激光与光电子学进展, 2020, 57 (24): 241008, 网络出版: 2020-11-19
一种基于改进SiameseRPN的全景视频目标跟踪算法 下载: 1029次
Algorithm for Panoramic Video Tracking Based on Improved SiameseRPN
图 & 表
图 3. 卷积网络提取的深度特征的可视化结果。 (a) Conv 1; (b) Relu 1; (c) Pool 1; (d) Conv 3; (e) Relu 3; (f) Pool 3; (g) Conv 5; (h) Relu 5; (i)原图
Fig. 3. Visualization results of deep features in convolutional neural networks. (a) Conv 1; (b) Relu 1; (c) Pool 1; (d) Conv 3; (e) Relu 3; (f) Pool 3; (g) Conv 5; (h) Relu 5; (i) original image
图 6. Relu6、h-swish、swish的激活函数对比
Fig. 6. Comparison among Relu6, h-swish, and swish activation functions
图 8. SiameseRPN与改进网络的实验结果对比
Fig. 8. Comparison of experiment results by SiameseRPN and improved network
图 9. 四个不同场景下不同算法的结果对比
Fig. 9. Comparison of results by different algorithms in four different scenarios
表 1全景数据集的跟踪难点分布
Table1. Distribution of tracking difficulties in panoramic data sets
|
表 2各算法的性能对比
Table2. Performance comparison of all algorithms
|
王殿伟, 方浩宇, 刘颖, 姜静, 任新成, 许志杰, 覃泳睿. 一种基于改进SiameseRPN的全景视频目标跟踪算法[J]. 激光与光电子学进展, 2020, 57(24): 241008. Dianwei Wang, Haoyu Fang, Ying Liu, Jing Jiang, Xincheng Ren, Zhijie Xu, Yongrui Qin. Algorithm for Panoramic Video Tracking Based on Improved SiameseRPN[J]. Laser & Optoelectronics Progress, 2020, 57(24): 241008.