In-situ detection system and device for moisture of corn kernels in pustulation period

Abstract

The invention relates to the technical field of corn test, in particular to a pustulation stage corn kernel moisture in-situ detection system and device, the pustulation stage corn kernel moisture in-situ detection device is utilized, the height, angle and the like of a spectrum detector are adjusted according to actual conditions to determine the position of a corncob tip, and near infrared spectrum data of the corncob tip are collected through the spectrum detector; the method comprises the following steps: carrying out water content determination on corn kernels in a pustulation period, establishing a detection model and evaluating the detection model through construction of a general model based on combination of a direct correction algorithm, a genetic algorithm and a secondary wavelength screening method for iteratively retaining information variables, and finally completing rapid detection of the water content of the corn kernels in the pustulation period by using the detection model. According to the system and the device, the problem of wavelength redundancy in the whole spectrum of the near infrared spectrum is solved, the characteristic wavelength points can be screened out from the whole spectrum to form the characteristic subset, the problems of long time consumption and high working intensity of the technology for detecting the moisture content in the corn breeding process are solved, the efficiency and the productivity in a laboratory and a production environment are improved, and the labor intensity is reduced. The rapid detection of the grain moisture in the corn filling period can be realized.

Description

Technical Field

[0001] The invention relates to the technical field of corn seed testing, and in particular to an in-situ detection system and device for corn grain moisture during the grain filling period. Background Art

[0002] Spectral detection is one of the main methods to reflect moisture, an important indicator of corn quality. my country's corn planting will now and in the future develop in the direction of standardized, mechanized and automated planting. Corn biological detection is an important link in distinguishing high-quality corn seeds, and corn quality is also an important guarantee for mechanized corn harvesting. The annual corn biological breeding volume can reach tens of thousands or even hundreds of thousands. The detection of corn moisture is a very important link in corn testing and requires precise processing. The traditional determination of moisture content requires a large number of corn test samples, equipment and manual operation time. In corn breeding, due to the constraints of objective conditions such as the planting area for breeding new varieties, the number of corn plants that can be planted per square meter, and the number of effective test ears, the number of samples sampled and the sampling cost when measuring the moisture content of corn kernels during the filling period are subject to certain restrictions. Based on the current problems of slow speed, seed damage, and the need for a large number of sample sets in the detection of corn grain moisture during the filling period, as well as the fact that corn grain detection is limited to the laboratory, the in-situ rapid detection of corn grain moisture based on near-infrared spectroscopy is carried out to reduce the number of samples in the corn breeding process, save the cost of corn breeding, and provide technical support for moisture detection in corn breeding and variety improvement. It is helpful to improve the rapid development of large-scale automation of corn seed breeding and promote the advancement of corn mechanization. Therefore, it is of great significance to carry out in-situ real-time detection of corn moisture content.

[0003] Near Infrared Spectroscopy (NIRS) is a technology that has developed rapidly in the late 1980s. It has the characteristics of fast analysis speed, no sample destruction, no need for sample preparation, simultaneous determination of multiple sample components, and online detection. It has been widely used in the rapid detection of agricultural products. However, in the study of rapid determination of near-infrared spectral components, the sample size is generally between 100 and 200. In practical applications, the minimization of the correction sample set, reliance on empirical sample detection, and low modeling rate of new samples of near-infrared spectroscopy analysis methods have greatly hindered the application of near-infrared spectroscopy analysis. It can not only improve the modeling rate of the quantitative analysis model of near-infrared spectroscopy, but also expand the application field of near-infrared spectroscopy analysis methods; Summary of the invention

[0004] The present invention aims to overcome the deficiencies of the prior art and provides an in-situ detection system and device for moisture in corn kernels during the filling period, which can not only improve the modeling rate of the quantitative analysis model of near-infrared spectroscopy and expand the application field of near-infrared spectroscopy analysis methods; reduce the number of samples in the corn breeding process, save parents in corn breeding, and provide moisture detection technical support for corn breeding, variety improvement, etc.; help to improve the rapid development of large-scale automation of corn seed breeding and promote the advancement of corn mechanization.

[0005] The in-situ detection system for corn grain moisture during the grain filling period of the present invention is implemented by the following steps:

[0006] S1: Sample collection and preparation

[0007] The position of the corn ear tip was determined by using a field detection device, and the near-infrared spectral data of the corn ear tip were collected. X corn varieties suitable for planting in northern reclamation areas were selected. Sampling was carried out every 7 days starting from the late grain filling period of corn. A spectral sample was obtained each time, and a spectral sample was collected b times, and a×b spectral samples were obtained in total, which was named SZL sample set. Then, the near-infrared spectral data of 200 kernels in the middle of the same corn ear were collected, which was named ZL sample set. The two sample sets corresponded one to one, and contained a×b spectral samples and 1845 wavelength variables respectively.

[0008] S2: Determination of moisture content

[0009] The moisture content of the samples corresponding to the ZL sample set obtained in step S1 is measured to obtain the moisture content of the corn kernels;

[0010] S3: Construction of a general model based on a direct correction algorithm combined with a genetic algorithm and a secondary wavelength screening method with iterative information retention variables

[0011] Step 1: Generate the initial population of ZL and SZL sample data;

[0012] Step 2: Use the tournament selection method to randomly select a number of individuals from the initial population and calculate the probability P of the individual being inherited to the next generation population. i , select P i The higher individuals are selected as the winners for crossover and mutation, and the winners retained after the iteration are selected as the new individual population;

[0013] Step 3: Preset variable threshold f * = 0.5, select all wavelength variables in the individual population with a probability exceeding the threshold, and form candidate subsets T corresponding to the ZL and SZL sample data respectively. ZL and T SZL , where T ZL A total of 793 wavelength variables, T SZLA total of 800 wavelength variables;

[0014] Step 4: For the candidate subset T ZL and T SZL Generate binary matrices of size n×m, where n is the number of samples and m is the number of wavelength variables. Build partial least squares models for each row of variables in the matrix, and use the root mean square error (RMSECV) of 5-fold cross validation to evaluate the performance of variable subsets. In each iteration, calculate the RMSECV value of each variable when it is included and removed, and get the difference between the two. Introduce the P value of the non-parametric test algorithm Mann-Whiney U test to judge the importance of the variable. According to the IRIV classification rule, divide the variables into 4 categories, retain the strong information variables and weak information variables that are beneficial to the model, and finally form the best candidate subsets Xs and Xt corresponding to the ZL and SZL sample data, where Xs has a total of 53 wavelength variables and Xt has a total of 25 wavelength variables.

[0015] Step 5: For the ZL spectral data Xs and SZL spectral data Xt, the KS algorithm is used to divide the two groups of samples in a ratio of 3:1; 3 of them form the ZL transformation set Xs trans and SZL transformation set Xt trans , 1 set constitutes the training set;

[0016] Step 6: Calculate the formula F = Xs through the transformation matrix trans ·Xt trans -1 , get the transformation matrix F, and combine F with the SZL prediction set Xt val Matrix multiplication was performed to eliminate the main errors between samples with different traits. After spectral correction was completed, a general model for quantitative analysis of corn moisture based on partial least squares regression (PLSR) was established.

[0017] S4: Detection model establishment and evaluation

[0018] The spectral data of the prediction set is used to use the universal model for quantitative analysis of corn moisture established in S3 to screen the characteristic wavelengths and spectrum correction of corn kernel moisture in the filling period and establish a partial least squares regression model, and the determination coefficient is used to evaluate the accuracy of the regression model; if the evaluation result of the prediction set does not meet the actual detection accuracy requirements, S3 is re-executed; when the detection accuracy of the established partial least squares regression model meets the requirements, the corresponding model is output to complete the construction of the universal model for rapid detection of corn kernel moisture by near-infrared spectroscopy;

[0019] S5: Application of detection model

[0020] By performing near-infrared spectroscopy scanning on the tip of the corn ear whose kernel moisture content is to be tested during the grain-filling period and inputting the detection model, the rapid detection of the moisture content of the corn kernels during the grain-filling period can be completed.

[0021] The in-situ detection device for moisture in corn grains during the filling period of the present invention comprises a crawler-type mobile trolley, a lifting device is fixed at the center position of the upper end of the mobile trolley, a mounting seat is fixed at the top of the lifting device, a mechanical arm is rotatably mounted on the upper end of the mounting seat, a light shield is rotatably mounted on the other end of the mechanical arm, and a spectrometer is fixed in the light shield; a storage table is fixed on the middle outer wall of the lifting device, a clamping device is provided on the storage table, and a computer is clamped in the clamping device; batteries are respectively fixed at the front and rear ends of the mobile trolley, a handle is fixed at the rear end of the mobile trolley, and a control button is provided on one side end face of the storage table.

[0022] As a further improvement of the present invention, the lifting device is composed of a support tube, a lifting rod, a motor and a threaded rod. The support tube is fixed at the upper center position of the mobile trolley, the lower end of the lifting rod is placed in the support tube, the motor is fixed at the bottom of the support tube, the threaded rod is fixed at the output end of the motor, a threaded groove is opened in the lifting rod, and the threaded rod is placed in the threaded groove and threadedly connected thereto; a storage platform is fixed to the outer wall of the upper end of the support tube, and the mounting seat is fixed to the top of the lifting rod.

[0023] As a further improvement of the present invention, a countersunk hole is provided at the center position of the upper end of the mounting seat, a bearing is provided in the countersunk hole, a rotating shaft is fixed in the inner ring of the bearing, a supporting plate is fixed at the upper end of the rotating shaft, an annular groove is provided at the upper end of the mounting seat, a plurality of balls are provided in the annular groove, and the bottom end of the supporting plate abuts against the surface of the balls; a positioning hole coaxial with the countersunk hole is provided at the center position of the upper end of the mounting seat, an annular pressure cover is fixed in the positioning hole by bolts, a pressure ring is coaxially provided at the bottom end of the annular pressure cover, and the pressure ring presses the outer ring of the bearing into the countersunk hole.

[0024] As a further improvement of the present invention, the mechanical arm is composed of a fixed arm, a connecting arm and a supporting arm, the fixed arm is fixed on the supporting plate, the connecting arm is rotatably connected to the upper end of the fixed arm, the supporting arm is rotatably connected to the other end of the connecting arm, and the light shield is rotatably connected to the front end surface of the supporting arm; an electric telescopic rod is rotatably connected to one side end surface of the fixed arm through a connecting piece, and the extended end of the electric telescopic rod is rotatably connected to the bottom end of the supporting arm through a connecting piece.

[0025] As a further improvement of the present invention, the clamping device is composed of a bidirectional threaded rod, a clamping plate and a handle. A long strip-shaped groove is provided in the storage table, and a bidirectional threaded rod is rotatably connected in the groove. One end of the bidirectional threaded rod passes through the storage table and is placed outside it, and a handle is fixedly connected to its end. Sliders are threadedly connected to both ends of the outer wall of the bidirectional threaded rod, and clamping plates are respectively fixed at the upper ends of the two sliders. The computer is clamped between the two clamping plates.

[0026] As a further improvement of the present invention, the storage battery is electrically connected to the spectrum detector, the motor and the electric telescopic rod respectively.

[0027] As a further improvement of the present invention, the spectrum detector is connected to computer data.

[0028] The in-situ detection system and device of corn grain moisture in the grain filling period of the present invention uses the spectral data collected by the detection device to establish a partial least squares corn moisture quantitative analysis model by combining the spectral data processing strategy of the direct correction algorithm with the genetic algorithm and the secondary wavelength screening method of iteratively retaining information variables, which not only solves the problem of wavelength redundancy in the full spectrum of the near-infrared spectrum, but also can screen out characteristic wavelength points from the full spectrum to form a characteristic subset. The prediction process of the sample to be tested involves inputting its new spectral data into the trained model, and the model will perform spectral correction according to the characteristics of these spectral data to predict the corresponding moisture content, effectively solving the problem of long time and high work intensity in the detection technology of moisture content in the corn breeding process, completing the construction of the corn grain moisture rapid detection model, and can also be carried out without destructive analysis of samples, greatly improving the efficiency and productivity in the laboratory and production environment, and realizing the rapid detection of corn grain moisture during the grain filling period. At the same time, the detection device of the present design has a reasonable structure and is easy and convenient to operate. The moisture content of field corn can be detected in real time in the field through the field detection device. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 It is a model flow chart of the in-situ detection system of corn grain moisture in the grain filling period of the present invention;

[0030] Figure 2 This is the near infrared absorption spectrum of the tip of the corn ear and 200 kernels in the middle of the corn ear;

[0031] Figure 3 Comparison of the test results of the general model with the moisture content obtained by the drying method;

[0032] Figure 4 This is a schematic structural diagram of an in-situ detection device for moisture in corn grains during the grain filling period of the present invention;

[0033] Figure 5 This is a schematic diagram of the storage table structure;

[0034] Figure 6 for Figure 4 Enlarged view of part A in the middle;

[0035] Figure 7 This is a schematic diagram of the mechanical arm structure;

[0036] Figure 8 for Figure 7Enlarged view of part B in the middle. DETAILED DESCRIPTION

[0037] Example 1

[0038] The in-situ detection system for corn grain moisture during the grain filling period of the present invention is implemented by the following steps:

[0039] S1: Sample collection and preparation

[0040] The in-situ detection device for moisture in corn grains during the filling period prepared in Example 2 was used to adjust the height and angle of the spectrometer 7 according to actual conditions to determine the position of the corn ear tip, and the near-infrared spectral data of the corn ear tip was collected by the spectrometer 7. 2-3 corn varieties suitable for planting in northern reclamation areas were selected, and sampling was carried out every 7 days starting from the late corn filling period. 29 spectral samples were obtained each time, and a total of four spectral samples were collected, with a total of 116 spectral samples, which were named SZL sample set; then the near-infrared spectral data of 200 kernels in the middle of the same corn ear were collected, which was named ZL sample set; the two sample sets corresponded one to one, and each contained 116 spectral samples and 1845 wavelength variables. Figure 2 These are the near-infrared absorption spectra of the tip of the corn ear and 200 kernels in the middle of the corn ear;

[0041] S2: Determination of moisture content

[0042] The moisture content of the samples corresponding to the ZL sample set obtained in step S1 is measured to obtain the moisture content of the corn kernels;

[0043] S3: Construction of a general model based on a direct correction algorithm combined with a genetic algorithm and a secondary wavelength screening method with iterative information retention variables

[0044] Step 1: Generate the initial population of ZL and SZL sample data;

[0045] Step 2: Use the tournament selection method to randomly select a number of individuals from the initial population and calculate the probability P of the individual being inherited to the next generation population. i , select P i The higher individuals are selected as the winners for crossover and mutation, and the winners retained after the iteration are selected as the new individual population;

[0046] Step 3: Preset variable threshold f * = 0.5, select all wavelength variables in the individual population with a probability exceeding the threshold, and form candidate subsets T corresponding to the ZL and SZL sample data respectively. ZL and T SZL , where T ZL A total of 793 wavelength variables, T SZL A total of 800 wavelength variables;

[0047] Step 4: For the candidate subset T ZL and T SZL Generate binary matrices of size n×m, where n is the number of samples and m is the number of wavelength variables. Build a least squares model for each row of variables in the matrix, and use the root mean square error (RMSECV) of 5-fold cross validation to evaluate the performance of variable subsets. In each iteration, calculate the RMSECV value of each variable when it is included and removed, and get the difference between the two. Introduce the P value of the non-parametric test algorithm Mann-Whiney U test to judge the importance of the variable. According to the IRIV classification rule, divide the variables into 4 categories, retain the strong information variables and weak information variables that are beneficial to the model, and finally form the best candidate subsets Xs and Xt corresponding to the ZL and SZL sample data, where Xs has a total of 53 wavelength variables and Xt has a total of 25 wavelength variables.

[0048] Step 5: For the ZL spectral data Xs and SZL spectral data Xt, the two groups of samples are divided into a ratio of 3:1 using the KS algorithm; the ZL transformation set Xs trans and SZL transformation set Xt trans Each contains 87 samples, the training set is the remaining 29 samples of ZL, and the prediction set is the remaining 29 samples of SZL;

[0049] Step 6: Calculate the formula F = Xs through the transformation matrix trans ·Xt trans -1 , get the transformation matrix F, and combine F with the SZL prediction set Xt val Matrix multiplication was performed to eliminate the main errors between samples with different traits. After spectral correction was completed, a general model for quantitative analysis of corn moisture based on partial least squares regression (PLSR) was established.

[0050] S4: Detection model establishment and evaluation

[0051] The prediction set spectral data were combined with the direct correction algorithm, genetic algorithm and iterative retention information variables to screen the characteristic wavelengths of corn grain moisture in the filling period to establish a partial least squares regression model, and the prediction set determination coefficient R was used. 2 The accuracy of the regression model is evaluated, and the modeling stipulates that when the prediction determination coefficient exceeds 0.9, it is considered that the model can meet the requirements of actual detection. In this embodiment, the determination coefficient of the constructed moisture detection general model is 0.9715, which meets the requirements of the modeling regulations, and the model construction is completed.

[0052] S5: Application of detection model

[0053] After sampling corn samples for testing kernel moisture during the grain filling period, near infrared spectroscopy is performed, and the spectral data is preprocessed and then input into the detection model according to the preferred characteristic wavelength, which can complete the rapid detection of the moisture content of corn kernels during the grain filling period. The test results of the moisture content of corn kernels are between 43.74% and 60.83%, with an average deviation of less than 0.04%. Figure 3 Comparison between the model test results and the moisture content obtained by the drying method.

[0054] Example 2

[0055] The in-situ detection device for moisture content of corn kernels during the grain filling period of the present invention is as follows: Figure 4 As shown, it includes a crawler-type mobile trolley 1, a lifting device 2 is fixed at the upper center position of the mobile trolley 1, and the lifting device 2 is composed of a support tube 13, a lifting rod 14, a motor 15 and a threaded rod 16. The support tube 13 is fixed at the upper center position of the mobile trolley 1, and the lower end of the lifting rod 14 is placed in the support tube 13. A motor 15 is fixed at the bottom of the support tube 13, and a threaded rod 16 is fixed at the output end of the motor 15. A threaded groove is opened in the lifting rod 14, and the threaded rod 16 is placed in the threaded groove and threadedly connected thereto. By starting the motor 15, the threaded rod 16 can be driven to rotate, so that the lifting rod 14 can move up and down in the support tube 13, so that the height of the device can be adjusted to adapt to different plants and ear heights.

[0056] like Figure 6 As shown, a mounting seat 4 is fixed at the top of the lifting rod 14, a countersunk hole is opened at the center position of the upper end of the mounting seat 4, a bearing 18 is arranged in the countersunk hole, a rotating shaft 19 is fixed in the inner ring of the bearing 18, a supporting plate 20 is fixed at the upper end of the rotating shaft 19, an annular groove 21 is opened at the upper end of the mounting seat 4, a plurality of balls 22 are arranged in the annular groove 21, and the bottom end of the supporting plate 20 is against the surface of the ball 22; a positioning hole 23 coaxial with the countersunk hole is also opened at the center position of the upper end of the mounting seat 4, an annular pressure cover 24 is fixed in the positioning hole 23 by bolts, a pressure ring 25 is coaxially provided at the bottom end of the annular pressure cover 24, and the pressure ring 25 presses the outer ring of the bearing 18 into the countersunk hole.

[0057] like Figure 7 As shown, a mechanical arm 5 is fixed at the center position of the upper end of the support plate 20, and the mechanical arm 5 is composed of a fixed arm 26, a connecting arm 27 and a supporting arm 28. The fixed arm 26 is fixed to the upper end of the support plate 20, and the connecting arm 27 is rotatably connected to the upper end of the fixed arm 26, and the supporting arm 28 is rotatably connected to the other end of the connecting arm 27. Figure 8As shown, two symmetrical mounting seats 34 are fixed on the front end surface of the support arm 28, and through holes are provided on the two mounting seats 34. The through hole on one of the mounting seats 34 is composed of an inner circular hole and an outer hexagonal hole coaxial therewith, and a hexagonal nut is embedded in the hexagonal hole. A connecting block 35 is provided between the two mounting seats 34, and a through hole I is provided on the connecting block 35 corresponding to the through holes on the mounting seats 34 on both sides, and bolts are provided in the two through holes and through hole I, one end of the bolt is screwed tightly with the hexagonal nut thread, and a knob 36 is fixed on the other end, and a light shield 6 is fixed on one side end surface of the connecting block 35, and a spectrometer 7 is fixed in the light shield 6 for collecting near-infrared spectrum data of corn ear tips. An electric telescopic rod 29 is rotatably connected to one side end face of the fixed arm 26 through a connecting piece, and the protruding end of the electric telescopic rod 29 is rotatably connected to the bottom end of the support arm 28 through a connecting piece. By manually rotating the fixed arm 26, the horizontal detection position of the spectrometer 7 can be adjusted, and by starting the electric telescopic rod 29 to extend or retract its protruding end, the height of the spectrometer 7 can be adjusted so that it can adapt to the detection of corn kernels at different heights. By turning the knob 36 forward, the two mounting seats 34 can clamp the connecting block 35 to fix the position of the light shield 6. By turning the knob 36 backward, the fastening limit of the two mounting seats 34 on the connecting block 35 can be released. At this time, the position of the light shield 6 in the vertical direction can be adjusted at will, thereby adjusting the detection angle of the spectrometer 7 in the vertical direction.

[0058] A storage platform 8 is fixed on the outer wall of the upper end of the support tube 13. Figure 5 As shown, a clamping device 9 is provided on the storage table 8, and the clamping device 9 is composed of a bidirectional threaded rod 30, a clamping plate 31 and a handle 32. A long strip-shaped groove 33 is provided in the storage table 8, and the bidirectional threaded rod 30 is rotatably connected in the groove 33. One end of the bidirectional threaded rod 30 passes through the storage table 8 and is placed outside it, and a handle 32 is fixedly connected to its end. Slide blocks are threadedly connected to both ends of the outer wall of the bidirectional threaded rod 30, and clamping plates 31 are respectively fixed at the upper ends of the two sliders. The computer 10 is clamped between the two clamping plates 31. The bidirectional threaded rod 30 can be driven to rotate by manually rotating the handle 32, so that the two sliders can move toward or away from each other, and then, the two clamping plates 31 can be moved closer to or farther away from each other, so as to clamp or release the computer 10.

[0059] A battery 11 is fixed at the front and rear ends of the mobile cart 1, respectively. The battery 11 is electrically connected to the spectrum detector 7, the motor 15 and the electric telescopic rod 29, respectively. The spectrum detector 7 is data-connected to the computer 10, and the spectrum detector 7 can transmit the detection results to the computer 10, so that the staff can save the detection results conveniently; a control button 3 is provided on one side end face of the storage table 8, and the control button 3 can be used to control the start and stop of the motor 15 and the electric telescopic rod 29; a handle 12 is fixed at the rear end of the mobile cart 1, and the handle 12 is used to facilitate the staff to push the device to walk.

Claims

1. The in-situ detection system of corn grain moisture during the grain filling period is realized through the following steps: S1: Sample collection and preparation The position of the corn ear tip was determined by using a field detection device, and the near-infrared spectral data of the corn ear tip were collected. X corn varieties suitable for planting in northern reclamation areas were selected. Sampling was carried out every 7 days starting from the late grain filling period of corn. A spectral sample was obtained each time, and a spectral sample was collected b times, and a×b spectral samples were obtained in total, which was named SZL sample set. Then, the near-infrared spectral data of 200 kernels in the middle of the same corn ear were collected, which was named ZL sample set. The two sample sets corresponded one to one, and contained a×b spectral samples and 1845 wavelength variables respectively. S2: Determination of moisture content The moisture content of the samples corresponding to the ZL sample set obtained in step S1 is measured to obtain the moisture content of the corn kernels; S3: Construction of a general model based on a direct correction algorithm combined with a genetic algorithm and a secondary wavelength screening method with iterative information retention variables Step 1: Generate the initial population of ZL and SZL sample data; Step 2: Use the tournament selection method to randomly select a number of individuals from the initial population and calculate the probability P of the individual being inherited to the next generation population. i , select P i The higher individuals are selected as the winners for crossover and mutation, and the winners retained after the iteration are selected as the new individual population; Step 3: Preset variable threshold f * = 0.5, select all wavelength variables in the individual population with a probability exceeding the threshold, and form candidate subsets T corresponding to the ZL and SZL sample data respectively. ZL and T SZL , where T ZL A total of 793 wavelength variables, T SZL A total of 800 wavelength variables; Step 4: For the candidate subset T ZL and T SZL Generate binary matrices of size n×m, where n is the number of samples and m is the number of wavelength variables. Build partial least squares models for each row of variables in the matrix, and use the root mean square error (RMSECV) of 5-fold cross validation to evaluate the performance of variable subsets. In each iteration, calculate the RMSECV value of each variable when it is included and removed, and get the difference between the two. Introduce the P value of the non-parametric test algorithm Mann-Whiney U test to judge the importance of the variable. According to the IRIV classification rule, divide the variables into 4 categories, retain the strong information variables and weak information variables that are beneficial to the model, and finally form the best candidate subsets Xs and Xt corresponding to the ZL and SZL sample data, where Xs has a total of 53 wavelength variables and Xt has a total of 25 wavelength variables. Step 5: For the ZL spectral data Xs and SZL spectral data Xt, the KS algorithm is used to divide the two groups of samples in a ratio of 3:1; 3 of them form the ZL transformation set Xs trans and SZL transformation set Xt trans , 1 set constitutes the training set; Step 6: Calculate the formula F = Xs through the transformation matrix trans ·Xt trans -1 , get the transformation matrix F, and combine F with the SZL prediction set Xt val Matrix multiplication was performed to eliminate the main errors between samples with different traits. After spectral correction was completed, a general model for quantitative analysis of corn moisture based on partial least squares regression (PLSR) was established. S4: Detection model establishment and evaluation The spectral data of the prediction set is used in the universal model for quantitative analysis of corn moisture established by S3 to screen the characteristic wavelengths and spectral correction of corn kernel moisture in the filling period and establish a partial least squares regression model, and the determination coefficient is used to evaluate the accuracy of the regression model; if the evaluation result of the prediction set does not meet the actual detection accuracy requirements, S3 is re-executed to perform characteristic wavelength screening, spectral correction and regression model establishment; when the detection accuracy of the established partial least squares regression model meets the requirements, the corresponding model is output to complete the construction of the universal model for rapid detection of corn kernel moisture by near-infrared spectroscopy; S5: Application of detection model By performing near-infrared spectroscopy scanning on the tip of the corn ear whose kernel moisture content is to be tested during the grain-filling period and inputting the detection model, the rapid detection of the moisture content of the corn kernels during the grain-filling period can be completed.

2. The in-situ detection device for corn grain moisture during the filling period is characterized by The invention comprises a crawler-type mobile vehicle (1), a lifting device (2) is fixed at the center position of the upper end of the mobile vehicle (1), a mounting seat (4) is fixed at the top of the lifting device (2), a mechanical arm (5) is rotatably mounted on the upper end of the mounting seat (4), a light shield (6) is rotatably mounted on the other end of the mechanical arm (5), and a spectrum detector (7) is fixed in the light shield (6); a storage platform (8) is fixed on the middle outer wall of the lifting device (2), a clamping device (9) is provided on the storage platform (8), and a computer (10) is clamped in the clamping device (9); batteries (11) are respectively fixed at the front and rear ends of the mobile vehicle (1), a handle (12) is fixed at the rear end of the mobile vehicle (1), and a control button (3) is provided on one side end face of the storage platform (8).

3. The in-situ detection device for moisture in corn grains during the filling period according to claim 2 is characterized in that The lifting device (2) is composed of a support tube (13), a lifting rod (14), a motor (15) and a threaded rod (16); the support tube (13) is fixed at the upper center position of the moving vehicle (1); the lower end of the lifting rod (14) is placed in the support tube (13); the motor (15) is fixed at the bottom of the support tube (13); the output end of the motor (15) is fixed with a threaded rod (16); a threaded groove is provided in the lifting rod (14); the threaded rod (16) is placed in the threaded groove and is threadedly connected thereto; a storage platform (8) is fixed on the outer wall of the upper end of the support tube (13); and the mounting seat (4) is fixed at the top of the lifting rod (14).

4. The in-situ detection device for moisture in corn grains during the filling period according to claim 2 is characterized in that A countersunk hole is provided at the center position of the upper end of the mounting seat (4), a bearing (18) is provided in the countersunk hole, a rotating shaft (19) is fixed in the inner ring of the bearing (18), a supporting plate (20) is fixed at the upper end of the rotating shaft (19), an annular groove (21) is provided at the upper end of the mounting seat (4), a plurality of balls (22) are provided in the annular groove (21), and the bottom end of the supporting plate (20) abuts against the surface of the balls (22); a positioning hole (23) coaxial with the countersunk hole is provided at the center position of the upper end of the mounting seat (4), an annular pressure cover (24) is fixed in the positioning hole (23) by bolts, a pressure ring (25) is coaxially provided at the bottom end of the annular pressure cover (24), and the pressure ring (25) presses the outer ring of the bearing (18) into the countersunk hole.

5. The in-situ detection device for moisture in corn grains during the filling period according to claim 2, characterized in that The mechanical arm (5) is composed of a fixed arm (26), a connecting arm (27) and a supporting arm (28); the fixed arm (26) is fixed on the supporting plate (20); the connecting arm (27) is rotatably connected to the upper end of the fixed arm (26); the supporting arm (28) is rotatably connected to the other end of the connecting arm (27); and the light shield (6) is rotatably connected to the front end surface of the supporting arm (28); an electric telescopic rod (29) is rotatably connected to one end surface of the fixed arm (26) via a connecting piece; the extended end of the electric telescopic rod (29) is rotatably connected to the bottom end of the supporting arm (28) via a connecting piece.

6. The in-situ detection device for moisture in corn grains during the filling period according to claim 2, characterized in that The clamping device (9) is composed of a bidirectional threaded rod (30), a clamping plate (31) and a handle (32). A long strip-shaped groove (33) is provided in the storage table (8), and the bidirectional threaded rod (30) is rotatably connected in the groove (33). One end of the bidirectional threaded rod (30) passes through the storage table (8) and is placed outside it, and the handle (32) is fixedly connected to its end. Slide blocks are respectively threadedly connected at both ends of the outer wall of the bidirectional threaded rod (30), and clamping plates (31) are respectively fixed at the upper ends of the two sliders. The computer (10) is clamped between the two clamping plates (31).

7. The in-situ detection device for moisture in corn grains during the filling period according to claim 2, characterized in that The storage battery (11) is electrically connected to the spectrum detector (7), the motor (15) and the electric telescopic rod (29) respectively.

8. The in-situ detection device for moisture in corn grains during the filling period according to claim 2, characterized in that The spectrum detector (7) is data-connected to a computer (10).

Images