Research On Calculation Method Of Soybean Canopy Wilting Index Based On Multispectral Image Processin

Soybean growth process needs a lot of water.Rapid detection of soybean canopy wilting characteristics under drought stress is of great significance for soybean variety selection,cultivation regulation and fine management.In view of the problems of cumbersome and time-consuming test methods when traditional chemical techniques were used to determine soybean wilting index,this paper took soybean in the northeast region as the research object,organically integrated multispectral imaging processing technology with Fourier transform and fractal dimension methods,and proposed the soybean canopy wilting index calculation method based on Fourier transform and fractal dimension,in order to achieve fast non-destructive detection of soybean wilting state and provide theoretical basis and technical support.The main research contents were as follows:(1)The soybean canopy was extracted from multispectral image.Four kinds of soybean multispectral images,including GRE,RED,REG and NIR channel,were collected by Sequoia multispectral camera.Based on the soybean multispectral reflection image pretreated by median filter and mean filter,the target canopy area of soybean multispectral image was extracted by iterative threshold method and affine transformation algorithm.The average effective segmentation rate was 0.9702,the average under-segmentation rate was 0.0264,and the average over-segmentation rate was 0.0183.The effective extraction of soybean canopy provided a reliable data basis for the calculation of wilting index.(2)The calculation model of wilting index was constructed based on Fourier Transform(FT)for the soybean canopy.Fourier transform was used to transform the time domain information of soybean multispectral image into the frequency domain,and analyzed the spectral characteristics of soybean canopy multispectral image,such as amplitude spectrum,phase spectrum and energy spectrum.When the spectral radius of each channel was 50,the energy reached more than 98%,and was concentrated in the low-frequency region of the spectral center.According to the change difference between the low-frequency DC component in the normal and drought soybean multispectral image and the total energy in the spectral radius,the calculation method of wilting index based on Fourier transform was proposed for the soybean canopy.The validity of wilting index was verified and analyzed by the average leaf inclination.The determination coefficient R~2of the four channels was above 0.8507.(3)The calculation model of wilting index was constructed based on Fractal Dimension(FD)for the soybean canopy.According to the characteristics of correlation between fractal dimension and roughness of object surface,the multispectral imaging technology and fractal dimension method were organically integrated,and the method for calculating wilting index was proposed for the soybean canopy by using the fractal dimension algorithms such as box-counting dimension,differential box-counting dimension and double blanket dimension.The validity of the method was verified by the average leaf inclination,and its determination coefficient R~2 was above 0.8486.It showed that the wilting index of the three calculation methods based on fractal dimension were applicable to four channel spectral images,among which the correlation between wilting index based on double blanket dimension and average leaf inclination was 0.8877.It extracted fractal dimension features from the surface covering the whole soybean multispectral image,which was more universal for calculating wilting index.(4)The wilting index calculation system was developed for the multispectral image of soybean canopy.According to the Py Charm and MATLAB,a wilting index calculation system was designed and implemented for the multispectral image of soybean canopy.The system mainly dealt with soybean multispectral images,integrated the calculation method of soybean wilting index based on Fourier transform and fractal dimension,and also covered the basic functions of image processing,providing a reference for building a human-computer interactive visual processing system.The system can characterize the degree of changes in canopy ecology and morphology of soybean under stress,and provided technical support for scientific regulation of plant traits in the process of soybean variety selection,cultivation and management.