CN103018196A - Fast detection method for rape water demand information - Google Patents

Abstract

The invention relates to a rapid detection method for water demand information of rapeseed. The near-infrared hyperspectral imaging device used is composed of a light box, a halogen tungsten light source, a controller, a computer, a near-infrared camera, a spectrograph, a lens, a Y-branch linear and Composition of electronically controlled displacement stage. Use the near-infrared hyperspectral imaging device to collect the hyperspectral data cube of rape leaves; compare the difference between the target and the background in the hyperspectral images at different wavelengths, and use the 970nm image with a large background difference to segment the target background; extract the optimal wavelength of the moisture content of rapeseed , use ENVI software to extract the principal component image data at 720nm, 960nm, and 1450nm; establish the grayscale, texture and reflection intensity distribution feature space of the principal component image, combine the leaf moisture content data, and based on the partial least squares method, establish the moisture content prediction of rapeseed The model, referring to the standardized control group samples and real-time environmental temperature and humidity data, gives the water demand and irrigation reference information of rapeseed, and realizes the rapid and non-destructive detection of rapeseed water demand information.

Description

A Rapid Detection Method of Rapeseed Water Requirement Information

technical field

The invention relates to a non-destructive detection method for water demand information of rapeseed, especially refers to non-destructive detection of water stress state of rapeseed based on near-infrared hyperspectral imaging technology, combined with real-time environmental information detection, evaluation of water demand information of rapeseed and irrigation decision-making management Methods.

Background technique

Rapeseed is one of the most important oil crops in China, with an annual planting area of more than 100 million mu and an annual output of nearly 10 million tons of rapeseed, accounting for 40%-45% of the total output of oil crops. Rapeseed is a crop that needs more water. Water stress causes physiological obstacles to the growth and development of rapeseed, which inhibits plant protein synthesis; it also has a great impact on the absorption and utilization of boron. Rapeseed is a boron-sensitive crop, and boron deficiency often leads to rape " Flowers but not fruit", which will reduce the yield of rapeseed and affect the quality of rapeseed. Drought and water shortage are common in northern my country and hilly rapeseed areas; it is of great significance to quickly diagnose the water shortage situation of plants, scientifically and accurately guide irrigation, effectively use limited water resources, and ensure high-quality and high-yield rapeseed.

At present, the water demand and water stress of crops are mainly obtained indirectly by measuring the stomatal conductance of plant leaves, leaf water potential, canopy temperature, transpiration rate, and changes in plant stem diameter, or by using the experience of producers and routine dry conditions in the laboratory. Wet weight measurement is the main method. These traditional testing methods will damage crops, affect crop growth, and consume a lot of manpower and material resources, and have poor timeliness. Crop water stress detection technology based on reflection spectrum, computer vision and infrared temperature detection has the advantages of fast, convenient and non-destructive. At present, there are some related research patents in China based on spectrum, computer vision and infrared temperature detection. Publication, application number 200510088935.0 patent application for invention, discloses a portable non-destructive detection method and measuring instrument for plant nitrogen and water content, by detecting the spectral reflection intensity information of plant leaves at four characteristic wavelengths to carry out plant nutrition Diagnosis, using the inversion of the vegetation index of four wavelengths to obtain plant nitrogen and water content information; the invention patent application with the application number 200710178192.5 discloses an online crop canopy temperature difference irrigation decision-making monitoring system, through a group of The monitoring devices such as the infrared canopy temperature sensor installed inside the high-speed platform and the ambient temperature sensor installed on the support pole can realize the monitoring of the canopy temperature of the crops in the plot. Non-destructive diagnostic methods based on spectral technology usually use point source sampling, which cannot reflect the difference in light reflection characteristics of the entire leaf or canopy area. The visual sensor has a high resolution and a large field of view. The image segmentation technology can effectively remove the influence of factors such as the background, which overcomes the shortcomings of the small test range of the spectroscopic method and the strict requirements on the test site. However, the traditional The spectral resolution of computer vision technology is very low, and what is usually obtained is a synthetic image of the target in the visible light region of 400-700nm or a multispectral image of a few bands, while the characteristic bands of water molecules are in the 960nm and 1450nm regions, which leads to the traditional It is difficult to apply the advanced visual image technology to the non-destructive detection of crop water stress. However, the water stress detection based on infrared temperature has a single feature and is highly sensitive to environmental temperature and humidity. Therefore, it can only judge the trend of crop water stress at present, and it is difficult to carry out accurate quantitative analysis.

The near-infrared hyperspectral imaging technology adopted in the present invention obtains the water stress information of rapeseed. This technology is a new technology integrating near-infrared reflection spectrum and hyperspectral imaging technology. The high resolution of 3.5nm is continuously imaged in 256 bands, forming an image data cube arranged in order of wavelength, which has the advantages of reflection spectrum and visual image technology, and can not only detect the color, texture and shape changes caused by water stress in plants It can also conduct a comprehensive analysis of the changes in the near-infrared spectral reflectance distribution information of water molecules caused by water stress on plant leaves. At present, there are no relevant patents and reports on the use of near-infrared hyperspectral imaging technology for crop water demand information detection in domestic and foreign literature.

Contents of the invention

The purpose of the invention is to provide a rapid detection method of rapeseed water demand information. Through the self-built near-infrared hyperspectral imaging acquisition device, the hyperspectral data cube of rape leaves is collected; the optimal characteristic wavelength and principal component image of rape water stress are extracted, and the near-infrared spectral distribution of water stress and the color and texture of the characteristic image are constructed , Morphological feature space; Based on multi-information fusion technology, a diagnostic model of water stress in rapeseed is established to realize rapid non-destructive detection and quantitative analysis of water stress information in rapeseed. It provides a scientific basis for rapeseed water management and irrigation decision-making.

In order to achieve the above object, the present invention adopts the near-infrared hyperspectral imaging device built by itself to collect the near-infrared hyperspectral data of the rape sample. The near-infrared hyperspectral imaging device includes the following components: light box, tungsten halogen light source, controller, computer, A hyperspectral imaging sensor, a Y branch linear lamp and an electronically controlled displacement stage, the hyperspectral imaging sensor is composed of a near-infrared camera, a spectrograph and a lens connected in sequence, and is fixed at the top center of the light box; the electronically controlled displacement The table is fixed at the geometric center of the bottom surface in the light box, located directly below the hyperspectral imaging sensor; the Y-branch linear lamp is symmetrically installed on the left and right sides of the middle of the light box; the halogen tungsten light source and the Y-branch linear lamp Connected by two glass optical fibers; the hyperspectral imaging sensor and the electronically controlled displacement stage are connected to the controller through a data line; the computer is connected to the controller through a data line, and the controller accepts the control instructions of the computer to control the The travel speed of the electronically controlled displacement stage and the scanning speed, exposure time, focal length and other parameters during information collection of the hyperspectral imaging sensor are described.

The hyperspectral imaging sensor composed of near-infrared camera, spectrograph and lens is the lowest end of which is the lens, and then connects the spectrograph and near-infrared camera upwards in turn; the near-infrared camera is indium gallium Arsenic imaging camera with a spectral range of 900-1700nm.

The spectral resolution of the hyperspectral imaging sensor described therein is 3.5nm, and the hyperspectral data of the sample is collected in a line scanning mode; In the near-infrared wavelength range of -1700nm, a scanning process can simultaneously acquire near-infrared hyperspectral images in 256 bands.

The interior of the light box described therein is electrostatically painted in black.

In order to realize the purpose of the invention, a kind of rapid detection method of rapeseed water demand information of the present invention is carried out according to the following steps:

(1) Fix the rapeseed sample on the electronically controlled displacement platform, adjust the focal length of the lens and the resolution, gain, and exposure time of the near-infrared camera to ensure the clarity of the image, and set the speed of the electronically controlled displacement platform to avoid distortion of the scanned image; The hyperspectral imaging sensor calibrates the black and white field, sets the effective reflection intensity interval of the hyperspectral image, and eliminates the dark current noise of the camera;

(2) Collect the near-infrared hyperspectral data cube of the rapeseed sample, digitally filter through the second-order Butterworth filter to remove the camera "bad point" data of the near-infrared hyperspectral data, and compare the difference between the target and the background in the hyperspectral image at different wavelengths , using 970nm images with large background differences for background segmentation;

(3) Extract the optimal wavelength of water content of rapeseed, and use ENVI software to extract the principal component image data at 720nm, 960nm, and 1450nm;

(4) Based on the obtained principal component images, the grayscale, texture and reflection intensity distribution feature space of the principal component images was established, combined with the leaf moisture content data, based on the partial least squares method, a prediction model of rapeseed moisture content was established, referring to the comparison of standardized management Group samples and real-time ambient temperature and humidity data, giving rapeseed water demand and irrigation reference information.

Beneficial effects of the present invention: (1) The present invention uses a near-infrared hyperspectral imaging device to quantitatively analyze the water demand information of rapeseed, which has not been involved in previous documents. (2) The present invention obtains the comprehensive information of rapeseed leaves synchronously, integrates the near-infrared image grayscale, texture and near-infrared reflection intensity distribution characteristics to predict and analyze the water content of rapeseed, and realizes the rapid detection of rapeseed water demand information, which is different from the traditional Compared with the single detection method of spectrum and visual image, the recognition accuracy and stability have been significantly improved, and the measurement error of rapeseed moisture content is less than 5%.

Description of drawings

Fig. 1 is a kind of rapid detection method flowchart of rapeseed water demand information of the present invention;

Figure 2 is a schematic diagram of the structure of a near-infrared hyperspectral imaging device; 1-light box; 2-halogen tungsten light source; 3-controller; 4-computer; 5-near-infrared camera; 6-spectrograph; 7-lens; 8 -Y branch linear lamp; 9-electrically controlled displacement stage; 10-rape sample.

Detailed ways

The present invention will be described in further detail below in conjunction with the accompanying drawings.

In order to achieve the above object, the present invention adopts a self-constructed near-infrared hyperspectral imaging device to collect near-infrared hyperspectral data of rape samples. The near-infrared hyperspectral imaging device includes the following components: light box 1, tungsten halogen light source 2, controller 3 , a computer 4, a near-infrared camera 5, a spectrograph 6, a lens 7, a Y-branch linear lamp 8 and an electronically controlled displacement stage 9. The function of light box 1 is to shield external interference and provide a stable light source and detection environment for near-infrared hyperspectral data collection. The top center of light box 1 is fixed with a hyperspectral system composed of near-infrared camera 5, spectrograph 6 and lens 7. Imaging sensor; the geometric center of the bottom surface of the light box is fixed with an electronically controlled displacement platform 9, and Y branch linear lights 8 are symmetrically installed on the left and right sides of the middle of the light box; wherein the hyperspectral imaging sensor is vertically installed directly above the electrically controlled displacement platform 9; The tungsten halogen light source 2 is connected with the Y branch linear lamp 8 through two glass optical fibers; the hyperspectral imaging sensor and the electronically controlled displacement stage 9 are connected with the controller 3 through data lines; the computer 4 is connected with the controller 3 through data lines , the controller 3 accepts the control instructions from the computer 4, controls the travel speed of the electronically controlled displacement stage 9 and parameters such as the scanning speed, exposure time, and focal length of the hyperspectral imaging sensor for information collection, and implements information collection control.

The hyperspectral imaging sensor that wherein said near-infrared camera 5, spectrograph 6 and lens 7 are jointly formed, its lowermost end is lens 7, connects spectrograph 6, near-infrared camera 5 successively upwards thereafter; Wherein said near The infrared camera 5 is an InGaAs imaging camera with a spectral range of 900-1700nm.

The spectral resolution of the hyperspectral imaging sensor described therein is 3.5nm, and the hyperspectral data of the sample is collected by a line scanning mode; the line scanning mode is completed by driving the sample to move along the scanning direction through the electronically controlled displacement stage 9 directly below it Yes, within the near-infrared wavelength range of 900-1700nm, a scanning process can simultaneously acquire near-infrared hyperspectral images in 256 bands.

The wavelength range of the halogen tungsten light source is 400-2600nm.

The interior of the light box described therein is electrostatically painted in black.

In actual measurement, follow the steps below:

(1) Fix the rapeseed sample 10 on the electronically controlled displacement stage 9, adjust the focal length of the lens to 10mm and the resolution of the near-infrared camera to 717×525, the shutter speed to 1/1000s, and the frame rate to 26 to ensure the clarity of the image , set the speed parameter of the electronically controlled displacement stage 9 to 9 to avoid distortion of the scanned image; calibrate the black and white field of the hyperspectral imaging sensor, and collect black field and white field information, wherein the black field is performed by turning off the light source and scanning the lens cover, The white field is carried out by scanning the standard white board to obtain the relative reference value of the wavelength range of 900-1700nm. The reference value of the black field is 0, and the relative reference value of the white field is 255. The difference between the white field and the black field is used as the denominator to calculate The imaging gray value of each pixel is used to set the effective reflection intensity interval of the hyperspectral image to eliminate the dark current noise of the camera; the relative intensity value of each pixel is calculated by the following formula:

　　　　　　    　　　   (1)

(1)

In the formula: x is the intensity detection value of the pixel point; x _min and x _max are the mean value of the intensity of the black field and the mean value of the intensity of the white field.

(2) Collect the near-infrared hyperspectral data cube of rapeseed sample 10, digitally filter through the second-order Butterworth filter to remove the camera "bad point" data of the near-infrared hyperspectral data, and compare the distance between the target and the background in the hyperspectral image at different wavelengths. difference, using 970nm images with large background differences for background segmentation;

(3) Extract the optimal wavelength of water content of rapeseed, and use ENVI software to extract the principal component image data at 720nm, 960nm, and 1450nm;

(4) Based on the acquired principal component images, the grayscale, texture and reflection intensity distribution feature space of the principal component images was established, combined with the leaf moisture content data, based on the partial least squares method, the rapeseed moisture content prediction model was established:

W=621.75+171.54ANS ₁₄₅₀ -169.28ANS ₉₆₀ +68.55ANS ₇₂₀ (2)

In the formula: W water content of rapeseed (%), ANS ₁₄₅₀ , ANS ₉₆₀ , ANS ₇₂₀ are the characteristic parameters of the gray scale, texture and reflection intensity distribution characteristic variables under the corresponding wavelength after normalization and weighted fusion; The moisture content of seedling rape was predicted, and the average absolute error between the predicted value and the measured value was 3.87%, the average relative error was 5.08%, the correlation coefficient was 0.96, and the mean square error was 2.97. The prediction accuracy is significantly improved compared with the existing methods. Referring to the standardized management control group samples and real-time environmental temperature and humidity data, the water demand and irrigation reference information of rapeseed is given.

The above is only in conjunction with a specific embodiment to illustrate and help further understanding of the present invention, but the specific details of the embodiment are only to illustrate the present invention, and do not represent all technical embodiments under the concept of the present invention. In the eyes of the skilled person, some insubstantial changes that do not deviate from the inventive concept, such as simple changes or replacements with technical features having the same or similar technical effects, all fall within the protection scope of the present invention.

Claims (5)

1. A fast detection method of rapeseed water demand information, utilizes near-infrared hyperspectral imaging device to gather the hyperspectral data of rapeseed sample and carries out analysis and processing, it is characterized in that, the step that comprises is: (1) Fix the rape sample on the electronically controlled displacement platform, adjust the focal length of the lens and the resolution, gain and exposure time of the near-infrared camera to ensure the clarity of the image, and set the speed of the electrically controlled displacement platform to avoid distortion of the scanned image; The hyperspectral imaging sensor calibrates the black and white field, sets the effective reflection intensity interval of the hyperspectral image, and eliminates the dark current noise of the camera; (2) Collect the near-infrared hyperspectral data cube of the rapeseed sample, digitally filter through the second-order Butterworth filter to remove the camera "bad point" data of the near-infrared hyperspectral data, and compare the difference between the target and the background in the hyperspectral image at different wavelengths , using 970nm images with large background differences for background segmentation; (3) Extract the optimal wavelength of water content of rapeseed, and use ENVI software to extract the principal component image data at 720nm, 960nm, and 1450nm; (4) Based on the acquired principal component images, the grayscale, texture and reflection intensity distribution feature space of the principal component images was established, combined with the moisture content data of rapeseed samples, and based on the partial least squares method, a prediction model of moisture content of rapeseed was established, referring to standardized management Control samples and real-time ambient temperature and humidity data, giving rapeseed water demand and irrigation reference information. 2. The fast detection method of a kind of rapeseed water demand information according to claim 1, is characterized in that, described near-infrared hyperspectral imaging device comprises following parts: light box, tungsten halogen light source, controller, computer, superspectral imaging device A spectral imaging sensor, a Y-branch linear lamp, and an electronically controlled displacement platform. The hyperspectral imaging sensor is composed of a near-infrared camera, a spectrograph, and a lens connected in sequence, and is fixed at the top center of the light box; the electronically controlled displacement platform The geometric center of the bottom surface is fixed in the light box, located directly below the hyperspectral imaging sensor; the Y-branch linear lamp is symmetrically installed on the left and right sides of the middle of the light box; the halogen tungsten light source and the Y-branch linear lamp pass through Two glass optical fibers are connected; the hyperspectral imaging sensor and the electric displacement stage are connected to the controller through a data line; the computer is connected to the controller through a data line, and the controller accepts the control instructions of the computer to control the The travel speed of the electronically controlled displacement stage and the scanning speed, exposure time, focal length and other parameters when collecting information from the hyperspectral imaging sensor. 3 . The method for quickly detecting water demand information of rapeseed according to claim 2 , wherein the near-infrared camera is an InGaAs imaging camera with a spectral range of 900-1700 nm. 4. the rapid detection method of a kind of rapeseed water demand information according to claim 3, is characterized in that, described hyperspectral imaging sensor spectral resolution is 3.5nm, adopts the hyperspectral data of line scan mode to collect sample, and described The line scanning method refers to the electric displacement stage located directly below the hyperspectral imaging sensor, which drives the rape sample to move along a straight line during data collection. In the near-infrared wavelength range of 900-1700nm, a scanning process can be synchronized Acquire near-infrared hyperspectral images in 256 bands. 5 . The rapid detection method for water demand information of rapeseed according to claim 4 , wherein the inside of the light box is painted with black electrostatic spraying. 6 .

Images