Journal of Jilin University(Engineering and Technology Edition) ›› 2022, Vol. 52 ›› Issue (4): 925-933.doi: 10.13229/j.cnki.jdxbgxb20200912

Previous Articles    

Pedestrian segmentation and detection in multi-scene based on G-UNet

Xue-yun CHEN(),Xue-yu BEI,Qu YAO,Xin JIN   

  1. School of Electrical Engineering,Guangxi University,Nanning 530004,China
  • Received:2020-11-26 Online:2022-04-01 Published:2022-04-20

Abstract:

Current semantic segmentation methods can obtain the outline of pedestrians, but when pedestrians block each other, the information such as the number, height and central position of the pedestrians in the figure can not be directly obtained. To solve this problem, we propose the G-UNet model algorithm that, in addition to the semantic segmentation backbone, a Gaussian ellipse density kernel detection branch of the pedestrian area is added, so that the center position, height and width of the pedestrian are detected respectively by the maximum point of the kernel, vertical axis and horizontal axis scale. The uniqueness of the maximum point of density kernel solves the detection problem of pedestrian occlusion. Besides, UNet stiffly concatenates the features of the bottom layer and the top layer in a spatial symmetric way so that 50% fixed errors are directly transmitted to the bottom layer. Then, we propose a trainable Soft Connection to obtain the optimal error distribution propagation method. Finally, to solve the problem that the value of the traditional loss function is proportional to the demarcated area of pedestrian, which makes small-scale pedestrians easy to be undetected, an Objective Enhanced Loss was proposed to improve the ability of detecting small-scale pedestrians of network. In the self-built pedestrian segmentation dataset, the experimental results show that the proposed method is effective and superior to other methods.

Key words: computer application, pedestrian semantic segmentation, Gaussian kernel, soft connection, target enhancement loss

CLC Number: 

  • TP391.41

Fig.1

Schematic diagram of pedestrians blocking each other"

Fig.2

G-UNet network structure diagram"

Fig.3

Schematic diagram of soft connection"

Fig.4

Visualization and processing of label information"

Fig.5

G-UNet loss curve"

Fig.6

Segmentation results of different algorithms for small scale pedestrians"

Table 1

Accuracy comparison of different semantic segmentation algorithms"

NetIoUF1?ScoreRecallPrecision
FCN?8S68.1280.5571.4192.38
PSPNet71.5881.9675.2290.03
UNet73.2483.6377.0691.43
本文76.0284.8079.0691.45

Fig.7

Detection results of different algorithms for pedestrians"

Table 2

Accuracy comparison of pedestrian detection"

NetIoUF1?ScoreRecallPrecision
YOLO?v367.0179.8469.5193.78
Faster R?CNN72.3380.9675.4387.37
FRCN70.8280.5573.5689.01
本文74.6484.8078.6989.67

Table 3

Experimental results of soft connection"

αIoUF1?ScoreRecallPrecision
070.8280.3374.2387.52
0.172.1681.8076.0288.45
0.574.0382.7177.2289.03
173.6681.9576.3688.43
273.0281.4976.0687.75
(0~2.5)*74. 4183.5077.8690. 01

Table 4

Experimental results of target enhancement loss"

aIoUF1?ScoreRecallPrecision
-1070.8681.0772.3292.22
-572.5682.4374.7291.05
072.9382.3675.4290.71
1073.2382.8076.2290.63
3074.5483.3977.3690.43
5075.0283.8778.2690.35
7074.6283.6178.3589.63
9074.0283.4178.1389.45
1 Saeidi M, Ahmadi A. Deep learning based on CNN for pedestrian detection: an overview and analysis[C]∥The 9th International Symposium on Telecommunications (IST), Tehran, Iran, 2018: 108-112.
2 Dalal N, Triggs B. Histograms of oriented gradients for human detection[C]∥IEEE Computer Society Conference on Computer Vision and Pattern Recognition,San Diego, USA, 2005 .
3 Girshick R, Donahue J, Darrell T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]∥IEEE Conference on Computer Vision and Pattern Recognition, Columbus, USA, 2014.
4 Girshick R. Fast R-CNN[C]∥IEEE International Conference on Computer Vision (ICCV), Santiago, Chile, 2015.
5 Ren S, He K, Girshick R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2017, 39(6): 1137-1149.
6 Redmon J, Divvala S, Girshick R, et al. You only look once: unified, real-time object detection[C]∥IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, USA, 2016.
7 Redmon J, Farhadi A. YOLO9000: Better, faster, stronger[C]∥IEEE Conference on Computer Vision and Pattern Recognition (CVPR),Honolulu,USA, 2017: 6517-6525.
8 Redmon J, Farhadi A. YOLOv3: An Incremental Improvement[EB/OL].[2018-02-08]..
9 Tayara H, Kim G S. Vehicle detection and counting in high-resolution aerial images using convolutional regression neural network[J]. IEEE Access, 2017, 6: 2220-2230.
10 Chen X Y, Lin J Y, Xiang S M, et al. Detecting maneuvering target accurately based on a two-phase approach from remote sensing imagery[J]. IEEE Geoence and Remote Sensing Letters, 2020, 17(5): 849-853.
11 罗会兰, 张云. 基于深度网络的图像语义分割综述[J].电子学报, 2019, 47(10): 2211-2220.
Luo Hui-lan, Zhang Yun. A survey of image semantic segmentation based on deep network[J]. Acta Electronica Sinica, 2019, 47(10): 2211-2220.
12 Shelhamer E, Long J, Darrell T. Fully convolutional networks for semantic segmentation[J]. IEEE Computer Society, 2017, 39(4): 640-651.
13 Zhao H, Shi J, Qi X, et al. Pyramid scene parsing network[C]∥IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, USA, 2017: 6230-6239.
14 Ronneberger O, Fischer P, Brox T. U-net: convolutional networks for biomedical image segmentation[J]. Computer Vision and Pattern Recognition,2015, 9351: 234-241.
15 Chen L C, Papandreou G, Schroff F, et al. Rethinking atrous convolution for semantic image segmentation[DB/OL].[2017-02-05]..
[1] Ming LIU,Yu-hang YANG,Song-lin ZOU,Zhi-cheng XIAO,Yong-gang ZHANG. Application of enhanced edge detection image algorithm in multi-book recognition [J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(4): 891-896.
[2] Xiao-hui WEI,Yan-wei MIAO,Xing-wang WANG. Rhombus sketch: adaptive and more accurate sketch for streaming data [J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(4): 874-884.
[3] Shi-min FANG. Multiple source data selective integration algorithm based on frequent pattern tree [J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(4): 885-890.
[4] Da-xiang LI,Meng-si CHEN,Ying LIU. Spontaneous micro-expression recognition based on STA-LSTM [J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(4): 897-909.
[5] Xue-zhi WANG,Qing-liang LI,Wen-hui LI. Spatio⁃temporal model of soil moisture prediction integrated with transfer learning [J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(3): 675-683.
[6] Su-ming KANG,Ye-e ZHANG. Hadoop⁃based local timing link prediction algorithm across social networks [J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(3): 626-632.
[7] Xue WANG,Zhan-shan LI,Ying-da LYU. Medical image segmentation based on multi⁃scale context⁃aware and semantic adaptor [J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(3): 640-647.
[8] Ji-hong OUYANG,Ze-qi GUO,Si-guang LIU. Dual⁃branch hybrid attention decision net for diabetic retinopathy classification [J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(3): 648-656.
[9] Lin MAO,Feng-zhi REN,Da-wei YANG,Ru-bo ZHANG. Two⁃way feature pyramid network for panoptic segmentation [J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(3): 657-665.
[10] You QU,Wen-hui LI. Single-stage rotated object detection network based on anchor transformation [J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(1): 162-173.
[11] Hong-wei ZHAO,Dong-sheng HUO,Jie WANG,Xiao-ning LI. Image classification of insect pests based on saliency detection [J]. Journal of Jilin University(Engineering and Technology Edition), 2021, 51(6): 2174-2181.
[12] Zhou-zhou LIU,Qian-yun ZHANG,Xin-hua MA,Han PENG. Compressed sensing signal reconstruction based on optimized discrete differential evolution algorithm [J]. Journal of Jilin University(Engineering and Technology Edition), 2021, 51(6): 2246-2252.
[13] Sheng-sheng WANG,Jing-yu CHEN,Yi-nan LU. COVID⁃19 chest CT image segmentation based on federated learning and blockchain [J]. Journal of Jilin University(Engineering and Technology Edition), 2021, 51(6): 2164-2173.
[14] Dong-ming SUN,Liang HU,Yong-heng XING,Feng WANG. Text fusion based internet of things service recommendation for trigger⁃action programming pattern [J]. Journal of Jilin University(Engineering and Technology Edition), 2021, 51(6): 2182-2189.
[15] Jun-cong LIN,Jun LEI,Meng CHEN,Shi-hui GUO,Xing GAO,Ming-hong LIAO. Real⁃time camera planning for dynamic multiple targets considering cinematographic visual properties [J]. Journal of Jilin University(Engineering and Technology Edition), 2021, 51(6): 2154-2163.
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!