Study On Vision-based Crop Growth Diagnostic Mechanism And Method In Greenhouse

As industrialized agriculture technology progress rapidly, the automatization becomes more important to the crop management. The nondestructive diagnosis is a very hot agricultural production technology. With a view to this topic, this thesis studied the development of a vision-based crop growth diagnostic technology through the analyses on the characteristics of cucumber's leaf images and growth-point images in greenhouse environment.A planting experimentation was made in a greenhouse. The relationship between the leaves and the nutrient content was investigated by analyzing the color properties of leaves' images and the leaf nutrition content.This research studied the correlation between the color properties of cucumber leaf image, based on the RGB and HSI mode. In addition, the nitrogen content of the plant was analyzed. The G weight of the image showed the highest linear correlation (R=0.88) with the Nitrogen content of the cucumber leaf. And multi-regress analysis and PCA was made between color property and chlorophyll content based on the correlation between leaf color characters and contents. A multi-linear-regress model was found about leaf chlorophyll content and color property.An artificial light-box was designed for taking the leaves image in order to reduce the influence of image color from the sunlight. The diffuse and symmetrical light could be obtained by the light-box. The noise of the image could be reduced by further special data processing.The growth-point is an important index of conformation diagnostic for the cucumber management in the greenhouse. As an object of this paper, a gray co-occurrence matrix, as main index, was developed as an estimate criterion based on obtained alive growth-point as texture property to support the diagnosis. A clustering character center could be found by cluster analysis for typical sample, and it could estimate cucumber's status: health, sub-health, or ill-health. And the health and the ill-health were distinguished more obviously than sub-health.Based on the obtained nutrition index, time-series of leaf chlorophyll was analyzed. The error of the prediction of the chlorophyll content was about 0.085. The principle of gray-system-forecast model could be applied to design a diagnostic system model of the cucumber growth, which combines the linear regression model of leaf nutrition contents and the time-series-forecast model of chlorophyll contents and fused with the information of health-evaluation obtained by using texture property cluster analysis at leaf growth-point.