Dynamic Shadow Detection and Tracking of Online Video from Mobile View

Authors

Youpeng Zhang, 1. National Key Laboratory of Fundamental Science on Synthetic Vision, Sichuan University, Chengdu 610065, China; ;
Wang Chun, 2. College of Computer Science, Sichuan University, Chengdu 610065, China;
Yanli Liu, 2. College of Computer Science, Sichuan University, Chengdu 610065, China;

Abstract

Abstract: A shadow edge detection and tracking framework based on optical flow tracking in outdoor scenes is proposed. The SVM is trained by using the edge information features extracted from the known results; the optical flow tracking is performed on the two frames before and after; and the SVM model is used to identify the points corresponding to the shadow features from the neighborhood Canny confidence edges of the unstable tracking points. A method for dynamically updating the SVM is designed for the new material problem of the scene caused by the viewpoint movement. The complexity of the shadow projection area may invalidate the SVM. For this problem, a region comparison algorithm is used to improve the accuracy of the results. The experimental results show that the algorithm can accurately detect and track the projected shadows of moving objects such as moving humans in the video under the mobile view.

Recommended Citation

Zhang, Youpeng; Chun, Wang; and Liu, Yanli (2019) "Dynamic Shadow Detection and Tracking of Online Video from Mobile View," Journal of System Simulation: Vol. 31: Iss. 7, Article 22.
DOI: 10.16182/j.issn1004731x.joss.18-VR0714
Available at: https://dc-china-simulation.researchcommons.org/journal/vol31/iss7/22

First Page

1439

Revised Date

2018-10-25

DOI Link

https://doi.org/10.16182/j.issn1004731x.joss.18-VR0714

Last Page

1447

CLC

TP391

Recommended Citation

Zhang Youpeng, Wang Chun, Liu Yanli. Dynamic Shadow Detection and Tracking of Online Video from Mobile View[J]. Journal of System Simulation, 2019, 31(7): 1439-1447.

DOI

10.16182/j.issn1004731x.joss.18-VR0714