HYPERSPECTRAL OBJECT TRACKING USING SPECTRAL ADAPTIVE IMAGING MODEL

Hyperspectral video data, with its high spectral resolution and unified imaging and spectroscopy, holds great potential for fine-grained object recognition and tracking. However, traditional hyperspectral imaging sensors are large in size and heavy in weight, hindering the integration on small platforms. Moreover, high spectral similarity exists between adjacent narrow-band video frames, leading to significant information redundancy. Based on a spectral adaptive imaging model, this paper proposes a task-solving framework for typical video object tracking. By establishing an imaging quality index evaluation criterion and time constraints, the proposed method can adaptively select the optimal detection and tracking band from spectral video data. The paper conducted experimental validation and analysis on multiple hyperspectral object tracking videos. Results show the proposed spectral band adaptive video object tracking method can screen out the optimal spectral detection and tracking narrow band. The method proposed in this paper can effectively improve processing and analysis efficiency and reduce data acquisition cost, facilitating the design of a simplified spectral imaging detection system.