Study On The Computer Vision-based Nutrition Non-contact Monitoring System For Cucumber Seedlings In Greenhouse

When in nutrition stress, or there is little nutrition in soil, chlorophyll content of plants' leaves will vary, photosynthesis efficiency of them will lower, which will significantly influence the output. Traditional nutrition measurement methods are series of chemic analysis methods, which are destructive. In recent years, more and more scholars dedicate on the study of automatic and continuous monitoring plants growth and nutrition status in order to improve the precision of nutrition management. In this paper, computer vision technology, as a non-contact measurement method, was applied in plantsproduction to realize the plant-oriented management of nutrition liquor. Based on above goals, the following work was finished in this paper:1. A computer-vision cucumber nutrition monitoring system was designed. This system is consisted of plant growth system, image acquisition and analysis system, which can perform non-destructive measurement of plants.2. With image processing and statistics, the color features of cucumber leaves' images were extracted. By correlation analysis and eliminating abnormal points, linear regression equations were established between some color features and nitrogen content of cucumber leaves by destructive measurement, which basically realized the prediction of nitrogen content.3. With back propagation neural network, prediction model of nitrogen content of leaves was established. In this network, the means of R, G, B were input nodes and nitrogen content was the only output node. The result of validation showed, the correlation coefficient between prediction nitrogen content by this network and nitrogen content by destructive measurement was up to 0.871, which was enough for prediction.4. The non-contact plant nutrition monitoring and image analysis software was programmed using LabWindows/CVI and the IMAQ Vision image processing library which had been developed by National Instrument company. This software consists of these modules: image acquisition, image enhancement, system calibration, color features extraction, nitrogen content prediction and data management, etc.This research has great significance in computer vision application in precise management of greenhouse plants production, non-destructive measurement and prediction of nitrogen content of plants, and ultimately realizing the automatic and intelligent control of industrialized agriculture.