本文選題：深度學(xué)習(xí) + 深度置信網(wǎng)絡(luò)�。� 參考：《南昌大學(xué)》2017年博士論文

【摘要】：隨著圖像拍攝設(shè)備、智能手機(jī)和互聯(lián)網(wǎng)技術(shù)的發(fā)展,紋身圖像的采集、傳播變得越來(lái)越容易。伴隨著突發(fā)事件的發(fā)展,紋身同其它生物特征一樣,成為對(duì)罪犯嫌疑人識(shí)別的有力證據(jù)。如何對(duì)紋身圖像進(jìn)行檢測(cè)和語(yǔ)義解讀,并為相關(guān)部門和人員提供有力的證據(jù),已引起安全部門的重視。紋身圖像具有明顯的圖案信息的局部性、內(nèi)容的復(fù)雜性、紋理的清晰性、顏色的單一性、圖案Logo標(biāo)志性、大小形狀多樣性等特點(diǎn)。這些特點(diǎn)使得紋身圖像的檢測(cè)與識(shí)別相對(duì)比較困難,同時(shí)也使得很難用單一的特征對(duì)其描述。深度學(xué)習(xí)通過(guò)逐層追疊形成多層的網(wǎng)絡(luò)結(jié)構(gòu),這種結(jié)構(gòu)可以從底層到高層逐層提取到圖像的高層特征,從而有效對(duì)圖像進(jìn)行表示。這使得深度學(xué)習(xí)逐漸成為學(xué)術(shù)界、企業(yè)界研究的熱點(diǎn),同時(shí)也為紋身圖像檢測(cè)提供了一種新的途徑。本文的主要工作是緊密圍繞紋身圖像的特點(diǎn)和現(xiàn)有的深度學(xué)習(xí)理論,展開深入的研究,提出針對(duì)紋身圖像檢測(cè)的若干改進(jìn)算法。本文的主要工作和貢獻(xiàn)包括以下幾個(gè)方面:1.四種主要深度學(xué)習(xí)算法在紋身圖像檢測(cè)中的比較研究。通過(guò)分析紋身圖像的特點(diǎn),探討哪種算法比較適用于紋身檢測(cè)的研究工作。實(shí)驗(yàn)結(jié)果表明,在紋身檢測(cè)方面,四種深度學(xué)習(xí)算法比傳統(tǒng)的方法的性能更好,其中深度卷積神經(jīng)網(wǎng)絡(luò)和深度置信網(wǎng)絡(luò)的性能更為突出。2.基于多特征融合的深度置信網(wǎng)絡(luò)紋身圖像檢測(cè)改進(jìn)算法(MF-DBN)。紋身圖像的內(nèi)容復(fù)雜性、紋理的清晰性、顏色的單一性等諸多特點(diǎn)決定了單一特征很難對(duì)其進(jìn)行準(zhǔn)確描述,同時(shí)像素又比較高。傳統(tǒng)的DBN算法比較適中小尺寸的圖像識(shí)別任務(wù),針對(duì)這些問(wèn)題,從多特征融合的視角,設(shè)計(jì)了一個(gè)基于紋身圖像的多特征融合深度置信網(wǎng)絡(luò)改進(jìn)算法(MF-DBN),有效地解決了紋身圖像高維性和單一特征描述不足的問(wèn)題。在NIST紋身數(shù)據(jù)集上正確率達(dá)到96.89%,比NIST公布的最好正確率提高了0.59%。3.基于視覺詞包的深度置信網(wǎng)絡(luò)紋身圖像檢測(cè)改進(jìn)算法(BOVW-DBN)。該算法不僅解決了輸入高維性的問(wèn)題,還針對(duì)紋身圖案的局部性、大小不一性等特點(diǎn),利用SIFT算法,建立中層詞包語(yǔ)義模型,有效的實(shí)現(xiàn)對(duì)紋身圖像的檢測(cè)與識(shí)別。在NIST紋身數(shù)據(jù)集上正確率達(dá)到97.57%,比NIST公布的最好正確率提高了1.17%;在Flickr 10K上的正確率也達(dá)到79.27%。4.基于空間金字塔的深度置信網(wǎng)絡(luò)紋身圖像檢測(cè)改進(jìn)算法(SP-DBN)。該算法利用SPM模型,針對(duì)紋身圖像的大小及空間分布等問(wèn)題,解決了紋身圖像的空間信息特征提取問(wèn)題,實(shí)現(xiàn)紋身圖像檢測(cè)。實(shí)驗(yàn)結(jié)果顯示,在NIST紋身數(shù)據(jù)集上正確率達(dá)到97.23%,比NIST公布的最好正確率提高了0.93%;在Flickr 10K上的正確率達(dá)到80.46%。5.基于三通道融合的卷積神經(jīng)網(wǎng)絡(luò)紋身圖像檢測(cè)改進(jìn)算法(CFT-CNN)。針對(duì)全連接層在不同尺度下的特征抽取能力,首先根據(jù)紋身圖像的檢測(cè)問(wèn)題設(shè)計(jì)了一個(gè)簡(jiǎn)單的單通道T-CNN模型;在單通道T-CNN模型的基礎(chǔ)上,又設(shè)計(jì)出一個(gè)三通道連接層的卷積神經(jīng)網(wǎng)絡(luò)模型(CFT-CNN),并應(yīng)用到紋身圖像檢測(cè)的任務(wù)中。同時(shí)針對(duì)紋身圖像的特點(diǎn)做了相應(yīng)的預(yù)處理。在NIST數(shù)據(jù)集上,CFT-CNN的正確率達(dá)到97.87%,比NIST公布的最好結(jié)果提高了1.57%,在Flickr 10K數(shù)據(jù)集上的正確率也達(dá)到85.61%。6.基于三通道融合的Faster R-CNN紋身圖案檢測(cè)改進(jìn)算法(CFT Faster RCNN)。該算法針對(duì)紋身圖像大小、尺度變化大等問(wèn)題,在Faster R-CNN的基礎(chǔ)上,充分考慮到全連接層在不同尺度下的特征提取能力,在ROI池化后增加一個(gè)三通道的全連接層,解決了紋身圖像位置識(shí)別困難的問(wèn)題。實(shí)驗(yàn)結(jié)果表明,在NIST數(shù)據(jù)集上,本算法比Faster R-CNN算法在MAP上提同了3.35%,在IOU上提高了4.59%。最后,對(duì)基于DBN和CNN的紋身圖像檢測(cè)改進(jìn)算法進(jìn)行了對(duì)比研究。結(jié)果表明改進(jìn)算法在NIST上性能有所提升,其中改進(jìn)的CNN算法的效果更好。在改進(jìn)的DBN方法中,由于詞包模型是在SIFT特征基礎(chǔ)上對(duì)紋身圖像進(jìn)行表示,在小數(shù)據(jù)集上更好。SP-DBN算法由于考慮到空間信息,對(duì)大一些的數(shù)據(jù)集的效果相對(duì)較好。從NIST和Flickr兩個(gè)數(shù)據(jù)集的實(shí)驗(yàn)結(jié)果來(lái)看,NIST數(shù)據(jù)集存在一定局限性,Flickr數(shù)據(jù)集更接近現(xiàn)實(shí)環(huán)境。
[Abstract]:With the development of image photographing equipment, smart phone and Internet technology, the collection of tattoo images has become more and more easy. With the development of unexpected events, the tattoo, like other biological features, has become a powerful evidence for the identification of criminal suspects. How to detect and interpret tattoo images and to be related departments and people The staff provide strong evidence, which has aroused the attention of the security department. The tattoo image has the features of obvious pattern information, the complexity of the content, the clarity of the texture, the singleness of the color, the Logo logo of the pattern, the diversity of the size and shape, which make the detection and recognition of the tattoo image relatively difficult and also make it possible It is difficult to describe it with a single feature. Deep learning forms a multilayer network structure by overlapping layer by layer. This structure can be extracted from the bottom to the high level to the high level feature of the image, thus effectively expressing the image. This makes the deep learning gradually become a hot topic in the academic circle, and also for the tattoo image inspection. The main work of this paper is to focus on the characteristics of the tattoo image and the existing depth learning theory, and to carry out an in-depth study and propose some improved algorithms for the tattoo image detection. The main work and contributions of this paper include the following aspects: 1. the four main depth learning algorithms are in tattoo image detection. By analyzing the characteristics of the tattoo image, this paper discusses which algorithm is suitable for the research of tattoo detection. The experimental results show that, in the aspect of tattoo detection, the performance of the four depth learning algorithms is better than the traditional method, and the performance of the deep convolution neural network and the depth confidence network is more prominent based on the.2. based on the tattoo detection. The improved algorithm for multi feature fusion of deep confidence network tattoo image detection (MF-DBN). The complexity of the content of the tattoo image, the clarity of the texture, the singleness of the color, and so on, determine that the single feature is difficult to accurately describe it, and the pixels are relatively high. The traditional DBN algorithm compares the image recognition task with the small size and size. To these problems, from the perspective of multi feature fusion, a multi feature fusion depth confidence network improvement algorithm (MF-DBN) based on the tattoo image is designed, which effectively solves the problem of the lack of high dimension and single feature description of tattoo images. The accuracy of the NIST tattoo data set is 96.89%, which is 0. higher than the best correct rate published by NIST. 59%.3. based on the visual word packet, the improved algorithm (BOVW-DBN) of the depth confidence network tattoo image detection (BOVW-DBN). This algorithm not only solves the problem of high dimension of input, but also aims at the features of the locality and size of the tattoo pattern, and uses the SIFT algorithm to establish the middle word packet semantic model, effectively realizing the detection and recognition of the tattoo image. In NIST The correct rate of the tattoo data set is 97.57%, which is 1.17% better than the best correct rate published by NIST; the correct rate on the Flickr 10K also reaches 79.27%.4. based on the improved algorithm of the depth confidence network tattoo image detection based on space Pyramid (SP-DBN). The algorithm uses the SPM model to solve the problem of the size and spatial distribution of the tattoo image. The experimental results show that the correct rate of the NIST tattoo data set is 97.23%, which is 0.93% better than the best correct rate published by NIST; the correct rate on the Flickr 10K has reached the improved algorithm of the convolution neural network image detection based on the three channel fusion of 80.46%.5.. CFT-CNN). In view of the feature extraction ability of all connected layers at different scales, a simple single channel T-CNN model is designed according to the detection of tattoo image. On the basis of single channel T-CNN model, a convolution neural network model (CFT-CNN) for three channel connection layer is designed, and the task is applied to the task of tattoo image detection. In the NIST data set, the correct rate of CFT-CNN is 97.87%, which is 1.57% higher than the best result published by NIST, and the correct rate on the Flickr 10K data set also reaches the 85.61%.6. based Faster R-CNN tattoo pattern detection improvement algorithm based on the three channel fusion (CFT Faster RCNN). In view of the size of the tattoo image and the large scale change, on the basis of Faster R-CNN, we fully consider the feature extraction ability of the full connection layer at different scales, and add a three channel full connection layer after the ROI pool, which solves the problem of the difficult position recognition of the tattoo image. The experimental results show that this calculation is based on the NIST data set. The method is compared with the Faster R-CNN algorithm on MAP, which is 3.35%, the 4.59%. is improved on IOU, and the improved algorithm based on DBN and CNN is compared. The results show that the improved algorithm improves the performance on NIST, and the improved CNN algorithm has better effect. In the improved DBN method, the word packet model is in SIFT. On the basis of features, the image of tattoo is expressed. On the small data set, the better.SP-DBN algorithm has a better effect on the larger data sets because of the consideration of spatial information. From the experimental results of two data sets of NIST and Flickr, the NIST dataset has some limitations, and the Flickr data set is closer to the real environment.

【學(xué)位授予單位】：南昌大學(xué)
【學(xué)位級(jí)別】：博士
【學(xué)位授予年份】：2017
【分類號(hào)】：TP391.41

【相似文獻(xiàn)】

相關(guān)期刊論文 前10條

1 曹玉東;;基于融合特征的近似圖像檢測(cè)方法[J];計(jì)算機(jī)技術(shù)與發(fā)展;2012年08期

2 敖琪;痕跡圖像檢測(cè)系統(tǒng)的研制與應(yīng)用[J];警察技術(shù);2001年04期

3 焦圣喜;韓立強(qiáng);岳曉峰;賈瑞慶;;一種智能圖像檢測(cè)的方法[J];制造業(yè)自動(dòng)化;2006年12期

4 吳春亞;劉獻(xiàn)禮;王玉景;王鵬;劉軍;趙彥玲;;機(jī)械加工表面粗糙度的圖像檢測(cè)方法[J];哈爾濱理工大學(xué)學(xué)報(bào);2007年03期

5 顏紅金;張永林;;醫(yī)用眼睛圖像檢測(cè)系統(tǒng)的構(gòu)建[J];計(jì)算機(jī)工程與設(shè)計(jì);2010年24期

6 高波;張曉桂;王瑜;;一種折頁(yè)錯(cuò)位的圖像檢測(cè)方法的探討[J];中國(guó)印刷與包裝研究;2010年S1期

7 姚彩虹;蔡樂(lè)才;祝志威;;嵌入式圖像檢測(cè)的一種解決方案及應(yīng)用前景[J];四川理工學(xué)院學(xué)報(bào)(自然科學(xué)版);2012年01期

8 張海偉;紀(jì)玉波;;人臉圖像檢測(cè)與正規(guī)化的研究[J];電子設(shè)計(jì)工程;2012年20期

9 姚金良;羅蘭;;近似拷貝圖像檢測(cè)綜述[J];杭州電子科技大學(xué)學(xué)報(bào);2013年04期

10 張正華;余松煜;喬艷濤;方志軍;;一種新穎的抗失真的圖像檢測(cè)算法[J];計(jì)算機(jī)應(yīng)用與軟件;2007年04期

相關(guān)會(huì)議論文 前5條

1 劉鐵根;鄧集杰;朱均超;李晉申;江俊峰;;嵌入式圖像檢測(cè)一體機(jī)研究[A];2006年全國(guó)光電技術(shù)學(xué)術(shù)交流會(huì)會(huì)議文集（D 光電信息處理技術(shù)專題）[C];2006年

2 梁振新;張建勛;鄒林;裴怡;;焊接操作模擬中焊條位置圖像檢測(cè)方法[A];第十一次全國(guó)焊接會(huì)議論文集（第2冊(cè)）[C];2005年

3 劉鐵根;朱均超;鄧集杰;王雙;楊永;;基于高速DSP嵌入式圖像檢測(cè)系統(tǒng)研究[A];中國(guó)光學(xué)學(xué)會(huì)2006年學(xué)術(shù)大會(huì)論文摘要集[C];2006年

4 鄭嶸;方志宏;夏勇;;板坯縱向裂紋的圖像檢測(cè)方法[A];中國(guó)計(jì)量協(xié)會(huì)冶金分會(huì)2012年會(huì)暨能源計(jì)量與節(jié)能降耗經(jīng)驗(yàn)交流會(huì)論文集[C];2012年

5 顏建強(qiáng);高新波;田春娜;李英;李亮;;多媒體信息服務(wù)中色情圖像檢測(cè)算法[A];第十五屆全國(guó)圖象圖形學(xué)學(xué)術(shù)會(huì)議論文集[C];2010年

相關(guān)博士學(xué)位論文 前5條

1 許慶勇;基于深度學(xué)習(xí)理論的紋身圖像檢測(cè)研究[D];南昌大學(xué);2017年

2 蔡波;橋梁形變的圖像檢測(cè)關(guān)鍵技術(shù)研究[D];中國(guó)工程物理研究院;2016年

3 李力行;基于圖像檢測(cè)的空間激光通信鏈路誤差校正技術(shù)研究[D];中國(guó)科學(xué)院光電技術(shù)研究所;2017年

4 喬鬧生;PCB自動(dòng)光學(xué)檢測(cè)的圖像檢測(cè)與處理關(guān)鍵技術(shù)研究[D];電子科技大學(xué);2010年

5 劉思遠(yuǎn);基于高維視覺特征模型的目標(biāo)圖像檢測(cè)與圖像分割技術(shù)研究[D];電子科技大學(xué);2006年

相關(guān)碩士學(xué)位論文 前10條

1 吳維起;基于FPGA的高分辨率線陣CCD圖像檢測(cè)系統(tǒng)的設(shè)計(jì)與實(shí)現(xiàn)[D];國(guó)防科學(xué)技術(shù)大學(xué);2014年

2 吳坤;近似拷貝圖像檢測(cè)[D];中國(guó)計(jì)量學(xué)院;2015年

3 胡越浪;基于神經(jīng)網(wǎng)絡(luò)的水稻紋枯病圖像檢測(cè)[D];四川農(nóng)業(yè)大學(xué);2014年

4 孫珂琪;圖像檢測(cè)中圓錐螺紋牙型輪廓重構(gòu)研究[D];太原科技大學(xué);2015年

5 王勛爵;基于圖像檢測(cè)的自校準(zhǔn)水平儀設(shè)計(jì)與實(shí)現(xiàn)[D];東南大學(xué);2015年

6 馬登輝;出租車視頻監(jiān)控的異常圖像檢測(cè)與乘客識(shí)別研究[D];長(zhǎng)安大學(xué);2016年

7 胡幫義;基于數(shù)字圖像檢測(cè)和虛擬現(xiàn)實(shí)的橋梁施工控制技術(shù)研究[D];廣州大學(xué);2016年

8 趙龍飛;微光圖像檢測(cè)方法的研究[D];天津大學(xué);2014年

9 楊凱;基于局部圖像特征的圖像檢測(cè)方法研究及應(yīng)用[D];山東中醫(yī)藥大學(xué);2016年

10 李先斌;近似重復(fù)圖像檢測(cè)及其應(yīng)用[D];中國(guó)科學(xué)技術(shù)大學(xué);2010年

，

本文編號(hào)：1890357