Page 113 - 《广西植物》2026年第6期
P. 113

1 0 2 8                                广  西  植  物                                         46 卷
                            1          2                   1           2               1               2
                LI Shengfan ꎬ LÜ Shi ꎬ ZHENG Renhao ꎬ LIN Hui ꎬ DU Hongwei ꎬ YAN Huihui ꎬ
                                 1              1                2              2                 2∗
                     ZHOU Ran ꎬ WANG Ziyi ꎬ LIN Dongmei ꎬ LIN Zhanxi ꎬ LIU Fengshan
               ( 1. Forestry College ( Carbon Neutrality College )ꎬ Fujian Agriculture and Forestry Universityꎬ Fuzhou 350002ꎬ Chinaꎻ 2. International
                              College of Juncao Science/ National Engineering Research Center of Juncao Technologyꎬ Fujian
                                        Agriculture and Forestry Universityꎬ Fuzhou 350002ꎬ China )

                 Abstract: Aiming to address the challenges of morphological similarity among different species of Juncao and the low
                 efficiency of traditional identification methodsꎬ this study employed spectral feature analysis and machine learning
                 modeling to achieve efficient identification and classification of eight Juncao species. Spectral reflectance data of leaf
                 samples from eight speciesꎬ including Cenchrus fungigraminusꎬ Pennisetum purpureumꎬ P. purpureum cv. Laimu ̄1ꎬ
                 P. purpureum cv. Redꎬ P. alopecuroidesꎬ P. glaucum × purpureumꎬ P. americanumꎬ and Saccharum officinarum
                 cv. PNGꎬ were collected within the 400 - 900 nm wavelength range. By processing the original spectra along with their
                 reciprocal taking logarithm transformationsꎬ the first derivativesꎬ and the second derivativesꎬ spectral characteristics
                 were analyzedꎬ and multiple vegetation indices were extracted. Based on six types of feature sets — original spectrumꎬ
                 reciprocal taking logarithm transformationꎬ the first derivativeꎬ the second derivativeꎬ three ̄edge parametersꎬ and
                 vegetation index — twenty graded feature combinations were constructed. Classification models were developed using
                 support vector machine (SVM) and random forest (RF) algorithmsꎬ with model accuracy evaluated accordingly. The
                 results were as follows: (1) The eight Juncao species exhibited typical vegetation spectral characteristics in the visible
                 light regionꎬ with reflectance rising sharply in the red ̄edge region (700 - 750 nm). (2) Different spectral processing
                 methods significantly amplified inter ̄species spectral differences in specific bandsꎬ such as reciprocal taking logarithm
                 transformation within 570 - 650 nmꎬ the first derivatives around 730 nmꎬ and the second derivatives in the 670 - 760
                 nm range. The red ̄edge amplitudeꎬ red ̄edge areaꎬ and simple ratio (SR) vegetation index demonstrated the strongest
                 discriminative power among species. (3) Model performance indicated that the SVM algorithm generally outperformed
                 RF. The combination of “original spectra + three ̄edge parameters” achieved the highest accuracy of 70.56% under the
                 SVMꎬ which was 14.03% higher than the same combination under the RF. The optimal performance for RF was observed
                 with the “ original spectra + reciprocal taking logarithm transformation ” combinationꎬ though it showed limited
                 adaptability to high ̄dimensional features. In conclusionꎬ integrating spectral features from visible and near ̄infrared bands
                 with spectral transformation techniques and vegetation indicesꎬ coupled with the SVM algorithmꎬ provides an effective
                 theoretical foundation and technical support for the rapid identification and accurate classification of Juncao species.
                 Key words: species identificationꎬ spectral reflectanceꎬ feature extractionꎬ support vector machineꎬ random forest



                菌草是一类具有重要经济与生态价值的多年                            用需 求 日 益 凸 显 ( 陈 钟 佃 等ꎬ 2023ꎻ 张 丽 丽 等ꎬ
            生草本植物ꎬ20 世纪 80 年代福建农林大学国家菌                         2024ꎻ桂干北等ꎬ2024)ꎮ 但是ꎬ物种精准识别仍是
            草工程技术研究中心首创“以草代木” 栽培食药用                            产业瓶颈ꎬ不同菌草物种在营养生长期形态高度
            菌的“菌草技术”ꎬ改变了传统菌业依赖木材的生                             相似ꎬ传统形态学识别依赖经验、效率低且在幼苗
            产模式(林占熺等ꎬ2019)ꎮ 其因生长迅速、抗逆性                         期或逆境生长状态下易误判ꎬ影响物种布局与应
            强、适应性广等优势ꎬ已广泛应用于饲料生产、生                             用效果ꎮ
            态修复及生物能源开发等领域ꎬ成为推动农业循                                  光谱反射率技术作为植物生理生态研究的核
            环经济与生态治理的关键作物( 郑金英等ꎬ2011)ꎮ                         心手段ꎬ凭借非破坏性、快速高效、信息丰富的优
            随 着 菌 草 产 业 规 模 化 发 展ꎬ 杂 交 狼 尾 草                   势ꎬ已成 为 解 决 植 物 物 种 识 别 难 题 的 重 要 工 具
            ( Pennisetum glaucum × purpureum )、 紫 象 草 ( P.     (Semyalo et al.ꎬ2024)ꎮ 其核心原理在于植物叶片
            purpureum cv. Red)、象草( P. purpureum)、巨菌草           的光谱反射率特征由叶绿素、水分、氮素等内部生
            (Cenchrus fungigraminus) 等主流物种的差异化应                化组分与叶片结构共同决定ꎬ不同物种因生化与
   108   109   110   111   112   113   114   115   116   117   118