基于python开发的高光谱数据预处理方法+源码+文档+代码解析（高分优秀项目）

# 算法列表 * 标准正态变换 MSC * 多元散射校正 SNV * Savitzky-Golay平滑滤波函数 SG * 滑动平均滤波 move_avg * 一阶差分 D1 * 二阶差分 D2 * 小波变换 wave * 均值中心化 mean_centralization * 标准化 standardlize * 最大最小归一化 max_min_normalization * 矢量归一化 vector_normalization 感谢 @[panxy0826](https://blog.csdn.net/Joseph__Lagrange/article/details/95302398)的开源，本Tool 部分搬运修改于该作者 # 快速使用 ## 1. 导入数据 ```python # 导入 pandas 读取数据 import pandas as pd import numpy as np # 读取数据 data = pd.read_csv("./data/peach_spectra_brix.csv") # m * n print("数据矩阵 data.shape：",data.shape) # 50个样本， 600个 波段 第一列是 桃子糖度值 需要分离开 X = data.values[:,1:] ``` ## 2. 数据绘制 ```python from pretreatment import Pretreatment as pre p = pre() # 该方法为快速示例 而编写 # 测试用例 图片名 波段起始点 波段间距 p.PlotSpectrum(X, '演示', 0, 5).show() ## 为保证数据可以正确绘制 请将 矩阵转化为 numpy.ndarray 格式 type(X) ``` <img src="assets/image-20211018212653707.png" alt="image-20211018212653707" style="zoom: 33%;" /> ## 3. 数据预处理 ### 标准正态变换 MSC ```python msc = p.msc(X) p.PlotSpectrum(msc, 'msc', 0, 5).show() ``` <img src="assets/image-20211018212754142.png" alt="image-20211018212754142" style="zoom: 33%;" /> ### 多元散射校正 SNV ```python snv = p.snv(X) p.PlotSpectrum(snv, 'snv', 0, 5).show() ``` <img src="assets/image-20211018212821032.png" alt="image-20211018212821032" style="zoom: 33%;" /> ### Savitzky-Golay平滑滤波函数 SG ```python # 此处参数为参考 具体 后续 详述 sg = p.SG(X, 4*5+1,2*3,2) p.PlotSpectrum(sg, 'sg', 0, 5).show ``` <img src="assets/image-20211018212847272.png" alt="image-20211018212847272" style="zoom: 33%;" /> ### 滑动平均滤波 move_avg ```python move_avg = p.move_avg(X) p.PlotSpectrum(move_avg, 'move_avg', 0, 5).show ``` <img src="assets/image-20211018212913863.png" alt="image-20211018212913863" style="zoom:33%;" /> ### 一阶差分 D1 ```python D1 = p.D1(X) p.PlotSpectrum(D1, 'D1', 0, 5).show ``` <img src="assets/image-20211018212939808.png" alt="image-20211018212939808" style="zoom:33%;" /> ### 二阶差分 D2 ```python D2 = p.D2(X) p.PlotSpectrum(D2, 'D2', 0, 5).show ``` <img src="assets/image-20211018212958282.png" alt="image-20211018212958282" style="zoom:33%;" /> ### 小波变换 wave ```python wave = p.wave(X) p.PlotSpectrum(wave, 'wave', 0, 5).show ``` <img src="assets/image-20211018213018343.png" alt="image-20211018213018343" style="zoom:33%;" /> ### 均值中心化 mean_centralization ```python mean_centralization = p.mean_centralization(X) p.PlotSpectrum(mean_centralization, 'mean_centralization', 0, 5).show ``` <img src="assets/image-20211018213038106.png" alt="image-20211018213038106" style="zoom:33%;" /> ### 标准化 standardlize ```python standardlize = p.standardlize(X) p.PlotSpectrum(standardlize, 'standardlize', 0, 5).show ``` <img src="assets/image-20211018213059006.png" alt="image-20211018213059006" style="zoom:33%;" /> ### 最大最小归一化 max_min_normalization ```python max_min_normalization = p.max_min_normalization(X) p.PlotSpectrum(max_min_normalization, 'max_min_normalization', 0, 5).show ``` <img src="assets/image-20211018213117138.png" alt="image-20211018213117138" style="zoom:33%;" /> ### 矢量归一化 vector_normalization ```python vector_normalization = p.vector_normalization(X) p.PlotSpectrum(vector_normalization, 'vector_normalization', 0, 5).show ``` <img src="assets/image-20211018213156529.png" alt="image-20211018213156529" style="zoom:33%;" /> # 注意事项 示例数据来源：[nirpyresearch.com](https://nirpyresearch.com/) 源代码参考：https://gitee.com/aBugsLife/spectral-pretreatment-method