基于RetinaNet及优化损失函数的夜间车辆检测方法

doi:10.19562/j.chinasae.qcgc.2021.08.010

Abstract

Abstract:

In view of the high false detection rate and poor detection results for distant small target in vehicle detection at night for intelligent driving system， the loss functions are comprehensively optimized based on RetinaNet. For classification loss function， both the generation mechanism and the influence on training results of the IoUs of negative and positive samples are analyzed， the classification loss function of correlated IoU is constructed， the IoU of negative sample is used to make network lay emphasis on the training of negative samples， which are hard to classify， meanwhile the IoU of positive sample is used to enhance the locating accuracy of detection frame. For locating loss function， the normalization way of traditional L1 loss is improved and the ability of small target detection is enhanced. In addition， a design optimization is conducted on network structure for the vehicle features in night scene， and a test verification is performed on vehicle data at night. The results show that the average detection accuracy of the model optimized increases by 14.6 percentage point.

Key words: vehicle detection at night, loss functions, IoU, RetinaNet, small target detection

Bingli Zhang,Haoran Qin,Shang Jiang,Jieyu Zheng,Zhenghai Wu. A Method of Vehicle Detection at Night Based on RetinaNet and Optimized Loss Functions[J].Automotive Engineering, 2021, 43(8): 1195-1202.

Figures/Tables 13

References 22

1	王美钰. 夜间高速公路车辆车灯的反光检测的研究［J］. 现代计算机， 2020（12）：97-101.
	WANG M Y. Research on the reflection detection of vehicle lights on highway at night［J］. Modern Computer， 2020（12）：97-101.
2	郭君斌，王建强，易世春，等. 基于单目视觉的夜间前方车辆检测方法［J］. 汽车工程， 2014，36（5）：573-579，585.
	GUO J B， WANG J Q， YI S C， et al. Preceding vehicle detection at night based on monocular vision［J］. Automotive Engineering， 2014，36（5）：573-579，585.
3	董天阳，朱浩楠，王浩. 基于视频的夜间车辆检测与跟踪算法研究［J］. 计算机科学， 2017，44（S2）：233-237.
	DONG T Y， ZHU H N， WANG H. Video-based nighttime vehicle detection and tracking algorithm［J］. Computer Science， 2017，44（S2）：233-237.
4	许益成，谭文安，陈丽婷. 基于Hough变换检测前大灯的夜间车辆检测［J］. 控制工程， 2018，25（8）：1360-1363.
	XU Y C， TAN W A， CHEN L T. A nighttime vehicle detection method using hough transform to detect headlights［J］. Control Engineering of China， 2018，25（8）：1360-1363.
5	于莉媛，郭云雷，牛萍娟，等.基于尾灯的夜间前方车辆检测与跟踪方法［J］. 天津工业大学学报， 2019，38（1）：61-68.
	YU L Y， GUO Y L， NIU P J， et al. Method of preceding vehicle detection and tracking at night based on taillights［J］. Journal of Tianjin Polytechnic University， 2019，38（1）：61-68.
6	乔瑞萍，董员臣，王方，等. 基于颜色通道混合特征的夜间车辆检测仿真［J］. 计算机仿真， 2020，37（12）：107-110.
	QIAO R P， DONG Y C， WANG F， et al. Simulation of night vehicle detection based on mixed features of color channels［J］. Computer Integrated Manufacturing Systems， 2020，37（12）：107-110.
7	ALCANTARILLA P， BERGASA L， JIMÉNEZ P， et al. Night time vehicle detection for driving assistance light beam controller［C］. Intelligent Vehicles Symposium. IEEE， 2008.
8	KIM H， KUK S， KIM M， et al. An effective method of head lamp and tail lamp recognition for night time vehicle detection［J］. World Academy of Science Engineering & Technology， 2013， 126（21）：384-392.
9	文学志，赵宏，王楠，等. 基于知识和外观方法相结合的后方车辆检测［J］. 东北大学学报（自然科学版）， 2007（3）：333-336.
	WEN X Z， ZHAO H， WANG N， et al. Rear-vehicle detection combining both knowledge-based and appearance-based methods［J］. Journal of Northeastern University （Natural Science）， 2007（3）：333-336.
10	陈艳，严腾，宋俊芳，等. 基于高斯混合模型和AdaBoost的夜间车辆检测［J］. 计算机应用， 2018，38（1）：260-263，283.
	CHEN Y， YAN T， SONG J F， et al. Night-time vehicle detection based on Gaussian mixture model and AdaBoost［J］. Journal of Computer Applications， 2018，38（1）：260-263，283.
11	REN S， HE K， GIRSHICK R， et al. Faster R-CNN： towards real-time object detection with region proposal networks［J］. Advances in Neural Information Processing Systems， 2015， 28： 91-99.
12	LI Y， HE K， SUN J. R-FCN： object detection via region-based fully convolutional networks［C］. Advances in Neural Information Processing Systems， 2016： 379-387.
13	CAI Z， VASCONCELOS N. Cascade R-CNN： delving into high quality object detection［C］.Proceedings of the IEEE conference on computer vision and pattern recognition， 2018： 6154-6162.
14	REDMON J， DIVVALA S， GIRSHICK R， et al. You only look once： Unified， real-time object detection［C］. Proceedings of the IEEE conference on computer vision and pattern recognition. 2016： 779-788.
15	REDMON J， FARHADI A. YOLO9000： better， faster， stronger［C］. IEEE Conference on Computer Vision & Pattern Recognition. IEEE， 2017：6517-6525.
16	REDMON J， FARHADI A. Yolov3： an incremental improvement［J］. arXiv Preprint arXiv：， 2018.
17	LIU W， ANGUELOV D， ERHAN D， et al. Ssd： single shot multibox detector［C］. European Conference on Computer Vision. Springer， Cham， 2016： 21-37.
18	LIN T Y， GOYAL P， Girshick R， et al. Focal loss for dense object detection［J］. IEEE Transactions on Pattern Analysis & Machine Intelligence， 2017， PP（99）：2999-3007.
19	HE K， ZHANG X， REN S， et al. Deep residual learning for image recognition［C］. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition， 2016： 770-778.
20	JIANG B， LUO R， MAO J， et al. Acquisition of localization confidence for accurate object detection［C］. Proceedings of the European Conference on Computer Vision （ECCV），2018： 784-799.
21	LI X， WANG W， WU L， et al. Generalized focal loss： learning qualified and distributed bounding boxes for dense object detection［J］. arXiv Preprint arXiv：， 2020.
22	ERHAN D， SZEGEDY C， TOSHEV A， et al. Scalable object detection using deep neural networks［C］. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition， 2014： 2147-2154.

模型	特征层选择	预设框设置	子网络简化	AP	检测速度/fps
RetinaNet				32.2	15
改进后的RetinaNet	√			38.5	11
	√	√		43.9	17
	√	√	√	43.1	23

序号	优化方法	AP	AP⁵⁰	AP^S
1	改进后的RetinaNet	43.1	81.8	28.4
2	定位损失	44.6	84.2	31.5
3	分类损失（仅正样本）	44.2	83.1	30.0
4	分类损失（仅负样本）	45.1	85.6	32.1
5	分类损失	45.5	86.4	32.8
6	分类损失+定位损失	46.8	87.9	34.7

γ	AP	AP⁵⁰
0.5	44.1	83.4
1.0	45.6	84.8
1.5	46.5	87.5
2.0	46.8	87.9
2.5	44.1	83.4
5.0	38.6	75.4

γ	α=1	α=1.25	α=1.5	α=1.75	α=2
0.8	43.4	45.2	45.9	46.4	45.9
0.9	43.8	46.2	46.5	46.8	46.2
1.0	44.1	46.0	46.5	47.0	46.8
1.1	44.0	46.1	46.4	47.2	46.9
1.2	44.0	45.8	46.0	46.9	46.6

[1]	Zheng Zuo,Yunpeng Wang,Bin Ma,Bosong Zou,Yaoguang Cao,Shichun Yang. Quantitative Evaluation and Analysis of On-board Network Components Risk Rate Based on AFC-TARA [J]. Automotive Engineering, 2023, 45(9): 1553-1562.
[2]	Huiyan Zhang,Xuyang Tang,Lei Shi,Kangyao Deng. Study on Altitude Self-adaption and Adjustment Capacity of Turbocharging System [J]. Automotive Engineering, 2022, 44(2): 256-263.
[3]	Teng Cheng,Lei Sun,Dengchao Hou,Qin Shi,Junning Zhang,Jiong Chen,He Huang. Multi-level and Multi-modal Target Detection Based on Feature Fusion [J]. Automotive Engineering, 2021, 43(11): 1602-1610.

A Method of Vehicle Detection at Night Based on RetinaNet and Optimized Loss Functions

RichHTML

PDF (PC)

Like

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 13

References 22

Related Articles 3

Metrics

Comments

Recommended 10