Hyperspectral video (HSV) encompasses a wealth of material information across its bands, offering significant potential over traditional RGB color video. HSV maintains robust object tracking capabilities under rapid appearance changes, showcasing its immense potential. However, most hyperspectral object tracking (HOT) methods still adopt an information selection strategy to generate a single false-color image for tracking, which fails to fully utilize the HSV feature and thus limits performance. Inspired by multi-view learning, we consider each spectral band of hyperspectral data as representing a different view of the same object, since the differentiated information reflected by different wavelengths points to the same semantic meaning. The joint representation of these information can be beneficial for a deeper understanding of the characteristics of the tracked object. To address this issue, this paper proposes a multi-view cross-attention network (MCAN) for HOT. Firstly, based on the differences in information across different wavelengths, we propose a multi-view learning framework to fully exploit the semantic information of HSV. Furthermore, we designed a triple cross-attention fusion module that can effectively fuse the HSV feature obtained from multi-view to fully utilize the advantages of HSV, thereby improving the efficiency and accuracy of tracking and recognition. The experimental results show that the method proposed in this paper is able to better utilize the information in the HSV domain and achieve more accurate and robust performance than most existing methods. The source code will be obtained from https://github.com/cccc-1111/MCAN.