Key points and visible part fusion attention network for occluded pedestrian detection in traffic environments

Abstract

Aiming at the problem of low detection accuracy of occluded pedestrian in traffic environments, this paper proposes a key points and visible part fusion network for occluded pedestrian detection. The proposed algorithm constructs two attention modules by introducing human key points and the bounding box of visible parts respectively, which suppresses the occluded parts in the channel features and spatial features of pedestrian features respectively. Experimental results on CityPersons and Caltech datasets demonstrate the effectiveness of the proposed algorithm. The missing rate (MR) is reduced to 40.78 on the Heavy subset of the CityPersons dataset and surpasses many outstanding methods.

References

[1] LiuM Y, JiangJ, ZhuC, et al.. VLPD: context-aware pedestrian detection via vision-language semantic self-supervision[C], 2023, New York, IEEE: 6662-6671

[2] AbdulK, MohammedN, AndreasD. Localized semantic feature mixers for efficient pedestrian detection in autonomous driving[C], 2023, New York, IEEE: 5476-5485

[3] TangY, LiuM, LiB P, et al.. OTP-NMS: toward optimal threshold prediction of NMS for crowded pedestrian detection[J]. IEEE transactions on image processing, 2023, 32: 3176-3187

[4] SundararamanR, DeA C, MarchandE, et al.. Tracking pedestrian heads in dense crowd[C], 2021, New York, IEEE: 3865-3875

[5] LiuL, ChenJ, WuH, et al.. Cross-modal collaborative representation learning and a large-scale RGBT benchmark for crowd counting[C], 2021, New York, IEEE: 4823-4833

[6] CAO J, PANG Y, XIE J, et al. From handcrafted to deep features for pedestrian detection: a survey [J]. IEEE transactions on pattern analysis and machine intelligence, 2021: 4913–4934.

[7] WangL, XuL, YangM H. Pedestrian detection in crowded scenes via scale and occlusion analysis[C], 2016, New York, IEEE: 1210-1214

[8] ZhangH, YanC, LiX L, et al.. MSAGNet: multi-stream attribute-guided network for occluded pedestrian detection[J]. IEEE signal processing letter, 2022, 29: 2163-2167

[9] TangY, LiuM, LiB Q, et al.. OTP-NMS: toward optimal threshold prediction of NMS for crowded pedestrian detection[J]. IEEE transactions on image processing, 2023, 32: 3176-3187

[10] XueP, ChenH J, LiY F, et al.. Multi-scale pedestrian detection with global-local attention and multi-scale receptive field context[J]. IET computer vision, 2023, 17: 13-25

[11] QingY, QianY Y, WeiH Y, et al.. FE-CSP: a fast and efficient pedestrian detector with center and scale prediction[J]. The journal of supercomputing, 2023, 79(4):4084-4104

[12] RedmonJ, FarhadiA. YOLO9000: better, faster, stronger[C], 2017, New York, IEEE: 7263-7271

[13] ZhouC, YuanJ. Bi-box regression for pedestrian detection and occlusion estimation[C], 2018, Berlin, Heidelberg, Springer: 135-151

[14] ZhangS, WenL, BianX, et al.. Occlusion-aware R-CNN: detecting pedestrians in a crowd[C], 2018, Berlin, Heidelberg, Springer: 637-653

[15] WangX, XiaoT, JiangY, et al.. Repulsion loss: detecting pedestrians in a crowd[C], 2018, New York, IEEE: 7774-7783

[16] LiuS, HuangD, WangY. Adaptive NMS: refining pedestrian detection in a crowd[C], 2019, New York, IEEE: 6459-6468

[17] WooS, ParkJ, LeeJ Y, et al.. CBAM: convolutional block attention module[C], 2018, Berlin, Heidelberg, Springer: 3-19

[18] YaoC, YangJ, CeylanD, et al.. Learning visibility for robust dense human body estimation[C], 2022, Berlin, Heidelberg, Springer: 412-428

[19] ChengB, XiaoB, WangJ, et al.. Higher HRNet: scale-aware representation learning for bottom-up human pose estimation[C], 2020, New York, IEEE: 5386-5395

[20] LinT Y, MaireM, BelongieS, et al.. Microsoft coco: common objects in context[C], 2014, Berlin, Heidelberg, Springer: 740-755

[21] ZhangS, BenensonR, SchieleB. Citypersons: a diverse dataset for pedestrian detection[C], 2017, New York, IEEE: 3213-3221

[22] CordtsM, OmranM, RamosS, et al.. The city-scapes dataset for semantic urban scene understanding[C], 2016, New York, IEEE: 3213-3223

[23] DollárP, WojekC, SchieleB, et al.. Pedestrian detection: a benchmark[C], 2009, New York, IEEE: 304-311

[24] SongT, SunL, XieD, et al.. Small-scale pedestrian detection based on somatic topology localization and temporal feature aggregation[C], 2018, Berlin, Heidelberg, Springer: 1-16

[25] ZhangL, LinL, LiangX, et al.. Is faster R-CNN doing well for pedestrian detection[C], 2016, Berlin, Heidelberg, Springer: 443-457

[26] XieJ, PangY, KhanM H, et al.. Mask-guided attention network and occlusion-sensitive hard example mining for occluded pedestrian detection[J]. IEEE transactions on image processing, 2020, 30: 3872-3884

[27] LiuW, LiaoS, RenW, et al.. High-level semantic feature detection: a new perspective for pedestrian detection[C], 2019, New York, IEEE: 5187-5196

[28] ZhouP, ZhouC, PengP, et al.. NOH-NMS: improving pedestrian detection by nearby objects hallucination[C], 2020, New York, ACM: 1967-1975

[29] M, ZhuC, WangJ, et al.. Adaptive pattern-parameter matching for robust pedestrian detection[C]//Proceedings of the AAAI Conference on Artificial Intelligence, February 2–9, 2021, online. Maynard, Massachusetts: AAAI, 2021, 35(3):2154-2162

[30] ZhangY, HeH, LiJ, et al.. Variational pedestrian detection[C], 2021, New York, IEEE: 11622-11631

[31] WuJ, ZhouC, YangM, et al.. Temporal-context enhanced detection of heavily occluded pedestrians[C], 2020, New York, IEEE: 13430-13439

[32] ZouF M, LiX, XuQ M, et al.. Correlation-and-correction fusion attention network for occluded pedestrian detection[J]. IEEE sensors journal, 2023, 23(6):6061-6073