References
Chen C L P, Liu Z L. Broad learning system: an effective and efficient incremental learning system without the need for deep architecture. IEEE Trans Neural Netw Learn Syst, 2018, 29: 10–24
Bay H, Ess A, Tuytelaars T, et al. Speeded-up robust features (SURF). Comput Vision Image Underst, 2008, 110: 346–359
Zhang K H, Yang M H. Real-time compressive tracking. In: Proceedings of the 12th European Conference on Computer Vision, Florence, 2012. 864–877
Henriques J F, Caseiro R, Martins P, et al. High-speed tracking with kernelized correlation filters. IEEE Trans Pattern Anal Mach Intell, 2015, 37: 583–596
Kalal Z, Mikolajczyk K, Matas J. Tracking-learning-detection. IEEE Trans Pattern Anal Mach Intell, 2012, 34: 1409–1422
Lecun Y, Bottou L, Bengio Y, et al. Gradient-based learning applied to document recognition. Proc IEEE, 1998, 86: 2278–2324
Danelljan M, Robinson A, Khan F S, et al. Beyond correlation filters: learning continuous convolution operators for visual tracking. In: Proceedings of the 14th European Conference on Computer Vision, Amsterdam, 2016. 472–488
Hyeonseob N, Bohyung H. Learning multi-domain convolutional neural networks for visual tracking. In: Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, Las Vegas, 2016. 4293–4302
Wu Y, Lim J, Yang M H. Object tracking benchmark. IEEE Trans Pattern Anal Mach Intell, 2015, 37: 1834–1848
Acknowledgements
This work was supported in part by National Natural Science Foundation of China (Grant Nos. 51939001, 61976033, U1813203, 61803064, 61751202), Natural Foundation Guidance Plan Project of Liaoning (Grant No. 2019-ZD-0151), Science & Technology Innovation Funds of Dalian (Grant No. 2018J11CY022), and Fundamental Research Funds for the Central Universities (Grant No. 3132019345).
Author information
Authors and Affiliations
Corresponding author
Additional information
Conclusions and discussion
Based on our experimental results, the following two aspects will be discussed. (1) Tracking adaptability: First, we trained an accurate information evaluator. BLS is the process of acquiring sparse features. Sparse feature learning models are attractive for exploring the essential characteristics of tracking data. Based on statistical target occlusion and loss, we can adjust the candidate region search and SURF feature matching. By using such a tracking strategy, we can effectively enhance tracking adaptability. (2) Time consumption: Our method has a small time overhead because BLS has few parameters and is solved using ridge regression. Experimental results demonstrated the effectiveness of the proposed algorithm, but it still has some deficiencies. Taking tracking speed as an example, there is still room for further improvement. Robustness to long-term occlusion and loss, as well as scale variation, must be improved. Additionally, online tracking must be implemented. Follow-up research will focus on these issues and additional studies will be required to develop faster and more robust online target tracking systems.
Electronic Supplementary Material
Supplementary material, approximately 31 MB.
Rights and permissions
About this article
Cite this article
Zhang, D., Li, T., Chen, C.L.P. et al. Target tracking algorithm based on a broad learning system. Sci. China Inf. Sci. 65, 154201 (2022). https://doi.org/10.1007/s11432-020-3272-y
Received:
Revised:
Accepted:
Published:
DOI: https://doi.org/10.1007/s11432-020-3272-y