Rice Seed Vigor Detection Method Based on Multispectral Image Analysis

Abstract

This invention discloses a method for detecting rice seed vigor based on multispectral image analysis. The method includes collecting seeds of the same rice variety with different vigor levels; simultaneously acquiring spectral and image information of the rice seeds using a multispectral imaging system in a spectral image acquisition room to obtain spectral images; conducting a standard germination experiment on each seed to detect its vigor; removing the background from the spectral images and performing normalized standard discriminant analysis, using different visual markers to mark regions of interest, and generating a visual analysis image as a vigor prediction model; using the vigor prediction model to predict the vigor of seeds of the same variety, and evaluating the prediction effect in conjunction with the results of the standard germination experiment. This invention establishes a vigor prediction model with quantitative detection and analysis by combining the spectral images of rice seeds acquired by a multispectral imaging system with the results of standard germination experiments. Using this vigor prediction model, highly accurate quantitative indicators and non-destructive detection of rice seeds can be achieved.

Description

Technical Field

[0001] This invention relates to the field of rice seed detection, and in particular to a method for detecting rice seed vigor based on multispectral image analysis. Background Technology

[0002] Rice seed vigor is one of the important indicators for evaluating seed quality, and it is a comprehensive reflection of seed germination rate and growth. High-vigor rice seeds have significant growth advantages and production potential, which is of great importance to seed management and agricultural production. Rice seed vigor is highest at physiological maturity, and it gradually decreases as the storage time increases due to natural and irreversible seed aging.

[0003] Existing methods for detecting rice seed vigor typically rely on enzyme activity assays, adenosine triphosphate (ATP) content assays, seedling growth assays, germination rate assays, and hyperspectral imaging techniques. However, these methods are complex, time-consuming, and can damage rice seeds, failing to simultaneously satisfy both non-destructive and quantitative detection requirements.

[0004] Therefore, how to provide a method for detecting rice seed vigor that combines non-destructive testing with high-accuracy quantitative detection is a technical problem that urgently needs to be solved. Summary of the Invention

[0005] The main objective of this invention is to provide a method for detecting rice seed vigor based on multispectral image analysis, which aims to solve the technical problems of current rice seed vigor detection methods being complex, time-consuming, and damaging to rice seeds, and failing to simultaneously meet the requirements of non-destructive testing and quantitative detection.

[0006] To achieve the above objectives, this invention provides a method for detecting rice seed vigor based on multispectral image analysis, the method comprising the following steps:

[0007] S1: Collect seeds of the same rice variety with different vigor levels;

[0008] S2: In the spectral image acquisition room, rice seeds are evenly arranged in a petri dish, and the petri dish is placed on the sampling stage of the multispectral imaging system. The multispectral imaging system is used to simultaneously acquire the spectral and image information of the rice seeds to obtain a spectral image.

[0009] S3: Perform a standard germination test on each seed to check its viability;

[0010] S4: Remove the background from the spectral image, perform normalized standard discriminant analysis, use different visual markers to mark the region of interest, and generate a visual analysis image as a vitality prediction model;

[0011] S5: Use the vigor prediction model to predict the seed vigor of the same variety, and evaluate the prediction effect in combination with the results of the standard germination experiment.

[0012] Optionally, seeds of the same rice variety with different vigor levels can be collected. Specifically, this includes collecting seeds of the same rice variety and performing different artificial accelerated aging treatments to obtain seeds of the same rice variety with different vigor levels.

[0013] Optionally, seeds of the same rice variety can be collected and subjected to different artificial accelerated aging treatments, including: placing rice seeds in a single layer in a sealed bag; aging in a 42°C water bath for 5, 10, and 20 days; and after the aging treatment, placing them at 25°C for one week to allow the moisture content to return to the level before the aging treatment.

[0014] Optionally, before collecting the spectral and image information of rice seeds, the method further includes:

[0015] The multispectral imaging system was calibrated, and the light source was set according to the sample characteristics of rice seeds.

[0016] The calibration includes: using a white plate for reflectivity calibration, using a black plate for background calibration, and using a circular dot plate for geometric pixel position calibration.

[0017] Optionally, a standard germination experiment can be performed on each seed to detect seed vigor. Specifically, the seeds are placed in petri dishes containing two sheets of moistened filter paper, placed in an incubator, and cultured for 10 days at 28°C with 14 hours of light and 25°C with 10 hours of darkness. The germination standard is defined as a radicle length of ≥2 mm, and the germination status is recorded as seed vigor.

[0018] Optionally, the background of the spectral image is removed, normalized standard discriminant analysis is performed, different visual markers are used to mark the region of interest, and a visual analysis image is generated as a vitality prediction model, specifically including:

[0019] Based on the seed vigor detection, the background of several multispectral images with different germination rates was removed, and the target area image where the rice seed is located was retained, thus completing the image background homogenization.

[0020] Normalized standard discriminant analysis was performed on several multispectral images. Based on the characteristics of seeds in different multispectral images, different visual markers were used to mark the regions of interest. The data results were then transformed into visual analysis images as a viability prediction model.

[0021] Optionally, seed vigor prediction models can be used to predict the vigor of the same variety, specifically including:

[0022] The spectral images of the sample to be tested are acquired using a multispectral imaging system;

[0023] Remove the background from the spectral image of the sample to be tested, and use the viability prediction model to determine the visual markers in the spectral image of the sample to be tested.

[0024] Based on the judgment results, the seed vigor of the sample to be tested is obtained.

[0025] Optionally, the visual marker is a color.

[0026] Optionally, the visual marker is identification information of color pixel values.

[0027] Optionally, the method further includes:

[0028] Based on the three color parameters of the color pixel value, a three-dimensional model is established, and several points corresponding to the color pixel values ​​of seeds with different germination rates are marked in the three-dimensional model.

[0029] A seed vitality characteristic curve is generated based on fitting several points with respect to color pixel values; wherein, the characteristic curve has several curve segments, and each curve segment corresponds to a range of seed vitality values;

[0030] Using the color pixel values ​​in the spectral image of the sample to be tested and the results of the standard germination experiment, the points in the characterization curve are corrected to generate a corrected characterization curve;

[0031] The modified characterization curves were used to predict seed vigor of the same variety.

[0032] The beneficial effects of this invention are as follows: It proposes a method for detecting rice seed vigor based on multispectral image analysis. The method includes collecting seeds of the same rice variety with different vigor levels; in a spectral image acquisition room, rice seeds are evenly arranged in a petri dish, and the petri dish is placed on the sampling stage of a multispectral imaging system. The multispectral imaging system simultaneously acquires the spectral and image information of the rice seeds to obtain a spectral image; a standard germination experiment is performed on each seed to detect its vigor; the background of the spectral image is removed, and normalized standard discriminant analysis is performed. Different visual markers are used to mark the region of interest, generating a visual analysis image as a vigor prediction model; the vigor prediction model is used to predict the vigor of seeds of the same variety, and the prediction effect is evaluated by combining the results of the standard germination experiment. This invention establishes a vigor prediction model with quantitative detection and analysis by combining the spectral images of rice seeds acquired by a multispectral imaging system with the results of the standard germination experiment. Using this vigor prediction model, highly accurate quantitative indicators and non-destructive detection of rice seeds can be achieved. Attached Figure Description

[0033] Figure 1This is a schematic flowchart of an embodiment of the rice seed vigor detection method based on multispectral image analysis of the present invention;

[0034] Figure 2 These are multispectral images acquired in this invention;

[0035] Figure 3 This is an image with a uniform background in this invention;

[0036] Figure 4 This is a visual representation of rice seeds after aging treatment for 0, 5, 10, and 20 days in this invention.

[0037] Figure 5 This is a schematic diagram of the characterization curve plotted by fitting RGB color values ​​in this invention;

[0038] Figure 6 This is a schematic diagram illustrating the optimization of the characterization curve using newly added points in this invention.

[0039] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0041] This invention provides a method for detecting rice seed vigor based on multispectral image analysis, referring to... Figure 1 , Figure 1 This is a schematic flowchart of an embodiment of the rice seed vigor detection method based on multispectral image analysis of the present invention.

[0042] In this embodiment, a method for detecting rice seed vigor based on multispectral image analysis is provided, the method comprising the following steps:

[0043] S1: Collect seeds of the same rice variety with different vigor levels;

[0044] S2: In the spectral image acquisition room, rice seeds are evenly arranged in a petri dish, and the petri dish is placed on the sampling stage of the multispectral imaging system. The multispectral imaging system is used to simultaneously acquire the spectral and image information of the rice seeds to obtain a spectral image.

[0045] S3: Perform a standard germination test on each seed to check its viability;

[0046] S4: Remove the background from the spectral image, perform normalized standard discriminant analysis, use different visual markers to mark the region of interest, and generate a visual analysis image as a vitality prediction model;

[0047] S5: Use the vigor prediction model to predict the seed vigor of the same variety, and evaluate the prediction effect in combination with the results of the standard germination experiment.

[0048] In a preferred embodiment, collecting seeds of the same rice variety with different vigor levels specifically includes: collecting seeds of the same rice variety and performing different artificial accelerated aging treatments to obtain seeds of the same rice variety with different vigor levels.

[0049] In a preferred embodiment, seeds of the same rice variety are collected and subjected to different artificial accelerated aging treatments, specifically including: placing rice seeds in a single layer in a sealed bag; aging in a 42°C water bath for 5 days, 10 days, and 20 days; and after the aging treatment, placing them at 25°C for one week to allow the moisture content to return to the level before the aging treatment.

[0050] In a preferred embodiment, before acquiring the spectral and image information of rice seeds, the method further includes: calibrating the multispectral imaging system and setting the light source according to the sample characteristics of the rice seeds; wherein the calibration includes: using a white plate for reflectivity calibration, using a black plate for background calibration, and using a circular spot plate for geometric pixel position calibration.

[0051] In a preferred embodiment, the standard germination experiment is performed on each seed to detect seed vigor. Specifically, the steps include: placing the seeds in petri dishes containing two sheets of moistened filter paper, placing them in an incubator, and culturing them for 10 days at 28°C for 14 hours of light and 25°C for 10 hours of darkness. The germination standard is defined as a radicle length of ≥2mm, and the germination status is statistically analyzed to determine the seed vigor.

[0052] In a preferred embodiment, the background of the spectral image is removed, and normalized standard discriminant analysis is performed. Different visual markers are used to mark the region of interest, and a visual analysis image is generated as a viability prediction model. Specifically, this includes: removing the background from several multispectral images with different germination rates according to the detected seed viability, retaining the target area image where the rice seed is located, and completing the image background uniformization; performing normalized standard discriminant analysis on several multispectral images, and using different visual markers to mark the region of interest according to the characteristics of the seeds in different multispectral images, and converting the data results into a visual analysis image as a viability prediction model.

[0053] In a preferred embodiment, the seed vigor prediction model is used to predict the seed vigor of the same variety, specifically including: acquiring the spectral image of the sample to be tested using a multispectral image system; removing the background of the spectral image of the sample to be tested; using the vigor prediction model to judge the visual markers in the spectral image of the sample to be tested; and obtaining the seed vigor of the sample to be tested based on the judgment result.

[0054] In one embodiment, the visual marker is a color.

[0055] In another embodiment, the visual marker is identification information of color pixel values. Based on this, the method further includes: establishing a three-dimensional model based on three color parameters of the color pixel values; marking several points in the three-dimensional model corresponding to the color pixel values ​​of seeds with different germination rates; fitting a characterization curve of seed vigor with respect to the color pixel values ​​based on the several points; wherein the characterization curve has several curve segments, each curve segment corresponding to a range of seed vigor values; correcting the points in the characterization curve using the color pixel values ​​in the spectral image of the sample to be tested and the results of a standard germination experiment, generating a corrected characterization curve; and using the corrected characterization curve to predict the seed vigor of the same variety.

[0056] In this embodiment, a method for detecting rice seed vigor based on multispectral image analysis is provided. A vigor prediction model with quantitative detection and analysis is established by using the spectral images of rice seeds acquired by a multispectral imaging system and the results of standard germination experiments. This vigor prediction model can be used to achieve highly accurate quantitative indicators and non-destructive detection of rice seeds.

[0057] To explain this application more clearly, the following provides a specific example of the rice seed vigor detection method based on multispectral image analysis:

[0058] (1) Collecting rice seeds

[0059] Collect seeds of the same rice variety with different vigor levels (or obtain seeds with different vigor gradients using artificial accelerated aging treatment). The specific method for artificial accelerated aging treatment is as follows: place rice seeds in a single layer in a sealed bag and age them in a 42℃ water bath for 5, 10, and 20 days. After aging treatment, place them at 25℃ for one week to allow the moisture content to return to pre-aging levels before acquiring spectral image information. (The aging time and method can be adjusted according to the seed type.)

[0060] (2) Acquiring spectral images

[0061] The spectral and image information of rice seeds was simultaneously acquired using the Videoometer Lab4™ multispectral imaging system (Videometer A / S, Herlev, Denmark). Specifically, before acquiring spectral image information, the system was warmed up for approximately 30 minutes for system initialization. Then, calibration was performed using three plates: a white plate for reflectivity calibration, a black plate for background calibration, and a circular spot plate for geometric pixel position calibration. The light source was set according to the sample characteristics.

[0062] After completing the system parameter settings, evenly arrange the rice seeds in a 9cm petri dish (50 seeds / dish). Place the petri dish on the sampling stage at the bottom of the integrating sphere of the multispectral imaging system to acquire spectral image information, obtaining 20 high-resolution spectral images of the seeds at 2192 × 2192 pixels (19 spectral images of different bands and 1 RGB image). Figure 2 As shown. During image acquisition, the sphere closes on the sampling stage, creating optical closure conditions.

[0063] (3) Detecting seed vigor

[0064] After aging treatment, rice seeds were subjected to seed vigor testing according to the national standard GB / T 3543.4-1995 to assess seed vigor. Specifically, after acquiring multispectral images, the seeds were sequentially placed in petri dishes containing two sheets of moistened filter paper and cultured in an incubator (28℃ for 14 hours under light and 25℃ for 10 hours in darkness) for 10 days. Germination was assessed based on a radicle length ≥2mm, and seed germination was statistically analyzed.

[0065] (4) Establish an nCDA analysis model

[0066] An nCDA analysis model was established using the MSI-Transformation Builder in VideoometerLab4 software. Specifically, after acquiring multispectral images, the background segmentation method was established using MSI-Segmentation Builder. Backgrounds were removed from two multispectral images showing high viability (100% germination rate) and no viability (0% germination rate), or from multiple multispectral images showing different viability (germination rates ranging from 0% to 100%). The target region (rice seeds) was preserved, thus achieving image background homogenization. Figure 3 As shown.

[0067] The two groups of seeds were then labeled, and nCDA analysis was performed using multispectral imaging system software. Based on the characteristics of different seeds, different visual markers were used to mark the regions of interest (ROIs). The data results were converted into visual analysis images, obtaining blue and red nCDA images. The model was saved and can be directly used to predict seed vigor for this variety. Simultaneously, correlation analysis can be performed between the obtained nCDA images and seed vigor indices (germination rate, vigor index, etc.) to achieve non-destructive prediction of rice seed vigor. It should be noted that in practical applications, the ROI can be a region in the spectral image that meets specific characteristics, thereby highlighting that region in the seed using visual markers.

[0068] (5) Using the nCDA analysis model

[0069] The vigor prediction model was used to predict the vigor of seeds of the same variety, and the prediction effect was evaluated in conjunction with the results of a standard germination experiment. Specifically, a multispectral imaging system was used to acquire the spectral and image information of the samples to be tested. After removing the background, the established model was used to directly predict vigor. At this point, seeds with different vigor levels were represented by different colors. As the germination rate decreased from 100% to 0%, the color gradually changed from blue to red, and then sequentially represented as: blue, green, yellow, and red. Vigor could then be distinguished based on color. Figure 4 As shown.

[0070] In another specific example, visual markers are identified by color pixel values, such as the three color pixel values ​​of RGB. As seen in the example above, the color changes according to a certain pattern as the germination rate decreases from 100% to 0%. Therefore, a three-dimensional model can be built using the three color pixel values ​​of RGB. Several points corresponding to the color pixel values ​​of seeds with different germination rates are marked in the three-dimensional model. These points are then used to fit and generate a seed vigor characteristic curve with respect to the color pixel values. The characteristic curve has several curve segments, each corresponding to a range of seed vigor values. Subsequently, the points in the characteristic curve are corrected using the color pixel values ​​from the spectral image of the sample to be tested and the results of a standard germination experiment, generating a corrected characteristic curve. The corrected characteristic curve is then used to predict the seed vigor of the same variety.

[0071] For example, such as Figure 5As shown, in the initial stage, multispectral images were acquired after performing 10 different aging treatments on seeds of the same variety for different durations to obtain 10 RGB images with different germination rates (viability values). Subsequently, for each RGB image with different viability values, the R, G, and B values ​​of each seed in the RGB image were extracted. After removing bad data, the average value of each color value was accumulated and used as the identification information of the seed with that viability value. In a three-dimensional coordinate system established by the three RGB color values, the identification information of the 10 different viability values ​​was used as coordinate points to perform curve fitting and plotting, resulting in 9 curve segments. Each curve segment represents the range between the viability values ​​corresponding to two adjacent points. For example, the germination rate is between 100% and 90% between the first point (24, 10, 228) and the second point (30, 62, 236), and between the second point (30, 62, 236) and the third point (22, 163, 225), and so on.

[0072] like Figure 6 As shown, after obtaining the initial curve plotted from 10 points, in actual testing, the points corresponding to the identifiers in the spectral images of the seeds to be tested cannot always coincide with the previously fitted 10 points. This means the actual vigor value of the seed cannot be determined, only an approximate range is given, and the accuracy of the initial curve obtained is low. Therefore, it is necessary to continuously optimize the fitted curve using subsequent spectral images and standard germination experiment results. For example, between the second and third points, the curve segment is optimized using newly obtained points (25, 121, 230) to obtain two germination rate ranges: 90%-88% and 88%-83%, thereby improving the accuracy of seed vigor value detection. Simultaneously, when the point of the seed to be tested does not fall into any curve segment, its vigor value is estimated by the distance from the point of the seed to the key point in the curve segment. Through continuous optimization of the curve, the accuracy of this estimation process can be improved, ultimately generating a more accurate vigor value or vigor value range for the seed to be tested.

[0073] It should be noted that, since the material content of different varieties of seeds usually varies, the correspondence between seed vigor values ​​and color values ​​is not completely the same when represented in multispectral images. Therefore, when fitting and drawing characterization curves, different varieties of seeds have completely different and unique characterization curves. Using the characterization curves corresponding to different varieties of seeds to detect the vigor values ​​of the corresponding varieties of seeds has high accuracy and reliability.

[0074] In this example, considering the insufficient amount of data used to establish the vigor prediction model, a characterization curve of seed vigor with respect to color pixel values ​​can be established. During subsequent detection, the characterization curve can be continuously corrected using the detected spectral images and standard germination experiment results to make it closer to the true curve. Once the characterization curve is sufficiently accurate, it can be compared with the color pixel values ​​of the test sample to determine which segment of the characterization curve the color pixel value is located in or is closest to. The seed vigor value corresponding to that segment is then used as the seed vigor detection value of the test sample.

[0075] It is understood that in the description of this specification, references to terms such as "one embodiment," "another embodiment," "other embodiments," or "first embodiment to Nth embodiment," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0076] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0077] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for detecting rice seed vigor based on multispectral image analysis, characterized in that, The method includes the following steps: S1: Collect seeds of the same rice variety with different vigor levels; S2: In the spectral image acquisition room, rice seeds are evenly arranged in a petri dish, and the petri dish is placed on the sampling stage of the multispectral imaging system. The multispectral imaging system is used to simultaneously acquire the spectral and image information of the rice seeds to obtain a spectral image. S3: Perform a standard germination test on each seed to check its viability; S4: Based on the seed vigor detection, remove the background from several multispectral images with different germination rates, retain the target area image where the rice seeds are located, and complete the image background homogenization; perform normalization standard discrimination analysis on several multispectral images, and use different visual markers to mark the regions of interest based on the characteristics of the seeds in different multispectral images, and transform the data results into a visual analysis image as a vigor prediction model. S5: Acquire spectral images of the sample to be tested using a multispectral imaging system; remove the background from the spectral images of the sample to be tested; use a vigor prediction model to judge the visual markers in the spectral images of the sample to be tested; obtain the seed vigor of the sample to be tested based on the judgment results; evaluate the prediction effect by combining the results of the standard germination experiment. Wherein, the visual marker is identification information of RGB color pixel values; the method further includes: Based on the three color parameters of RGB color pixel values, a three-dimensional model is established, and several points corresponding to the color pixel values ​​of seeds with different germination rates are marked in the three-dimensional model. A seed vitality characterization curve is generated based on fitting several points with respect to RGB color pixel values; wherein, the characterization curve has several curve segments, and each curve segment corresponds to a range of seed vitality values; Using the RGB color pixel values ​​in the spectral image of the sample to be tested and the results of the standard germination experiment, the points in the characterization curve are corrected to generate a corrected characterization curve. The modified characterization curves were used to predict seed vigor of the same variety.

2. The rice seed vigor detection method based on multispectral image analysis as described in claim 1, characterized in that, Collecting seeds of the same rice variety with different vigor levels involves: collecting seeds of the same rice variety and subjecting them to different artificial accelerated aging treatments to obtain seeds of the same rice variety with different vigor levels.

3. The rice seed vigor detection method based on multispectral image analysis as described in claim 2, characterized in that, Seeds of the same rice variety were collected and subjected to different artificial accelerated aging treatments, including: placing rice seeds in a single layer in a sealed bag; aging in a 42℃ water bath for 5, 10, and 20 days; and after the aging treatment, placing them at 25℃ for one week to allow the moisture content to return to the level before the aging treatment.

4. The rice seed vigor detection method based on multispectral image analysis as described in claim 1, characterized in that, Before collecting spectral and image information of rice seeds, the method further includes: The multispectral imaging system was calibrated, and the light source was set according to the sample characteristics of rice seeds. The calibration includes: using a white plate for reflectivity calibration, using a black plate for background calibration, and using a circular dot plate for geometric pixel position calibration.

5. The method for detecting rice seed vigor based on multispectral image analysis as described in claim 1, characterized in that, Each seed was subjected to a standard germination experiment to test its viability. The steps included: placing the seeds in petri dishes containing two sheets of moistened filter paper, placing them in an incubator, and incubating them for 10 days at 28°C with 14 hours of light and 25°C with 10 hours of darkness. The germination standard was defined as a radicle length of ≥2mm. The germination status of the seeds was then recorded as the seed viability.

Images