Target tracking algorithm based on a broad learning system
Zhang, Dan; Li, Tieshan; Chen, C. L. Philip; Yang, He
Sci China Inf Sci, 2022, 65(5): 154201
Deep learning can be used to solve the problem of deep information acquisition. Based on the successful application of convolutional neural networks (CNNs) in the image processing direction, CNNs were successfully applied to video tracking. Although deep learning performs well at target tracking, real-time tracking must be improved in terms of its computational cost. A broad learning system (BLS) was proposed. As an alternative to a deep network architecture, its calculation speed is very fast. Additionally, the BLS can extract sparse features from training data and sparse feature learning models are attractive for exploring essential characterization. Based on these advantages we propose a target tracking algorithm based on BLS using a candidate region search and SURF feature matching of multiple clues. This represents an attempt at applying broad learning to target tracking.