用場景。在基于圖像信息識別的室內(nèi)定位方法中,圖像信息識別算法中,圖像的匹配技術(shù)是最關(guān)鍵的一步,圖像匹配技術(shù)的速率和匹配成功率是影響圖像信別效果的重要因素。本文借鑒了伯克利大學(xué)的“基于圖像處理的城市環(huán)境定位研究”中使用的定位系統(tǒng)模型,提出了更適合室內(nèi)環(huán)境的定位系統(tǒng)模型和相對高效的解決辦法。本文根據(jù)此場景,提出了一系列解決方案,主要的研究方向有以下幾個方面:本文詳細分析SIFT和SURF算法的理論方法和實現(xiàn)方式,針對室內(nèi)環(huán)境的場景特點,對比和分析了直接使用SIFT和SURF匹配的優(yōu)點和缺點�；赟URF特征匹配提出了向量空間模型,并且引入了支持向量機算法,實現(xiàn)了圖像的分類,通過在地圖上標(biāo)注出匹配圖像的位置,實現(xiàn)了室內(nèi)定位系統(tǒng)。詳細對比和分析了基于SURF特征值匹配的系統(tǒng)和基于向量空間模型分類器系統(tǒng)的優(yōu)缺點。通過實驗證明了采用SURF特征匹配算法和向量空間模型的分類系統(tǒng)在室內(nèi)定位應(yīng)用中在匹配速率和配準(zhǔn)率上都有良好表現(xiàn)。綜上所述,本文研究了基于圖像信息識別的室內(nèi)定位系統(tǒng),針對室內(nèi)的應(yīng)用場景特點,提出基于SURF的向量空間模型分類系統(tǒng),詳細分析了基于SURF特征的匹配系統(tǒng)的優(yōu)點,并針對此算法進行了相應(yīng)的仿真實驗,證明了其準(zhǔn)確性和時效性。
[Abstract]:In recent years, more and more attention has been paid to the technology of image information recognition and processing. More and more achievements have been made in academic research. They have been widely used in daily life scenes, industry, medicine, aerospace and so on. Image information recognition technology still has many potential applications and research significance. At the same time, since the 21st century, the image acquisition performance and computing power of smartphone products, including Apple Samsung and major domestic mobile phone manufacturers, have grown rapidly and mobile smartphone devices have become popular. People are increasingly relying on devices like mobile phones to provide a variety of services. In this paper, it is analyzed that the location representation is very limited in public buildings, such as shopping malls, and the GPS signal can not be received indoors, so it becomes a hot application scene to realize indoor positioning by using smart phone. In the indoor localization method based on image information recognition, the image matching technology is the most important step in the image information recognition algorithm. The rate and success rate of image matching technology are the important factors that affect the effect of image information recognition. This paper draws lessons from the location system model used in the study of Urban Environment location based on Image processing of Berkeley University, and puts forward a more suitable location system model for indoor environment and a relatively efficient solution. According to this scenario, this paper puts forward a series of solutions, the main research directions are as follows: this paper analyzes the theoretical methods and implementation methods of sift and surf algorithm in detail, aiming at the characteristics of indoor environment. The advantages and disadvantages of direct use of sift and surf are compared and analyzed. A vector space model based on surf feature matching is proposed, and support vector machine (SVM) algorithm is introduced to realize the classification of images. The indoor positioning system is implemented by marking the location of the matching image on the map. The advantages and disadvantages of the system based on surf eigenvalue matching and vector space model classifier are compared and analyzed in detail. It is proved by experiments that the classification system based on surf feature matching algorithm and vector space model has good performance in both matching rate and registration rate in indoor localization applications. To sum up, this paper studies the indoor positioning system based on image information recognition. According to the characteristics of indoor application scene, a vector space model classification system based on SURF is proposed, and the advantages of the matching system based on SURF feature are analyzed in detail. The simulation results show that the algorithm is accurate and time-efficient.
【學(xué)位授予單位】：電子科技大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2017
【分類號】：TP391.41;TN929.53

【參考文獻】

相關(guān)期刊論文 前1條

1 楊曉敏;嚴(yán)斌宇;李康麗;蘇冰山;;一種基于詞袋模型的圖像分類方法[J];太赫茲科學(xué)與電子信息學(xué)報;2014年05期

相關(guān)碩士學(xué)位論文 前3條

1 劉宇;基于圖像處理的定位系統(tǒng)研究[D];電子科技大學(xué);2015年

2 劉程;基于詞袋模型的圖像分類技術(shù)研究[D];安徽大學(xué);2014年

3 朱道廣;基于視覺詞袋模型的圖像分類研究[D];解放軍信息工程大學(xué);2013年

，

本文編號：2017537