《電子技術(shù)應(yīng)用》
您所在的位置:首頁(yè) > 嵌入式技術(shù) > 設(shè)計(jì)應(yīng)用 > 基于深度神經(jīng)網(wǎng)絡(luò)的偽裝人臉識(shí)別
基于深度神經(jīng)網(wǎng)絡(luò)的偽裝人臉識(shí)別
2020年電子技術(shù)應(yīng)用第5期
張潤(rùn)生1,2,3,,賀 超1,2,,3,況朝青1,,2,,3
1.重慶郵電大學(xué) 通信與信息工程學(xué)院,,重慶400065,;2.重慶高校市級(jí)光通信與網(wǎng)絡(luò)重點(diǎn)實(shí)驗(yàn)室,重慶400065,; 3.泛在感知與互聯(lián)重慶市重點(diǎn)實(shí)驗(yàn)室,,重慶400065
摘要: 偽裝人臉識(shí)別在刑偵安防領(lǐng)域有著巨大的應(yīng)用價(jià)值。針對(duì)現(xiàn)階段對(duì)偽裝人臉識(shí)別的研究較少,、算法魯棒性不強(qiáng)等缺點(diǎn),,提出了基于深度神經(jīng)網(wǎng)絡(luò)的偽裝人臉識(shí)別算法。改進(jìn)了SqueezeNet網(wǎng)絡(luò)模型,,并將其與FaceNet網(wǎng)絡(luò)架構(gòu)進(jìn)行結(jié)合,,用于人臉圖像的身份識(shí)別。通過(guò)在訓(xùn)練數(shù)據(jù)集中引入偽裝人臉圖像,,讓網(wǎng)絡(luò)學(xué)習(xí)到偽裝的特征,。實(shí)驗(yàn)結(jié)果表明,該算法識(shí)別準(zhǔn)確率接近90%,,相較于其他網(wǎng)絡(luò)模型,,具有更好的識(shí)別效果。
中圖分類號(hào): TN911.73,;TP391.4
文獻(xiàn)標(biāo)識(shí)碼: A
DOI:10.16157/j.issn.0258-7998.191314
中文引用格式: 張潤(rùn)生,,賀超,況朝青. 基于深度神經(jīng)網(wǎng)絡(luò)的偽裝人臉識(shí)別[J].電子技術(shù)應(yīng)用,,2020,,46(5):27-30.
英文引用格式: Zhang Runsheng,He Chao,,Kuang Chaoqing. Disguised face recognition based on deep neural network[J]. Application of Electronic Technique,,2020,46(5):27-30.
Disguised face recognition based on deep neural network
Zhang Runsheng1,,2,,3,He Chao1,2,,3,,Kuang Chaoqing1,2,,3
1.School of Communication and Information Engineering,,Chongqing University of Posts and Telecommunications, Chongqing 400065,,China,; 2.Optical Communication and Networks Key Laboratory of Chongqing,Chongqing 400065,,China,; 3.Ubiquitous Sensing and Networking Key Laboratory of Chongqing,Chongqing 400065,,China
Abstract: Camouflage face recognition has great application value in the field of criminal investigation and security. Aiming at the shortcomings of few researches on camouflage face recognition and weak robustness at present, a camouflaged face recognition algorithm based on deep neural network is proposed. The SqueezeNet network model has been improved and combined with the FaceNet network architecture for identity recognition of face images. By adding camouflage face images in the training data set, the network can learn the characteristics of the camouflages. The experimental results show that the recognition accuracy of the algorithm is close to 90%, which is better than other network models.
Key words : deep neural network,;disguised face recognition;SqueezeNet network model,;FaceNet network architecture

0 引言

    近年來(lái),,視頻監(jiān)控領(lǐng)域下的人臉識(shí)別得到了廣泛關(guān)注,通過(guò)監(jiān)控抓捕到犯罪嫌疑人的新聞時(shí)有出現(xiàn),,大大提高了案件的偵破率,。但是監(jiān)控拍攝到的圖像中很可能會(huì)存在遮擋,使得識(shí)別率下降,,錯(cuò)失抓捕嫌疑人的機(jī)會(huì),。遮擋一般分為兩種,即自然遮擋和人為偽裝[1],。自然遮擋包括樹(shù)葉,、欄桿等,人為偽裝包括帽子,、墨鏡,、圍巾等。通過(guò)偽裝,,犯罪分子可以逃避監(jiān)控的追蹤,,增大了案件的偵破難度。

    針對(duì)這些問(wèn)題,,文獻(xiàn)[2]提出了一種基于遮擋模式的稀疏表示分類的方法,,構(gòu)建的解析詞典與測(cè)試圖像具有相同的遮擋,提高了分類性能,;利用稀疏字典學(xué)習(xí)的判別性來(lái)處理人臉識(shí)別問(wèn)題中的連續(xù)遮擋,。文獻(xiàn)[3]使用Gabo小波,、PCA和SVM來(lái)解決遮擋檢測(cè)問(wèn)題,將人臉圖像分成兩個(gè)相等的分量,,從每個(gè)分量中提取Gabor小波特征,,用于降維主成分分析,最后使用局部二值模式來(lái)完成識(shí)別過(guò)程,;在識(shí)別期間,,權(quán)重被分配給測(cè)試圖像的每個(gè)局部區(qū)域,與給定未被遮擋的訓(xùn)練示例的每個(gè)區(qū)域的可能性成比例,。文獻(xiàn)[4]在人臉圖像的每個(gè)點(diǎn)上找到最大匹配區(qū)域,,提取其傅里葉幅度譜作為特征,最后使用余弦相似度進(jìn)行識(shí)別,。但是這些方法都是針對(duì)一定類型的遮擋通過(guò)建模來(lái)完成識(shí)別的,,泛化性能較差;而神經(jīng)網(wǎng)絡(luò)能夠通過(guò)大量數(shù)據(jù)的訓(xùn)練來(lái)學(xué)習(xí)到相關(guān)特征,,獲得更好的識(shí)別性能,,在信號(hào)調(diào)制,、計(jì)算機(jī)視覺(jué),、文本分析、故障檢測(cè)等領(lǐng)域均有廣泛的應(yīng)用[5-8],。




文章詳細(xì)內(nèi)容請(qǐng)下載http://forexkbc.com/resource/share/2000002785




作者信息:

張潤(rùn)生1,,2,3,,賀  超1,,2,3,,況朝青1,,2,3

(1.重慶郵電大學(xué) 通信與信息工程學(xué)院,,重慶400065,;2.重慶高校市級(jí)光通信與網(wǎng)絡(luò)重點(diǎn)實(shí)驗(yàn)室,重慶400065,;

3.泛在感知與互聯(lián)重慶市重點(diǎn)實(shí)驗(yàn)室,,重慶400065)

此內(nèi)容為AET網(wǎng)站原創(chuàng),未經(jīng)授權(quán)禁止轉(zhuǎn)載,。