Method and device for predicting the germination capacity of seeds

Abstract

A method for predicting the germination capacity of seeds (10) is described, comprising: providing unsprouted seeds (10); irradiating the unsprouted seeds (10) with a radiation source (11) with an initial intensity I0; detecting a reduced transmittance I transmitted after partial absorption of the initial radiation I0 trans depending on at least one angle; recording at least one further spectral physical quantity (20) by means of a spectral measurement; and subsequent evaluation of a structure of the recorded transmitted attenuated transmit intensities I trans and the relationships of these structures to each other and the at least one further spectral physical quantity recorded, in order to predict a probability of the germination capacity of the unsprouted seed (10). Furthermore, a device (100) for predicting the germination capacity of seed (10) is described.

Description

[0001] The present invention relates to a method and a device for predicting the germination capacity of seeds, in particular by means of X-ray technology.

[0002] To assess the germination capacity of seeds, they are regularly germinated and the percentage of a given quantity of seeds that remains viable is evaluated. However, this requires germinating seeds each time, after which they are no longer available.

[0003] In the case of optical evaluation, the number of germinated seeds is recorded manually or using automated methods at specific intervals. The overall germination rate is then derived from this data. Additionally, methods using computed tomography exist that characterize germination in special planting containers. However, with all of the aforementioned methods, the seeds are no longer available afterward, meaning that each germination test reduces the stored seed quantity or destroys stored seeds through germination. Currently, assessing germination capacity using methods known from the state of the art is only possible through destructive techniques.

[0004] Especially in seed banks, determining the germination capacity of seeds is of great importance, as many different genotypes must be permanently available and capable of germination there.

[0005] One object of the present invention is therefore to provide a non-destructive method and a non-destructive device for predicting the germination capacity of seeds, in particular with which it is possible to indicate a probability of germination without causing the seeds to germinate.

[0006] This problem is solved by a method for predicting the germination capacity of seeds according to claim 1 and a device for predicting the germination capacity of seeds according to claim 6.

[0007] The proposed method for predicting seed germination capacity comprises providing unsprouted seed; irradiating the unsprouted seed with a radiation source with an initial intensity I0; and detecting a reduced transmittance I transmitted after partial absorption of the initial radiation I0. transdepending on at least one angle; the acquisition of at least one further spectral physical quantity by means of a spectral measurement; and subsequent evaluation of a structure of the acquired transmitted attenuated transmission intensities I trans and the ratios of these structures to each other and at least one other spectral physical quantity detected, in order to predict the probability of germination of the ungerminated seed. At least one structure can be detected per seed of seed 10, so that the ratios of the structures to each other can be related. For example, the transmitted attenuated transmission intensity I can be transDepending on at least one angle, at least one grayscale value can be recorded. This recording can be achieved using a radiation detector. In particular, many different grayscale values ​​can be recorded, forming a plurality of structures. The proposed method allows for the analysis of this at least one grayscale value by recording the transmitted attenuated transmission intensity I. trans Depending on at least one angle and by measuring at least one other spectral physical quantity, the germination capacity of ungerminated seeds can be analyzed two-dimensionally or three-dimensionally—in the case of measurements taken at several different angles. The wavelength(s) of the radiation source is / are known.

[0008] Another aspect of the present invention relates to a device for predicting the germination capacity of seeds, wherein the device comprises a radiation source for irradiating the ungerminated seeds with an initial intensity I0. The device also comprises a radiation detector for detecting a reduced transmission intensity I transmitted after partial absorption of the initial radiation I0. trans depending on at least one angle; a further spectral measuring device for detecting at least one further spectral physical quantity by means of a spectral measurement; and wherein the device is equipped with an evaluation unit for subsequently evaluating a structure of the detected transmitted attenuated transmittance I transand the relationships between these structures and the detected at least one further spectral physical quantity can be connected or linked to predict the probability of germination of the unsprouted seed. The proposed device can be used to carry out the proposed method, or the proposed method can be carried out with the proposed device.

[0009] When radiation, such as X-rays, penetrates an object like seeds, the radiation is attenuated according to Lambert-Beer's law. This attenuation depends on the radiation energy (e.g., X-ray energy), the path length, and the density and atomic number of the material being penetrated. If these parameters (material, energy) are known, the product of length and density can be determined by measuring the transmitted intensity along the radiation path behind the object. The attenuated intensity, depending on the angle of the recorded projection, can then correspond to grayscale information from a calibrated detector, such as an X-ray detector. Furthermore, by varying the angle during the acquisition of numerous individual projections, a complete 3D reconstruction can be generated.Since plant components differ only very slightly in their physical structure, it is possible to determine the biomass and water content of the seed, as well as its structural composition, from a radiographic image, particularly one that can be generated, for example, by different grayscale values ​​from a detector, or from a 3D reconstruction produced by a line or area detector. The inventors have thus discovered how the ratios of the grayscale values ​​in the seed, or its shape / structure, must be to predict germination.

[0010] The technical method proposed herein uses the analysis of radiation attenuation, such as X-rays, to predict the germination rate of seeds based on their internal structure and / or the homogeneity of their individual components. This can be achieved using trained machine learning models or statistical methods such as principal component analysis (PCA) or similar techniques. Additional modalities, such as color data from a camera or multi- or hyperspectral data, can also be employed. This allows for the non-destructive assignment of a germination probability to seeds without damaging them.

[0011] Preferred embodiments of the present invention are explained in detail below with reference to the accompanying drawings. These show: Fig. 1 a schematic view of a device for predicting the germination capacity of seeds; Fig. 2. A flowchart of a procedure for predicting the germination capacity of seeds; Fig. 3a, b a view of a device for predicting the germination capacity of seeds ( Fig. 3a) and a monitor which displays a recording of the device ( Fig. 3b) shows; Fig. 4 Another view of a device for predicting the germination capacity of seeds; Fig. 5 a recording of unsprouted seed which has been recorded with the device for predicting seed germination capacity, and Fig. 6. a further recording of unsprouted seed which was recorded with the device for predicting seed germination capacity; Fig. 7. A photograph of unsprouted maize kernels of genotype A, which was taken with the device for predicting the germination capacity of seeds; Fig. 8 a recording of unsprouted maize kernels of genotype B, which was recorded with the device for predicting the germination capacity of seeds; Fig. 9 a recording of unsprouted maize kernels of genotype A, which was recorded with the device for predicting the germination capacity of seed; Fig. 10. A PCA evaluation of the three genotypes A, B, and C from the Fig. 7 to 9 based on two main components; Fig. 11. A further PCA evaluation of the three genotypes A, B and C from the Fig. 7 to 9 based on two main components; Fig. 12 a radioscopy image of a kernel of corn from an angle; Fig. 13 a radioscopy image of another kernel of corn from a different angle; and Fig. 14 a CT image of a corn kernel.

[0012] Individual aspects of the invention described herein are set forth below. Fig. Figures 1 to 6 are described. In the present application, identical reference numerals refer to identical or equivalent elements, and it is not necessary for all reference numerals to be repeated in all drawings.

[0013] When a component is described herein as "designed to do something", it means that the component has been designed structurally and physically in such a way that the component performs what it is supposed to do.

[0014] Fig. 1 branches off a device 100 for predicting the germination capacity of seeds 10 and Fig. Figure 2 shows a flowchart of a procedure for predicting the germination capacity of seeds. Fig. 3a and Fig. Figure 4 each shows a device 100 for predicting the germination capacity of seeds 10 and Fig. Figure 3b shows a display device, such as a monitor 110, which displays measurement results of the ungerminated seed 10, the measurement results being recorded with the device 100 and the proposed method. Fig. 5 and Fig. Figure 6 shows images of the irradiated unsprouted seed 10, which were taken using the proposed method and the proposed device 100. Fig. Figures 7 to 9 show images of irradiated, ungerminated corn kernels of different genotypes A, B and C. Fig. 10 and Fig. Figure 11 shows two examples of a PCA analysis based on two main components of PCA analysis. In summary, the Fig. The principle of the present invention is explained below in sections 1 to 6.

[0015] As in Fig. As shown in Figure 2, the method 200 for predicting the germination capacity of seed 10 comprises, in step 210, providing unsprouted seed 10; in step 220, irradiating the unsprouted seed 10 with a radiation source 11 with an initial intensity I0. The radiation source can be an X-ray source, which, for example, is one of the types shown in Figure 2. Fig. 5 and Fig. The images shown in Figure 6 of the irradiated seed 10 lead to the conclusion that the method 200 comprises, in step 230, the detection of a reduced transmission intensity I transmitted after partial absorption of the initial radiation I0. trans depending on at least one angle. The transmittance intensity I trans is detected using a radiation detector 60. Preferably, the transmission intensity I is measured. trans depending on a large number of different angles, with each recorded measurement of the transmittance intensity I transDepending on an angle, a gray value information is obtained at the radiation detector 60. By acquiring a multitude of different angles, a multitude of gray value information is acquired, particularly by the radiation detector 60. This multitude of gray value information can be analyzed; that is, the structure of the gray value information is analyzed. In step 240, the method 200 includes acquiring at least one further spectral physical quantity 20 by means of a spectral measurement. A multitude of spectral physical quantities 20 can be acquired, which can be detected with different spectral measurements. Finally, in step 250, the method 200 includes a statistical evaluation of the acquired transmitted attenuated transmission intensities I. transand at least one other detected spectral physical quantity 20 to predict the probability of germination of the unsprouted seed 10. When carrying out the procedure 200, the wavelength or wavelengths of the radiation source 11 are known. The seed 10 can be irradiated with a variety of different wavelengths. The light from the radiation source can be polychromatic, so that the seed is illuminated with a variety of different wavelengths at once. However, the radiation source 11 can also be set to emit monochromatic light for illuminating the seed 10. In this case, the seed 10 can, for example, be irradiated with a variety of different wavelengths successively. To convert a polychromatic light source 10 into a monochromatic light source 10, filters can be used, for example.Steps 210, 220, and 230 are performed in this order. Step 240, however, can be performed before, during, or after any of steps 210, 220, or 230. Step 240 is therefore independent of steps 210, 220, and 230. Preferably, however, step 240 is performed after positioning the seed 10, so that steps 210 to 240 are completed with a single positioning of the seed 10 in a sample chamber. The spectral measurement must, however, be attributable to the seed in the transmission measurement. A spectral measurement can be attributed to the corresponding transmitted seed if the position of the transmitted seed on the X-ray detector is known and the spectral measurement is performed on the same seed or at the same position of the seed.In particular, a spectral measurement can be clearly assigned to the corresponding seed examined using X-ray technology if the measurements in steps 210, 220, and 230 are performed with spatial resolution and the seeds are not repositioned, or if the measurements in steps 210, 220, and 230 are performed simultaneously. Step 250 is performed last, after all measurement results for seed 10 have been recorded.

[0016] Preferably, the method comprises 200 steps after detecting the transmitted attenuated transmission intensity I trans Determining, in particular measuring, a product µ*d from a traversed transmission length d of the transmitted attenuated transmission intensity I trans and an absorption coefficient for biological material µ of the seed 10 according to the Lambert-Beer law I trans= I0*exp(-µd), where the absorption coefficient µ for biological material, i.e., for seed 10, is almost always the same. The absorption coefficient µ for biological material is essentially constant. Consequently, the product µ*d corresponds to the biomass of the illuminated seed 10. In other words, the attenuated radiation corresponds to the natural logarithm of the quotient I. trans / I0, i.e. µ*d = In (I trans / I0). In principle, the attenuated radiation I trans already the product of transmission length and absorption coefficient, since I trans=I0*exp(-µd) and µ and d are the respective transmission length and essentially constant absorption coefficients, respectively, as already described in DE10 2015 218 504 A1. DE10 2015 218 504 A1 describes a method for determining biomass, which can be used here to determine the absorption coefficient µ of the seed 10. If the absorption coefficient of the seed 10 is not known, it can be determined in step 230 of method 200 by, for example, determining the volume of the seed 10 by means of transmission at several different angles (i.e., performing a CT scan of the seed 10) and determining the mass of the seed 10 using a scale. By determining the product µ*d, the weight of the seed 10 is essentially determined. By measuring the volume of the seed 10, the density of the seed 10 can finally be deduced.If the volume and mass of the seed are known, particularly averaged to, for example, 1000 g seed weight, the absorption coefficient of the seed can be determined. For biological materials, one element can be assumed, such as hydrocarbons. The attenuation is then directly proportional to the path length d. Either the path length d can be approximated / measured using an imaging technique, or the absorption coefficient can be used as a priori information based on the 1000 g seed weight. In the simplest case, the seed can be assumed to be a cylinder in the detector image to determine its volume. The path length d, also called the transmission length d, is defined as the path length of a seed of the seed 10 that is traversed, for example, by the radiation source 10.

[0017] Preferably, the method 200 comprises acquiring a plurality of projections of the transmitted attenuated transmission intensity I trans depending on a multitude of different angles between a detection point of the transmitted attenuated transmission intensity I trans and a position of the seed 10, in order to perform a two-dimensional or three-dimensional detection of the seed 10 by means of the angle-dependent detected transmitted attenuated transmission intensities I trans to enable. The detection point of the transmitted attenuated transmission intensity I transThis corresponds to a detector surface 62 of the detector 60. The position of the seed 10 lies between the radiation source 11 and the detector 60 on an imaginary connecting line between the radiation source 11 and the detector 60. In particular, the position of the seed 10 defines a distance D from the detector. The detector 60 is arranged, in particular, perpendicular to the radiation source 11. The detector has, in particular, an extent perpendicular to the radiation source 11, such that the detector detects the attenuated transmission intensity I trans Depending on the detection location on detector 60, it records at a different angle α1, α2 in relation to the imaginary connecting line. Fig. Figure 1 shows, for example, two different angles α1 and α2, which are recorded in step 230. Preferably, an angular range of 360° is recorded, which in particular enables three-dimensional recording of the seed 10. With an angular range of 180° and an opening angle of the radiation source 11, for example an X-ray source 10, the throughput can be increased. With a scanned angular range of 180°, a two-dimensional recording of the seed 10 can in particular be achieved.

[0018] Preferably, the method includes an evaluation of a structure of the detected transmitted attenuated transmission intensity I transand the relationships of these structures to each other and the recorded at least one further spectral physical quantity 20 by using a classical algorithm and / or a machine learning method and / or principal component analysis, which is designed to use further known quantities of the investigated seed 10 for evaluation. Machine learning is an umbrella term for methods of evaluation using artificial intelligence. A neural network, for example, would be a specific method of machine learning. In the narrowest sense, even principal component analysis (PCA) would be an aspect of unsupervised learning. But Random Forest also falls under the term machine learning and can be used here.

[0019] In the Fig. 5 and Fig. 6. For example, the homogeneity or the differences (i.e., the inhomogeneity) between the individual grains can be used to determine the structure or structures of the detected transmitted attenuated transmission intensity I. trans to record. Fig. Figure 12 shows a radioscopy image of a single starchy corn kernel, i.e., a seed 10, from an angle; Fig. Figure 13 shows a radiographic image of a single other starchy corn kernel from a different angle. In the Fig. 12 and Fig. For example, 13 shows the different shades of gray and thus the homogeneity or inhomogeneity of two different may kernels. Fig. Figure 14 shows a CT image of the single corn kernel from Fig. 13. The CT image from Fig. Figure 14 is a cross-sectional image from a 3D reconstruction made from multiple radioscopy images taken from different angles of the corn kernel. The corn kernel is made of Fig. His embryo 12 and his endosperm 14 can be seen.

[0020] What contributes to germination, or does not, depends on the type of seed 10. Therefore, the proposed solution requires a predictive model for evaluating the structure of each type of seed 10. For example, in starch-containing seeds 10, such as maize, the differentiation between endosperm and embryo is crucial, which indicates a combination of attenuation and variance (also referred to here as inhomogeneity in the grayscale) and color as characteristics. The color can be determined using spectral measurement along with the other spectral physical quantity 20. In oil-containing seeds 10 (not shown), it is more the homogeneity in the grayscale and attenuation, as well as the recorded color information, that indicate germination capacity.For this reason, as proposed, for each type of seed 10, the evaluation of the recorded structures, PCA analysis, and / or another model are used to determine the main components for the respective seed. The analyzed main components of a seed 10 are used to predict its germination capacity. The proposed method can be applied to unknown seed 10. The evaluation of the structure of the recorded transmitted attenuated transmission intensities I. trans and the relationships of these structures to each other and the at least one further spectral physical quantity recorded, in order to predict a probability of the germination capacity of the unknown, unsprouted seed 10, includes creating a model using PCA and / or other models of a known seed. The created model can then be applied to unknown seed 10 to determine the germination capacity of the unknown seed.

[0021] The more inhomogeneous the grayscale values ​​recorded in the irradiated starch-containing seed 10, i.e., the more precisely the structures can be detected, the higher the probability that the irradiated seed 10 is viable. Conversely, the more homogeneous the grayscale values ​​recorded in the irradiated oil-containing seed 10, i.e., the less precisely the structures can be detected, the higher the probability that the irradiated seed 10 is viable. This purely X-ray-based classification of germination capacity depends on the resolution of the X-ray system and has limitations (e.g., damage to the seed from insects or severe desiccation). For this reason, the addition of at least one further modality is necessary (i.e., the acquisition of at least one further spectral physical quantity by means of a spectral measurement) in order to determine germination capacity non-destructively.

[0022] The acquisition of at least one further spectral physical quantity can include the acquisition of color data from a camera (32) and / or the acquisition of multispectral data and / or the acquisition of hyperspectral data and / or the acquisition of the mass of the irradiated ungerminated seed using a scale (40) and / or the acquisition of the height of the ungerminated seed 10 using LiDAR 50 (Light Detecting and Ranging), laser sectioning, or another optical measurement method to acquire a 3D morphology of the ungerminated seed 10. The height of a seed grain 10 can be determined using LiDAR (see Fig. 1) From the grain height measured using LiDAR (or another optical 3D measurement method), a transmission length d can then be derived. LiDAR measures the distance between a surface of the LiDAR device and the sensor. If a grain lies on the surface, the measured distance to the LiDAR sensor is reduced by the grain height. Thus, if the positions of the X-ray sensor and the source are known, the transmission length d, relevant for the X-ray radiation, can be calculated.

[0023] Fig. 1 and Fig. 3 ( Fig. 3a and Fig. 3b) and Fig. Figure 4 shows views of the proposed device 100 for predicting the germination capacity of seeds 10.

[0024] The proposed device 100 comprises a radiation source 11 for irradiating the ungerminated seed 10 with an initial intensity I0; a radiation detector 60 for detecting a reduced transmission intensity I transmitted after partial absorption of the initial radiation I0 trans depending on at least one angle α1, α2; a further spectral measuring device 30 for detecting at least one further spectral physical quantity by means of a spectral measurement; and wherein the device 100 is equipped with an evaluation unit 90 for subsequently evaluating the detected transmitted attenuated transmission intensity I trans and the recorded at least one further spectral physical quantity can be connected or linked to in order to predict a probability of the germination capacity of the unsprouted seed 10 (see Fig. 1) The radiation source 11 can be an X-ray source and / or a terahertz radiation source and / or an ultrasound radiation source and / or radar, with which the seed 10 can be irradiated. The radiation source 11 can emit rays in a first wavelength range λ1, which strikes the seed to be examined. For example, in the case of X-rays, the first wavelength range λ1 lies between 10 -9 m to 10 -11Since the radiation source 11 can also use other radiation ranges, as described herein, the first radiation range λ1 can be different. The at least one further spectral physical quantity can, for example, be determined by using LiDAR 50. The further spectral measuring device 30 can therefore be a LiDAR device 50. The seed 10 can be irradiated with wavelengths in the LiDAR range. Typical LiDAR wavelengths lie in a second wavelength range of λ2 between 905 nm and 1550 nm. Two different wavelength ranges λ1, λ2 can be used for both the radiation source 11 and the at least one further spectral physical quantity. The further spectral measuring device 30 can also be a camera 32 and / or a spectral analysis device 33. The further spectral measuring device 30 can also be a scale 40.The additional spectral measuring device 30 may use a different second wavelength range λ2 than the one specified for LiDAR. However, it is immediately and unambiguously apparent to a person skilled in the art in which wavelength range the respective spectral measuring devices 30 operate, which is why it is omitted here to specify them in detail. Fig. Reference numerals 32, 33, 40 and 50 are found in reference numerals 1, where the devices with reference numerals 32, 33, 40 and 50 are to be understood as optional. The device 100 necessarily includes at least one further spectral measuring device 30, i.e., one spectral measuring device 30 or several further spectral measuring devices 30 (=32, 33, 40 and / or 50).

[0025] Depending on the radiation source 11 used, the radiation detector 60 is preferably an X-ray detector and / or an ultrasound detector and / or a terahertz detector. It is conceivable that several different radiation sources 11 are used, so that several different radiation detectors must be used to detect the corresponding radiation. Fig. Figure 1 shows only a radiation detector 60 symbolically. However, it is understandable to an expert that for each additional spectral measuring device 30 a suitable detector is also available, for example a LiDAR detector, a detector for light in the visible range such as a camera 32, in order to detect the rays from the corresponding wavelength range.

[0026] The scale 40 itself is not in itself a spectral measuring instrument 30. For the sake of simplicity, however, the scale 40 is included here under another spectral measuring instrument 30.

[0027] Preferably, the radiation detector 60 is configured to measure the transmitted attenuated transmission intensity I trans depending on a multitude of angles α1, α2 as a multitude of gray value information between a detection point 62 of the transmitted attenuated transmission intensity I trans and to detect a position 13 of the seed 10 in order to detect the seed 10 in two dimensions or three dimensions using the angle-dependent detected transmitted attenuated transmission intensity I trans to enable. The detection location 62 corresponds to a detecting detector area 62 (see Fig. 1) The multitude of grayscale information forms the structure of the recorded transmitted attenuated transmission intensities I trans .

[0028] The radiation detector 60 can be a line detector or an area detector. With the area detector, two-dimensional or three-dimensional detection of the seed 10 is possible using the angle-dependent transmitted attenuated transmission intensity I. trans The line detector enables the acquisition / determination of measured values ​​in conveyor belt systems. This generates a 2D image for evaluation from the line detector and the feed of a conveyor belt.

[0029] Preferably, the further spectral measuring device 30 is designed to perform a spectral analysis of the seed 10 in order to determine a major component and / or several further components of the seed 10. Further components of the seed 10 can include, for example, the mean gray value of the seed 10, the maximum gray value of the seed 10, the minimum gray value of the seed 10, and / or the size of the seed, which can be determined by measuring gray value homogeneity and / or the intensity of additional spectral channels.

[0030] The additional spectral measuring device 30 can be a camera 32 for capturing color data of the seed 10 and / or a spectral analysis device 33 for capturing multispectral data and / or hyperspectral data of the seed 10. Multispectral data primarily refers to measured data acquired in the visible range (RGB) and / or in additional individual bands from the IR or UV range. Hyperspectral data refers to data acquired through a substantially continuous scanning of the long-wave infrared (FIR), short-wave infrared (SWIR), and RGB ranges up to the low UV.

[0031] Preferably, the evaluation unit 90 is configured to perform a classical algorithm and / or a machine learning method and / or a principal component analysis of the seed 10. The evaluation unit 90 is configured to calculate the size of a seed of the seed 10, the shape of the seed 10, the homogeneity or inhomogeneity of the recorded grayscale values ​​of the seed 10, as well as mean values ​​and standard deviations of the aforementioned quantities. The evaluation unit 90 may include a screen 110 to display the images of the seed 10, as shown, for example, in Fig. Figure 3b shows that the evaluation unit 90 can also include a computer connected to the screen 110. Alternatively, the evaluation unit 90 could be an external touchpad or an external smartphone, which can be connected to the device 100 to receive and subsequently evaluate the measurement results acquired by the device 100. Fig. 3a and Fig. Figure 4 each shows a device 100, wherein the evaluation unit 90 is shown according to Fig. 3a, Fig. 3b and Fig. 4 is not part of the device 100, while after Fig. 1. The evaluation unit 90 is part of the device 100. In Fig. Figure 1 shows screen 110 with evaluation unit 90 in a rectangle, where screen 110 can be external or internal to evaluation unit 90. Both options are shown in Fig. 1, Fig. 3a, Fig. 3b, and Fig. The 4 options shown are possible. Fig. 3b in conjunction with Fig. Figure 3a indicates that the device 100 is connected to an evaluation unit 90, namely a computer with a screen 110, to enable evaluation. Evaluation includes performing the classical algorithm and / or the classical learning method and / or principal component analysis (PCA). For this purpose, the measurement results acquired by the device 100 are transferred to the evaluation unit 90.

[0032] The evaluation unit 90 can be a neural network that uses further known parameters relating to the investigated seed 10 for evaluation. In particular, these further known parameters can be introduced into the neural network, thereby training the neural network to learn. The evaluation unit 90 can include a classical algorithm for evaluating the measurement results. The classical algorithm can determine computational units that are not based on principal component analysis, statistical evaluation, or a machine learning (ML) approach.

[0033] In this context, classical algorithms encompass all functions used in digital image analysis. This includes, for example, calculating averages, corral spheres, lengths, homogeneity, ratios, etc. In other words, things that can be easily calculated using a mathematical formula.

[0034] In this context, the term ML approach refers to the field of machine learning. This includes supervised learning and unsupervised learning. It also encompasses neural networks, principal component analysis, random forests, and support vector machines.

[0035] Statistical analysis allows for the comparison of a value to known classes. If a value falls within the range of a known class, statistical analysis can determine, for example, how far the value deviates from that class. Is it one sigma, two sigmas, or is the value an outlier?

[0036] Fig. 5 and Fig. Figure 6 shows measurement results, i.e., images or radiographic images, of seed 10, which were taken with the device 100. Fig. Figure 5 shows images of rice grains which were placed in the device 100 and irradiated with the radiation source 11. Fig. Figure 6 shows images of peanuts that were placed in the device 100 and irradiated with the radiation source 11. While the peanuts were used as seeds 10 in Fig. 6 each show a homogeneous grey value of 120, some of the rice grains show 10 as seed. Fig. Rice grains 5 show a homogeneous gray value of 120, while other rice grains used as seed 10 do not show a gray value of 120. The rice grains exhibiting a strongly absorbing gray value of 120 are used to predict their germination capacity. The result of at least one further spectral measurement is used to predict germination capacity. Consequently, germination capacity is predicted according to method 200.

[0037] To determine the germination capacity of seed 10, the characteristics / dimensions are determined using classical algorithms and spectral / RGB measurements. For example, parameters such as color and color homogeneity (across different wavelength ranges) are recorded. This information on the recorded parameters / dimensions was compared with the results of actual germination trials, allowing the germination probability to be derived for a specific variety of seed 10. By comparing these results with actual germination trials, it was possible to determine which of these calculated parameters / dimensions are crucial for germination in a given seed variety 10. The terms "dimensions" and "dimensions" are used synonymously here.Additionally, these features from the classical algorithms and the results from the germination experiments can be used to train a machine learning (ML) algorithm, which can then provide a probability regarding germination based solely on the raw data. In other words, the ML algorithm can execute a proposed procedure 200. After training, the proposed procedure 200 can predict a germination probability based on the information acquired by procedure 200, without using any further actual germination experiments.

[0038] The rice grains from Fig. 5, which show an inhomogeneous gray value of 120, are likely to be viable according to method 200. The peanuts from Fig. Seeds 6, which show a homogeneous gray value, are probably no longer viable. The homogeneity or inhomogeneity of the recorded gray levels of the irradiated seed 10 can indicate a probability for the germination capacity of the seed 10.

[0039] In the present invention, the characteristics that determine germination capacity vary for each seed. However, to make generally valid statements about germination capacity, the following characteristics and their specific expression can be used: endosperm size, embryo size (if visible), voids (gray value variations) in the endosperm / embrio, grain fill level, and variations within the grain. Furthermore, additional measurement methods can be used to assess germination capacity.

[0040] Fig. Figures 7 to 9 each show a recording of a spectral measurement of unsprouted maize kernels of a single genotype A ( Fig. 7), B ( Fig. 8) and C ( Fig. 9), which has been recorded with the device 100 for predicting the germination capacity of seed 10, in order to record at least one further physical quantity 20 or a further feature 1 and / or feature 2. Fig. Figures 7 to 9 each show a false-color X-ray image to highlight measured contrasts. Fig. 10 and Fig. Figures 11 each show a PCA evaluation of the three genotypes A, B and C from the Fig. 7 to 9 based on two main components. On the axes of the Fig. 10 and Fig. Figure 11 shows a first principal component PC1 against a second principal component PC2 (PC for principal component).

[0041] According to Fig. In equation 10, the first principal component PC1 has a mass fraction of 98.56% relative to the second principal component PC2. Accordingly, the second principal component PC2 has a mass fraction of 1.44% relative to the first principal component PC1. Therefore: PC1 + PC2 = 1.44% + 98.56% = 100%.

[0042] According to Fig. In equation 10, the first principal component PC1 has a mass fraction of 65.73% relative to the second principal component PC2. Correspondingly, the second principal component PC2 has a mass fraction of 34.03% relative to the first principal component PC1. Thus: PC1 + PC2 = 34.03% + 65.73% = 100%.

[0043] In the Fig. 10 and Fig. Figure 11 shows that the recorded data for genotypes AC can be separated according to various traits using PCA analysis. Fig. In addition to germination capacity (indicated by arrow 500), another characteristic 2 (arrow 520) is indicated in 10, while in Fig. In addition to the germination rate 500, feature 1 (arrow 510) and feature 2 (arrow 520) are indicated. Starting from the origin (0,0), both show Fig. 10 and Fig. 11. Arrow 500 points to the area which has a high probability of germination. In the Fig. 10 and Fig. 11 shows that the corn kernels of genotype A (from Fig. 7) coincide with the area of ​​arrow 500, while the maize of genotype C (from Fig. 9) do not overlap with the area indicated by arrow 500 at all. The PCA evaluation therefore shows that maize of genotype A is viable, while maize of genotype C is not. From the Fig. 10 and Fig. 11 further indicates that maize of genotype B (from Fig.8) is to be positioned closer to arrow 500 than the maize of genotype C. The germination probability of maize of genotype B is therefore significantly higher than that of maize of genotype C, but lower than that of maize of genotype A.

[0044] The principle described herein for predicting the germination capacity of seeds can be summarized as follows.

[0045] When radiation, such as X-rays, penetrates an object, it is attenuated according to Lambert-Beer's law. This attenuation depends on the radiation energy (e.g., X-ray energy), the path length, and the density and atomic number of the material being penetrated. If these parameters (material, energy) are known, the product of length and density can be determined by measuring the transmitted intensity along the path of the X-rays behind the object. The attenuated intensity corresponds to the grayscale information of a calibrated radiation detector, such as an X-ray detector. Furthermore, by varying the angle during the acquisition of numerous individual projections, a complete 3D reconstruction can be generated.

[0046] Since plant components differ only very slightly in their physical structure, it is possible to determine the biomass and water content of seeds, as well as their structural composition, from a transmission image (grayscale information from the detector) or a 3D reconstruction generated by a line or area detector. This is described, for example, in DE10 2015 218 504 A1. The inventors of the present application have discovered the following: The solution outlined here predicts germination capacity by evaluating the attenuation of X-rays, the internal structure of the seed, or the homogeneity of individual components. This is achieved using trained machine learning models or statistical methods such as principal component analysis (PCA). Additional physical parameters can also be used, such as color data from a camera or multi- or hyperspectral data. This makes it possible to assign a germination probability to seed 10 non-destructively, without destroying it.

[0047] The method 200 or the device 100 described herein could also be carried out by using technologies other than X-rays as described herein.

[0048] The technical teaching proposed herein finds application in both plant breeding and the storage of seeds in so-called gene banks. At the same time, all related fields can benefit from the determination of germination capacity. For example, the proposed method or device could be used in seed production or distribution to prevent seed waste, thereby increasing sales by ensuring a larger quantity of seed is available.

[0049] Although some aspects have been described in connection with a device or a method, it is understood that these aspects also constitute a description of a corresponding method or device, such that a block or component of a device or system is also to be understood as a corresponding method step or as a feature of a method step, and vice versa. For reasons of redundancy, a complete description of the present invention in the form of method steps or device features is omitted here.

[0050] In the preceding detailed description, various features were sometimes grouped together in examples to streamline the disclosure. This type of disclosure should not be interpreted as indicating that the claimed examples have more features than are expressly stated in each claim. Rather, as the following claims reflect, the subject matter may consist of fewer than all the features of a single disclosed example. Consequently, the following claims are hereby incorporated into the detailed description, with each claim potentially representing a separate, independent example.While each claim can stand as its own separate example, it should be noted that, although dependent claims refer back to a specific combination with one or more other claims, other examples also include a combination of dependent claims with the subject matter of any other dependent claim, or a combination of any feature with other dependent or independent claims. Such combinations are included unless it is stated that a specific combination is not intended. Furthermore, it is intended that a combination of features of a claim with any other independent claim is also included, even if that claim is not directly dependent on the independent claim. QUOTES INCLUDED IN THE DESCRIPTION

[0000] This list of documents cited by the applicant was automatically generated and is included solely for the reader's convenience. The list is not part of the German patent or utility model application. The DPMA accepts no liability for any errors or omissions. Cited patent literature

[0000] DE 10 2015 218 504 A1 [0016, 0046]

Claims

[1] Methods for predicting the germination capacity of seeds (10) comprising: Providing unsprouted seed (10); Irradiation of the ungerminated seed (10) using a radiation source (11) with an initial intensity I0; Detecting a reduced transmission intensity I transmitted after partial absorption of the initial radiation I0 trans depending on at least one angle; Determine at least one further spectral physical quantity (20) by means of a spectral measurement; and subsequent Evaluating a structure of the recorded transmitted attenuated transmission intensities I trans and the relationships of these structures to each other and the at least one further spectral physical quantity recorded in order to predict a probability of the germination capacity of the unsprouted seed (10). [2] The method of claim 1, comprising: after detecting the transmitted attenuated transmission intensity I trans Determining a product µ*d from a traversed transmission length d of the transmitted attenuated transmission intensity I trans and an absorption coefficient µ of the seed (10) according to the Lambert-Beer law I trans =i I0*exp(-µd), where the absorption coefficient of the seed (10) is known or is determined beforehand by a spectral measurement. [3] Method according to claim 1 or 2, wherein the method comprises: Capturing a multitude of projections of the transmitted attenuated transmission intensity I trans depending on a multitude of different angles between a detection point of the transmitted attenuated transmission intensity I transand a position of the seed (10) to enable two-dimensional or three-dimensional detection of the seed (10) by means of the angle-dependent detected transmitted attenuated transmission intensity I trans to enable. [4] Method according to any one of claims 1 to 3, wherein the method comprises: Evaluating the structure of the detected transmitted attenuated transmission intensity I trans and the relationships of these structures to each other and the recorded at least one further spectral physical quantity (20) by using a classical algorithm and / or a machine learning method and / or a principal component analysis which is designed to use further known quantities relating to the investigated seed (10) for evaluation. [5] Method according to any one of claims 1 to 4, wherein the acquisition of the at least one further spectral physical quantity comprises acquiring color data from a camera (32) and / or acquiring multispectral data and / or acquiring hyperspectral data and / or acquiring the mass of the irradiated ungerminated seed (10) with a scale (40) and / or acquiring the height of the ungerminated seed (10) by means of LiDAR (50), laser sectioning or another optical measurement method to acquire a 3D morphology of the ungerminated seed. [6] Device (100) for predicting the germination capacity of seed (10), wherein the device (100) comprises: a radiation source (11) for irradiating the ungerminated seed (10) with an initial intensity I0; a radiation detector (60) for detecting a reduced transmission intensity I transmitted after partial absorption of the initial radiation I0 transdepending on at least one angle; a further spectral measuring device (30) for detecting at least one further spectral physical quantity by means of a spectral measurement; and wherein the device (100) is equipped with an evaluation unit (90) for subsequent evaluation of the recorded transmitted attenuated transmission intensity I trans and the detected at least one other spectral physical quantity is connectable or linked to in order to predict a probability of the germination capacity of the unsprouted seed (10). [7] Device (100) according to claim 6, wherein the radiation source (11) is an X-ray source and / or a terahertz radiation source and / or an ultrasound radiation source and / or a radar. [8] Device according to one of claims 6 or 7, wherein the radiation detector (60) is an X-ray detector and / or an ultrasound detector and / or a terahertz detector. [9] Device (100) according to one of claims 6 to 8, wherein the radiation detector (60) is configured to detect the transmitted attenuated transmission intensity I trans depending on a multitude of angles as a multitude of gray value information between a detection location (12) of the transmitted attenuated transmission intensity I trans and to detect a position (13) of the seed (10) in order to detect the seed (10) in two dimensions or three dimensions using the angle-dependent detected transmitted attenuated transmission intensity I trans to enable. [10] Device (100) according to any one of claims 6 to 9, wherein the radiation detector (60) is a line detector or an area detector. [11] Device (100) according to one of claims 6 to 10, wherein the further spectral measuring device (30) is designed to perform a spectral analysis of the seed (10) in order to determine a main component and / or several further components of the seed (10). [12] Device (100) according to one of claims 6 to 11, wherein the further spectral measuring device (30) a camera (32) for capturing color data of the seed (10) and / or a spectral analysis device (33) for capturing multispectral data and / or hyperspectral data of the seed (10). [13] Device (100) according to one of claims 6 to 12, wherein the evaluation unit (90) is configured to perform a classical algorithm and / or a machine learning method and / or a principal component analysis of the seed (10). [14] Device (100) according to one of claims 6 to 13, wherein the evaluation unit (90) is a neural network which uses further known parameters relating to the investigated seed (10) for evaluation.

Images