本文選題：卷積神經(jīng)網(wǎng)絡(luò) + 行人檢測　； 參考：《哈爾濱工業(yè)大學(xué)》2017年碩士論文

【摘要】：交通環(huán)境中的行人檢測問題對于算法的檢測精度和速度都有較高要求。傳統(tǒng)方法能夠滿足速度要求,但在精度上差距較遠(yuǎn),基于卷積神經(jīng)網(wǎng)絡(luò)的方法精度較高,但是計算量巨大。本文主要在基于卷積神經(jīng)網(wǎng)絡(luò)的基礎(chǔ)上構(gòu)建精度更高速度更快的行人檢測方法。首先,本文分別從網(wǎng)絡(luò)結(jié)構(gòu)設(shè)計,損失函數(shù)設(shè)計,正則化方法和優(yōu)化策略四個方面總結(jié)了卷積網(wǎng)絡(luò)應(yīng)用相關(guān)的基礎(chǔ)知識;然后以MNIST數(shù)據(jù)集上的手寫數(shù)字識別為例,重點比較了梯度下降優(yōu)化算法不同變體之間的差別,為優(yōu)化算法的選擇提供了實踐依據(jù)。其次,在目標(biāo)檢測領(lǐng)域內(nèi)領(lǐng)先的Faster RCNN方法基礎(chǔ)上,基于總結(jié)的設(shè)計準(zhǔn)則和行人的尺度特性,調(diào)整了網(wǎng)絡(luò)的錨點窗口設(shè)置和區(qū)域生成網(wǎng)絡(luò)方式,添加了環(huán)境區(qū)域池化層。然后基于開源深度學(xué)習(xí)框架對該網(wǎng)絡(luò)在加州理工行人數(shù)據(jù)集上進(jìn)行了實現(xiàn),實驗結(jié)果表明該方法可以實現(xiàn)高效的行人檢測。然后,針對單尺度區(qū)域生成網(wǎng)絡(luò)由于輸入特征圖固定造成的無法兼顧大小行人檢測精度的問題,設(shè)計了基于特征圖分層的多尺度區(qū)域生成網(wǎng)絡(luò),并為之設(shè)計了相應(yīng)的隨機(jī)縮放裁剪數(shù)據(jù)擴(kuò)增方法應(yīng)對訓(xùn)練不均衡問題。實驗結(jié)果表明,該網(wǎng)絡(luò)能夠在輸入分辨率較低的情況下實現(xiàn)比單尺度方法在高分辨率輸入情況下更高的檢測精度。最后,針對測試過程中檢測速度慢的問題,設(shè)計了基于奇異值分解和Tucker-2分解的全連接層和卷積層的壓縮方法,分別將高維的全連接層和卷積層近似為級聯(lián)的低維全連接層和卷積層。結(jié)果表明,通過“訓(xùn)練-分解-調(diào)優(yōu)”的三段壓縮方式,該方法能夠在不明顯損失檢測精度的情況下實現(xiàn)單層4倍、總體1.6倍的加速和總體4倍的模型大小壓縮。壓縮后的網(wǎng)絡(luò)在GTX1080顯卡加速下能夠達(dá)到30幀每秒。
[Abstract]:Pedestrian detection in traffic environment requires high detection accuracy and speed. The traditional method can meet the requirement of speed, but the accuracy is far behind. The method based on convolution neural network has high accuracy, but the computation is huge. In this paper, a more accurate and faster pedestrian detection method is constructed based on convolution neural network. Firstly, this paper summarizes the basic knowledge of convolution network application from four aspects of network structure design, loss function design, regularization method and optimization strategy, and then takes handwritten digit recognition on MNIST dataset as an example. The differences between different variants of gradient descent optimization algorithm are compared, which provides a practical basis for the selection of optimization algorithm. Secondly, on the basis of the leading Faster RCNN method in the field of target detection, based on the summary design criteria and pedestrian scale characteristics, the network anchor window setting and area generation network mode are adjusted, and the environmental area pool layer is added. Then the network is implemented on the California Institute of Technology pedestrian data set based on open source deep learning framework. The experimental results show that the proposed method can achieve efficient pedestrian detection. Then, aiming at the problem that the single scale region generation network can not take into account the size of pedestrian detection accuracy due to the fixed input feature map, a multi-scale region generation network based on feature graph layer is designed. And designed the corresponding random scaling clipping data amplification method to deal with the problem of uneven training. The experimental results show that the network can achieve higher detection accuracy than the single-scale method in the case of high resolution input with low input resolution. Finally, aiming at the problem of slow detection speed in the testing process, a compression method of full connection layer and convolution layer based on singular value decomposition and Tucker-2 decomposition is designed. The high dimensional full connection layer and convolution layer are approximated as cascade low dimensional full connection layer and convolution layer respectively. The results show that by using the three-stage compression method of "training-decomposition-tuning", the method can achieve four times of single layer, 1.6 times of total acceleration and 4 times of model size compression without obvious loss of detection accuracy. The compressed network can reach 30 frames per second with the acceleration of GTX 1080 graphics card.
【學(xué)位授予單位】：哈爾濱工業(yè)大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2017
【分類號】：TP391.41;TP183

【參考文獻(xiàn)】

相關(guān)期刊論文 前2條

1 李海龍;吳震東;章堅武;;基于區(qū)域卷積神經(jīng)網(wǎng)絡(luò)的行人檢測[J];通信技術(shù);2017年04期

2 丁文祥;朱孔凡;;基于卷積神經(jīng)網(wǎng)絡(luò)的實時視頻下行人檢測[J];信息技術(shù);2016年12期

相關(guān)博士學(xué)位論文 前1條

1 任少卿;基于特征共享的高效物體檢測[D];中國科學(xué)技術(shù)大學(xué);2016年

相關(guān)碩士學(xué)位論文 前2條

1 楊眷玉;基于卷積神經(jīng)網(wǎng)絡(luò)的物體識別研究與實現(xiàn)[D];電子科技大學(xué);2016年

2 王斌;基于深度學(xué)習(xí)的行人檢測[D];北京交通大學(xué);2015年

，

本文編號：2109442