A method and system for predicting the germination rate of peas based on hyperspectral reflectance

Abstract

The present invention provides a method and system for predicting the germination rate of peas based on hyperspectral reflectance, relating to the field of hyperspectral technology. The present invention obtains hyperspectral image data of historical peas and the corresponding germination conditions; generates a germ recognition coefficient according to the reflectance meters at 680 nm and 1730 nm, and separates the germ region and other regions of the image through this coefficient; analyzes the reflectance of multiple bands in the germ region to generate a germ activity evaluation coefficient; calculates the average value of the germ activity evaluation coefficients of each pixel point in the germ region to obtain an average germ activity evaluation coefficient; studies the spatial distribution of the germ region to generate a reasonable coefficient for the germ spatial distribution; constructs and trains a pea germination prediction model with the reasonable coefficient for the germ spatial distribution and the average germ activity evaluation coefficient as inputs and the pea germination condition as a label; selects a certain sample for germination prediction and calculates the germination rate. When predicting the germination rate, the present invention takes into account the unique structure, composition and spatial distribution of peas, making the prediction result more accurate.

Description

Technical Field

[0001] The present invention relates to the field of hyperspectral technology, and particularly to a method and system for predicting the germination rate of pea seeds based on hyperspectral reflectance. Background Art

[0002] The existing detection methods for the germination rate of pea seeds mainly rely on traditional artificial germination test methods. Usually, pea seeds need to be cultivated under specific temperature and humidity conditions for a period of time (usually 5 to 7 days), and the germination rate is statistically calculated according to the actual germination situation. Although the results of this method are intuitive, the detection cycle is long, the labor intensity is high, and the rapid detection requirements for a large number of seeds cannot be met. In addition, this type of method is somewhat destructive and does not have non-destructiveness and real-time performance. With the development of agricultural intelligence and precision, the non-destructive detection of crop seeds using hyperspectral technology has become a research hotspot. Some studies have tried to use hyperspectral reflectance data to predict the moisture content or viability state of crop seeds, but the existing methods generally have problems such as insufficient detection accuracy, poor generalization ability of algorithm models, and inability to accurately distinguish the germ and non-germ regions of pea seeds.

[0003] In the prior art, the publication number CN103636315A discloses an on-line detection device and method for seed germination rate based on hyperspectral technology, including: S1 collecting hyperspectral images of seed training samples, and extracting characteristic parameters of the hyperspectral images of the seed training samples in characteristic bands; S2 using traditional methods to conduct experiments on the seed training samples to obtain the germination results of the seed training samples; S3 taking the characteristic parameters and spectral reflectance values of the hyperspectral images of the seed training samples in characteristic bands as inputs, and the seed germination results as outputs, to establish a prediction model for seed germination rate based on hyperspectral technology; S4 collecting hyperspectral images of the seeds to be detected, and using the seed germination rate prediction model for detection to obtain the germination rate of the seeds to be detected.

[0004] However, the prior art still has deficiencies. The prior art only collects hyperspectral data of whole seeds and conducts overall spectral feature analysis, and basically does not conduct comprehensive and refined analysis on the special internal moisture, components and spatial distribution of peas. For seeds like peas with thick seed coats and complex internal structures, the germ activity is the key factor determining the germination rate, and the overall spectral signal is easily interfered by the reflection of the seed coat, affecting the prediction accuracy.

[0005] The above information disclosed in the background art section is only used to enhance the understanding of the background of the present disclosure, and therefore it may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0006] The object of the present invention is to provide a method and system for predicting the germination rate of peas based on hyperspectral reflectance to solve the problems raised in the above-mentioned background technology.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A method for predicting the germination rate of peas based on hyperspectral reflectance, the specific steps include:

[0009] 1. A method for predicting the germination rate of peas based on hyperspectral reflectance, characterized in that the specific steps include:

[0010] Step 1: Obtain the hyperspectral image data of historical peas and the corresponding germination conditions; the hyperspectral image data of the peas is the hyperspectral image data of a single pea, and the corresponding germination conditions include germination and non-germination;

[0011] Step 2: Calculate the germ identification coefficient of each pixel point according to the reflectance of the peas at wavelengths of 680 nm and 1730 nm in the hyperspectral image data, and divide the hyperspectral image data of the peas into a germ region and other regions through the germ identification coefficient;

[0012] Step 3: Analyze the reflectance of the germ region of the peas in the hyperspectral image data at wavelengths of 970 nm, 1450 nm, and 1940 nm to generate a moisture evaluation coefficient, and analyze the reflectance of the germ region of the peas in the hyperspectral image data at wavelengths of 680 nm, 1200 nm, and 1730 nm to generate a component activity evaluation coefficient; generate a germ activity evaluation coefficient according to the moisture evaluation coefficient and the component activity evaluation coefficient; calculate the average value of the germ activity evaluation coefficients of all pixel points in the germ region of the peas as the average germ activity evaluation coefficient;

[0013] Step 4: Calculate the center of the peas in the hyperspectral image data, and calculate the reasonable coefficient of germ spatial distribution according to each position in the pixel points of the center of the peas in the hyperspectral image data and the germ region of the peas; use the reasonable coefficient of germ spatial distribution and the average germ activity evaluation coefficient as inputs, and the pea germination situation as a label to construct and train a pea germination prediction model;

[0014] Step 5: Randomly select N peas from a batch of peas whose germination rate is to be predicted, and calculate the reasonable coefficient of germ spatial distribution and the average germ activity evaluation coefficient of each pea according to Steps 1-4; input the reasonable coefficient of germ spatial distribution and the average germ activity evaluation coefficient of the randomly selected peas into the pea germination prediction model to obtain the germination situation of the randomly selected peas, and calculate the pea germination rate according to the predicted germination situation.

[0015] Further, the hyperspectral image obtained by the hyperspectral imaging system includes peas and the background. The pea area is manually annotated, and the hyperspectral image of the pea area is cropped as the hyperspectral image data of the peas. The hyperspectral imaging system includes a visible light + near-infrared detector and a short-wave infrared detector. Among them, the visible light + near-infrared detector can cover the 400 - 1000 nm band, the short-wave infrared detector can cover the 1000 - 2500 nm band, the hyperspectral imaging system can cover the 400 - 2500 nm band, and the wavelength resolution of the hyperspectral imaging system is 1 - 5 nm.

[0016] Further, the specific formula for calculating the germ recognition coefficient is:

[0017] Fa i =R i (680)*R i (1730)

[0018] Among them, Fa i is the germ recognition coefficient of the i-th pixel of the peas in the hyperspectral image data, R i (680) is the reflectance of the i-th pixel of the peas in the hyperspectral image data at a wavelength of 680 nm, R i (1730) is the reflectance of the i-th pixel of the peas in the hyperspectral image data at a wavelength of 1730 nm, and i is the index of the pixel of the hyperspectral image data of the peas;

[0019] The specific logic for dividing the hyperspectral image data of the peas into the germ area and other areas is: a preset germ recognition threshold Fa0 is set. All pixel points where Fa i <Fa0 form the germ area, and all pixel points where Fa i ≥Fa0 form other areas.

[0020] Further, the specific formula for generating the moisture evaluation coefficient is:

[0021] WR j =R j (970)*R j (1940)*lnR j (1450)

[0022] Among them, WR j is the moisture evaluation coefficient of the j-th pixel of the germ area of the peas, R j (970) is the reflectance of the j-th pixel of the germ area of the peas at a wavelength of 970 nm, R j (1450) is the reflectance of the j-th pixel of the germ area of the peas at a wavelength of 1450 nm, R j(1940) is the reflectance of the j-th pixel in the germ region of the pea at a wavelength of 1940 nm, where j is the index of the pixel in the germ region of the pea;

[0023] The specific formula for generating the component activity evaluation coefficient is:

[0024] GR j = αR j (680) + βR j (1200) + γR j (1730)

[0025] Where, GR j is the component activity evaluation coefficient of the j-th pixel in the germ region of the pea, and R j (680) is the reflectance of the j-th pixel in the germ region of the pea at a wavelength of 680 nm, and R j (1200) is the reflectance of the j-th pixel in the germ region of the pea at a wavelength of 1200 nm, and R j (1730) is the reflectance of the j-th pixel in the germ region of the pea at a wavelength of 1730 nm; α, β, and γ are weight coefficients, and α + β + γ = 1, with α > γ > β;

[0026] The specific formula for generating the germ activity evaluation coefficient is:

[0027] ZAR j = GR j * WR j

[0028] Where, ZAR j is the germ activity evaluation coefficient of the j-th pixel in the germ region of the pea.

[0029] Furthermore, the specific formula for calculating the center of the pea is:

[0030]

[0031] Where, is the center position of the pea, and (x i , y i ) is the position of the i-th pixel of the pea in the hyperspectral image data;

[0032] The specific formula for generating the reasonable coefficient of germ spatial distribution is:

[0033]

[0034] Where, SP is the reasonable coefficient of germ spatial distribution, and (x j , y j) is the position of the j-th pixel in the germ region of the pea in the hyperspectral image data. J is the total number of pixels in the germ region of the pea, and j is the index of the pixels in the germ region of the pea.

[0035] Further, the specific logic for calculating the pea germination rate is as follows: count the number of peas with a germination status of germinated in the prediction results, and divide the number of peas with a germination status of germinated in the prediction results by the total number of selected peas to obtain the germination rate. The specific formula for calculating the germination rate is as follows:

[0036]

[0037] Among them, FY is the germination rate, M is the number of peas with a germination status of germinated in the prediction results, and N is the total number of selected peas.

[0038] The present invention further provides a pea germination rate prediction system based on hyperspectral reflectance. The system is used to implement the pea germination rate prediction based on hyperspectral reflectance, and specifically includes:

[0039] A data acquisition module for obtaining hyperspectral image data of historical peas and the corresponding germination status. The hyperspectral image data of the peas is the hyperspectral image data of a single pea, and the corresponding germination status includes germinated and non-germinated.

[0040] A region recognition module for calculating the germ recognition coefficient of each pixel according to the reflectance of the pea at wavelengths 680nm and 1730nm in the hyperspectral image data, and dividing the hyperspectral image data of the pea into a germ region and other regions through the germ recognition coefficient.

[0041] An activity evaluation module for analyzing the reflectance of the germ region of the pea at wavelengths 970nm, 1450nm, and 1940nm in the hyperspectral image data to generate a moisture evaluation coefficient, and analyzing the reflectance of the germ region of the pea at wavelengths 680nm, 1200nm, and 1730nm in the hyperspectral image data to generate a component activity evaluation coefficient. Generating a germ activity evaluation coefficient according to the moisture evaluation coefficient and the component activity evaluation coefficient; calculating the average value of the germ activity evaluation coefficients of all pixels in the germ region of the pea as the average germ activity evaluation coefficient.

[0042] A space evaluation module for calculating the center of the pea in the hyperspectral image data, and calculating the germ space distribution rationality coefficient according to the center of the pea in the hyperspectral image data and the position of each pixel in the germ region of the pea. Using the germ space distribution rationality coefficient and the average germ activity evaluation coefficient as inputs, and constructing and training a pea germination prediction model with the pea germination status as a label.

[0043] A germination prediction module is used to randomly select N peas from a batch of peas whose germination rate is to be predicted, and evaluate the reasonable coefficient of the germ spatial distribution and the average germ activity evaluation coefficient for each pea; input the reasonable coefficient of the germ spatial distribution and the average germ activity evaluation coefficient of the randomly selected peas into the pea germination prediction model to obtain the germination situation of the randomly selected peas, and calculate the pea germination rate based on the predicted germination situation.

[0044] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0045] By introducing the reflectance in the 680nm and 1730nm bands to construct the germ recognition coefficient, the present invention accurately identifies the germ area in pea seeds, effectively improving the model's discrimination ability for different tissue structures of peas. Further, by analyzing the reflectance at 970nm, 1450nm, and 1940nm to generate the moisture evaluation coefficient, and combining the reflectance at 680nm, 1200nm, and 1730nm to generate the component activity evaluation coefficient, a multi-dimensional evaluation of the germ vitality state is realized, improving the accuracy and scientificity of the model in predicting the seed germination rate.

[0046] The present invention also calculates the reasonable coefficient of the germ spatial distribution, introduces the spatial distribution information in the hyperspectral image data into the prediction model, comprehensively considers the activity evaluation and spatial distribution of the germ area of pea seeds, and further improves the adaptability and generalization ability of the model to pea seeds of different varieties and batches. This method realizes the rapid and non-destructive detection of the germination rate of pea seeds, has high prediction accuracy and real-time application ability, significantly improves the seed quality control and intelligent grading and screening levels, and meets the requirements of modern agricultural intelligent and precise seed management. Description of the Drawings

[0047] Figure 1 It is a schematic diagram of the overall method flow of the present invention.

[0048] Figure 2 It is a schematic diagram of the overall system structure of the present invention. Detailed Embodiments

[0049] To make the objectives, technical solutions, and advantages of the present invention clearer and more understandable, the following further details the present invention in combination with specific embodiments.

[0050] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should have the ordinary meanings understood by those with ordinary skills in the field to which the present invention pertains. The "first", "second" and similar terms used in the present invention do not denote any order, quantity or importance, but are only used to distinguish different components. Words such as "comprising" or "including" mean that the elements or objects appearing before this word cover the elements or objects listed after this word and their equivalents, without excluding other elements or objects. Words such as "connected" or "linked" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Upper", "lower", "left", "right", etc. are only used to indicate relative position relationships, and when the absolute position of the object being described changes, the relative position relationship may also change accordingly.

[0051] Embodiment:

[0052] Please refer to Figure 1 , the present invention provides a technical solution:

[0053] A prediction of pea germination rate based on hyperspectral reflectance, the specific steps include:

[0054] Step 1: Obtain the hyperspectral image data of historical peas and the corresponding germination conditions; the hyperspectral image data of the peas is the hyperspectral image data of a single pea, and the corresponding germination conditions include germination and non-germination;

[0055] The hyperspectral image obtained by the hyperspectral imaging system includes peas and the background, and the pea area is manually marked, and the hyperspectral image of the pea area is cut out as the hyperspectral image data of the peas. The hyperspectral imaging system includes a visible light + near-infrared detector and a short-wave infrared detector. Among them, the visible light + near-infrared detector can cover the 400 - 1000 nm band, the short-wave infrared detector can cover the 1000 - 2500 nm band, the hyperspectral imaging system can cover the 400 - 2500 nm band, and the wavelength resolution of the hyperspectral imaging system is 1 - 5 nm.

[0056] In the hyperspectral image, each pixel point contains complete spectral information. In this embodiment, for the hyperspectral image taken by the hyperspectral imaging system, each pixel point contains the reflectance information of all bands from 400 to 2500 nm. For the same pixel point, the reflectance of all bands from 400 to 2500 nm can be obtained.

[0057] Whiteboard calibration plays a crucial role in the processing of hyperspectral reflectance data. By means of standardization processing, the influence of the light source and the instrument is eliminated, ensuring the accuracy and consistency of the data. For constructing a pea germination rate prediction model with high precision and strong adaptability, whiteboard calibration is an essential basic step.

[0058] Step 2: Calculate the germ identification coefficient of each pixel according to the reflectance of peas at wavelengths of 680 nm and 1730 nm in the hyperspectral image data, and divide the hyperspectral image data of peas into the germ region and other regions through the germ identification coefficient;

[0059] The specific formula for calculating the germ identification coefficient is as follows:

[0060] Fa i =R i (680)*R i (1730)

[0061] Among them, Fa i is the germ identification coefficient of the i-th pixel of peas in the hyperspectral image data, R i (680) is the reflectance of the i-th pixel of peas in the hyperspectral image data at a wavelength of 680 nm, and R i (1730) is the reflectance of the i-th pixel of peas in the hyperspectral image data at a wavelength of 1730 nm, and i is the index of the pixel of the hyperspectral image data of peas;

[0062] Divide the hyperspectral image data of peas into the germ region and other regions; the specific logic is as follows: preset the germ identification threshold Fa0, and all pixel points with Fa i <Fa0 form the germ region, and all pixel points with Fa i ≥Fa0 form other regions;

[0063] Chlorophyll strongly absorbs at 680 nm. If the chlorophyll content of a pixel point is high, more light at 680 nm will be absorbed, and the reflectance of light at 680 nm will be low. The chlorophyll content on the surface of peas usually has a small difference, and the difference in chlorophyll is mainly determined by the internal substances of peas. Usually, the chlorophyll content of the germ inside peas is much higher than that of other regions; 1730 nm is the absorption region of fatty acid groups. If the content of fatty acid groups is high, the reflectance of light at 1730 nm will be low, and the fatty acid groups inside the germ of peas are much higher than those of other regions; the germ identification coefficient generated by coupling the reflectance at 680 nm and the reflectance at 1730 nm can well reflect the difference between the germ region and other regions in the hyperspectral image data of peas. The smaller its value, the more the pixel point absorbs chlorophyll and fatty acid groups, and the greater the probability of being in the germ region;

[0064] Step 3: Analyze the reflectance of the germ region of peas in the hyperspectral image data at wavelengths of 970 nm, 1450 nm, and 1940 nm to generate a moisture evaluation coefficient, and analyze the reflectance of the germ region of peas in the hyperspectral image data at wavelengths of 680 nm, 1200 nm, and 1730 nm to generate a component activity evaluation coefficient; generate a germ activity evaluation coefficient based on the moisture evaluation coefficient and the component activity evaluation coefficient; calculate the average value of the germ activity evaluation coefficients of all pixel points within the germ region of peas as the average germ activity evaluation coefficient;

[0065] The specific formula for generating the moisture evaluation coefficient is:

[0066] WR j =R j (970)*R j (1940)*ln R j (1450)

[0067] Where, WR j is the moisture evaluation coefficient of the j-th pixel point in the germ region of peas, R j (970) is the reflectance of the j-th pixel point in the germ region of peas at a wavelength of 970 nm, R j (1450) is the reflectance of the j-th pixel point in the germ region of peas at a wavelength of 1450 nm, R j (1940) is the reflectance of the j-th pixel point in the germ region of peas at a wavelength of 1940 nm, and j is the index of the pixel points in the germ region of peas;

[0068] The moisture evaluation coefficient reflects the activity of pea germs from the moisture level. The smaller the value, the greater the activity of pea germs and the greater the probability of pea germination. 970 nm is the absorption peak of free water. The greater its reflectance, the less free water, the drier the pea germs, the lower the activity, and it is not conducive to pea germination. 1940 nm reflects the moisture gradient of pea germs. The smaller its reflectance, the more sufficient the overall moisture of pea germs, not dehydrated and not aged, which is conducive to pea germination; 1450 nm reflects the absorption of bound water and cell water in pea germs. The smaller its reflectance, the more bound water and cell water in pea germs, the stronger the cell integrity and activity, and the better the cell membrane function, which is conducive to pea germination. ln R j (1450) enhances the influence of 1450 nm on the overall moisture evaluation through a logarithmic function.

[0069] The specific formula for generating the component activity evaluation coefficient is:

[0070] GR j =αR j(680) + βR j (1200) + γR j (1730)

[0071] Among them, GR j is the component activity evaluation coefficient of the j-th pixel in the germ region of the pea, and R j (680) is the reflectance of the j-th pixel in the germ region of the pea at a wavelength of 680 nm, and R j (1200) is the reflectance of the j-th pixel in the germ region of the pea at a wavelength of 1200 nm, and R j (1730) is the reflectance of the j-th pixel in the germ region of the pea at a wavelength of 1730 nm; α, β, and γ are weighting coefficients, and α + β + γ = 1, α > γ > β;

[0072] The component activity evaluation coefficient reflects the germ activity from the component level. The smaller its value, the greater the activity of the pea germ and the greater the probability of germination. Chlorophyll strongly absorbs at 680 nm. If the chlorophyll content of the pixel is high, more light at 680 nm will be absorbed, and the reflectance of light at 680 nm will be low. The chlorophyll content on the surface of the pea usually has a small difference, and the difference in chlorophyll is mainly determined by the internal substances of the pea. Usually, the chlorophyll content of the germ inside the pea is much higher than that of other regions; and the higher the chlorophyll content of the pea germ, the better the activity of the pea germ and the easier it is to germinate; 1200 nm can penetrate relatively thin cell walls and cytoplasm. The smaller its reflectance, the more complete the pea cell structure and the better the tissue uniformity, indicating that the germ cell wall is thin, the substances are rich, and the metabolism is active, which is manifested as a high overall activity of the pea seedlings; 1730 nm is the absorption region of fatty acid groups. The amount of fatty acid groups can reflect the amount of proteins and lipids stored by the germ for germination. If the content of fatty acid groups is high, the reflectance of light at 1730 nm will be low, and the more proteins and lipids stored by the germ for germination, the easier it is to germinate; because the smaller the reflectance, the more corresponding substances there are and the greater the probability of germination, so the smaller the weighting coefficient, the more important the corresponding component. Among chlorophyll content, pea cell structure integrity, and fatty acid groups, the pea cell structure integrity has a greater impact on pea germination, followed by the impact of fatty acid groups, and then the impact of chlorophyll content. Therefore, α > γ > β is set.

[0073] The specific formula for generating the germ activity evaluation coefficient is:

[0074] ZAR j = GR j * WR j

[0075] Among them, ZAR jis the evaluation coefficient of the germ activity of the j-th pixel in the germ area of the pea; it reflects the activity of the pea germ, and the larger the value, the greater the activity of the pea germ; a single evaluation coefficient of germ activity can only reflect the germ activity at the position of a single pixel, and the evaluation coefficient of germ activity comprehensively analyzes the germ area of the pea and comprehensively reflects the overall activity of the pea germ.

[0076] Step 4: Calculate the center of the pea in the hyperspectral image data of the pea, and calculate the reasonable coefficient of germ spatial distribution according to the center of the pea in the hyperspectral image data and the positions of each pixel in the germ area of the pea; use the reasonable coefficient of germ spatial distribution and the average evaluation coefficient of germ activity as inputs, and the pea germination situation as a label to construct and train a pea germination prediction model.

[0077] The specific formula for calculating the center of the pea is:

[0078]

[0079] where is the position of the center of the pea, and (x i , y i ) is the position of the i-th pixel of the pea in the hyperspectral image data;

[0080] The specific formula for generating the reasonable coefficient of germ spatial distribution is:

[0081]

[0082] where SP is the reasonable coefficient of germ spatial distribution, and (x j , y j ) is the position of the j-th pixel in the germ area of the pea in the hyperspectral image data, J is the total number of pixels in the germ area of the pea, and j is the index of the pixel in the germ area of the pea;

[0083] The reasonable coefficient of germ spatial distribution reflects the reasonable situation of the spatial distribution of the pea germ inside the pea. The smaller the value, the more reasonable the spatial distribution, and the more conducive to pea germination; is the average distance from the pixels in the germ area of the pea to the center of the pea. For peas, the sizes of the germs inside are usually not very different. The closer to the center, the more reasonable the spatial distribution.

[0084] The pea germination prediction model uses a feedforward neural network, with the reasonable coefficient of the germ spatial distribution and the evaluation coefficient of the average germ activity as inputs, and the corresponding pea germination situation as the label. The existing technologies can be used to train the pea germination prediction model, which specifically includes: an input layer, a hidden layer, an output layer, and an activation function. The input layer is responsible for receiving the reasonable coefficient of the germ spatial distribution and the evaluation coefficient of the average germ activity; the hidden layer is used to process the data of the reasonable coefficient of the germ spatial distribution and the evaluation coefficient of the average germ activity; it consists of multiple layers, each layer contains 4 time nodes, and the time nodes of each hidden layer are connected to the previous layer through weights, which are used to perform feature abstraction and non-linear transformation on the input reasonable coefficient of the germ spatial distribution and the evaluation coefficient of the average germ activity; by using the ReLu activation function, a non-linear relationship is introduced to enable the model to fit complex feature relationships; an independent neuron is set in the output layer, which is responsible for converting the local and high-level feature representations extracted by the hidden layer for outputting the pea germination situation; the root mean square error loss function is adopted; the input data is calculated through the network once to obtain the output result, the loss function is calculated according to the predicted value and the true value, the gradient of the loss function with respect to each weight and bias is calculated through the chain rule, and the weights and biases of the network are updated using the gradient descent algorithm to minimize the loss function.

[0085] Step 5: Randomly select N peas from a batch of peas whose germination rate is to be predicted, and calculate the reasonable coefficient of the germ spatial distribution and the evaluation coefficient of the average germ activity for each pea according to Steps 1 - 4; input the reasonable coefficient of the germ spatial distribution and the evaluation coefficient of the average germ activity of the randomly selected peas into the pea germination prediction model to obtain the germination situation of the randomly selected peas, and calculate the pea germination rate according to the predicted germination situation.

[0086] The specific logic for calculating the pea germination rate is: count the number of peas with a germination situation of germination in the prediction results, and divide the number of peas with a germination situation of germination in the prediction results by the total number of selected peas to obtain the germination rate; the specific formula for calculating the germination rate is:

[0087]

[0088] Among them, FY is the germination rate, M is the number of peas with a germination situation of germination in the prediction results, and N is the total number of selected peas.

[0089] Please refer to Figure 2 For this, the present invention further provides a pea germination rate prediction system based on hyperspectral reflectance. The system is used to implement the pea germination rate prediction based on hyperspectral reflectance, and specifically includes:

[0090] A data acquisition module for obtaining hyperspectral image data of historical peas and the corresponding germination status; the hyperspectral image data of the peas is the hyperspectral image data of a single pea, and the corresponding germination status includes germination and non-germination;

[0091] A region recognition module for calculating the germ recognition coefficient of each pixel point based on the reflectance of peas at wavelengths of 680 nm and 1730 nm in the hyperspectral image data, and dividing the hyperspectral image data of the peas into a germ region and other regions through the germ recognition coefficient;

[0092] An activity evaluation module for analyzing the reflectance of the germ region of peas in the hyperspectral image data at wavelengths of 970 nm, 1450 nm, and 1940 nm to generate a moisture evaluation coefficient, and analyzing the reflectance of the germ region of peas in the hyperspectral image data at wavelengths of 680 nm, 1200 nm, and 1730 nm to generate a component activity evaluation coefficient; generating a germ activity evaluation coefficient based on the moisture evaluation coefficient and the component activity evaluation coefficient; calculating the average value of the germ activity evaluation coefficients of all pixel points in the germ region of the peas as the average germ activity evaluation coefficient;

[0093] A space evaluation module for calculating the center of the peas in the hyperspectral image data, and calculating the reasonable coefficient of germ space distribution according to each position in the pixel points of the center of the peas and the germ region of the peas in the hyperspectral image data; constructing and training a pea germination prediction model with the reasonable coefficient of germ space distribution and the average germ activity evaluation coefficient as inputs and the pea germination status as a label;

[0094] A germination prediction module for randomly selecting N peas from a batch of peas whose germination rate is to be predicted, and calculating the reasonable coefficient of germ space distribution and the average germ activity evaluation coefficient of each pea according to the steps; inputting the reasonable coefficient of germ space distribution and the average germ activity evaluation coefficient of the randomly selected peas into the pea germination prediction model to obtain the germination status of the randomly selected peas, and calculating the pea germination rate according to the predicted germination status.

[0095] The above formulas are all calculated by taking the numerical value after dimensionless, and the formula is a formula obtained by software simulation of a large amount of collected data to approximate the real situation. The preset parameters in the formula are set by those skilled in the art according to the actual situation.

[0096] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those skilled in the art will appreciate that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0097] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units. They may be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the objectives of the solution of this embodiment.

[0098] As described above, the above is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed in this application should not easily think of changes or substitutions, and all should be covered within the protection scope of this application.

Claims

1. A method for predicting pea germination rate based on hyperspectral reflectance, characterized in that: The specific steps include: Step 1: Acquire historical hyperspectral image data of peas and corresponding germination conditions; the hyperspectral image data of peas is hyperspectral image data of a single pea, and the corresponding germination conditions include germination and non-germination; Step 2: Calculate the germ recognition coefficient of each pixel point according to the reflectance of peas at wavelengths of 680nm and 1730nm in the hyperspectral image data, and divide the hyperspectral image data of peas into germ areas and other areas according to the germ recognition coefficient; Step 3: Analyze the reflectance of the pea germ region at wavelengths of 970nm, 1450nm and 1940nm in the hyperspectral image data to generate a moisture evaluation coefficient, and analyze the reflectance of the pea germ region at wavelengths of 680nm, 1200nm and 1730nm in the hyperspectral image data to generate a component activity evaluation coefficient; generate a germ activity evaluation coefficient based on the moisture evaluation coefficient and the component activity evaluation coefficient; calculate the average value of the germ activity evaluation coefficients of all pixels in the pea germ region as the average germ activity evaluation coefficient; Step 4: Calculate the pea center of the pea in the hyperspectral image data, and calculate the reasonable coefficient of germ spatial distribution according to the position of each pixel point in the pea center and the pea germ area of ​​the hyperspectral image data; use the reasonable coefficient of germ spatial distribution and the average germ activity evaluation coefficient as input, and the pea germination situation as a label to build and train a pea germination prediction model; Step 5: Randomly select N peas from a batch of peas whose germination rate is to be predicted, and calculate the reasonable coefficient of germ spatial distribution and the average germ activity evaluation coefficient of each pea according to steps 1 to 4; input the reasonable coefficient of germ spatial distribution and the average germ activity evaluation coefficient of the randomly selected peas into the pea germination prediction model to obtain the germination situation of the randomly selected peas, and calculate the pea germination rate according to the predicted germination situation.

2. The method for predicting pea germination rate based on hyperspectral reflectance according to claim 1, characterized in that: The hyperspectral image acquired by the hyperspectral imaging system includes peas and background, and the pea area is manually marked, and the hyperspectral image of the pea area is cut out as the hyperspectral image data of the peas. The hyperspectral imaging system includes a visible light + near-infrared detector and a short-wave infrared detector, wherein the visible light + near-infrared detector can cover the 400-1000nm band, the short-wave infrared detector can cover the 1000-2500nm band, the hyperspectral imaging system can cover the 400-2500nm band, and the wavelength resolution of the hyperspectral imaging system is 1-5nm.

3. The method for predicting pea germination rate based on hyperspectral reflectance according to claim 2, characterized in that: The specific formula for calculating the germ recognition coefficient is: Fa i =R i (680)*R i (1730) Among them, Fa i is the germ recognition coefficient of the ith pixel of pea in the hyperspectral image data, R i (680) is the reflectance of the i-th pixel of the pea at a wavelength of 680nm in the hyperspectral image data, R i (1730) is the reflectance of the i-th pixel of the pea in the hyperspectral image data at a wavelength of 1730 nm, and i is the index of the pixel in the hyperspectral image data of the pea; The specific logic for dividing the hyperspectral image data of peas into the germ region and other regions is as follows: A preset germ recognition threshold Fa0 is set, and all pixel points where Fa i < Fa0 form the germ region, and all pixel points where Fa i ≥ Fa0 form the other regions.

4. The method for predicting pea germination rate based on hyperspectral reflectance according to claim 2, characterized in that: The specific formula used to generate the moisture assessment factor is: WR j =R j (970)*R j (1940)*ln R j (1450) Among them, WR j is the moisture assessment coefficient of the jth pixel in the pea germ area, R j (970) is the reflectance of the jth pixel in the pea germ region at a wavelength of 970 nm, R j (1450) is the reflectance of the jth pixel in the pea germ region at a wavelength of 1450nm, R j (1940) is the reflectance of the j-th pixel in the pea germ region at a wavelength of 1940 nm, and j is the index of the pixel in the pea germ region; The specific formula for generating the component activity evaluation coefficient is: GR j =αR j (680)+βR j (1200)+γR j (1730) Among them, GR j is the component activity evaluation coefficient of the jth pixel in the pea germ area, R j (680) is the reflectance of the jth pixel in the pea germ region at a wavelength of 680nm, R j (1200) is the reflectivity of the jth pixel in the pea germ region at a wavelength of 1200 nm, R j (1730) is the reflectance of the j-th pixel in the pea germ region at a wavelength of 1730 nm; α, β and γ are weight coefficients, and α+β+γ=1, α>γ>β; The specific formula for generating the germ activity evaluation coefficient is: ZAR j =GR j *WR j Among them, ZAR j is the germ activity evaluation coefficient of the j-th pixel in the germ area of ​​pea.

5. The method for predicting pea germination rate based on hyperspectral reflectance according to claim 2, characterized in that: The specific formula used to calculate the center of the pea is: in, is the center position of the pea, (x i ,y i ) is the position of the i-th pixel of the pea in the hyperspectral image data; The specific formula for generating a reasonable coefficient of germ spatial distribution is: Among them, SP is the rational coefficient of germ spatial distribution, (x j ,y j ) is the position of the j-th pixel in the pea germ region in the hyperspectral image data, J is the total number of pixels in the pea germ region, and j is the index of the pixel in the pea germ region.

6. The method for predicting pea germination rate based on hyperspectral reflectance according to claim 1, characterized in that: The specific logic for calculating the pea germination rate is: the number of peas with germination status in the randomly selected peas predicted by statistics is divided by the total number of peas selected to obtain the germination rate; the specific formula for calculating the germination rate is: Among them, FY is the germination rate, M is the number of peas with germination status as germinated in the prediction results, and N is the total number of peas selected.

7. A pea germination rate prediction system based on hyperspectral reflectance, characterized in that: The system is used to realize the prediction of pea germination rate based on hyperspectral reflectance as described in any one of claims 1 to 6, and specifically comprises: A data acquisition module, used to obtain historical hyperspectral image data of peas and corresponding germination conditions; the hyperspectral image data of peas is the hyperspectral image data of a single pea, and the corresponding germination conditions include germination and non-germination; A region recognition module is used to calculate the germ recognition coefficient of each pixel point according to the reflectance of peas at wavelengths of 680nm and 1730nm in the hyperspectral image data, and divide the hyperspectral image data of peas into germ regions and other regions according to the germ recognition coefficient; The activity evaluation module is used to analyze the reflectance of the pea germ region at wavelengths of 970nm, 1450nm and 1940nm in the hyperspectral image data to generate a moisture evaluation coefficient, and analyze the reflectance of the pea germ region at wavelengths of 680nm, 1200nm and 1730nm in the hyperspectral image data to generate a component activity evaluation coefficient; generate a germ activity evaluation coefficient according to the moisture evaluation coefficient and the component activity evaluation coefficient; calculate the average value of the germ activity evaluation coefficients of all pixels in the pea germ region as the average germ activity evaluation coefficient; The spatial evaluation module is used to calculate the pea center of the pea in the hyperspectral image data, and calculate the reasonable coefficient of germ spatial distribution according to the position of each pixel point in the pea center and the pea germ area of ​​the hyperspectral image data; the reasonable coefficient of germ spatial distribution and the average germ activity evaluation coefficient are used as input, and the pea germination situation is used as a label to build and train a pea germination prediction model; The germination prediction module is used to randomly select N peas from a batch of peas whose germination rate is to be predicted, calculate the reasonable coefficient of germ spatial distribution and the average germ activity evaluation coefficient of each pea; input the reasonable coefficient of germ spatial distribution and the average germ activity evaluation coefficient of the randomly selected peas into the pea germination prediction model to obtain the germination situation of the randomly selected peas, and calculate the pea germination rate according to the predicted germination situation.

Images