为提高无人驾驶汽车感知系统硬件资源利用率,构建了一种基于特征融合的无人驾驶多任务感知算法。采用改进的CSPDarknet53作为模型的主干网络,通过构建特征融合网络与特征融合模块对多尺度特征进行提取与融合,并以7种常见道路物体的检测与可行驶区域的像素级分割两任务为例,设计多任务模型DaSNet(Detection and Segmentation Net)进行训练与测试。使用BDD100K数据集对YOLOv5s、Faster R-CNN以及U-Net模型进行训练,并对mAP、Dice系数以及检测速度等性能指标做出对比分析。研究结果表明:DaSNet多任务模型在道路物体检测任务上,mAP值分别比YOLOv5s和Faster RCNN高出0.5%和4.2%,在RTX2080Ti GPU上达到121FPS的检测速度;在占优先权与不占优先权的可行驶区域上分割的Dice值相较于U-Net网络分别高出了4.4%与6.8%,有较明显的提升。
孙传龙
,
赵红
,
崔翔宇
,
牟亮
,
徐福良
,
路来伟
,
孙传龙
,
赵红
,
崔翔宇
,
牟亮
,
徐福良
,
路来伟
. 基于特征融合的无人驾驶多任务感知算法[J]. 复杂系统与复杂性科学, 2023
, 20(3)
: 103
-110
.
DOI: 10.13306/j.1672-3813.2023.03.014
In order to improve the utilization of hardware resources of driverless vehicle perception system, a multi-task driverless vehicle perception algorithm based on feature fusion is constructed. The improved CSPDarknet53 is used as the backbone network of the model, and multi-scale features are extracted and fused by constructing feature fusion network and feature fusion module. The detection of 7 common road objects and pixel-level segmentation of the driving area are taken as examples. Multi-task DaSNet (Detection and Segmentation Net) is designed for training and testing. In order to compare model performance, BDD100K data set is used to train YOLOv5s, Faster R-CNN and U-NET models, and comparative analysis is made on mAP, Dice coefficient and detection speed and other performance indicators. The results showed that DaSNet multi-task model′s mAP value is 0.5% and 4.2% higher than YOLOv5s and Faster RCNN, respectively, and the detection speed of 121FPS can be achieved on RTX2080Ti GPU. Compared with U-NET network, Dice value of segmentation in priority and non-priority drivable are 4.4% and 6.8% higher, showing an obvious improvement.
[1] 王俊. 无人驾驶车辆环境感知系统关键技术研究[D]. 合肥:中国科学技术大学, 2016.
WANG J. Research on keytechnologies of environment awareness system for unmanned vehicle [D]. Hefei: University of Science and Technology of China, 2016.
[2] 王世峰, 戴祥, 徐宁, 等. 无人驾驶汽车环境感知技术综述[J]. 长春理工大学学报(自然科学版), 2017,40(1): 16.
WANG S F, DAI X, XU N, et al. Overview on environment perception technology for unmanned ground vehicle[J]. Journal of Changchun University of Science and Technology (Natural Science Edition), 2017,40(1): 16.
[3] CHEN Q, XIE Y, GUO S, et al. Sensingsystem of environmental perception technologies for driverless vehicle: a review of state of the art and challenges[J]. Sensors and Actuators A Physical, 2021, 319: 112566.
[4] GAYATHRIK D, MAMATA R, NGUYEN T D L. Artificial intelligence trends for data analytics using machine learning and deep learning approaches[M]. USA: Calabasas: CRC Press, 2020.
[5] 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: IEEE, 2014: 580587.
[6] GIRSHICK R. Fast r-cnn [C] // IEEE International Conference on Computer Vision. Santiago: IEEE, 2015: 14401448.
[7] 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): 11371149.
[8] 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. Las Vegas: IEEE, 2016: 779788.
[9] LIU S, QI X, SHI J, et al. Multi-scalepatch aggregation (MPA) for simultaneous detection and segmentation[C] // 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Seattle :IEEE, 2016: 31413149.
[10] PINHEIRO P, COLLOBERT R, DOLLAR P. Learning tosegments objects candidates[C] // Advances in Neural Information Processing Systems. Montreal:NIPS,2015.
[11] CHEN L C, PAPANDREOU G, KOKKINOS I, et al. Semanticimage segmentation with deep convolutional nets and fully connected CRFs[J]. Computer Science, 2014(4): 357361.
[12] ZHAO H SH, QI X J, SHEN X Y, et al. ICNet forreal-time semantic segmentation on high-resolution images[J]. Lecture Notes in Computer Science, 2017, 11207: 418434.
[13] RONNEBERGER O, FISCHER P, BROX T. U-net: convolutional networks for biomedical image segmentation[J]. Lecture Notes in Computer Science, 2015, 9351: 234241.
[14] 李亚. 多任务学习的研究[D].合肥:中国科学技术大学, 2018.
LI Y. Research on multi-task learning [D]. Hefei: University of Science and Technology of China, 2018.
[15] LIU S H, QI L, QIN H F, et al. Path aggregation network for instance segmentation[J/OL]. IEEE.[20211001]. DOI:10.1109/CVPR.2018.00913.
[16] REZATOFIGHI H, TSOI N, J Y GWAK, et al. Generalized intersection over union: a metric and a loss for bounding box regression[C] // 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Long Beach:IEEE, 2019: 658666.
