基于光饱和影响校正的作物叶绿素分布光谱成像检测

Spectral Imaging Detection of Crop Chlorophyll Distribution Based on Optical Saturation Effect Correction

叶绿素含量是作物光合能力与营养评价的重要指标,因此快速检测作物叶绿素含量与分布可为作物营养动态分析与长势评估提供支持。基于RGB(Red,Green,Blue)和NIR(Near Infrared)多光谱图像的获取,开展玉米作物营养状态分布光谱学成像检测。构建了多光谱图像采集平台获取RGB和NIR图像,研究了基于光饱和校正算法的RGB图像的光饱和校正与NIR图像去噪方法,通过图像的匹配分割,冠层的提取校正,建立了基于冠层图像的作物SPAD值检测模型与分布成图。采集15株玉米植株RGB-NIR图像,并同步获取不同植株,不同位置共68个叶绿素含量指标SPAD值。首先对RGB图像进行光饱和校正,再对NIR图像进行滤波和图像增强,其次对RGB和NIR图像进行了SURF(speeded-up robust features)和RANSAC(random sample consensus)图像匹配,利用RGB图像的颜色特征,采用ExG(Extra Green)和OTSU算法生成分割掩模,对RGB图像和NIR图像进行分割提取,提取图像的R,G,B和NIR分量,利用4阶灰度板进行反射率校正,然后计算作物图像中像素级P SPAD值,并建立图像P SPAD值与叶绿素仪SPAD值的拟合模型,最后绘制作物SPAD分布图。通过HSI(Hue,Saturation,Intensity)彩色模型中的I分量直方图对比去饱和前后光分布范围,以作物SPAD值分布图验证光饱和校正算法对作物叶绿素含量分布检测提升的效果。RGB图像光饱和校正前I分量集中在[140~180]之间,光饱和校正后的RGB图像I分量集中在[85~130]之间,校正了相机成像时产生模糊和RGB图像饱和。对分割后的RGB图像和NIR图像提取R,G,B,NIR分量进行4阶灰度板校正,相关系数分别为0.829,0.828,0.745和0.994,进而生成R,G,B和NIR四波段的反射率伪彩色图像,反射率R NIR>R G>R R>R B。体现了作物的在蓝光和红光区域吸收光,在绿光区域和近红外区域反射光的光谱特性。校正前后的R和NIR分量反射率计算图像P SPAD值拟合叶绿素含量指标SPAD值的模型结果显示,校正前R 2为0.3326,校正后R 2为0.6193,绘制作物的SPAD特征分布图,可为作物的营养动态快速分析与分布检测提供技术支持。

Chlorophyll content is an important indicator of photosynthesis capability and nutrient content in crops.Measuring chlorophyll content of crops is considered to be the most effective method for detecting crop growth status.In this paper,a multi-spectral camera was built to capture images of maize plant in RGB(Red,Green,Blue)and NIR(Near Infrared)band,which was the fundamentalfor the distribution analysisof nutritional status with rapidand non-destructive method.The RGB and NIR images were acquired by image acquisition platform.Light saturation correction of RGB images based on light saturation correction algorithm was studied.The crop SPAD distribution map was established following the image matching and segmentation,image information extraction and correction.In the experiment,images of 15 maize plants were acquired by RGB-NIR camera,and 68 SPAD values were measured at different positions of the plants.Firstly,the RGB images were corrected by light saturation correction algorithm.At the same time,the NIR images were filtered and enhanced.Secondly,the RGB and NIR images were matched with SURF(Speeded-Up Robust Features)and RANSAC(Random Sample Consensus)algorithm.Used RGB images color feature,the mask was generated with ExG(Extra Green)and OTSU algorithm,and applied in the RGB-NIR images segmentation.The R,G,B and NIR components of the image were extracted and the reflectance were corrected by the fourth-order gray-scale plate.The Intensity(I)component histogram and crop SPAD value distribution were compared to verify the effect of the optical saturation correction algorithm.The results show that I component of RGB image concentrates between[140~180]before optical saturation correction,and between[85~130]after optical saturation correction because of correction in image blurring and RGB image saturation.The correlation coefficients between the image components(R,G,B,NIR)and gray scale reflectance were 0.829,0.828,0.745 and 0.994,respectively.Then the pseudo-color images of R,G,B and NIR bands were generated.The reflectance results(R NIR>R G>R R>R B)indicated the spectral characteristics of crops which absorbed light in blue and red regions,reflected light in green and near infrared regions.Thirdly,the SPAD values at pixel level were calculated.The accuracy of chlorophyll content fitted in SPAD formula with R and NIR component reflectance before and after correction were compared.The R^2 was 0.3326 before correction and the R^2 after correction was 0.6193.Finally,the SPAD distribution map of crops was drawn,which could provide technical support for analyzing and monitoring the nutritional distribution of crops.