Research on Band Selection of Visual Attention Mechanism for Object
Detection
YANG Guang1, JIN Chun-bai1, REN Chun-ying2*, LIU Wen-jing1, CHEN Qiang1
1. Aviation University of Air Force, Changchun 130022, China
2. Northeast Institute of Geography and Agroecology, Chinese Academy of Sciences, Changchun 130102, China
摘要： 近年来，波段选择在高光谱图像降维处理中得到了广泛地应用，然而常用的数据降维方法并没能将与人类视觉系统相关的信息进行有效利用，如果将人类与生俱来的视觉注意机制能力应用到高光谱图像中目标的视觉显著性特征的增强或识别，对于高光谱图像的目标检测研究无疑会产生相当的促进作用。研究提出引入视觉注意机制理论应用于波段选择研究，构建面向目标检测应用的视觉注意机制波段选择模型。通过分析计算波段图幅的目标与背景的可识别程度，量化所在波段对地物目标与背景的判别能力，提出了基于目标视觉可识别度的波段选择方法；利用LC显著性算法进行空间域的视觉显著性目标分析，计算背景与目标的显著性差异绝对值，提出基于LC显著目标结构分布的波段选择方法。将这两种方法结合提出的改进子空间划分方法，建立面向目标检测的视觉注意机制波段选择模型，并经高光谱遥感AVIRIS San Diego公开数据集进行目标检测实验验证，结果表明所提出的基于视觉注意机制的波段选择模型对于目标检测应用具有较好的检测效果，实现了数据降维和高效的计算处理。
关键词：波段选择；视觉注意机制；可识别度；显著性算法；目标检测
Abstract：In recent years, band selection has been widely used in hyperspectral image dimensionality reduction processing. However, the commonly used data dimensionality reduction methods have not effectively utilized the information related to the human visual system. The research on target detection of hyperspectral images will undoubtedly have a considerable role in promoting. This paper proposes to apply the theory of visual attention mechanism to the study of band selection and constructs a band selection model of visual attention mechanism for target detection applications. By analyzing and calculating the identifiability of the target and background in the band map and quantifying the discrimination ability of the band to the ground object target and background, a band selection method based on the target visual identifiability is proposed. The LC saliency algorithm is used to analyze the visual saliency targets in the spatial domain, calculate the absolute value of the significance difference between the background and the target, and propose a band selection method based on the structural distribution of LC saliency targets. These two methods are combined with the improved subspace partition method proposed to establish a band selection model of visual attention mechanism for target detection. The model is verified by target detection experiments on hyperspectral remote sensing AVIRIS San Diego public dataset. The results show that the proposed band selection model based on the visual attention mechanism has a good detection effect for target detection applications and realizes data reduction and efficient computing processing.
杨 桄，金椿柏，任春颖，刘文婧，陈 强. 面向目标检测的视觉注意机制波段选择研究[J]. 光谱学与光谱分析, 2024, 44(01): 266-274.
YANG Guang, JIN Chun-bai, REN Chun-ying, LIU Wen-jing, CHEN Qiang. Research on Band Selection of Visual Attention Mechanism for Object
Detection. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2024, 44(01): 266-274.
[1] Cheng S, Wang L, Du A. Springer Science and Business Media LLC, 2021.
[2] Chen X, Liu L, Kong X. Optics & Laser Technology, 2022, 149: 107787.
[3] Dai S, Li D. Sensors, 2020, 20(2): 459.
[4] Itti L, Koch C, Niebur E. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1998, 20(11): 1254.
[5] Le Moan S, Deger F, Mansouri A, et al. International Conference on Image & Signal Processing. Agadir, Morocco, 2012.
[6] Cao Y, Zhang J, Tian Q, et al. IEEE China Summit & International Conference on Signal & Information Processing. IEEE, 2015.
[7] Zhang J, Sclaroff S, Zhe L, et al. IEEE International Conference on Computer Vision. IEEE, 2015.
[8] Kadir T, Brady M. Visual Information Engineering, 2003. VIE 2003. International Conference on. 2003.
[9] Liang J, Zhou J, Bai X, et al. International Conference on Image Processing. IEEE, 2013.
[10] Su P, Liu D, Li X, et al. IEEE Geoscience and Remote Sensing Letters, 2018, 15(4): 572.
[11] Wang Q, Li Q, Li X. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2019, 12(2): 4940.
[12] CAO Peng-fei, PENG Chang-ning(曹鹏飞, 彭昌宁). Optical Instruments (光学仪器), 2018, 40(3): 33.
[13] Yun Z, Shah M. Proceedings of the 14th ACM International Conference on Multimedia, Santa Barbara, CA, USA, October 23-27, 2006. ACM, 2006.
[14] Green R O, Eastwood M L, Sarture C M, et al. Remote Sensing of Environment, 1998, 65(3): 227.
