(34.237.124.210) 您好!臺灣時間:2021/02/25 19:26
字體大小: 字級放大   字級縮小   預設字形  
回查詢結果

詳目顯示:::

我願授權國圖
: 
twitterline
研究生:江岳勳
研究生(外文):Yueh-Hsing Chiang
論文名稱:使用深度卷積神經網路與混合圖像資料擴增進行睡意駕駛辨識
論文名稱(外文):Drowsy Drivers Recognition using Deep CNNs and Mixing Images Data Augmentation
指導教授:陳煥陳煥引用關係
指導教授(外文):Huan Chen
口試委員:許志義陳牧言
口試委員(外文):Jyh-Yih HsuMu-Yen Chen
口試日期:2019-12-05
學位類別:碩士
校院名稱:國立中興大學
系所名稱:資訊科學與工程學系所
學門:工程學門
學類:電資工程學類
論文種類:學術論文
論文出版年:2019
畢業學年度:108
語文別:中文
論文頁數:71
中文關鍵詞:深度卷積神經網路資料擴增混合圖像睡意駕駛汽車事故
外文關鍵詞:Deep Convolutional NetworksData AugmentationDrowsy DriversAutomobile AccidentsMixing Images
相關次數:
  • 被引用被引用:0
  • 點閱點閱:80
  • 評分評分:系統版面圖檔系統版面圖檔系統版面圖檔系統版面圖檔系統版面圖檔
  • 下載下載:0
  • 收藏至我的研究室書目清單書目收藏:0
睡意駕駛辨識是現在許多行車場景中需要的警戒功能。因此在這種情形下,我們針對駕駛的睡意行為關注並加以辨識,藉此提醒駕駛並降低汽車事故發生率。本論文主要目的為採用結合混合圖像資料擴增方式與深度卷積神經網路方法對於睡意與非睡意情形進行辨識分類。近年此資料集相關研究主要著重於使用影像處理演算法或特徵工程複雜前處理方法來提升辨識準確度。這些前處理方法在處理過程中或許會因挑選指標、方法丟失重要的分類特徵。為了避免特徵擷取所產生的分類偏差,本論文使用混合圖像資料擴增為主要技術,利用圖像互相混合情形產生新型未知樣本,再以五種深度卷積神經網路進行搭配分類。藉由擴大原始資料樣本特徵訊息量範圍以靠近真正檢測目標測試樣本特徵訊息量範圍來提升分類準確率。經實驗結果顯示,在使用約90%訓練集與10%測試集時的驗證方式其準確度可達92.3%,而在使用80%訓練集與20%驗證集的驗證方式中其驗證準確度高達99.2%。跟過往相關研究趨勢相對應是有顯著成果,因此證明我們所使用的方法是具有有效性。而且此方式避免過往複雜前處理部份,提升整體流程的簡易性,使此作法更具搭配優勢。
Drowsy drivers recognition is a warning function needed in many driving scenarios. Therefore, in this case, we pay attention to and identify drowsiness behaviors of drivers, thereby reminding drivers and reducing the incidence of automobile accidents. The main purpose of this paper is to identify and classify drowsiness and non-drowsiness by combining mixing images data augmentation and deep convolutional neural network. In recent years, researches on this dataset have focused on using image processing algorithms or complex pre-processing methods of feature engineering to improve the accuracy. These pre-processing methods may lose important classification features due to the selection of indicators and methods. In order to avoid the classification bias caused by feature selection. This paper uses mixing images data augmentation as the main technique. The Mixing images methods generate new unknown samples. We uses five deep convolutional neural networks for classification. The accuracy can be improved by expanding the characteristic information range of the original data samples to be closer to the real detection target. The results showed that the accuracy is up to 92.3% when 90% training set and 10% test set were used, while the accuracy was up to 99.2% when 80% training set and 20% verification set were used. There are results corresponding to the related research trends in the past years, so the method we used is effective. In addition, this method avoids the complicated preprocessing part and improves the simplicity of the overall process, so that this method has more matching advantages.
目錄
致謝辭 ⅰ
摘要 ii
Abstract iii
目錄 iv
圖目錄 vi
表目錄 ix
第一章 緒論 1
1.1 研究動機 1
1.2 研究目的 6
1.3 主要貢獻 7
1.4 論文架構 7
第二章 理論背景與文獻探討 9
2.1 台灣國立清華大學睡意駕駛資料集 9
2.2 卷積神經網路 10
2.2.1 卷積神經網路簡介 10
2.2.2 AlexNet 14
2.2.3 VGGNet 15
2.2.4 ResNet 18
2.2.5 DenseNet 20
2.2.6 SqueezeNet 23
2.3 資料擴增 27
2.3.1 資料擴增簡介 27
2.3.2 Cutout 30
2.3.3 Mixup 30
2.3.4 RICAP 31
2.3.5 CutMix 32
2.4 OpenCV 33
2.5 過往相關研究 33
第三章 研究方法與架構 35
3.1 方法架構與流程 35
3.2 影片圖像擷取 35
3.3 圖像標籤分類 37
3.4 單一圖像資料擴增 38
3.5 混合圖像資料擴增 41
3.5.1 Mixup 41
3.5.2 RICAP 42
3.5.3 CutMix 44
3.6 深度卷積神經網路 45
第四章 實驗結果與分析 47
4.1 實驗設定 47
4.1.1 實驗設備 47
4.1.2 實驗資料集 47
4.2 實驗結果 49
4.2.1 評估標準 49
4.2.2 本文方法實驗結果 50
4.3 實驗總結 64
第五章 結論與後續研究 65
參考文獻 66
[1]中華民國法務部全國法規資料庫,”民法184條”
https://law.moj.gov.tw/LawClass/LawSingle.aspx?pcode=B0000001&flno=184. Accessed on September 30 2019.
[2]中華民國法務部全國法規資料庫,”刑法284條” https://law.moj.gov.tw/LawClass/LawSingle.aspx?pcode=C0000001&flno=284. Accessed on September 30 2019
[3]中華民國法務部全國法規資料庫,”刑法276條” https://law.moj.gov.tw/LawClass/LawSingle.aspx?pcode=C0000001&flno=276. Accessed on September 30 2019
[4]中華民國交通部高速公路局,”交通安全資訊,交通事故分析” https://www.freeway.gov.tw/Publish.aspx?cnid=516&p=128.
Accessed on September 30 2019
[5]中央社,”國道取締違規遭撞,兩警與貨車駕駛死亡” https://www.cna.com.tw/news/firstnews/201804230012.aspx.
Accessed on September 30 2019
[6]Wikipedia,”2017年臺北南港遊覽車事故” https://zh.wikipedia.org/wiki/2017%E5%B9%B4%E8%87%BA%E5%8C%97%E5%8D%97%E6%B8%AF%E9%81%8A%E8%A6%BD%E8%BB%8A%E4%BA%8B%E6%95%85.
Accessed on September 30 2019
[7]國王車訊,”開車的你累了嗎?淺談《疲勞駕駛》症狀與如何預防” https://www.kingautos.net/204550?utm_source=facebook&utm_medium=kafb&fbclid=IwAR3uv1O1Fc6RznIyE68oLuNaVzaBnVAoZkj7HU4JyYksctygU76Iw_NDf6Y%20Accessed%20July%2025%202019.
Accessed on September 30 2019
[8]National Highway Traffic Safety Administration,” TRAFFIC SAFETY FACTS Crash Stats Drowsy Driving 2015” https://crashstats.nhtsa.dot.gov/Api/Public/ViewPublication/812446.
Accessed on September 30 2019
[9]American Automobile Association Foundation for Traffic and Safety,”MISSING 1-2 HOURS OF SLEEP DOUBLES CRASH RISK” https://newsroom.aaa.com/2016/12/missing-1-2-hours-sleep-doubles-crash-risk/.
Accessed on September 30 2019
[10]The Governors Highway Safety Association,” New Report Spotlights Dangers of Drowsy Driving”
https://www.ghsa.org/resources/new-report-spotlights-dangers-drowsy-driving. Accessed on September 30 2019
[11]American Automobile Association Foundation for Traffic and Safety,” Prevalence of Motor Vehicle Crashes Involving Drowsy Drivers, United States, 2009-2013” https://aaafoundation.org/prevalence-motor-vehicle-crashes-involving-drowsy-drivers-united-states-2009-2013/.
Accessed on September 30 2019
[12]Federal Rail Administration,” Fatigue Status of the U.S. Railroad Industry” https://www.fra.dot.gov/eLib/details/L04320.
Accessed on September 30 2019
[13]Insurance Information Institute,” Facts + Statistics: Drowsy driving”
https://www.iii.org/fact-statistic/facts-statistics-drowsy-driving
Accessed on September 30 2019
[14]ETtoday新聞雲,”台灣疲勞駕駛肇事率僅0.4%?” https://forum.ettoday.net/news/868252?redirect=1.
Accessed on September 30 2019
[15]科技大觀園, ”巨量資料協助預防疲勞駕駛” https://201812.scitechvista.nat.gov.tw/c/sgdG.htm.
Accessed on September 30 2019
[16]Weng, C. H., Lai, Y. H., & Lai, S. H. (2016, November). Driver drowsiness detection via a hierarchical temporal deep belief network. In Asian Conference on Computer Vision (pp. 117-133). Springer, Cham.
[17]GoodAudience, Amit Naik,” Introduction to Convolution Neural Networks”
https://blog.goodaudience.com/introduction-to-convolution-neural-networks-ef4d03a5ca69
Accessed on September 30 2019
[18]LeCun, Y., Bottou, L., Bengio, Y., & Haffner, P. (1998). Gradient-based learning applied to document recognition. Proceedings of the IEEE, 86(11), 2278-2324.
[19]LeCun, Y., Boser, B. E., Denker, J. S., Henderson, D., Howard, R. E., Hubbard, W. E., & Jackel, L. D. (1990). Handwritten digit recognition with a back-propagation network. In Advances in neural information processing systems (pp. 396-404).
[20]Sik-Ho Tsang,” Review: LeNet-1, LeNet-4, LeNet-5, Boosted LeNet-4 (Image Classification)”
https://medium.com/@sh.tsang/paper-brief-review-of-lenet-1-lenet-4-lenet-5-boosted-lenet-4-image-classification-1f5f809dbf17
Accessed on September 30 2019
[21]Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2012). Imagenet classification with deep convolutional neural networks. In Advances in neural information processing systems (pp. 1097-1105).
[22]Learn OpenCV,” Understanding AlexNet”
https://www.learnopencv.com/understanding-alexnet/
Accessed on September 30 2019
[23]Simonyan, K., & Zisserman, A. (2014). Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556.
[24]Coinmonks, Sik-Ho Tsang,” Review: VGGNet — 1st Runner-Up (Image Classification), Winner (Localization) in ILSVRC 2014”
https://medium.com/coinmonks/paper-review-of-vggnet-1st-runner-up-of-ilsvlc-2014-image-classification-d02355543a11
Accessed on September 30 2019
[25]Ioffe, S., & Szegedy, C. (2015). Batch normalization: Accelerating deep network training by reducing internal covariate shift. arXiv preprint arXiv:1502.03167.
[26]Simon, M., Rodner, E., & Denzler, J. (2016). Imagenet pre-trained models with batch normalization. arXiv preprint arXiv:1612.01452.
[27]Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., ... & Rabinovich, A. (2015). Going deeper with convolutions. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 1-9).
[28]He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 770-778).
[29]Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory. Neural computation, 9(8), 1735-1780.
[30]Towards Data Science, Vincent Fung,” An Overview of ResNet and its Variants”
https://towardsdatascience.com/an-overview-of-resnet-and-its-variants-5281e2f56035
Accessed on September 30 2019
[31]Xie, S., Girshick, R., Dollár, P., Tu, Z., & He, K. (2017). Aggregated residual transformations for deep neural networks. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 1492-1500).
[32]Targ, S., Almeida, D., & Lyman, K. (2016). Resnet in resnet: Generalizing residual architectures. arXiv preprint arXiv:1603.08029.
[33]Zagoruyko, S., & Komodakis, N. (2016). Wide residual networks. arXiv preprint arXiv:1605.07146.
[34]Huang, G., Liu, Z., Van Der Maaten, L., & Weinberger, K. Q. (2017). Densely connected convolutional networks. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 4700-4708).
[35]Towards Data Science, Sik-Ho Tsang,” Review: DenseNet — Dense Convolutional Network (Image Classification)”
https://towardsdatascience.com/review-densenet-image-classification-b6631a8ef803
Accessed on September 30 2019
[36]機器學習算法全棧工程師(Jeemy110),” DenseNet:比ResNet更優的CNN模型”
https://zhuanlan.zhihu.com/p/37189203.
Accessed on September 30 2019
[37]Iandola, F. N., Han, S., Moskewicz, M. W., Ashraf, K., Dally, W. J., & Keutzer, K. (2016). SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and<0.5 MB model size. arXiv preprint arXiv:1602.07360.
[38]机器之心,”纵览轻量化卷积神经网络:SqueezeNet、MobileNet、ShuffleNet、Xception”
https://zhuanlan.zhihu.com/p/32746221.
Accessed on September 30 2019
[39]Howard, A. G., Zhu, M., Chen, B., Kalenichenko, D., Wang, W., Weyand, T., ... & Adam, H. (2017). Mobilenets: Efficient convolutional neural networks for mobile vision applications. arXiv preprint arXiv:1704.04861.
[40]Zhang, X., Zhou, X., Lin, M., & Sun, J. (2018). Shufflenet: An extremely efficient convolutional neural network for mobile devices. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (pp. 6848-6856).
[41]Summers, C., & Dinneen, M. J. (2019, January). Improved mixed-example data augmentation. In 2019 IEEE Winter Conference on Applications of Computer Vision (WACV) (pp. 1262-1270). IEEE.
[42]Nanonet,” Data Augmentation | How to use Deep Learning when you have Limited Data — Part 2.”
https://nanonets.com/blog/data-augmentation-how-to-use-deep-learning-when-you-have-limited-data-part-2/.
Accessed on September 30 2019
[43]The Verge,” Google ‘fixed’ its racist algorithm by removing gorillas from its image-labeling tech.”
https://www.theverge.com/2018/1/12/16882408/google-racist-gorillas-photo-recognition-algorithm-ai. .
Accessed on September 30 2019
[44]Goodfellow, I., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., ... & Bengio, Y. (2014). Generative adversarial nets. In Advances in neural information processing systems (pp. 2672-2680).
[45]Radford, A., Metz, L., & Chintala, S. (2015). Unsupervised representation learning with deep convolutional generative adversarial networks. arXiv preprint arXiv:1511.06434.
[46]Zhu, J. Y., Park, T., Isola, P., & Efros, A. A. (2017). Unpaired image-to-image translation using cycle-consistent adversarial networks. In Proceedings of the IEEE international conference on computer vision (pp. 2223-2232).
[47]Heartbeat,” AutoML: The Next Wave of Machine Learning.” https://heartbeat.fritz.ai/automl-the-next-wave-of-machine-learning-5494baac615f.
Accessed on September 30 2019
[48]Cubuk, E. D., Zoph, B., Mane, D., Vasudevan, V., & Le, Q. V. (2018). Autoaugment: Learning augmentation policies from data. arXiv preprint arXiv:1805.09501.
[49]Insight,” AutoML for Data Augmentation.” https://blog.insightdatascience.com/automl-for-data-augmentation-e87cf692c366
Accessed on September 30 2019
[50]Ho, D., Liang, E., Stoica, I., Abbeel, P., & Chen, X. (2019). Population Based Augmentation: Efficient Learning of Augmentation Policy Schedules. arXiv preprint arXiv:1905.05393.
[51]Gong, X., Chang, S., Jiang, Y., & Wang, Z. (2019). AutoGAN: Neural Architecture Search for Generative Adversarial Networks. arXiv preprint arXiv:1908.03835.
[52]Tremblay, J., Prakash, A., Acuna, D., Brophy, M., Jampani, V., Anil, C., ... & Birchfield, S. (2018). Training deep networks with synthetic data: Bridging the reality gap by domain randomization. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops (pp. 969-977).
[53]Zhang, H., Cisse, M., Dauphin, Y. N., & Lopez-Paz, D. (2017). mixup: Beyond empirical risk minimization. arXiv preprint arXiv:1710.09412.
[54]CSDN,z小白,”深度学习Mixup: Beyond Empirical Risk Minimization”
https://blog.csdn.net/zzc15806/article/details/80696787
Accessed on September 30 2019
[55]DeVries, T., & Taylor, G. W. (2017). Improved regularization of convolutional neural networks with cutout. arXiv preprint arXiv:1708.04552.
[56]Takahashi, R., Matsubara, T., & Uehara, K. (2019). Data augmentation using random image cropping and patching for deep cnns. IEEE Transactions on Circuits and Systems for Video Technology.
[57]Yun, S., Han, D., Oh, S. J., Chun, S., Choe, J., & Yoo, Y. (2019). Cutmix: Regularization strategy to train strong classifiers with localizable features. arXiv preprint arXiv:1905.04899.
[58]OpenCV team, “Wikipedia:opencv.”
https://zh.wikipedia.org/wiki/OpenCV
Accessed on September 30 2019
[59]Liu, W., Qian, J., Yao, Z., Jiao, X., & Pan, J. (2019). Convolutional Two-Stream Network Using Multi-Facial Feature Fusion for Driver Fatigue Detection. Future Internet, 11(5), 115.
[60]Chiou, C. Y., Wang, W. C., Lu, S. C., Huang, C. R., Chung, P. C., & Lai, Y. Y. (2019). Driver Monitoring Using Sparse Representation With Part-Based Temporal Face Descriptors. IEEE Transactions on Intelligent Transportation Systems.
[61]Ed-doughmi, Y., & Idrissi, N. (2019, March). Driver Fatigue Detection using Recurrent Neural Networks. In Proceedings of the 2nd International Conference on Networking, Information Systems & Security (p. 44). ACM.
[62]Jabbar, R., Al-Khalifa, K., Kharbeche, M., Alhajyaseen, W., Jafari, M., & Jiang, S. (2018). Real-time driver drowsiness detection for android application using deep neural networks techniques. Procedia computer science, 130, 400-407.
[63]Celona, L., Mammana, L., Bianco, S., & Schettini, R. (2018, September). A Multi-Task CNN Framework for Driver Face Monitoring. In 2018 IEEE 8th International Conference on Consumer Electronics-Berlin (ICCE-Berlin) (pp. 1-4). IEEE.
[64]Shorten, C., & Khoshgoftaar, T. M. (2019). A survey on image data augmentation for deep learning. Journal of Big Data, 6(1), 60.
[65]Smith, L. N. (2018). A disciplined approach to neural network hyper-parameters: Part 1--learning rate, batch size, momentum, and weight decay. arXiv preprint arXiv:1803.09820.
[66]Liu, L., Jiang, H., He, P., Chen, W., Liu, X., Gao, J., & Han, J. (2019). On the variance of the adaptive learning rate and beyond. arXiv preprint arXiv:1908.03265.
[67]Zhang, M. R., Lucas, J., Hinton, G., & Ba, J. (2019). Lookahead Optimizer: k steps forward, 1 step back. arXiv preprint arXiv:1907.08610.
[68]Jiang, A. H., Wong, D. L. K., Zhou, G., Andersen, D. G., Dean, J., Ganger, G. R., ... & Pillai, P. (2019). Accelerating Deep Learning by Focusing on the Biggest Losers. arXiv preprint arXiv:1910.00762.
[69]Razzaq, S., Ahmad, M. N., Hamayun, M. M., ur Rahman, A., & Fraz, M. M. (2017, November). A hybrid approach for fatigue detection and quantification. In 2017 International Multi-topic Conference (INMIC) (pp. 1-7). IEEE.
電子全文 電子全文(網際網路公開日期:20221227)
QRCODE
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                               
第一頁 上一頁 下一頁 最後一頁 top
系統版面圖檔 系統版面圖檔