国产欧美精品一区二区,中文字幕专区在线亚洲,国产精品美女网站在线观看,艾秋果冻传媒2021精品,在线免费一区二区,久久久久久青草大香综合精品,日韩美aaa特级毛片,欧美成人精品午夜免费影视

In recent years, object tracking technology has gradually become a research hotspot. The deep feature representation ability of the convolutional neural network model is strong, which can make full use of image information, and can distinguish categories. Meanwhile, the kernel correlation filter tracking algorithm converts the convolution in the time domain to the dot product in the frequency domain, which reduces the calculation complexity and significantly improves the calculation speed. Complementing the advantages of the two algorithms, a new tracking algorithm that combines convolutional neural network and kernel correlation filter algorithm is proposed. Specifically, the deep features of the Siamese network are extracted hierarchically to construct training the model offline. Then kernel correlation filter algorithm quickly calculates the target and predicts the location of the target. Therefore, the proposed algorithm can improve the performance of tracking accuracy.