WO2016000088A1 - Hyperspectral waveband extraction method based on optimal index factor-correlation coefficient method - Google Patents

Hyperspectral waveband extraction method based on optimal index factor-correlation coefficient method Download PDF

Info

Publication number
WO2016000088A1
WO2016000088A1 PCT/CN2014/000676 CN2014000676W WO2016000088A1 WO 2016000088 A1 WO2016000088 A1 WO 2016000088A1 CN 2014000676 W CN2014000676 W CN 2014000676W WO 2016000088 A1 WO2016000088 A1 WO 2016000088A1
Authority
WO
WIPO (PCT)
Prior art keywords
index
correlation coefficient
data
band
bands
Prior art date
Application number
PCT/CN2014/000676
Other languages
French (fr)
Chinese (zh)
Inventor
程志庆
张劲松
郑宁
王鹤松
李春友
Original Assignee
中国林业科学研究院林业研究所
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 中国林业科学研究院林业研究所 filed Critical 中国林业科学研究院林业研究所
Publication of WO2016000088A1 publication Critical patent/WO2016000088A1/en

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands

Definitions

  • the invention relates to the field of spectral data processing, in particular to a method for extracting a hyperspectral remote sensing band.
  • Hyperspectral technology is the use of light from objects Spectral characteristics, the spectral information of the object is obtained by a high-resolution spectroscopic instrument, and the characteristic band of the object is analyzed by means of analysis Extract and distinguish to obtain useful spectral information. Due to the high resolution of the hyperspectral, the amount of data obtained is large and More redundant information, therefore, the ability to extract useful spectral information from hyperspectral data for hyperspectral techniques has Great significance.
  • the main analytical tool is to reduce the spectral information and extract the useful band.
  • the correlation coefficient analysis method is more commonly used and applied to remote sensing image processing.
  • the Optimum Index Factor can obtain the band combination with the most information. It has the advantages of rich information and small redundancy of band information, which can provide important reference for hyperspectral data processing. Comprehensive In addition, if the above two methods can be combined, it will be beneficial to improve the detection and simulation capabilities of hyperspectral data.
  • the object of the present invention is to overcome the conventional hyperspectral data band selection method and use a single band to meet the target requirements. Inversion is very susceptible to interference from other factors, and the use of multiple bands lacks the relationship between each band and the target. A method for selecting a band of hyperspectral remote sensing data is provided.
  • the invention comprises the following steps: a method for extracting hyperspectral bands based on an optimal exponential-correlation coefficient method, including the following step:
  • Step A Sort and sort the original hyperspectral data. The specific steps are as follows:
  • the unwanted information in the single raw hyperspectral data obtained is removed, and then the reflectivity of all the individual data is The data is integrated into the same file as the basic database for the following processing;
  • Step B Perform the optimal index calculation process after classifying and sorting the original hyperspectral data.
  • the specific method is as follows:
  • the optimal combination band needs to select three relevant bands for calculation at the same time, and the best index OIF is used as the evaluation index of the optimized combination.
  • the calculation formula is: Where: S i is the standard deviation of any i-band of the three bands, R ij is the correlation coefficient of any i and j bands of the three bands, and r is the combination number of any i and j bands;
  • Step C The 3-band correlation coefficient is simultaneously selected to calculate the maximum value, and the calculation method is as follows:
  • R std Rr std , where: R std is the correlation coefficient evaluation index of the 3 bands. The larger the R stf is, the larger the correlation coefficient values of the three bands are at the same time; R is the sum of the three bands and the target data, and r std is the standard deviation of the correlation coefficient between each of the three bands and the target data;
  • Step D Using the calculation method of the commercial power index to establish a comprehensive index evaluation system for the best index and correlation coefficient of the data, And on this basis, the selection of the hyperspectral band is as follows:
  • the calculation method of the commercial power index is to use the best index calculation result and the correlation coefficient calculation result as the two input index values.
  • Negative indicator data optimization processing formula is where: r ij ′ is the optimization index evaluation value of the i-th item under the j-th index, and max(r ij ) and min(r ij ) are the maximum and minimum values of the optimization index evaluation of all i items in the j-th index;
  • L 1 is the selection index of the hyperspectral band extraction method based on the optimal exponential-correlation coefficient method.
  • this experiment selects 5 nm interval hyperspectral data for processing and analysis, with 108 Spectral samples are modeled.
  • the measured data of chlorophyll data is directly observed by instruments, and the observation instrument is Konica's SPAD. Type chlorophyll meter.
  • Spectral data is processed by Matlab R2012b programming, and the extracted band is used to minimize the second Multiplication is used for regression analysis, and the accuracy ratio is compared with the regression model obtained by using the best index method or the correlation coefficient method alone. Compared.
  • the OIF value of the three-band combination of Figure 1, 2 is obtained by calculating the optimal index (OIF) of the pre-processed spectral data, OIF
  • the maximum value is mainly distributed in three band intervals: the first band is located at 740nm-1115nm, and the second band is located at 1850nm-1860nm, the third band is located at 1930nm-2010nm.
  • the three-band combination of the maximum values obtained by the OIF algorithm is: 745 nm, 1860 nm, 1950 nm and 750 nm, 1860 nm, 1950 nm.
  • the numbers are -0.679 and -0.692, respectively, and the bands with the largest correlation coefficients in the two groups are: 696 nm (correlation coefficient -0.728) And 1890nm (correlation coefficient -0.775); chlorophyll content and its original spectral reflectance at the 740-1140nm band Significant positive correlation, but the correlation coefficient is small, where the position correlation coefficient is the largest at 770 nm (correlation coefficient is 0.46). Studies have shown that the main absorption peak of chlorophyll is blue and red light, and in the green light area is absorption low. Therefore in the selection of the relationship When modeling the number of chlorophyll, the band with the largest correlation coefficient between 350 nm and 800 nm is used at 696 nm.
  • the model of the various methods obtained in Table 1 and the decision coefficients of the model show that the decision method of the selected three methods is The number has reached extremely significant, and the coefficient of determination from large to small is OIFC>OIF>MCC, and the 3-band pass obtained by the OIFC method.
  • the coefficient of determination for PLS modeling reached 0.739, which was 0.027 and 0.1383 higher than OIF and MCC, respectively.
  • the verification data randomly selected 22 sets of measured data in different growth periods of wheat to predict the predicted values of the above five groups of models. Line verification ( Figure 6).
  • the predicted values of the three sets of models have a significant linear correlation with the measured values.
  • the model prediction effect established by the extraction band of the invention has high precision.

Abstract

A hyperspectral waveband extraction method based on an optimum index factor-correlation coefficient method, comprising the following steps: step A: conducting processing such as sorting and organization of original hyperspectral data; step B: conducting optimum index factor calculation after sorting and organizing the original hyperspectral data; step C: calculating the maximum 3 waveband correlation coefficients; step D: selecting a hyperspectral waveband based on the optimum index factor calculation, correlation calculation and weight-quotient index calculation of the data. The method overcomes the defects of a traditional hyperspectral data waveband selection method of being susceptible to interference from other factors when inverting a target requirement using a single waveband, and lacking a relation between each waveband and a target when using a multiband.

Description

一种基于最佳指数-相关系数法的高光谱波段提取方法 Hyperspectral band extraction method based on optimal exponential-correlation coefficient method 技术领域 Technical field
本发明涉及光谱数据处理领域，尤其涉及一种高光谱遥感波段提取方法。 The invention relates to the field of spectral data processing, in particular to a method for extracting a hyperspectral remote sensing band.
背景技术 Background technique
随着光谱技术的发展，高光谱技术被广泛的应用到各领域。高光谱技术是利用物体的光 谱特性，通过高分辨率的光谱仪器获得物体的光谱信息，并使用分析手段对物体的特征波段 进行提取与区分，从而获取有用的光谱信息。由于高光谱的高分辨了导致获得的数据量大且 冗余信息多，因此，能够更好地从高光谱数据中提取有用光谱信息对高光谱技术的应用具有 重大意义。在光谱分析领域中，主要分析手段是对光谱信息进行降低维度与有用波段提取。 其中，相关系数分析法较为常用，并应用于遥感图像处理中。但该方法仅提取与目标要求相 关性最大的波段，而使用单个波段对目标要求进行反演时很容易受到其他因子的干扰。在遥 感图像处理中，最佳指数法(Optimum Index Factor，OIF)能够获得信息量最多的波段组合， 具有信息量丰富、波段信息冗余度小等优点，可为高光谱数据处理提供重要的借鉴思路。综 上，如能将上述两种方法结合到一起，将有益于提升高光谱数据探测、模拟能力。 With the development of spectroscopy technology, hyperspectral technology has been widely applied to various fields. Hyperspectral technology is the use of light from objects Spectral characteristics, the spectral information of the object is obtained by a high-resolution spectroscopic instrument, and the characteristic band of the object is analyzed by means of analysis Extract and distinguish to obtain useful spectral information. Due to the high resolution of the hyperspectral, the amount of data obtained is large and More redundant information, therefore, the ability to extract useful spectral information from hyperspectral data for hyperspectral techniques has Great significance. In the field of spectral analysis, the main analytical tool is to reduce the spectral information and extract the useful band. Among them, the correlation coefficient analysis method is more commonly used and applied to remote sensing image processing. But this method only extracts the target requirements The most critical band, while using a single band to invert the target requirements is vulnerable to interference from other factors. In the distance In the image processing, the Optimum Index Factor (OIF) can obtain the band combination with the most information. It has the advantages of rich information and small redundancy of band information, which can provide important reference for hyperspectral data processing. Comprehensive In addition, if the above two methods can be combined, it will be beneficial to improve the detection and simulation capabilities of hyperspectral data.
发明内容 Summary of the invention
本发明目的在于克服传统高光谱数据波段选择方法所存在的使用单个波段对目标要求进 行反演时很容易受到其他因子干扰的不足，而使用多波段时缺少对各波段与目标之间的关系， 提供一种高光谱遥感数据波段选择方法。 The object of the present invention is to overcome the conventional hyperspectral data band selection method and use a single band to meet the target requirements. Inversion is very susceptible to interference from other factors, and the use of multiple bands lacks the relationship between each band and the target. A method for selecting a band of hyperspectral remote sensing data is provided.
本发明包括以下步骤：一种基于最佳指数-相关系数法的高光谱波段提取方法，包括以下 步骤： The invention comprises the following steps: a method for extracting hyperspectral bands based on an optimal exponential-correlation coefficient method, including the following step:
步骤A：对原始高光谱数据进行分类、整理预处理，具体步骤如下： Step A: Sort and sort the original hyperspectral data. The specific steps are as follows:
首先对获得的单个原始高光谱数据中无用信息进行剔除，然后将所有单个数据中反射率 数据综合到同一个文件中，作为以下处理的基本数据库； First, the unwanted information in the single raw hyperspectral data obtained is removed, and then the reflectivity of all the individual data is The data is integrated into the same file as the basic database for the following processing;
步骤B：在原始高光谱数据分类、整理后进行最佳指数计算处理，具体方法如下： Step B: Perform the optimal index calculation process after classifying and sorting the original hyperspectral data. The specific method is as follows:
最优组合波段需要同时选取3个相关波段进行计算，采用最佳指数OIF作为优化组合的 评价指标，其计算公式为：
Figure PCTCN2014000676-appb-000001
其中：Si为选取3个波段中任意第i波段 的标准差，Rij为选取3个波段中任意i、j两波段的相关系数，r为任意i、j两波段的组合数； The optimal combination band needs to select three relevant bands for calculation at the same time, and the best index OIF is used as the evaluation index of the optimized combination. The calculation formula is:
Figure PCTCN2014000676-appb-000001
Where: S i is the standard deviation of any i-band of the three bands, R ij is the correlation coefficient of any i and j bands of the three bands, and r is the combination number of any i and j bands;
步骤C：3波段相关系数同时选取最大值的计算，计算方法如下： Step C: The 3-band correlation coefficient is simultaneously selected to calculate the maximum value, and the calculation method is as follows:
通过最佳指数计算后获得的所有3波段组合，计算每种波段与目标数据的相关性，然后 利用公式Rstd＝R-rstd计算最大相关系数，其中：Rstd为3波段的相关系数评价指标，Rstf越大 表示3个波段的相关系数值均同时最大；R为3个波段各自与目标数据相关系的总和，rstd为 3个波段各自与目标数据相关系数的标准差； Calculate the correlation between each band and the target data by using all the 3-band combinations obtained after the best index calculation, and then calculate the maximum correlation coefficient by using the formula R std =Rr std , where: R std is the correlation coefficient evaluation index of the 3 bands. The larger the R stf is, the larger the correlation coefficient values of the three bands are at the same time; R is the sum of the three bands and the target data, and r std is the standard deviation of the correlation coefficient between each of the three bands and the target data;
步骤D：采用商权指数计算方法建立数据的最佳指数与相关性系数综合指标评价体系， 并在此基础上进行高光谱波段的选择，具体如下： Step D: Using the calculation method of the commercial power index to establish a comprehensive index evaluation system for the best index and correlation coefficient of the data, And on this basis, the selection of the hyperspectral band is as follows:
商权指数的计算方法是将最佳指数计算结果与相关系数计算结果作为计算的两个输入指 标值，测评系统中目标数据对象集合为F＝(OIF，C)，基于最佳指数与相关性系数综合指标评 价体系计算参数集合c＝(c1，c2，…，cm)，得到原始评价信息矩阵R＝(rij)m×2，其中：rij为第j评 价指标下第i项目的优化指标评价值； The calculation method of the commercial power index is to use the best index calculation result and the correlation coefficient calculation result as the two input index values. The target data object set in the evaluation system is F=(OIF, C), based on the best index and correlation. The coefficient comprehensive index evaluation system calculates the parameter set c=(c 1 ,c 2 ,...,c m ), and obtains the original evaluation information matrix R=(r ij ) m×2 , where: r ij is the ith under the jth evaluation index. The evaluation value of the project's optimization index;
由于系统中各因素的量纲不一定相同，数值有时相差悬殊，从而造成数据的比较的难度 增大，需要对原始数据优化指标处理以及归一化处理，方法为正指标数据优化处理公式为
Figure PCTCN2014000676-appb-000002
负指标数据优化处理公式为
Figure PCTCN2014000676-appb-000003
其中：rij′为第j指 标下第i项目的优化指标评价值，max(rij)与min(rij)为第j指标里的所有i项目中优化指标评 价的最大值和最小值； Since the dimensions of various factors in the system are not necessarily the same, the values sometimes differ greatly, which makes the comparison of the data more difficult. The original data optimization index processing and normalization processing are required. The method is that the positive index data optimization processing formula is
Figure PCTCN2014000676-appb-000002
 Negative indicator data optimization processing formula is
Figure PCTCN2014000676-appb-000003
 Where: r ij ′ is the optimization index evaluation value of the i-th item under the j-th index, and max(r ij ) and min(r ij ) are the maximum and minimum values of the optimization index evaluation of all i items in the j-th index;
第j指标下第i项目的优化指标之间的比重值Pij计算公式为
Figure PCTCN2014000676-appb-000004
由商权法计 算的第j指标的商值
Figure PCTCN2014000676-appb-000005
其中：k＝1/ln2，当Pij＝0时，PijlnPij＝0，第j 指标的商权为
Figure PCTCN2014000676-appb-000006
 The specific gravity value P ij between the optimization indicators of the i-th item under the j-th index is calculated as
Figure PCTCN2014000676-appb-000004
The quotient of the jth indicator calculated by the commercial law
Figure PCTCN2014000676-appb-000005
Where: k=1/ln2, when P ij =0, P ij lnP ij =0, the trade term of the j-th indicator is
Figure PCTCN2014000676-appb-000006
综合权重计算公式为
Figure PCTCN2014000676-appb-000007
其中：
Figure PCTCN2014000676-appb-000008
为主观权重，w′j为综合权重；最佳指数 -相关系数法提取的总指标将可行方案集映射到“距离”空间，并将LP(w′j，j)作为综合评价 的总指标，其中
Figure PCTCN2014000676-appb-000009
取P＝1，此时
Figure PCTCN2014000676-appb-000010
L1称之为 海明距离，只注重偏差的总和，归一化处理公式为
Figure PCTCN2014000676-appb-000011
上述公式中i＝1，2，...，m；j＝1，2， L1越大则其综合评价值越高，这样就可以按照L1从小到大的顺序排序，从而获得提取波段结果， L1即为基于最佳指数-相关系数法的高光谱波段提取方法的选择指标。 The comprehensive weight calculation formula is
Figure PCTCN2014000676-appb-000007
among them:
Figure PCTCN2014000676-appb-000008
Subjective weight, w′ j is the comprehensive weight; the total index extracted by the best index-correlation coefficient method maps the feasible solution set to the “distance” space, and uses L P (w′ j , j) as the overall index of comprehensive evaluation. ,among them
Figure PCTCN2014000676-appb-000009
 Take P=1, at this time
Figure PCTCN2014000676-appb-000010
 L 1 is called the Hamming distance, and only the sum of the deviations is emphasized. The normalized processing formula is
Figure PCTCN2014000676-appb-000011
In the above formula, i=1, 2, ..., m; j=1, 2, the larger the L 1 is, the higher the comprehensive evaluation value is, so that the order of L 1 can be sorted from small to large, thereby obtaining the extraction band. As a result, L 1 is the selection index of the hyperspectral band extraction method based on the optimal exponential-correlation coefficient method.
附图说明 DRAWINGS
图1三波段组合的OIF值分布图 Figure 1 OIF value distribution of the three-band combination
图2三波段组合的OIF最大值分布图 Figure 2 OIF maximum distribution of the three-band combination
图3原始光谱与叶绿素含量的相关系数 Figure 3 Correlation coefficient between original spectrum and chlorophyll content
图4最优指数-相关系数选择指标(L1)分布图 Figure 4 Optimal index-correlation coefficient selection index (L 1 ) distribution map
图5最优指数-相关系数选择指标(L1)最大值分布图 Fig. 5 Optimal index-correlation coefficient selection index (L 1 ) maximum value distribution map
图6小麦叶片SPAD实测值与3种方法预测值的比较 Fig.6 Comparison of measured values of SPAD in wheat leaves and predicted values of three methods
具体实施方式 Detailed ways
以小麦活体叶绿素含量高光谱数据的波段的提取为例： Take the extraction of the band of hyperspectral data of chlorophyll content in wheat as an example:
由于高光谱数据量大，因此本实验选取5nm间隔的高光谱数据进行处理分析，以108个 光谱样本进行建模。叶绿素数据实测数据直接使用仪器观测，观测仪器为柯尼卡公司的SPAD 型叶绿素仪。采用Matlab R2012b编程对光谱数据进行处理，并对提取的波段利用偏最小二 乘法进行回归分析，并与单独使用最佳指数法或相关系数法所获得的回归模型进行精度的比 较。 Due to the large amount of hyperspectral data, this experiment selects 5 nm interval hyperspectral data for processing and analysis, with 108 Spectral samples are modeled. The measured data of chlorophyll data is directly observed by instruments, and the observation instrument is Konica's SPAD. Type chlorophyll meter. Spectral data is processed by Matlab R2012b programming, and the extracted band is used to minimize the second Multiplication is used for regression analysis, and the accuracy ratio is compared with the regression model obtained by using the best index method or the correlation coefficient method alone. Compared.
通过对预处理光谱数据进行最佳指数(OIF)计算后得到图1、2三波段组合OIF值，OIF 最大值主要分布在三个波段区间：第一波段位于740nm-1115nm，第二波段位于 1850nm-1860nm，第三波段位于1930nm-2010nm。OIF算法获得的最大值的三波段组合为： 745nm、1860nm、1950nm和750nm、1860nm、1950nm。 The OIF value of the three-band combination of Figure 1, 2 is obtained by calculating the optimal index (OIF) of the pre-processed spectral data, OIF The maximum value is mainly distributed in three band intervals: the first band is located at 740nm-1115nm, and the second band is located at 1850nm-1860nm, the third band is located at 1930nm-2010nm. The three-band combination of the maximum values obtained by the OIF algorithm is: 745 nm, 1860 nm, 1950 nm and 750 nm, 1860 nm, 1950 nm.
通过对预处理的高光谱数据进行叶绿素含量与其光谱反射率的相关系数(R)计算得出 结果，如图3，结果可见在620nm-700nm及1855-1920nm间达到相关系数峰值，平均相关系 数分别-0.679和-0.692，两组区间中各自相关系数最大的波段分别为：696nm(相关系数为-0.728) 和1890nm(相关系数为-0.775)；在740-1140nm波段位置叶绿素含量与其原始光谱反射率具有 显著的正相关性，但相关系数较小，其中在770nm(相关系数为0.46)波段位置相关系数最大。 研究表明叶绿素主要吸收峰是蓝光和红光区域，在绿光区域是吸收低谷。因此在选用相关系 数对叶绿素建模时，使用350nm-800nm之间相关系数最大的波段696nm。 Calculated by calculating the correlation coefficient (R) between the chlorophyll content and the spectral reflectance of the preprocessed hyperspectral data. As a result, as shown in Fig. 3, it can be seen that the correlation coefficient peak is reached between 620 nm and 700 nm and 1855 to 1920 nm, and the average phase relationship is obtained. The numbers are -0.679 and -0.692, respectively, and the bands with the largest correlation coefficients in the two groups are: 696 nm (correlation coefficient -0.728) And 1890nm (correlation coefficient -0.775); chlorophyll content and its original spectral reflectance at the 740-1140nm band Significant positive correlation, but the correlation coefficient is small, where the position correlation coefficient is the largest at 770 nm (correlation coefficient is 0.46). Studies have shown that the main absorption peak of chlorophyll is blue and red light, and in the green light area is absorption low. Therefore in the selection of the relationship When modeling the number of chlorophyll, the band with the largest correlation coefficient between 350 nm and 800 nm is used at 696 nm.
通过商权法结合最佳指数(OIF)值与相关系数值计算并利用海明距离作为评价指标获得 最优指数-相关系数选择指标图4、5，可知同时获得最大最优指数和叶绿素具有最大相关系 数三波段的组合第一波段位于670nm、740-1115nm，第二波段位于760nm、1850-1875nm，第 三波段位于1925-2500nm。由最大L1值决定的最佳指数-相关系数的三波段为：760nm、1860nm、 1970nm，其分别位于红光与近红外波段上。 By using the commercial rights method combined with the best index (OIF) value and the correlation coefficient value and using the Hamming distance as the evaluation index to obtain the optimal index-correlation coefficient selection index, Figures 4 and 5, it can be seen that the maximum optimal index and the chlorophyll are the largest. Correlation coefficient The combination of the three bands is at 670 nm, 740-1115 nm, the second band is at 760 nm, 1850-1875 nm, and the third band is at 1925-2500 nm. The three bands of the best exponential-correlation coefficient determined by the maximum L 1 value are: 760 nm, 1860 nm, and 1970 nm, which are respectively located in the red and near-infrared bands.
使用Matlab R2012b分别对108个光谱样本的OIF最大值的三波段745nm、1860nm、1950nm、 最佳指数-相关系数(OIFC)的三波段：760nm、1860nm和1970nm进行偏最小二乘法回归分 析，使用指数拟合相关系数最大的波段696nm，所得结果见表1。 Using Matlab R2012b, the three-band 745 nm, 1860 nm, 1950 nm of the OIF maximum of 108 spectral samples, respectively. The best index-correlation coefficient (OIFC) three-band: 760nm, 1860nm and 1970nm for partial least squares regression Analysis, using the index to fit the band with the largest correlation coefficient of 696 nm, the results are shown in Table 1.
表1回归模型与决定系数 Table 1 regression model and determination coefficient
Figure PCTCN2014000676-appb-000012
Figure PCTCN2014000676-appb-000012
注：表中**表示通过极显著统计检验 Note: ** in the table indicates that the statistical test is extremely significant.
表1中获得的各种方法的模型以及模型的决定系数可知，所选取的3种方法建模的决定系 数均已达到极显著，决定系数由大到小分别为OIFC>OIF>MCC，其中OIFC法获得的3波段通 过PLS建模的决定系数达到0.739，与OIF、MCC相比分别高出0.027、0.1383。 The model of the various methods obtained in Table 1 and the decision coefficients of the model show that the decision method of the selected three methods is The number has reached extremely significant, and the coefficient of determination from large to small is OIFC>OIF>MCC, and the 3-band pass obtained by the OIFC method. The coefficient of determination for PLS modeling reached 0.739, which was 0.027 and 0.1383 higher than OIF and MCC, respectively.
验证数据为小麦不同的生育时期内分别随机选取22组实测数据对上述5组模型预测值进 行验证(图6)。 The verification data randomly selected 22 sets of measured data in different growth periods of wheat to predict the predicted values of the above five groups of models. Line verification (Figure 6).
从模型预测值实测值之间的线性拟合图可知，3组模型预测值与实测值具有显著的线性相 关性，预测值与实测值的显著性OIFC最大(R2＝0.818)，其次为OIF(R2＝0.762)，最小的为相 关系数法选取的波段，同时OIFC预测值与实测值之间的均方根误差最小，可见相比单独的使 用最佳指数法或相关系数法获取的波段，本发明提取波段建立的模型预测效果具有较高的精 度。 From the linear fit between the measured values of the model predictions, the predicted values of the three sets of models have a significant linear correlation with the measured values. The significance of the predicted and measured values is the largest (IF 2 = 0.818), followed by OIF. (R 2 =0.762), the smallest is the band selected by the correlation coefficient method, and the root mean square error between the OIFC predicted value and the measured value is the smallest, which can be seen compared to the band obtained by using the best index method or correlation coefficient method alone. The model prediction effect established by the extraction band of the invention has high precision.

Claims (1)

  1. 一种基于最佳指数-相关系数法的高光谱波段提取方法，包括以下步骤： A method for extracting a hyperspectral band based on an optimal exponential-correlation coefficient method, comprising the following steps:
    步骤A：对原始高光谱数据进行分类、整理预处理，具体步骤如下： Step A: Sort and sort the original hyperspectral data. The specific steps are as follows:
    首先对获得的单个原始高光谱数据中无用信息进行剔除，然后将所有单个数据中反射率 数据综合到同一个文件中，作为以下处理的基本数据库； First, the unwanted information in the single raw hyperspectral data obtained is removed, and then the reflectivity of all the individual data is The data is integrated into the same file as the basic database for the following processing;
    步骤B：在原始高光谱数据分类、整理后进行最佳指数计算处理，具体方法如下： Step B: Perform the optimal index calculation process after classifying and sorting the original hyperspectral data. The specific method is as follows:
    最优组合波段需要同时选取3个相关波段进行计算，采用最佳指数OIF作为优化组合的 评价指标，其计算公式为：
    Figure PCTCN2014000676-appb-100001
    其中：Si为选取3个波段中任意第i波段 的标准差，Rij为选取3个波段中任意i、j两波段的相关系数，r为任意i、j两波段的组合数；     The optimal combination band needs to select three relevant bands for calculation at the same time, and the best index OIF is used as the evaluation index of the optimized combination. The calculation formula is:
    Figure PCTCN2014000676-appb-100001
    Where: S i is the standard deviation of any i-band of the three bands, R ij is the correlation coefficient of any i and j bands of the three bands, and r is the combination number of any i and j bands;
    步骤C：3波段相关系数同时选取最大值的计算，计算方法如下： Step C: The 3-band correlation coefficient is simultaneously selected to calculate the maximum value, and the calculation method is as follows:
    通过最佳指数计算后获得的所有3波段组合，计算每种波段与目标数据的相关性，然后 利用公式Rstd＝R-rstd计算最大相关系数，其中：Rstd为3波段的相关系数评价指标，Rstd越大 表示3个波段的相关系数值均同时最大；R为3个波段各自与目标数据相关系的总和，rstd为 3个波段各自与目标数据相关系数的标准差； Calculate the correlation between each band and the target data by using all the 3-band combinations obtained after the best index calculation, and then calculate the maximum correlation coefficient by using the formula R std =Rr std , where: R std is the correlation coefficient evaluation index of the 3 bands. The larger the R std is, the larger the correlation coefficient values of the three bands are at the same time; R is the sum of the three bands and the target data, and r std is the standard deviation of the correlation coefficient between each of the three bands and the target data;
    步骤D：采用商权指数计算方法建立数据的最佳指数与相关性系数综合指标评价体系， 并在此基础上进行高光谱波段的选择，具体如下： Step D: Using the calculation method of the commercial power index to establish a comprehensive index evaluation system for the best index and correlation coefficient of the data, And on this basis, the selection of the hyperspectral band is as follows:
    商权指数的计算方法是将最佳指数计算结果与相关系数计算结果作为计算的两个输入指 标值，测评系统中目标数据对象集合为F＝(OIF，C)，基于最佳指数与相关性系数综合指标评 价体系计算参数集合c＝(c1，c2，…，cm)，得到原始评价信息矩阵R＝(rij)m×2，其中：rij为第j评 价指标下第i项目的优化指标评价值； The calculation method of the commercial power index is to use the best index calculation result and the correlation coefficient calculation result as the two input index values. The target data object set in the evaluation system is F=(OIF, C), based on the best index and correlation. The coefficient comprehensive index evaluation system calculates the parameter set c=(c 1 ,c 2 ,...,c m ), and obtains the original evaluation information matrix R=(r ij ) m×2 , where: r ij is the ith under the jth evaluation index. The evaluation value of the project's optimization index;
    由于系统中各因素的量纲不一定相同，数值有时相差悬殊，从而造成数据的比较的难度 增大，需要对原始数据优化指标处理以及归一化处理，方法为正指标数据优化处理公式为
    Figure PCTCN2014000676-appb-100002
    负指标数据优化处理公式为
    Figure PCTCN2014000676-appb-100003
    其中：rij′为第j指 标下第i项目的优化指标评价值，max(rij)与min(rij)为第j指标里的所有i项目中优化指标评 价的最大值和最小值；     Since the dimensions of various factors in the system are not necessarily the same, the values sometimes differ greatly, which makes the comparison of the data more difficult. The original data optimization index processing and normalization processing are required. The method is that the positive index data optimization processing formula is
    Figure PCTCN2014000676-appb-100002
     Negative indicator data optimization processing formula is
    Figure PCTCN2014000676-appb-100003
     Where: r ij ′ is the optimization index evaluation value of the i-th item under the j-th index, and max(r ij ) and min(r ij ) are the maximum and minimum values of the optimization index evaluation of all i items in the j-th index;
    第j指标下第i项目的优化指标之间的比重值Pij计算公式为
    Figure PCTCN2014000676-appb-100004
    由商权法计 算的第j指标的商值
    Figure PCTCN2014000676-appb-100005
    其中：k＝1/ln2，当Pij＝0时，PijlnPij＝0，第j 指标的商权为
    Figure PCTCN2014000676-appb-100006
         The specific gravity value P ij between the optimization indicators of the i-th item under the j-th index is calculated as
    Figure PCTCN2014000676-appb-100004
    The quotient of the jth indicator calculated by the commercial law
    Figure PCTCN2014000676-appb-100005
    Where: k=1/ln2, when P ij =0, P ij lnP ij =0, the trade term of the j-th indicator is
    Figure PCTCN2014000676-appb-100006
    综合权重计算公式为
    Figure PCTCN2014000676-appb-100007
    其中：λj为主观权重，w′j为综合权重；最佳指数 -相关系数法提取的总指标将可行方案集映射到“距离“空间，并将LP(w′j，j)作为综合评价 的总指标，其中
    Figure PCTCN2014000676-appb-100008
    取P＝1，此时
    Figure PCTCN2014000676-appb-100009
    L1称之为 海明距离，只注重偏差的总和，归一化处理公式为
    Figure PCTCN2014000676-appb-100010
    上述公式中i＝1，2，...，m；j＝1，2， L1越大则其综合评价值越高，这样就可以按照L1从小到大的顺序排序，从而获得提取波段结 果，L1即为基于最佳指数-相关系数法的高光谱波段提取方法的选择指标。     The comprehensive weight calculation formula is
    Figure PCTCN2014000676-appb-100007
    Where: λ j is the subjective weight, w′ j is the comprehensive weight; the best index extracted by the best index-correlation coefficient method maps the feasible solution set to the “distance” space, and L P (w′ j , j) as the synthesis The overall indicator of the evaluation, of which
    Figure PCTCN2014000676-appb-100008
     Take P=1, at this time
    Figure PCTCN2014000676-appb-100009
     L 1 is called the Hamming distance, and only the sum of the deviations is emphasized. The normalized processing formula is
    Figure PCTCN2014000676-appb-100010
    In the above formula, i=1, 2, ..., m; j=1, 2, the larger the L 1 is, the higher the comprehensive evaluation value is, so that the order of L 1 can be sorted from small to large, thereby obtaining the extraction band. As a result, L 1 is a selection index of the hyperspectral band extraction method based on the optimal exponential-correlation coefficient method.
PCT/CN2014/000676 2014-07-02 2014-07-16 Hyperspectral waveband extraction method based on optimal index factor-correlation coefficient method WO2016000088A1 (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201410310957.6A CN104122210B (en) 2014-07-02 2014-07-02 Hyperspectral waveband extraction method based on optimum index factor-correlation coefficient method
CN201410310957.6 2014-07-02

Publications (1)

Publication Number Publication Date
WO2016000088A1 true WO2016000088A1 (en) 2016-01-07

Family

ID=51767707

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2014/000676 WO2016000088A1 (en) 2014-07-02 2014-07-16 Hyperspectral waveband extraction method based on optimal index factor-correlation coefficient method

Country Status (2)

Country Link
CN (1) CN104122210B (en)
WO (1) WO2016000088A1 (en)

Cited By (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109460712A (en) * 2018-10-12 2019-03-12 西北工业大学 A kind of EO-1 hyperion band selection method based on optimal neighbour reconstruct
CN109521437A (en) * 2018-12-05 2019-03-26 武汉大学 Multiplex physicotherapy laser radar wavelength selection method towards the detection of vegetation biochemical parameter
CN109657653A (en) * 2019-01-21 2019-04-19 安徽大学 A kind of wheat seed head blight recognition methods based on Imaging Hyperspectral Data
CN110309822A (en) * 2019-06-18 2019-10-08 哈尔滨工程大学 Hyperspectral image band selection method based on quantum evolution particle swarm algorithm
CN111145276A (en) * 2019-12-18 2020-05-12 河南大学 Hyperspectral image compression method based on deep learning and distributed source coding
CN111161362A (en) * 2020-01-15 2020-05-15 昆山小茶智能科技有限公司 Tea tree growth state spectral image identification method
CN111709637A (en) * 2020-06-11 2020-09-25 中国科学院西安光学精密机械研究所 Qualitative analysis method for interference degree of spectral curve
CN112163523A (en) * 2020-09-29 2021-01-01 北京环境特性研究所 Abnormal target detection method and device and computer readable medium
CN112858178A (en) * 2021-01-11 2021-05-28 武汉大学 Aviation thermal infrared hyperspectral image temperature and emissivity inversion method
CN113063739A (en) * 2021-02-25 2021-07-02 北京麦飞科技有限公司 Rice canopy nitrogen content monitoring method based on airborne hyperspectral sensor
CN113075129A (en) * 2021-02-07 2021-07-06 浙江师范大学 Hyperspectral image band selection method and system based on neighbor subspace division
CN113156452A (en) * 2021-04-07 2021-07-23 中国空间技术研究院 Satellite-borne differential absorption laser radar CO2Method for determining optimal wave band for profile detection
CN113155747A (en) * 2021-01-15 2021-07-23 西安科技大学 Method for constructing fractional order differential spectral index and monitoring wheat stripe rust thereof
CN113496218A (en) * 2021-07-27 2021-10-12 浙江省淡水水产研究所 Evaluation method and system for hyperspectral remote sensing sensitive band selection mode
CN114136445A (en) * 2021-12-03 2022-03-04 江苏天汇空间信息研究院有限公司 Method for improving airborne hyperspectral radiation correction precision of unmanned aerial vehicle
WO2023087630A1 (en) * 2021-11-22 2023-05-25 浙江大学 Method for estimating soil salinity of straw residue farmland by using remote sensing construction index
CN116297530A (en) * 2023-05-19 2023-06-23 广东正一包装股份有限公司 Barrier film surface quality detection method based on optical technology
CN116341223A (en) * 2022-12-07 2023-06-27 南京农业大学 Novel three-band spectrum index-based rice spike rot disease severity estimation method

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107421895A (en) * 2017-06-30 2017-12-01 中国水利水电科学研究院 A kind of water quality parameter retrieving concentration method and apparatus of multiband optimum organization
CN108280419B (en) * 2018-01-18 2020-06-30 中国地质科学院矿产资源研究所 Spatial feature detection method and system
CN111077093A (en) * 2020-01-10 2020-04-28 安徽理工大学 Method and device for quickly detecting coal gangue based on multispectral technology
CN111539409B (en) * 2020-04-09 2022-10-14 武汉大学 Ancient tomb question and character recognition method based on hyperspectral remote sensing technology
CN111507312A (en) * 2020-05-26 2020-08-07 长江勘测规划设计研究有限责任公司 Soil moisture extraction method based on hyperspectral data
CN112525346B (en) * 2020-12-03 2022-10-11 安徽理工大学 Method and system for selecting optimal waveband of spectral image based on improved OIF (optical image interface) and storage medium
CN112378866A (en) * 2021-01-18 2021-02-19 航天宏图信息技术股份有限公司 Water quality parameter inversion model training method, water quality monitoring method and device
CN113177488A (en) * 2021-05-07 2021-07-27 华北理工大学 Hyperspectral image waveband selection method based on optimal index waveband combination

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101131734A (en) * 2007-06-25 2008-02-27 北京航空航天大学 Automatic waveband selection method for high optical spectrum remote sensing image
CN101859383A (en) * 2010-06-08 2010-10-13 河海大学 Hyperspectral remote sensing image band selection method based on time sequence important point analysis
CN103150577A (en) * 2013-03-26 2013-06-12 河海大学 Method for selecting high spectrum remotely sensed image self-adaptive wave band based on particle swarm optimization
CN103268502A (en) * 2013-06-05 2013-08-28 中国科学院对地观测与数字地球科学中心 Hyperspectral image waveband selecting method and device
US20140029793A1 (en) * 2012-07-30 2014-01-30 Wei Chen Method of optimal out-of-band correction for multispectral remote sensing

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101131734A (en) * 2007-06-25 2008-02-27 北京航空航天大学 Automatic waveband selection method for high optical spectrum remote sensing image
CN101859383A (en) * 2010-06-08 2010-10-13 河海大学 Hyperspectral remote sensing image band selection method based on time sequence important point analysis
US20140029793A1 (en) * 2012-07-30 2014-01-30 Wei Chen Method of optimal out-of-band correction for multispectral remote sensing
CN103150577A (en) * 2013-03-26 2013-06-12 河海大学 Method for selecting high spectrum remotely sensed image self-adaptive wave band based on particle swarm optimization
CN103268502A (en) * 2013-06-05 2013-08-28 中国科学院对地观测与数字地球科学中心 Hyperspectral image waveband selecting method and device

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
DONG, GUANGJUN ET AL.: "Image Fusion for Hyperspectral Date of PHI and High-Resolution Aerial Image", JOURNAL OF INFRARED AND MILLIMETER WAVES, vol. 25, no. 2, 30 April 2006 (2006-04-30), pages 124 - 126, XP055249933 *

Cited By (30)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109460712A (en) * 2018-10-12 2019-03-12 西北工业大学 A kind of EO-1 hyperion band selection method based on optimal neighbour reconstruct
CN109460712B (en) * 2018-10-12 2024-01-12 西北工业大学 Hyperspectral wave band selection method based on optimal neighbor reconstruction
CN109521437A (en) * 2018-12-05 2019-03-26 武汉大学 Multiplex physicotherapy laser radar wavelength selection method towards the detection of vegetation biochemical parameter
CN109521437B (en) * 2018-12-05 2023-07-21 武汉大学 Multispectral laser radar wavelength selection method for vegetation biochemical parameter detection
CN109657653A (en) * 2019-01-21 2019-04-19 安徽大学 A kind of wheat seed head blight recognition methods based on Imaging Hyperspectral Data
CN109657653B (en) * 2019-01-21 2022-10-04 安徽大学 Wheat grain gibberellic disease identification method based on imaging hyperspectral data
CN110309822A (en) * 2019-06-18 2019-10-08 哈尔滨工程大学 Hyperspectral image band selection method based on quantum evolution particle swarm algorithm
CN110309822B (en) * 2019-06-18 2022-09-27 哈尔滨工程大学 Hyperspectral image band selection method based on quantum evolution particle swarm algorithm
CN111145276A (en) * 2019-12-18 2020-05-12 河南大学 Hyperspectral image compression method based on deep learning and distributed source coding
CN111145276B (en) * 2019-12-18 2023-03-24 河南大学 Hyperspectral image compression method based on deep learning and distributed source coding
CN111161362A (en) * 2020-01-15 2020-05-15 昆山小茶智能科技有限公司 Tea tree growth state spectral image identification method
CN111161362B (en) * 2020-01-15 2024-02-09 苏州伙伴实验设备有限公司 Spectral image identification method for growth state of tea tree
CN111709637A (en) * 2020-06-11 2020-09-25 中国科学院西安光学精密机械研究所 Qualitative analysis method for interference degree of spectral curve
CN111709637B (en) * 2020-06-11 2023-08-22 中国科学院西安光学精密机械研究所 Qualitative analysis method for interference degree of spectrum curve
CN112163523A (en) * 2020-09-29 2021-01-01 北京环境特性研究所 Abnormal target detection method and device and computer readable medium
CN112858178A (en) * 2021-01-11 2021-05-28 武汉大学 Aviation thermal infrared hyperspectral image temperature and emissivity inversion method
CN113155747A (en) * 2021-01-15 2021-07-23 西安科技大学 Method for constructing fractional order differential spectral index and monitoring wheat stripe rust thereof
CN113155747B (en) * 2021-01-15 2022-12-06 西安科技大学 Method for constructing fractional order differential spectral index and monitoring wheat stripe rust thereof
CN113075129A (en) * 2021-02-07 2021-07-06 浙江师范大学 Hyperspectral image band selection method and system based on neighbor subspace division
CN113075129B (en) * 2021-02-07 2023-03-31 浙江师范大学 Hyperspectral image band selection method and system based on neighbor subspace division
CN113063739A (en) * 2021-02-25 2021-07-02 北京麦飞科技有限公司 Rice canopy nitrogen content monitoring method based on airborne hyperspectral sensor
CN113156452A (en) * 2021-04-07 2021-07-23 中国空间技术研究院 Satellite-borne differential absorption laser radar CO2Method for determining optimal wave band for profile detection
CN113496218A (en) * 2021-07-27 2021-10-12 浙江省淡水水产研究所 Evaluation method and system for hyperspectral remote sensing sensitive band selection mode
CN113496218B (en) * 2021-07-27 2023-10-24 浙江省淡水水产研究所 Evaluation method and system for hyperspectral remote sensing sensitive wave band selection mode
WO2023087630A1 (en) * 2021-11-22 2023-05-25 浙江大学 Method for estimating soil salinity of straw residue farmland by using remote sensing construction index
CN114136445A (en) * 2021-12-03 2022-03-04 江苏天汇空间信息研究院有限公司 Method for improving airborne hyperspectral radiation correction precision of unmanned aerial vehicle
CN116341223A (en) * 2022-12-07 2023-06-27 南京农业大学 Novel three-band spectrum index-based rice spike rot disease severity estimation method
CN116341223B (en) * 2022-12-07 2024-01-16 南京农业大学 Novel three-band spectrum index-based rice spike rot disease severity estimation method
CN116297530B (en) * 2023-05-19 2023-07-25 广东正一包装股份有限公司 Barrier film surface quality detection method based on optical technology
CN116297530A (en) * 2023-05-19 2023-06-23 广东正一包装股份有限公司 Barrier film surface quality detection method based on optical technology

Also Published As

Publication number Publication date
CN104122210A (en) 2014-10-29
CN104122210B (en) 2017-01-25

Similar Documents

Publication Publication Date Title
WO2016000088A1 (en) Hyperspectral waveband extraction method based on optimal index factor-correlation coefficient method
Shiklomanov et al. Quantifying the influences of spectral resolution on uncertainty in leaf trait estimates through a Bayesian approach to RTM inversion
CN110132862B (en) Construction method and application of disease index special for wheat scab detection
CN105630743B (en) A kind of system of selection of spectrum wave number
US10481081B2 (en) Texture analysis of a coated surface using pivot-normalization
US20210404952A1 (en) Method for selection of calibration set and validation set based on spectral similarity and modeling
CN106596450B (en) Incremental method based on infrared spectrum analysis material component content
CN107679569A (en) Raman spectrum substance automatic identifying method based on adaptive hypergraph algorithm
CN108802002B (en) Silkworm egg Raman spectrum model construction method for rapidly identifying and removing diapause without damage
CN106951720A (en) Soil nutrient model transfer method based on canonical correlation analysis and linear interpolation
CN109870419B (en) Method for predicting nitrogen, phosphorus and potassium content of black soil by adopting aviation hyperspectral data
Liu et al. Identification of heat damage in imported soybeans based on hyperspectral imaging technology
CN106770005A (en) A kind of division methods of the calibration set for near-infrared spectrum analysis and checking collection
WO2023207453A1 (en) Traditional chinese medicine ingredient analysis method and system based on spectral clustering
CN113570538A (en) Blade RGB image skewed distribution parameter information acquisition and analysis method
CN106323466A (en) Leaf nitrogen content high spectral evaluation method for continuous wavelet transformation analysis
CN109991174A (en) A kind of Model transfer method between EO-1 hyperion camera and conventional spectrograph
CN109932335A (en) It is a kind of for the method for natural rubber assay in plant and measurement use LED near infrared spectrometer
He et al. Estimation of leaf chlorophyll content in winter wheat using variable importance for projection (VIP) with hyperspectral data
Poona et al. Reducing hyperspectral data dimensionality using random forest based wrappers
CN105651727B (en) The method that near-infrared spectrum analysis based on JADE and ELM differentiates apple shelf life
CN104181125A (en) Method for rapidly determining Kol-bach value of beer malt
Priedītis et al. Assessment of hyperspectral data analysis methods to classify tree species.
Zhang et al. Early-Stage Diagnosis of Panax Notoginseng Plant Blight Disease by Multispectral Imaging
CN108777019B (en) Near-infrared spectrum model transfer strategy optimization method and device

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 14896432

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 14896432

Country of ref document: EP

Kind code of ref document: A1