CN102788796B - Nutrient diagnosis device and nutrient diagnosis method for nitrogen of crops based on multi-information integration of high spectral images and fluorescent images - Google Patents

Abstract

The invention discloses a crop nitrogen nutrition diagnosis device and method based on multi-information fusion of hyperspectral images and fluorescent images. The device consists of a lens exchange switch a (1), a lens exchange switch b (2), a computer (3), and a main control unit of a fluorometer (4), MINI lens (5), LED light source (6), hyperspectral light source control unit (7), stepper motor controller (8), visible light lens (9), symmetrically placed visible light tubes (10), Visible light source (11), stage (12), moving stage (13), image acquisition card (14), motor (15), lighting room (16), stepping motor a (17), stepping motor b (18) COMPOSITION. The invention is expected to greatly improve the prediction accuracy of crop nutrient elements, detect the actual level of nitrogen in time and earlier, and will surely improve the scientific management level of cultivation.

Description

Crop nitrogen nutrition diagnosis device and method based on multi-information fusion of hyperspectral images and fluorescent images

technical field

The invention relates to the non-destructive detection of nitrogen nutrition of agricultural products, in particular to a device and method for non-destructive detection of nitrogen nutrition content of crops by using hyperspectral fluorescence image technology.

Background technique

Nitrogen plays an important role in crop physiological metabolism and growth, and is one of the most important limiting factors affecting crop growth. During crop growth, nitrogen fertilizer requirements are high. Nitrogen deficiency will inhibit leaf differentiation, reduce the number of leaves, and affect crop growth. Nutritional quality and yield will also have adverse effects; excessive nitrogen fertilizer application will easily cause groundwater pollution and soil pollution and degradation, so rational application of nitrogen fertilizer is very important to crop yield and quality. Precise nitrogen fertilizer management is based on accurate monitoring and detection of crop nutrient elements.

For a long time, the nitrogen nutrition diagnosis of crops has been based on routine laboratory tests, mainly including morphological diagnostic methods, leaf card methods, chemical diagnostic methods, and enzymatic diagnostic methods. These destructive methods have low test accuracy and poor timeliness. It affects the growth of crops and is not conducive to popularization and application. Computer vision technology mainly uses macroscopic physical characteristics such as leaf or canopy color, texture, and shape to diagnose. Spectral diagnosis technology mainly uses changes in spectral reflectance at characteristic wavelengths to invert the nutritional status of crops, which cannot fully characterize crop nutritional deficiencies. Lack of rich characteristic information, and the sampling method of point source makes it impossible to reflect the anisotropy of the plant well, so the error of the detection result is large, which restricts the application of high-precision non-destructive detection method of crop nutrition . the

Throughout the research status of crop nutrition diagnosis at home and abroad, regardless of spectral detection technology or computer vision technology, many studies are mainly aimed at the identification and diagnosis of symptoms of lack of nutrients, even if the quantitative diagnosis of nutritional levels , and the precision is also low.

The main reason for the low recognition accuracy of computer vision technology is that at present, computer vision technology usually only extracts the external characteristics of leaves such as color (gray scale), texture and morphology in the range of 400-1000nm. Moreover, it is not convincing enough to accurately invert crop nitrogen nutrition based on specific spectral images of several bands. For example, multi-spectral image acquisition equipment can only simultaneously collect 4 band components for the four channels of R, G, B, and IR. The content is far from enough to summarize the full picture of the message sent by crops.

The main reason for the low recognition accuracy of spectral technology is that the application of spectral technology to crop nitrogen detection has achieved relatively successful research results, mainly through the indirect detection of changes in spectral reflectance characteristics caused by changes in chlorophyll and internal organic tissues. There is also an interaction between leaf moisture and nutrition, and the spectral detection adopts the method of point source sampling, which reflects the statistical average of the spectrum of the sample area within the field of view, and inverts the nitrogen level of crops through the combination of spectral reflectance of several characteristic wavelengths , therefore, cannot reflect the anisotropic characteristics inside and outside the blade.

In addition, whether it is a spectrum or a visual image, the principle is actually the quantitative presentation of the reflected light on the leaf surface in various sensing instruments (including spectroscopic instruments and imaging instruments). The lack of nitrogen nutrition in crops will first cause physiological and biochemical changes in the internal tissues of the crops, especially the decrease in chlorophyll content, the surface morphology of leaves will change, the surface color of leaves will gradually turn yellow, and the texture will also begin to change. short time course. At the initial stage of crop nitrogen deficiency, the shape, color, and brightness of crop leaves are not abnormal, but when the symptoms of crop leaf morphology appear, nitrogen deficiency has already affected the growth and development of crops. Therefore, using spectral technology and image technology, it is more difficult to detect nitrogen levels in time and accurately in the early stages of crop nitrogen deficiency.

Contents of the invention

The purpose of the present invention is to solve the above existing problems and provide a crop nitrogen nutrition diagnosis device and method based on multi-information fusion of hyperspectral images and fluorescence images.

The purpose of the present invention is achieved in the following manner: a crop nitrogen nutrition diagnosis device based on hyperspectral image and fluorescence image multi-information fusion, characterized in that it is controlled by lens exchange switch a, lens exchange switch b, computer and fluorometer Unit, MINI lens, LED light source, hyperspectral light source control unit, stepper motor controller, visible light lens, symmetrically placed visible light tube, visible light source, stage, moving stage, frame acquisition card, motor, lighting room, Composed of stepping motor a and stepping motor b, the MINI lens is connected to the lens exchange switch a through the stepping motor a, the visible light lens is connected to the lens exchange switch b through the stepping motor b, and the LED light source is connected to the computer through the main control unit of the fluorescent instrument The stepper motor controller is connected to the moving stage through the motor, the distance between the visible light lens and the stage is 50cm, and the distance between the MINI lens and the stage is 7cm.

The detection method of described detection device, comprises the steps:

1) Collect fluorescence images: after dark adaptation of the leaves to be tested for 20 minutes, place them on the stage, press the lens exchange switch a, the MINI lens moves to the top of the stage, and run the computer. At this time, the LED light source emits a weak measurement Light, measure the maximum light quantum yield, then emit a saturated pulse light every 20s, and record the fluorescence parameters at this time, measure the fluorescence induction curve, and then emit actinic light every 10s, the actinic light changes from low to The high gradually rises to start measuring the fast light response curve;

2) Hyperspectral image acquisition: press the lens exchange switch b, the visible light lens moves to the top of the stage, run the computer, and measure the hyperspectral image of the blade to be tested;

3) Use the detection results to analyze, model and predict the nitrogen nutrition of crops: first, extract the image information part from the hyperspectral image data, and then analyze and find the characteristic wavelength and the characteristic wavelength under the characteristic wavelength that can best reflect the nitrogen content of the leaf to be measured. Hyperspectral images, and then use image processing methods to extract characteristic parameters that can reflect the nitrogen content from the hyperspectral images at these characteristic wavelengths, and finally establish a model for the nitrogen content of the leaves to be measured, and analyze and count the collected fluorescence parameter data. The change curve of fluorescence parameters with the amount of nitrogen application, find out the optimal nitrogen application amount of the leaves to be tested, and judge the nitrogen content of a vegetable to be tested according to the optimal nitrogen application amount obtained by the fluorescence parameters and the model obtained by hyperspectral And whether there is a lack of nitrogen nutrition.

The beneficial effects of the present invention are:

The invention can timely characterize the fluorescent image information of the internal dynamic physiological activity of crop leaves, and mutually assist and integrate with the hyperspectral image information that can reflect the crop surface texture, brightness, and texture, making full use of the advantages of hyperspectral images and fluorescent images, Expanding the effective feature space of nitrogen detection and seeking a high-precision comprehensive nitrogen detection method is expected to greatly improve the prediction accuracy of crop nutrient elements, and the actual level of nitrogen can be detected earlier and in a timely manner, which will definitely improve the scientific management level of cultivation, and also It provides a reference for the precise management of nitrogen nutrition in other crops.

Description of drawings

In order to make the content of the present invention easier to understand clearly, the present invention will be described in further detail below according to specific embodiments in conjunction with the accompanying drawings, wherein

Fig. 1 is the structural representation of detection device of the present invention

Fig. 2 is the flowchart of detection method of the present invention

Detailed ways:

The following examples describe the present invention in further detail.

The crop nitrogen nutrition diagnosis device based on multi-information fusion of hyperspectral images and fluorescent images in the present invention consists of a lens exchange switch a1, a lens exchange switch b2, a computer 3, a fluorometer main control unit 4, a MINI lens 5, an LED light source 6, a hyperspectral Light source control unit 7, stepping motor controller 8, visible light lens 9, visible light tube 10 placed symmetrically, visible light source 11, stage 12, moving stage 13, image acquisition card 14, motor 15, lighting room 16, Composed of stepping motor a17 and stepping motor b18, the MINI lens 5 is connected with the lens exchange switch a1 through the stepping motor a17, the visible light lens 9 is connected with the lens exchange switch b2 through the stepping motor b18, and the LED light source 6 is controlled by the fluorescent instrument The unit 4 is connected to the computer 3, the stepper motor controller 8 is connected to the moving stage 13 through the motor 15, the distance between the visible light lens 9 and the stage 12 is 50 cm, and the distance between the MINI lens 5 and the stage 12 is 7 cm.

The detection method includes the following steps: 1) collecting fluorescence images: put the lettuce on the stage 12 after dark adaptation for 20 minutes, press the lens exchange switch a1, move the MINI lens 5 to the top of the stage 12, and run it on the computer Software ImagingWin.exe, select the Setting tab on the upper part of the window, set the red gain, red intensity, and NIR intensity in Absorptivity respectively according to the parameter values on the imaging probe, and set the intensity and gain of Meas. Light to make the fluorescence in the AOI area The value is between 0.1-0.2. At this time, the LED light source emits weak measurement light, click the F0 and Fm buttons below the window to measure the maximum light quantum yield Fv/Fm. Select the Kinetic tab on the upper part of the window, click the Start button on the right side of the window, the LED light source emits a saturation pulse light every 20s, and record the fluorescence parameters at this time to measure the fluorescence induction curve. Select the Light Curve tab at the top of the window and click the Start button. At this time, the LED light source emits actinic light every 10s, and the actinic light gradually increases from low to high to start measuring the fast light response curve.

2) Hyperspectral image acquisition: press the lens exchange switch b2, the visible light lens 9 moves to the top of the stage 12, and run the computer software Spectral Image System Demo visible light software. Set the corresponding parameters and perform whiteboard calibration. Click move to measure the hyperspectral image of lettuce.

3) Use the detection results to analyze, model and predict the nitrogen nutrition of crops: the hyperspectral image acquisition is to take pictures of the samples on the stage through the camera, and transmit them to the computer through the image acquisition card. The samples on the computer are taken and sent to the computer through the main control unit. First, the image information part is extracted from the hyperspectral image data; then, various algorithms such as principal component analysis, wavelet analysis, and uneven second-order difference are used for analysis to find the characteristic wavelength and characteristic wavelength that best reflect the nitrogen content of lettuce Then, from the hyperspectral images at these characteristic wavelengths, image processing methods such as filter denoising and texture detection are used to extract characteristic parameters that can reflect the nitrogen content. Finally, multiple stepwise regression, partial least squares regression and neural network methods were used to establish a model for predicting the nitrogen content of lettuce. Statistically analyze the collected fluorescence parameter data with Excel/SPSS11.5 software, and calculate the change curve of fluorescence parameters with nitrogen application amount, find out the optimal nitrogen application amount of lettuce, and obtain the optimal nitrogen application amount according to the fluorescence parameters And the model obtained by hyperspectral, so as to judge the nitrogen content and whether there is a lack of nitrogen nutrition.

Claims (2)

1. the crop Nitrogen nutritional status device based on high spectrum image and fluoroscopic image Multi-information acquisition, it is characterized in that exchanging switch a by camera lens, switch b exchanged by camera lens, computing machine, luminoscope main control unit, MINI camera lens, LED light source, EO-1 hyperion light source control unit, controllor for step-by-step motor, visible light lens, the symmetrical visible fluorescent tube placed, visible light source, objective table, move thing platform, image pick-up card, motor, daylighting room, stepper motor a, stepper motor b forms, wherein MINI camera lens is exchanged switch a with camera lens be connected by stepper motor a, visible light lens is exchanged switch b by stepper motor b with camera lens and is connected, LED light source is connected with computing machine by luminoscope main control unit, controllor for step-by-step motor by motor with move thing platform and be connected, visible light lens and objective table distance are 50cm, MINI camera lens and objective table distance are 7cm.

2. the detection method of pick-up unit according to claim 1, is characterized in that comprising the steps:

1) fluoroscopic image is gathered: by blade dark adatpation to be measured after 20 minutes, be placed on objective table, press camera lens and exchange switch a, MINI camera lens moves to above objective table, moving calculation machine, now LED light source sends faint measurement light, measure maximum amount suboutput, then a saturation pulse light is sent every 20s LED light source, and record fluorescence parameter now, measure fluorescence induction curves, then send an actinic light every 10s LED light source, actinic light raises gradually from low to high and starts to measure fast light response curve;

2) high spectrum image collection: press camera lens and exchange switch b, visible light lens moves to above objective table, moving calculation machine, measures blade high spectrum image to be measured;

3) utilize Analysis of test results modeling and predict the nitrogen nutrition of crop: first, image information portion is extracted from hyperspectral image data, then the high spectrum image found out under the characteristic wavelength and characteristic wavelength that can reflect LTN content to be measured is analyzed, from the high spectrum image these characteristic wavelengths, then utilize image processing method to extract the characteristic parameter that can reflect nitrogen content, finally set up the model of LTN content to be measured, the fluorescence parameter data analysis statistical collected is gone out the change curve of fluorescence parameter with amount of nitrogen, find out the best amount of nitrogen of blade to be measured, the model that the best amount of nitrogen obtained according to fluorescence parameter and EO-1 hyperion obtain, thus judge the nitrogen content of vegetables to be measured and whether lack nitrogen nutrition.