基于机器视觉技术的尖椒冠层SPAD值测定仪的开发

为了快速准确地测量尖椒的冠层植被指数值,该文在理论分析的基础之上,开发了一套基于机器视觉技术的用于测量尖椒冠层SPAD(soil and plant analyzer development,SPAD)值的仪器。该仪器利用步进电机控制载有绿波段和近红外波段两片滤光片的转台,实现滤光片在CCD相机镜头前的切换,从而得到冠层的绿色波段图像和近红外波段图像。制作了近红外反射率在尖椒冠层和土壤背景之间的参照板,并以其为阈值结合冠层的近红外图像来分离尖椒冠层和土壤背景。通过建立图像灰度值和反射率之间的动态模型,计算作物冠层在相应波段的反射率,得到冠层植被指数,将作物冠层的SPAD值作为对比进行分析。经过试验建立该仪器测得的冠层植被指数GNDVI(green normalized difference vegetation index,GNDVI)值和冠层实际SPAD值之间的模型,结果表明两者之间具有较高的相关性,决定系数R2=0.8768。表明该仪器适用于尖椒冠层SPAD值的测定。

Development of SPAD value measure instrument for pepper canopy based on machine vision

Abstract: In order to rapidly and accurately acquire pepper canopy SPAD (soil and plant analyzer development) value, we developed an instrument based on machine vision. A steeper was used in the instrument to control a turntable which was mounted on the front of a CCD camera lens. There were a green band filter and a near infrared bands filter on the turntable. With the turntable turning, the two filters can be changed before the CCD camera lens. It allowed us to get a green band image and a near-infrared image of the crop canopy. As we know that the reflectance of crop and soil is very different. Based on it, we divided soil and canopy in the image. We created a reference plate with four different colors on it to help us perform image processing. First, we shot a green band image and a near-infrared image of a same crop canopy using the method previously described. Then we did gray processing for the near-infrared image. According to the gray value of the color on the reference plate between the soil and crop canopy on the near-infrared band, we can divide the background soil and the crop canopy in the near-infrared image. As the green band image and the near-infrared image were for the same crop canopy, we can divide the background soil and the crop canopy in the green band image easily using the result of the near-infrared image. In the same environment, the gray value was related to the reflectance. Based on this feature, we can use the reference plate to make the model of the relationship between the gray value and reflectance for each image. Then, we can determine the canopy reflectance in green band and near-infrared band by the model and the gray value of the canopy in each image. Finally, we determined GNDVI (green normalized difference vegetation index) value of the canopy and used it to determine the SPAD value of the canopy. We wrote a computer program corresponding to the entire image processing by VB (visual basic). In order to test the reliability of the instrument, we used the instrument to do the relevant experiments and analyzed the experimental results. In this study, we made a mathematical model for the relationship between the SPAD value of the pepper canopy and the GNDVI value which was measured by this instrument. We also built a validation set to verify the result. In the experiment, we collected a total of 40 pepper canopy images and used the instrument to get the images of pepper canopy and calculated the GNDVI value. Then, we used SPAD-502 chlorophyll meter to measure the leaf SPAD value which belonged to the canopy in the image. Depending on the size of the leaves, we made three to five points to measure. Then, we averaged the SPAD value of all measuring points as the SPAD value of the canopy, and randomly selected 30 samples from all the samples to make a mathematical model between the SPAD value of the pepper canopy and the GNDVI value. The test results showed there was good linear relationship between the SPAD value of the pepper canopy and the GNDVI value. The determination coefficient (R2) was 0.8768. It showed that the GNDVI value can be used to estimate the canopy SPAD value. We also used the remaining 10 samples to test the model. The result showed that the instrument had good reliability. In summary the instrument was suitable for measuring pepper canopy SPAD value.