Vehicle re-identification optimization algorithm based on high-confidence local features
-
摘要:
根据车辆重识别中区域置信度不同,提出了基于高置信局部特征的车辆重识别优化算法。首先,利用车辆关键点检测获得对应的多个关键点坐标信息,分割出车标扩散区域和其他重要的局部区域。根据车标扩散区域的高区分度特性,提升局部区域的置信度。使用多层卷积神经网络对输入图片进行处理,根据局部区域分割信息,对卷积得到的特征张量进行空间维度上的切割,获得代表全局信息和关键局部信息的特征张量。然后,通过全连接层特征张量转化为表示车辆个体的一维向量,计算损失函数。最后,在测试阶段使用全局特征,并利用训练好的车标扩散区域提取分支获得高置信局部特征,缩短局部识别一致的车辆目标距离。在典型车辆重识别数据集VehicleID上进行测试,验证了所提算法的有效性。
Abstract:In solving vehicle re-identification problems, different vehicle regions have different recognition degree of confidence. Based on this observation, we propose a vehicle re-identification optimization algorithm that takes advantage of the high-confidence local features. First, the vehicle key point detection algorithm is utilized to obtain the corresponding multiple key points' coordinate information of the vehicles, and to divide the vehicle brand extension regions and other prominent local regions. As the brand extension region is the most salient region, we propose to improve the degree of confidence of the local region in the testing phase. We also utilize a multi-layer convolutional neural network for processing the input images, cutting the convolutional features into several parts based on the obtained local regions, and acquiring feature tensors representing global and key regional information. Then, a fully connected layer is applied to combine the above features and output a one-dimensional vector for loss function calculating. In the testing phase, to reduce the target distances of vehicles with the same local identification, we propose to utilize the global features together with the high-confidence local features obtained by trained brand extension region extraction branch. Experiments on the widely used vehicle re-identification VehicleID dataset show that the proposed algorithm is effective.
-
表 1 VehicleID数据集信息
Table 1. VehicleID dataset information
数据集 图片数量 车辆ID数量 训练集 110 178 13 164 small测试集 7 332 800 medium测试集 12 995 1 600 large测试集 20 038 2 400 表 2 不同算法top1的识别准确率对比结果
Table 2. Top1 recognition accuracy results of different methods
表 3 top1的识别准确率消融对比结果
Table 3. Ablation comparison results of top1 recognition accuracy
算法 topl的识别准确率/% small
测试集medium
测试集large
测试集仅主分支 78.4 75.9 74.4 仅高置信局部特征分支 24.1 18.9 15.5 两个分支(本文算法) 79.2 77.1 75.2 -
[1] XUE G, ZHU H, HU Z, et al.Pothole in the dark:Perceiving pothole profiles with participatory urban vehicles[J].IEEE Transactions on Mobile Computing, 2017, 16(5):1408-1419. doi: 10.1109/TMC.2016.2597839 [2] LYU F, FANG L, XUE G, et al.Large-scale full WiFi coverage:Deployment and management strategy based on user spatio-temporal association analytics[J].IEEE Internet of Things Journal, 2019, 6(6):9386-9398. doi: 10.1109/JIOT.2019.2933266 [3] LIU Y, SHENG H, ZHENG Y, et al.GDMN:Group decision-making network for person re-identification[J].IEEE Access, 2018, 6:64169-64181. doi: 10.1109/ACCESS.2018.2877841 [4] LV K, DENG W, HOU Y, et al.Vehicle reidentification with the location and time stamp[C]//Proceedings of the IEEE International Conference on Computer Vision Workshops.Piscataway: IEEE Press, 2019: 399-406. [5] ZHANG X, ZHANG R, CAO J, et al.Part-guided attention learning for vehicle re-identification[EB/OL].(2020-02-17)[2020-03-01].https://arxiv.org/abs/1909.06023. [6] CHU R, SUN Y, LI Y, et al.Vehicle re-identification with viewpoint-aware metric learning[C]//Proceedings of the IEEE International Conference on Computer Vision.Piscataway: IEEE Press, 2019: 8282-8291. https://blog.csdn.net/jyhongjax/article/details/103922886 [7] 李熙莹, 周智豪, 邱铭凯.基于部件融合特征的车辆重识别算法[J].计算机工程, 2019, 45(6):12-20. http://www.cnki.com.cn/Article/CJFDTotal-JSJC201906003.htmLI X Y, ZHOU Z H, QIU M K.Vehicle re-identification algorithm based on component fusion feature[J].Computer Engineering, 2019, 45(6):12-20(in Chinese). http://www.cnki.com.cn/Article/CJFDTotal-JSJC201906003.htm [8] 王盼盼, 李玉惠, 李福卫.基于特征融合和度量学习的车辆重识别[J].电子科技, 2018, 31(9):29-31. http://www.cnki.com.cn/Article/CJFDTotal-DZKK201809009.htmWANG P P, LI Y H, LI F W.Vehicle re-identification based on multi-features and metric learning[J].Electronic Science and Technology, 2018, 31(9):29-31(in Chinese). http://www.cnki.com.cn/Article/CJFDTotal-DZKK201809009.htm [9] 王盼盼, 李玉惠.基于特征融合和L-M算法的车辆重识别方法[J].电子科技, 2018, 31(4):12-15. http://www.cnki.com.cn/Article/CJFDTotal-DZKK201804005.htmWANG P P, LI Y H.Vehicle re-identification method based on feature fusion and L-M algorithm[J].Electronic Science and Technology, 2018, 31(4):12-15(in Chinese). http://www.cnki.com.cn/Article/CJFDTotal-DZKK201804005.htm [10] SUPREM A, PU C.Looking GLAMORous: Vehicle Re-Id in heterogeneous cameras networks with global and local attention[EB/OL].(2020-02-06)[2020-03-01].https://arxiv.org/abs/2002.02256. [11] ZAPLETAL D, HEROUT A.Vehicle re-identification for automatic video traffic surveillance[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops.Piscataway: IEEE Press, 2016: 25-31. http://www.doc88.com/p-2823566922213.html [12] KHORRAMSHAHI P, KUMAR A, PERI N, et al.A dual-path model with adaptive attention for vehicle re-identification[C]//Proceedings of the IEEE International Conference on Computer Vision.Piscataway: IEEE Press, 2019: 6132-6141. https://blog.csdn.net/qq_32523711/article/details/103011596 [13] KANACI A, LI M, GONG S, et al.Multi-task mutual learning for vehicle re-identification[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops.Piscataway: IEEE Press, 2019: 62-70. [14] NEWELL A, YANG K, DENG J.Stacked Hourglass networks for human pose estimation[C]//European Conference on Computer Vision.Berlin: Springer, 2016: 483-499. https://blog.csdn.net/zhangjunhit/article/details/79467761 [15] 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.Piscataway: IEEE Press, 2016: 770-778. https://blog.csdn.net/weixin_37993251/article/details/87905239 [16] HE B, LI J, ZHAO Y, et al.Part-Regularized near-duplicate vehicle re-identification[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Piscataway: IEEE Press, 2019: 3997-4005. https://blog.csdn.net/qq_32523711/article/details/103172104 [17] LIU H, TIAN Y, YANG Y, et al.Deep relative distance learning: Tell the difference between similar vehicles[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Piscataway: IEEE Press, 2016: 2167-2175. https://blog.csdn.net/u010167269/article/details/51783446 [18] ZHONG Z, ZHENG L, KANG G, et al.Random erasing data augmentation[EB/OL].(2017-11-16)[2020-03-01].https://arxiv.org/abs/1708.04896. [19] WANG Z, TANG L, LIU X, et al.Orientation invariant feature embedding and spatial temporal regularization for vehicle re-identification[C]//Proceedings of the IEEE International Conference on Computer Vision.Piscataway: IEEE Press, 2017: 379-387. https://blog.csdn.net/sinat_42239797/article/details/104476940 [20] LIU X, LIU W, MEI T, et al.Provid:Progressive and multimodal vehicle reidentification for large-scale urban surveillance[J].IEEE Transactions on Multimedia, 2017, 20(3):645-658. http://cn.bing.com/academic/profile?id=ecc5f148d32f61affb579ecd3beb9dad&encoded=0&v=paper_preview&mkt=zh-cn [21] ZHOU Y, SHAO L.Aware attentive multi-view inference for vehicle re-identification[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Piscataway: IEEE Press, 2018: 6489-6498. https://www.zhangqiaokeyan.com/academic-conference-foreign_ieeecvf-cference-computer-visi-and-patt_thesis/020512457608.html [22] GUO H, ZHAO C, LIU Z, et al.Learning coarse-to-fine structured feature embedding for vehicle re-identification[C]//Thirty-Second AAAI Conference on Artificial Intelligence.Palo Alto: AAAI Press, 2018: 6853-6860. [23] LIU X, ZHANG S, HUANG Q, et al.RAM: A region-aware deep model for vehicle re-identification[C]//2018 IEEE International Conference on Multimedia and Expo.Piscataway: IEEE Press, 2018: 1-6. RAM: A region-aware deep model for vehicle re-identification