Study On Detection Method Of Tomato Leaf Nutrition Based On Hyperspectral Fusion Terahertz Imaging

At present,the planting area of facility vegetables in China has reached 4.1million hectares,ranking first in the world,and it has played a huge role in promoting China's rural and agricultural economy.However,due to the requirements of output and demand,facilities industry often apply a lot of fertilizers during production.This situation will not only cause serious waste of fertilizers and environmental non-point source pollution,but also cause the decline of crop quality and yield.Therefore,there is an urgent need to monitor the nutritional status of crops during the growth process,to achieve scientific management and control of nutrients,and to achieve high-quality and high-yield crops.The traditional chemical analysis method and the diagnosis of the nutritional status in plants relying on artificial experience are poor in timeliness,easily subjective,and cannot be judged systematically and scientifically.The non-destructive detection technology based on spectral images is fast,convenient,and non-destructive The advantages of sex have become the research hotspot of crop nutrition rapid diagnosis technology.This subject takes facility tomato as the object,and proposes a method of non-destructive detection of nitrogen,phosphorus and potassium elements of crops by using hyperspectral fusion of terahertz images and fusion of multi-source light information.This research mainly completed the following work:(1)The preparation experiment of the nondestructive testing method research was carried out.In order to accurately obtain test samples of different nitrogen,phosphorus and potassium nutrition levels,the method of soilless cultivation was used to cultivate tomato samples in proportion to gradients based on standard formulas.The nitrogen,phosphorus,and potassium contents of crops were measured using chemical analysis methods.On this basis,hyperspectral data cubes with different nitrogen,phosphorus,and potassium levels of crop leaves in the 390-1050 nm band range were collected using a hyperspectral imaging system;The terahertz imaging system acquired a terahertz tomographic image of tomato samples in the0-4.0THz range.(2)Study the characteristics of hyperspectral images and terahertz images with different levels of nitrogen,phosphorus and potassium.Based on the acquired hyperspectral and terahertz image data cubes,in order to accurately obtain the nutritional features,the area data is first segmented to remove the background,and the area spectral intensity data is averaged;then the averaged data is subjected to data preprocessing;then use KS algorithm,SPXY algorithm and RS algorithm to divide the correction set and verification set;use stability competitive adaptive weighting algorithm(SCARS),iterative reserved information variable algorithm(IRIV)for data dimensionality reduction and feature frequency selection;then in Based on the above two algorithms,the continuous projection algorithm(SPA)is used for re-screening to find the special spectral frequencies and characteristic images corresponding to their respective characteristic bands.On this basis,PLSR is used to establish the model.(3)Taking into account the possible limitations of a single detection method for non-destructive testing of crop nutrition,using the hyperspectral imaging technology and the terahertz spectral imaging technology to extract the nutritional characteristics of tomatoes,using neural network algorithm and multiple linear regression algorithm to establish tomato Fusion detection model of nitrogen,phosphorus,potassium hyperspectral and terahertz.