基于非局部操作的深度卷積神經(jīng)網(wǎng)絡(luò)車位占用檢測算法
doi: 10.11999/JEIT190349
-
1.
吉林大學(xué)計(jì)算機(jī)科學(xué)與技術(shù)學(xué)院 長春 130012
-
2.
吉林大學(xué)應(yīng)用技術(shù)學(xué)院 長春 130012
Deep Convolutional Neural Network for Parking Space Occupancy Detection Based on Non-local Operation
-
1.
College of Computer Science and Technology, Jilin University, Changchun 130012, China
-
2.
College of Applied Technology, Jilin University, Changchun 130012, China
-
摘要: 隨著城市交通智能化發(fā)展,準(zhǔn)確高效地獲取可用車位對(duì)于解決日益嚴(yán)峻的停車難問題至關(guān)重要。該文提出一種基于非局部操作的深度卷積神經(jīng)網(wǎng)絡(luò)車位占用檢測算法。針對(duì)停車位圖像特性,引入非局部操作,度量遠(yuǎn)距離像素間的相似性,直接獲取邊緣高頻特征;使用小卷積核獲取局部細(xì)節(jié)特征;以端到端的方式訓(xùn)練網(wǎng)絡(luò)。實(shí)驗(yàn)中,通過設(shè)置不同卷積核尺寸和非局部模塊層數(shù),優(yōu)化網(wǎng)絡(luò)結(jié)構(gòu)。實(shí)驗(yàn)結(jié)果表明,該文所提算法與傳統(tǒng)的基于紋理特征的車位占用檢測算法相比,無論在預(yù)測精度還是模型的泛化性能,均具有顯著的優(yōu)勢。與當(dāng)前廣泛應(yīng)用的基于局部特征提取的卷積神經(jīng)網(wǎng)絡(luò)相比,該算法具有較大的優(yōu)勢。在真實(shí)場景中,該算法同樣具有較高精度,具備實(shí)際應(yīng)用價(jià)值。
-
關(guān)鍵詞:
- 車位占用檢測 /
- 紋理特征 /
- 卷積神經(jīng)網(wǎng)絡(luò) /
- 非局部操作
Abstract: With the intelligent development of urban traffic, accurate and efficient access to available parking spaces is essential to solve the increasingly difficult problem of parking difficulties. Therefore, this paper proposes a deep convolutional neural network parking occupancy detection algorithm based on non-local operation. For the image characteristics of parking spaces, non-local operations are introduced, the similarity between distant pixels is measured, and the high-frequency features of the edges are directly obtained. The local details are obtained by using small convolution kernels, and the network is trained in an end-to-end manner. In the experiment, the network structure is optimized by setting different convolution kernel sizes and non-local module layers. The experimental results show that compared with the traditional texture feature-based parking space occupancy detection algorithm, the proposed algorithm has significant advantages in both prediction accuracy and generalization performance of the model. At the same time, compared with the currently widely used convolutional neural network based on local feature extraction, the algorithm also has great advantages. In real scenes, the algorithm also has high precision and has practical application value. -
表 1 不同卷積核尺寸的準(zhǔn)確率詳細(xì)對(duì)比(%)
卷積核尺寸 訓(xùn)練精度 測試精度 UFPR04 UFPR05 PUCPR 3 99.97 99.74 96.40 97.48 5 99.90 99.78 97.67 97.85 7 99.56 99.72 96.00 96.78 9 99.44 99.41 94.81 96.38 11 99.41 99.25 92.18 95.39 下載: 導(dǎo)出CSV
表 2 不同層數(shù)非局部模塊的準(zhǔn)確率詳細(xì)對(duì)比(%)
非局部模塊層數(shù) 訓(xùn)練精度 測試精度 UFPR04 UFPR05 PUCPR 1 99.90 99.78 97.67 97.85 2 99.96 99.81 97.65 97.55 3 99.95 99.85 98.55 98.35 下載: 導(dǎo)出CSV
表 3 不同方法的PKLot子數(shù)據(jù)集內(nèi)測試準(zhǔn)確率(%)
訓(xùn)練集 UFPR04 UFPR05 PUCPR 測試集 UFPR04 UFPR05 PUCPR 本文方法 99.85 99.62 99.92 mAlexnet 99.54 99.49 99.90 LPQu 99.50 98.90 99.58 Mean 99.64 99.30 99.61 下載: 導(dǎo)出CSV
表 4 不同方法的PKLot子數(shù)據(jù)集間測試準(zhǔn)確率(%)
訓(xùn)練集 測試集 方法 精度 UFPR04 UFPR05 本文方法 98.55 mAlexnet[14] 93.29 LPQg[18] 84.92 Max 88.33 PUCPR 本文方法 98.31 mAlexnet[14] 98.27 LPQg[18] 84.25 Mean 88.40 UFPR05 UFPR04 本文方法 94.45 mAlexnet[14] 93.69 LPQg[18] 85.76 Mean 85.53 PUCPR 本文方法 95.87 mAlexnet[14] 92.72 LPQu[17] 87.74 Mean 89.83 PUCPR UFPR04 本文方法 99.24 mAlexnet[14] 98.03 LPQg[18] 87.15 Mean 88.88 UFPR05 本文方法 98.89 mAlexnet[14] 96.00 LBPri[19] 82.78 Mean 84.20 下載: 導(dǎo)出CSV
-
CAICEDO F, BLAZQUEZ C, and MIRANDA P. Prediction of parking space availability in real time[J]. Expert Systems with Applications, 2012, 39(8): 7281–7290. doi: 10.1016/j.eswa.2012.01.091 DEL POSTIGO C G, TORRES J, and MENéNDEZ J M. Vacant parking area estimation through background subtraction and transience map analysis[J]. IET Intelligent Transport Systems, 2015, 9(9): 835–841. doi: 10.1049/iet-its.2014.0090 DAN N. Parking management system and method[P]. US, 20030144890, 2003. TSAI L W, HSIEH J W, and FAN K C. Vehicle detection using normalized color and edge map[J]. IEEE Transactions on Image Processing, 2007, 16(3): 850–864. doi: 10.1109/tip.2007.891147 HUANG C C, TAI Yushu, and WANG S J. Vacant parking space detection based on plane-based Bayesian hierarchical framework[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2013, 23(9): 1598–1610. doi: 10.1109/tcsvt.2013.2254961 DELIBALTOV D, WU Wencheng, LOCE R P, et al. Parking lot occupancy determination from lamp-post camera images[C]. The 16th International IEEE Conference on Intelligent Transportation Systems, The Hague, Netherlands, 2013: 2387–2392. doi: 10.1109/itsc.2013.6728584. LECUN Y, BENGIO Y, and HINTON G E. Deep learning[J]. Nature, 2015, 521(7553): 436–444. doi: 10.1038/nature14539 DE ALMEID P R L, OLIVEIRA L S, BRITTO JR A S, et al. PKLot–a robust dataset for parking lot classification[J]. Expert Systems with Applications, 2015, 42(11): 4937–4949. doi: 10.1016/j.eswa.2015.02.009 AMATO G, CARRARA F, FALCHI F, et al. Car parking occupancy detection using smart camera networks and deep learning[C]. 2016 IEEE Symposium on Computers and Communication, Messina, Italy, 2016: 1212–1217. doi: 10.1109/iscc.2016.7543901. BUADES A, COLL B, and MOREL J M. A non-local algorithm for image denoising[C]. 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, San Diego, USA, 2005: 60–65. doi: 10.1109/cvpr.2005.38. VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[C]. The 31st Conference on Neural Information Processing Systems, Long Beach, USA, 2017: 5998–6008. WANG Xiaolong, GIRSHICK R, GUPTA A, et al. Non-local neural networks[C]. 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Salt Lake City, USA, 2018: 7794–7803. doi: 10.1109/cvpr.2018.00813. HE Kaiming, ZHANG Xiangyu, REN Shaoqing, et al. Deep residual learning for image recognition[C]. 2016 IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, USA, 2016: 770–778. doi: 10.1109/cvpr.2016.90. AMATO G, CARRARA F, FALCHI F, et al. Deep learning for decentralized parking lot occupancy detection[J]. Expert Systems with Applications, 2017, 72: 327–334. doi: 10.1016/j.eswa.2016.10.055 KRIZHEVSKY A, SUTSKEVER I, and HINTON G E. ImageNet classification with deep convolutional neural networks[C]. The 25th International Conference on Neural Information Processing Systems, Red Hook, USA, 2012: 1097–1105. NURULLAYEV S and LEE S W. Generalized parking occupancy analysis based on dilated convolutional neural network[J]. Sensors, 2019, 19(2): 277. doi: 10.3390/s19020277 OJANSIVU V and HEIKKIL? J. Blur insensitive texture classification using local phase quantization[C]. The 3rd International Conference on Image and Signal Processing, Cherbourg-Octeville, France, 2008: 236–243. doi: 10.1007/978-3-540-69905-7_27. RAHTU E, HEIKKILA J, OJANSIVU V, et al. Local phase quantization for blur-insensitive image analysis[J]. Image and Vision Computing, 2012, 30(8): 501–512. doi: 10.1016/j.imavis.2012.04.001 OJALA T, PIETIKAINEN M, and MAENPAA T. Multiresolution gray-scale and rotation invariant texture classification with local binary patterns[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2002, 24(7): 971–987. doi: 10.1109/tpami.2002.1017623 -