| To judge crop nitrogen nutritional status, accurately, rapidly and timely is not only thebasis of rational application of nitrogen fertilizer but also the one of hot spot topics which theplant nutrition worker studies continuously. The maize has a great demand for nitrogen duringthe process of growth and development. Nitrogen deficiency will directly affect the growthand yield of maize while too much nitrogen will cause the loss of fertilizer utilization ratio,reduce crop quality, lead to groundwater pollution, and increase greenhouse gas emissions.Therefore, grasping the nutritional status of the plant accurately could let fertilizer inputsmore reasonable and prevent fertilizer waste and environment pollution. Hence, to establish asimple, rapid and accurate nutrition diagnosis method has become a top priority.In this article, maize leaf images were taken as the research object. The maize leafnitrogen content diagnosis method was studied based on plant physiology, colorimetricknowledge, Android mobile phone and image processing technology. The main researchcontents in this article are:(1) The maize breeding scheme and gradient fertilization scheme was made.Thecultivated maize samples of different nitrogen content were taken as a random sample. TheSPAD chlorophyll meter was used to test maize leaf SPAD value and image information ofliving maize leaf were collected by the Android mobile phone. The maize leaf nitrogencontent was measured by kjeldahl determination method.(2) The maize leaf image preprocessing method was studied. Including the leaf imagefiltering、 image color correction、and image segmentation. The linear filtering method wasused to image filter to eliminate the noise in the image; the new color correction methodbased on the calibration color piece was put forward to correct color information of imagesafter image filtering, which achieved a nice result. Images were properly cutting after thecolor correction image. Finally, combining with the color characteristics of the background,R/G threshold was used to image segmentation processing after image cutting.(3) The maize leaf color characteristics of image information were extracted. The RGBtrue color images of maize leaves were studied to extract six kind of color characteristic values from maize leaf area，such as R mean value、G mean value、B mean value、redstandardized value R/(R+G+B)、 green standardized value G/(R+G+B) and bluestandardized value B/(R+G+B).The relationship between six kinds of color characteristicinformation and corn leaf SPAD values or six kind of color characteristic information and thenitrogen content of maize leaf was analyzed. The correlation between corn leaf nitrogencontent and green light standardization value was the best. Finally, green standardized valueG/(R+G+B) was chosen as a color characterization value for nitrogen content of maize leaf.(4) Color characteristic values of40maize leaves with different nitrogen content wereextracted. The linear regression mathematical method was used to establish the mathematicalmodel between green standardized value and nitrogen content of maize leaf. Thedetermination coefficient of the mathematical model is0.7904. The mathematical model wasverified through experiment. The determination coefficient between calculation corn leafnitrogen content and actual corn leaf nitrogen content is0.7471, and the absolute error is-0.32to0.26%.(5) The software functions was analyzed according to requirement of corn leaf nitrogencontent detection method based on the Android mobile phone platform and image processingtechnology. According to its function, software interface and algorithm was designed. Thesoftware interface was designed by writing an XML file, and software algorithm was writtenwith the Java language. The software was tested in a variety of test environment with differenthardware and different system version. |
PDF Full Text Request