Correlation Filter Network Model Performance Analysis

Object tracking is an significant field of computer vision, which is extensive used in the fields of automatic driving and intelligent monitoring. However, target tracking still has difficulties in tracking caused by factors such as occlusion, illumination, and scale changes. In order to accurately track the target and increase the tracking performance of the model, the paper reproduces the two goals of the full convolution twin neural network and the correlation filtering neural network. The tracking algorithm is tested using the OTB100 data set. The target tracking overlap rate differs by 1.3%, and the success rate differs by 1.5%. Experiments show that, in the face of complex environments, the relevant filtering neural network updates the model in time and shows excellent tracking results.