Research of Moving Target Tracking Technology Based on LRCN
Jian Di, Hongyan Liu · 2017 International Conference on Computer Systems, Electronics and Control (ICCSEC) · 2017
Moving target tracking is a hot spot in computer vision in recent years. The tracking method, which is dominated by particle filter, has been widely used. The particle filter algorithm exhibits high complexity and poor real-time performance under large data processing conditions. With the development of neural networks and big data, a new LRCN network model combined with CNN and LSTM is proposed for moving target tracking in this paper. LRCN uses the deep learning framework to extract the characteristics of the video data. and uses the CNN to acquire the characteristics of the video sequence image, then make predictions in chronological order through the LSTM network structure. In addition, LRCN utilizes double deep learning to synchronize space convolution and time stream convolution. Using Matlab to experiment with standard data set VTB, remarkable results have been achieved in tracking accuracy and success rate with LRCN.