基于声光作用的树种高效识别和特征提取的研究
马晃伟

1医学光电科学与技术教育部重点实验室;

2福建师范大学 光电与信息工程学院，福建福州，350007；

摘要：光谱分析技术凭借快速、无损、稳定可靠的特性，在植被与树种识别等领域得以广泛应用。该技术通过采集树叶光谱，提取特征波段的特征，实现了树种的高效分类与识别。不过，传统光谱分析方法存在一定局限，其需要反复采集全光谱数据，数据处理过程复杂，耗费大量时间成本。而且，高维数据中常存在信息冗余与噪声，这些因素可能对模型性能产生不利影响。有鉴于此，本文提出一种将声光滤波器（Acousto-optic Tunable Filter，AOTF）与机器学习相结合的创新方法。AOTF具有亚纳米级的精准波长选择能力，能够快速获取特征波长的光谱数据。在初次识别时，先采集全光谱数据，再借助机器学习算法进行降维处理，筛选出有效的特征波长，进而构建高效的识别模型。这种特征选择方式，不仅能过滤掉无效信息与干扰波段，还能显著降低数据维度，提升模型性能。相较于传统方法，该方法无需再次采集全光谱数据，极大地优化了数据采集流程，既节省了时间，又简化了数据结构。因此，该方法在工业、农业等领域的快速识别任务中具有广阔的应用前景。

关键词：AOTF；机器学习；快速识别；特征波长

参考文献

[1]Cao J, Liu K, Liu L, et al. Identifying mangrove species using field close-range snapshot hyperspectral imaging and machine-learning techniques[J]. Remote Sensing, 2018, 10(12): 2047.

[2]Velasquez-Camacho L, Cardil A, Mohan M, et al. Remotely sensed tree characterization in urban areas: a review[J]. Remote Sensing, 2021, 13(23): 4889.

[3]Chaity M D, van Aardt J. Exploring the limits of species identification via a convolutional neural network in a complex forest scene through simulated imaging spectroscopy[J]. Remote Sensing, 2024, 16(3): 498.

[4]Nanni M R, Demattê J A M, Rodrigues M, et al. Mapping particle size and soil organic matter in tropical soil based on hyperspectral imaging and non-imaging sensors[J]. Remote Sensing, 2021, 13(9): 1782.

[5]Burger J, Gowen A. Data handling in hyperspectral image analysis[J]. Chemometrics and Intelligent Laboratory Systems, 2011, 108(1): 13-22.

[6]Li X, Xia R, Li J, et al. The application and challenges of spectral and image two-modal fusion techniques in coal gangue recognition[J]. International Journal of Coal Preparation and Utilization, 2024: 1-31.

[7]Wu H Y, Li B, Liu G, et al. Design and development of a mid-wave infrared imaging spectrometer based on AOTF[C]//Third International Computing Imaging Conference (CITA 2023). SPIE, 2023, 12921: 374-382.

[8]Mi Z, Zhao H, Guo Q. Thermal analysis of TeO2-based acousto-optic tunable filters for spectral imaging[C]//Sixth Conference on Frontiers in Optical Imaging and Technology: Novel Imaging Systems. SPIE, 2024, 13155: 257-268.