Research on Object Tracking and Target Recognition Based on Kalman Filter and YOLOV3

Xueming Zhai ； Jilei Jia

《International Core Journal of Engineering》 6卷11期 (2020/11) Pp. 164-169

https://doi.org/10.6919/ICJE.202011_6(11).0023

摘要

In order to solve the existing problems in moving target recognition and tracking, YOLOV3 is used to detect the target to be tracked in the current frame, and Kalman filter is used to predict the next position and the size of the bounding box according to the position of the current target. The improved Hungarian algorithm is used to correlate and match the data according to the intersection ratio and color histogram of the detected and predicted borders, and the target motion trajectory is obtained through continuous iteration of the system to complete the tracking. For the occluded target, a region based quality assessment network is introduced, which combines multiple high-quality detection images to recover the occluded part and improve the tracking accuracy.

關鍵字

YOLOV3 ； Kalman filter ； Hungarian algorithm

參考文獻

Girshick R,Donahue J,Darrell T,et al.Rich feature hierarchies for accurate obje- ct detection and semantic segmentation// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Columbus, USA,2014:58 0-587.

Wang L,Xu L,Min Y K, et al.Online multiple object tracking via flow and convolutional features[C]//IEEE International Conference on Image Processing. 2018.

Shen J, Liang Z, Liu J, et al. Multiobject Tracking by Submodular Optimization[J]. IEEE Transactions on Cybernetics, 2018, PP(99):1-12.

Nejhum S,Ho J,and Yang M H.Online visual tracking with histograms and articulating blocks[J]. Computer Vision and Image Understanding, 2010, 114(8): 901-914.

Kwon J,Lee K M.Tracking of a non-rigid object via patch-based dynamic appearance modeling and adaptive Basin Hopping Monte Carlo sampling[C]// Computer Vision and Pattern Recognition, 2009. CVPR 2009. IEEE Conference on. IEEE, 2009:1208-1215.