Grain grain storage duration prediction method and system
Through the combination of Kalman filtering algorithm and biophoton radiation data, the comprehensive problem of difficult to reflect the quality detection of grain storage in the existing technology is solved, and the accurate prediction of grain storage time is achieved to meet actual needs.
Patent Information
- Application Number
- CN202510335347.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-20
- Publication Date
- 2025-07-04
AI Technical Summary
The existing grain storage quality detection methods are difficult to comprehensively reflect the life state of the grain stored in grain, and it is difficult to describe the deterioration of grains from the time scale.
The Kalman filtering algorithm and biophoton radiation data are used to perform iterative prediction error correction through the Kalman model, the Kalman gain and state update equations are used to recurse the state vectors in succession, and the coefficients of the Kalman filtering state space model are determined in combination with the least squares linear model identification method to achieve accurate prediction of the storage time of grain grains.
It realizes an accurate prediction of the storage time of grain grains, which can reflect the changes in life state during storage, and provides sufficiently accurate prediction results for storage time.
Smart Images
Figure CN120258218A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of detection of storage quality of grain kernels, and particularly relates to a method and system for predicting the storage duration of grain kernels. Background Art
[0002] Grain kernels (such as wheat kernels, etc.) are living, non-rigid colloidal particles with complex physical and biological properties; viewed along the time axis, the life state of post-harvest grain kernels is a deteriorating dynamic process; as an organic whole, its essence is a process of the physical and biochemical properties of grain kernel components interacting and coupling with each other, and the relationships between its various index characteristics are interrelated rather than independent. That is to say, the coupled action of internal and external factors comprehensively determines the overall life state of grain kernels.
[0003] In actual situations, the deterioration of post-harvest grain kernels is caused by the over-limit of storage duration and subsequent qualitative changes in storage quality under certain storage conditions. Therefore, calibrating the storage duration information is essentially a comprehensive measurement problem of the overall life material state of grain kernels themselves. However, the current detection methods have certain limitations. On the one hand, the existing detection indicators usually cannot comprehensively reflect the life state of stored grain kernels; on the other hand, the existing detection indicators are also difficult to describe the deterioration situation of grain kernels from the time scale. Summary of the Invention
[0004] The purpose of the present invention is to provide a method and system for predicting the storage duration of grain kernels, which are used to solve the problems that the detection indicators of the existing methods for detecting the storage quality of grain kernels usually cannot comprehensively reflect the life state of stored grain kernels and are also difficult to describe the deterioration situation of grain kernels from the time scale.
[0005] To achieve the above purpose, the present invention provides a method for predicting the storage duration of grain kernels, including:
[0006] Taking the biophoton radiation data of the grain kernels to be predicted as the storage duration information for each iteration. Starting from the initial iteration, through the Kalman model of the grain kernel storage duration information, the state update equation corresponding to the prediction error, Kalman gain, and state vector obtained by using the Kalman filtering algorithm, and the time update equation corresponding to the prediction error, the state vector of the current iteration is obtained.
[0007] Updating the current iteration number, and then continuing to obtain the state vector of the current iteration until the prediction error corrected by the Kalman gain in the state update equation corresponding to the prediction error satisfies the iteration stop condition; according to the state vector obtained from the final iteration and the Kalman filtering state space model including the state vector and storage duration information of the current iteration, the predicted storage duration of the grain kernels is obtained.
[0008] The Kalman model is determined by identifying the coefficients of the Kalman filter state space model that includes the state vectors of the current iteration and the previous iteration and the storage duration information of the current iteration, using the known storage duration of the grain kernel samples and the biophoton radiation data as the state vector.
[0009] Further, starting from the first iteration, the method for obtaining the state vector of the current iteration through the prediction error, the state update equation corresponding to the Kalman gain, and the time update equation corresponding to the state vector and the prediction error includes:
[0010] Starting from the first iteration, at each iteration, through the time update equation corresponding to the prediction error, the prediction error before being corrected by the Kalman gain of the current iteration is obtained; the parameters in the time update equation corresponding to the prediction error include the prediction error before being corrected by the Kalman gain of the current iteration, the noise information of the previous iteration, and the prediction error after being corrected by the Kalman gain of the previous iteration.
[0011] Through the state update equation corresponding to the Kalman gain, the Kalman gain of the current iteration is obtained; the parameters in the state update equation corresponding to the Kalman gain include the Kalman gain of the current iteration, the noise information, and the prediction error before being corrected by the Kalman gain of the current iteration.
[0012] Through the time update equation corresponding to the state vector, the state vector of the current iteration is obtained; the parameters in the time update equation corresponding to the state vector include the state vector of the current iteration, the Kalman gain, the storage duration information, and the state vector of the previous iteration.
[0013] Through the state update equation corresponding to the prediction error, the prediction error after being corrected by the Kalman gain of the current iteration is obtained; the parameters of the state update equation corresponding to the prediction error include the Kalman gain of the current iteration and the prediction errors before and after being corrected by the Kalman gain of the current iteration.
[0014] Further, the method for obtaining the state vector of the current iteration through the time update equation corresponding to the state vector includes:
[0015] Substitute the state vector of the previous iteration, the storage duration information of the current iteration, and the obtained Kalman gain of the current iteration into the time update equation corresponding to the state vector to obtain the state vector of the current iteration.
[0016] In the case where the current iteration is the first iteration, the set initial value of the state vector is substituted into the time update equation corresponding to the state vector as the state vector of the previous iteration; in the case where the current iteration is not the first iteration, the state vector of the previous iteration is directly substituted into the time update equation corresponding to the state vector.
[0017] Further, the method for obtaining the Kalman gain of the current iteration through the state update equation corresponding to the Kalman gain includes:
[0018] Substitute the Kalman gain of the current iteration, the noise information, and the predicted error before being corrected by the Kalman gain of the current iteration into the state update equation corresponding to the Kalman gain to obtain the Kalman gain of the current iteration;
[0019] In the case where the current iteration is the first iteration, the set noise information value is substituted into the state update equation corresponding to the Kalman gain as the noise information of the current iteration, and in the case where the current iteration is not the first iteration, the noise information of the current iteration is directly substituted into the state update equation corresponding to the Kalman gain.
[0020] Further, the Kalman filter state space model including the state vectors of the current iteration and the previous iteration further includes a zero-mean normal input white noise sequence; the Kalman filter state space model including the state vector of the current iteration and the storage duration information further includes a zero-mean additive Gaussian white noise;
[0021] The noise information of the previous iteration in the time update equation corresponding to the predicted error is the noise variance matrix of the zero-mean normal input white noise sequence; the noise information of the current iteration in the state update equation corresponding to the Kalman gain is the noise variance matrix of the zero-mean additive Gaussian white noise.
[0022] Further, the identification of the coefficients of the Kalman filter state space model including the state vectors of the current iteration and the previous iteration and including the state vector of the current iteration and the storage duration information is carried out by using the least squares linear model identification method.
[0023] The above technical solution of the present invention provides a brand-new method for predicting the storage duration of grain kernels, and its beneficial effects include: This method utilizes the Kalman filter state space model identified by known samples and each time update equation and each state update equation obtained by the Kalman filter algorithm. Through each time update equation and each state update equation, the correlation relationship between the state vector of the current iteration and the state vector of the previous iteration can be reflected, and the difference between the predicted value and the true value of the storage duration of the grain kernels corresponding to the state vector characterized by the corrected prediction error after each iteration; combined with the setting of using the biophoton radiation data of the grain kernels to be predicted as the storage duration information for each iteration, the state vector is recursively derived through the iteration of the Kalman filter, so as to obtain the prediction error corrected by the Kalman gain for each iteration, which characterizes the relative error between the predicted result of the storage duration of the grain kernels corresponding to the state vector of each iteration and the biophoton radiation data of the grain kernels to be actually predicted, which is equivalent to obtaining the accuracy of the predicted result of the storage duration of the grain kernels compared with the actual storage duration of the grain kernels; thus, in the face of the grain kernels to be predicted, when the prediction error corrected by the Kalman gain satisfies the iteration stop condition, it indicates that the accuracy of the predicted result of the storage duration of the grain kernels corresponding to the state vector of the current iteration compared with the actual storage duration of the grain kernels meets the actual requirements, that is, the predicted result of the storage duration of the grain kernels is accurate enough; at this time, using the state vector obtained by the current iteration and the Kalman filter state space model, an accurate enough predicted result of the storage duration of the grain kernels can be obtained.
[0024] The present invention also provides a system for predicting the storage duration of grain kernels, including a processor, and executable program instructions are stored in the processor, and the executable program instructions are used to be executed to implement the above method for predicting the storage duration of grain kernels.
[0025] The technical solution of the above system for predicting the storage duration of grain kernels of the present invention can achieve the same beneficial effects as the above method for predicting the storage duration of grain kernels. Description of the Drawings
[0026] Figure 1 It is a principle block diagram of the method for predicting the storage duration of grain kernels in the embodiment of the method for predicting the storage duration of grain kernels of the present invention;
[0027] Figure 2 It is a sample schematic diagram of the BPE data of wheat kernels collected in the embodiment of the method for predicting the storage duration of grain kernels of the present invention;
[0028] Figure 3 It is a sample schematic diagram of the BPE data of wheat kernels collected in the embodiment of the method for predicting the storage duration of grain kernels of the present invention after being processed by wavelet threshold denoising;
[0029] Figure 4 This is an example block diagram of the prediction process of the grain storage duration in the embodiment of the grain storage duration prediction method of the present invention;
[0030] Figure 5 This is a schematic diagram of the principle of the BPE data acquisition system for completing the acquisition and storage of wheat grain BPE data in the embodiment of the grain storage duration prediction method of the present invention. Detailed implementation manners
[0031] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments.
[0032] Embodiment of the grain storage duration prediction method
[0033] This embodiment provides a technical solution for a grain storage duration prediction method. Through the Kalman model corresponding to the grain storage duration information and the Kalman filtering algorithm, this technical solution obtains an equation that can recursively calculate the state vector step by step, thereby recursively calculating the state vector at each iteration step and the accuracy of the predicted result of the grain storage duration corresponding to the state vector characterized by the corrected prediction error compared with the actual grain storage duration. When the accuracy meets the requirements in a certain iteration step, the storage duration of the grain to be predicted can be determined by using the state vector at this iteration step and the relationship between the state vector and the grain storage duration information.
[0034] Refer to Figure 1 , and the specific steps of this solution are as follows:
[0035] Taking the biophoton radiation data of the grain to be predicted as the storage duration information for each iteration, starting from the first iteration, using the prediction error, the state update equation corresponding to the Kalman gain, and the state vector, and the time update equation corresponding to the prediction error obtained by the Kalman model of the grain storage duration information through the Kalman filtering algorithm, the state vector of the current iteration is obtained;
[0036] Update the current iteration count, and then continue to obtain the state vector of the current iteration until the prediction error corrected by the Kalman gain in the state update equation corresponding to the prediction error meets the iteration stop condition; according to the state vector obtained in the last iteration and the Kalman filtering state space model including the state vector and storage duration information of the current iteration, the predicted grain storage duration is obtained;
[0037] The Kalman model for the storage duration information of grain kernels is determined by identifying the coefficients of the Kalman filter state space model that contains the state vectors of the current iteration and the previous iteration, and the state vector and storage duration information of the current iteration, using the known storage duration of the grain kernel samples and the biophoton radiation data as the state vector.
[0038] Considering that biophoton emission (BPE) is a universal life phenomenon, biophotons come from the non-local coherent electromagnetic field within living matter, which provides a comprehensive index of the basic characteristics of biological systems. A large number of studies at home and abroad have shown that the specific biological meaning of this index is related to the state of the measured sample itself and factors of the external environment, etc., and it has a very high sensitivity, which can reveal the internal detailed changes of biological systems and show the weak influence of the external environment. In addition, according to modern signal processing theory, the state space method is a good modeling method for describing the dynamic characteristics of systems. The dynamic characteristics of a system are described by a system of first-order differential equations composed of state variables. It can reflect the changes of all independent state variables of the system, so as to simultaneously determine all the internal motion states of the system.
[0039] Therefore, the method for predicting the storage duration of grain kernels in this embodiment utilizes the Kalman filter state space model identified from known samples and the time update equations and state update equations obtained by the Kalman filter algorithm. Through these time update equations and state update equations, the correlation between the state vector of the current iteration and the state vector of the previous iteration can be reflected, as well as the difference between the predicted value and the true value of the storage duration of the grain kernels corresponding to the state vector characterized by the corrected prediction error after each iteration. Combined with the setting of using the biophoton radiation data of the grain kernels to be predicted as the storage duration information for each iteration, the state vector is recursively derived through the iteration of the Kalman filter, thereby obtaining the prediction error corrected by the Kalman gain for each iteration, which characterizes the relative error between the predicted result of the storage duration of the grain kernels corresponding to the state vector of each iteration and the biophoton radiation data of the grain kernels to be actually predicted. This is equivalent to obtaining the accuracy of the predicted result of the storage duration of the grain kernels compared to the actual storage duration of the grain kernels. Thus, in the face of the grain kernels to be predicted, when the prediction error corrected by the Kalman gain satisfies the iteration stop condition, it indicates that the accuracy of the predicted result of the storage duration of the grain kernels corresponding to the state vector of the current iteration compared to the actual storage duration of the grain kernels meets the actual requirements, that is, the predicted result of the storage duration of the grain kernels is accurate enough. At this time, using the state vector obtained from the current iteration and the Kalman filter state space model, an accurate enough predicted result of the storage duration of the grain kernels can be obtained. In summary, this method can reflect the life state of the stored grain kernels through the state vector, and using this state vector and the characteristics of the Kalman filter can achieve the effect of determining the storage duration of the grain kernels accordingly.
[0040] Specifically, starting from the initial iteration, the method for obtaining the state vector of the current iteration through the time update equation corresponding to the prediction error, the state update equation corresponding to the Kalman gain, the time update equation corresponding to the state vector, and the state update equation corresponding to the prediction error includes:
[0041] Starting from the initial iteration, at each iteration, the prediction error before being corrected by the Kalman gain of the current iteration is obtained through the time update equation corresponding to the prediction error. The parameters in the time update equation corresponding to the prediction error include the prediction error before being corrected by the Kalman gain of the current iteration, the noise information of the previous iteration, and the prediction error corrected by the Kalman gain of the previous iteration.
[0042] The Kalman gain for the current iteration is obtained through the state update equation corresponding to the Kalman gain; the parameters in the state update equation corresponding to the Kalman gain include the Kalman gain for the current iteration, the noise information, and the prediction error before being corrected by the Kalman gain for the current iteration.
[0043] The state vector for the current iteration is obtained through the time update equation corresponding to the state vector; the parameters in the time update equation corresponding to the state vector include the state vector for the current iteration, the Kalman gain, the storage duration information, and the state vector for the previous iteration.
[0044] The prediction error after being corrected by the Kalman gain for the current iteration is obtained through the state update equation corresponding to the prediction error; the parameters in the state update equation corresponding to the prediction error include the Kalman gain for the current iteration and the prediction errors before and after being corrected by the Kalman gain for the current iteration.
[0045] In this embodiment, the Kalman filter state space model including the state vectors for the current iteration and the previous iteration further includes a zero-mean normal input white noise sequence; the Kalman filter state space model including the state vector for the current iteration and the storage duration information further includes a zero-mean additive Gaussian white noise; the noise information for the previous iteration in the time update equation corresponding to the prediction error is the noise variance matrix of the zero-mean normal input white noise sequence; the noise information for the current iteration in the state update equation corresponding to the Kalman gain is the noise variance matrix of the zero-mean additive Gaussian white noise; then the Kalman filter state space model including the state vectors for the current iteration and the previous iteration and the Kalman filter state space model including the state vector for the current iteration and the storage duration information are represented by the following formulas in this embodiment:
[0046] SC k = A·SC k-1 + WN k
[0047] AGE k = C·SC k + VN k
[0048] Where k represents the current iteration, k - 1 represents the previous iteration of the current iteration, and the grain kernel BPE data SC k is identified as the state vector for the current iteration, then SC k-1 is the state vector for the previous iteration; A is the gain matrix between state vectors, also known as the system matrix; WN k is a stationary zero-mean normal input white noise sequence, and its noise variance matrix is Q k; The noise of the grain storage time estimation system is stationary zero-mean additive white Gaussian noise VN k , and its noise variance matrix is R k ; The storage duration information of the grain is AGE k ; C is the gain matrix between the state vector and the output variable (i.e., the storage duration information of the grain).
[0049] The identification of the coefficients of the Kalman filter state space model that respectively contains the state vectors of the current iteration and the previous iteration and contains the state vector of the current iteration and the storage duration information is carried out by the least squares linear model identification method. Taking the case where the grain is wheat grains as an example, using the known storage duration (substituted as AGE k ) of the wheat grain samples and the biophoton radiation data as the state vector, the least squares linear model identification method is used to identify the coefficients A and C of the above Kalman filter state space model, thereby determining the Kalman model of the wheat grain storage duration, that is, the Kalman filter state space model after identifying the above coefficients A and C. The Kalman model of the wheat grain storage duration can obtain each time update equation and state update equation for realizing the prediction of the wheat grain storage duration through the Kalman filter algorithm; in other embodiments, the Kalman models of the storage durations of the corresponding types of grains can also be determined in the same way using other types of grain samples, and each time update equation and state update equation for realizing the prediction of the storage durations of the corresponding types of grains can be obtained through the Kalman filter algorithm.
[0050] Specifically, the time update equation corresponding to the state vector and the time update equation corresponding to the prediction error obtained through the Kalman filter algorithm according to the determined Kalman model of the grain storage duration information are respectively expressed by the following formulas:
[0051] SC k = A·SC k-1 + K k (AGE k - C·A·SC k-1 )
[0052] Pe' k = APe k-1 A T + Q k-1
[0053] The state update equation corresponding to the Kalman gain and the state update equation corresponding to the prediction error obtained through the Kalman filter algorithm according to the determined Kalman model of the grain storage duration information are respectively expressed by the following formulas:
[0054] K k = Pe' k C T (CPe' k C T + R k ) -1
[0055] Pe k = (I - K k C)Pe' k
[0056] Where: K k is the Kalman gain of the k-th iteration; Pe' k and Pe k are the prediction errors before and after being corrected by K k respectively in the k-th iteration; Q k-1 represents the noise information of the previous iteration in the time update equation corresponding to the prediction error; R k represents the noise information of the current iteration in the state update equation corresponding to the Kalman gain. Since the manner of obtaining the corresponding time update equation and state update equation according to the Kalman filter state space model through the Kalman filter algorithm belongs to the existing derivation method, it will not be elaborated here.
[0057] In this embodiment, considering the case where the current iteration is the first iteration, the method for obtaining the state vector of the current iteration through the time update equation corresponding to the state vector includes:
[0058] Substitute the state vector of the previous iteration, the storage duration information of the current iteration, and the obtained Kalman gain of the current iteration into the time update equation corresponding to the state vector to obtain the state vector of the current iteration; it should be noted that regardless of which iteration, the biophoton radiation data of the grain kernels to be predicted are used as the storage duration information for each iteration, that is, the BPE data of the grain kernels to be predicted are used as AGE k Substitute into the time update equation SC k = A·SC k-1 + K k (AGE k - C·A·SC k-1 ) to obtain the state vector of the current iteration.
[0059] In the case where the current iteration is the first iteration, the set initial value of the state vector is substituted into the time update equation corresponding to the state vector as the state vector of the previous iteration; in the case where the current iteration is not the first iteration, the state vector of the previous iteration is directly substituted into the time update equation corresponding to the state vector. Thus, the parameter values substituted into the time update equation for the first iteration are considered to handle the situation where there is no state vector of the previous iteration in the current iteration.
[0060] Similarly, the method for obtaining the Kalman gain of the current iteration through the state update equation corresponding to the Kalman gain includes:
[0061] Substitute the Kalman gain of the current iteration, the noise information, and the predicted error before being corrected by the Kalman gain of the current iteration into the state update equation corresponding to the Kalman gain to obtain the Kalman gain of the current iteration;
[0062] In the case where the current iteration is the first iteration, the set noise information value is substituted into the state update equation corresponding to the Kalman gain as the noise information of the current iteration; in the case where the current iteration is not the first iteration, the noise information of the current iteration is directly substituted into the state update equation corresponding to the Kalman gain. Thus, the parameter values substituted into the state update equation for the first iteration are considered to handle the situation where there is no noise information of the previous iteration in the current iteration.
[0063] In addition, in this embodiment, to minimize the noise contained in the biophoton radiation data of the grain kernels, the biophoton radiation data of the grain kernel samples is obtained by performing wavelet threshold denoising on the biophoton radiation sampling data corresponding to the grain kernel samples; the biophoton radiation data of the grain kernels to be predicted is obtained by performing wavelet threshold denoising on the biophoton radiation sampling data corresponding to the grain kernels to be predicted; in other embodiments, this wavelet threshold denoising process may not be performed; the method of performing wavelet threshold denoising includes:
[0064] Decompose the biophoton radiation data to the set number of layers through a set wavelet function; perform quantization processing on the decomposed high-frequency coefficients by the hard threshold method; then perform wavelet reconstruction based on the low-frequency coefficients of the bottom layer of the wavelet decomposition and the quantized high-frequency coefficients of each layer to output the denoised biophoton radiation data. Specifically, in this embodiment, the selected set wavelet function is the sym8 wavelet function, and the set number of layers is 5, that is, decomposed to the fifth layer; the quantization processing by the hard threshold method is specifically to first sort the high-frequency coefficients, set the three coefficients with the lowest sorting to zero, and keep the remaining coefficients unchanged. For Figure 2 the collected BPE data of wheat kernels shown, the data example after being processed by wavelet threshold denoising is as Figure 3As shown. Since the method of wavelet threshold denoising used in this embodiment specifically belongs to the prior art, it will not be elaborated here.
[0065] Moreover, the biophoton radiation sampling data corresponding to the above grain kernel samples is obtained by removing the background noise from the sampling data of biophoton radiation sampling of the grain kernel samples; the biophoton radiation sampling data corresponding to the grain kernels to be predicted is obtained by removing the background noise from the sampling data of biophoton radiation sampling of the grain kernels to be predicted. Thus, the background noise of the biophoton radiation sampling data brought about by the sampling process can be eliminated.
[0066] It should be noted that in this embodiment, in order to remove the influence of outdoor light on the grain kernels, the method of biophoton radiation sampling for the grain kernel samples includes: placing the grain kernel samples in the dark for a corresponding set duration of dark treatment, and then performing biophoton radiation sampling on the grain kernel samples after the dark treatment;
[0067] Similarly, the method of biophoton radiation sampling for the grain kernels to be predicted includes: placing the grain kernels to be predicted in the dark for a corresponding set duration of dark treatment, and then performing biophoton radiation sampling on the grain kernels to be predicted after the dark treatment. That is, the influence of outdoor light on the biophoton radiation data of the obtained grain kernels is eliminated through dark treatment.
[0068] In summary, taking the case where the grain kernels are wheat kernels as an example, referring to Figure 4 , the process of predicting the storage duration of grain kernels including the establishment process of the Kalman filter state space model in the early stage is described as follows:
[0069] 1) Select wheat kernel samples with known storage time information. Specifically, select 400 g of relatively strong, plump, and non-cracked wheat kernels, wash them 3 times with pure water, put them in a drying oven and dry them until the moisture content is (13.7 ± 0.2)%, and put them in a self-sealing bag for standby.
[0070] 2) Use a BPE data acquisition system including a biophoton ultra-weak luminescence measuring instrument to complete the acquisition and storage of the BPE data of the wheat kernel samples, and its principle is as Figure 5 shown.
[0071] Among them, the core component for realizing the acquisition of the BPE data of wheat kernels is a photomultiplier tube (PMT: Photomultiplier tube). The working voltage - HV of the PMT gain is 800 V to 1100 V, and the circuit adopts the working mode with the cathode K grounded; the intermediate stage D realizes the step-by-step amplification of the electron flow signal; the anode end A outputs the signal and is externally connected to a first-order filter circuit composed of a resistor RL and a capacitor C to reduce high-frequency noise. In this way, it can ensure sufficient amplification gain (>107 ) and can also limit the output of the anode dark current within a tolerable range (< nA level); the PMT is placed in a dark box and equipped with shielding and refrigeration devices to reduce the interference of high-frequency noise and background noise; two discrimination threshold levels V1 and V2 are externally added to the discriminator part; and it is required that V1 = 50 mV to 100 mV, V2 = a * V1, and a takes 10 to 50. During the measurement process, the photons emitted by the sample reach the photocathode K, exciting multiplied electrons, which are amplified by multiple stages D and then converge into a photocurrent pulse at the anode A for output. After being processed by circuits such as amplification, discrimination, and counting, the BPE data is finally output to the computer for storage.
[0072] Before data acquisition, turn on the ultra-weak bioluminescence measurement instrument, set the detection room temperature to 28 °C, adjust the high voltage to 1030 V, and preheat for 30 minutes; first weigh (5 ± 0.02) g of wheat grain samples from the self-sealing bag, place them in the dark for 30 min to remove the influence of outdoor light on the wheat to be measured. Then measure the system background noise for 60 s with a sampling interval of 1 s, and then put the dark-treated wheat samples into the sample cell for data acquisition, set the measurement time to 1024 s, and the sampling interval to 1 s; measure the background noise for 60 s again with a sampling interval of 1 s.
[0073] 3) For the biophoton radiation sampling data corresponding to the sampled wheat grain samples, remove the background noise of the BPE data acquisition system; the process of removing the background noise is as follows: to obtain the background data of the BPE data acquisition system twice and the BPE sampling data of wheat grains once; before and after measuring the BPE data of wheat grains, measure the equipment background without placing wheat grain samples once each; take the average of the two background data, and subtract this average from the BPE sampling data corresponding to the wheat grain samples to obtain the BPE data of wheat grains after removing the measurement system background as the biophoton radiation sampling data corresponding to the wheat grain samples.
[0074] After that, by performing wavelet threshold denoising processing on the biophoton radiation sampling data corresponding to the wheat grain samples, the biophoton radiation data of the wheat grain samples is obtained.
[0075] 4) According to the BPE data and storage time information of wheat grain samples with known storage times, use the least squares parameter identification method to identify the coefficients of the Kalman filter state space model (i.e., A and C in the above formula). Then use the Kalman filter algorithm to derive the time update equation corresponding to the prediction error, the state update equation corresponding to the Kalman gain, the time update equation corresponding to the state vector, and the state update equation corresponding to the prediction error with the criterion of minimizing the average power of the estimation error.
[0076] 5) For the wheat grains whose storage duration is to be predicted (i.e., the wheat grains to be predicted), the BPE data acquisition system is also used to collect and store the BPE data of these wheat grains. The background noise is removed and wavelet threshold denoising processing is performed according to the same process as the wheat grain samples to obtain the biophoton radiation data corresponding to the wheat grains to be predicted.
[0077] Take the biophoton radiation data of the wheat grains to be predicted as the storage duration information AGE for each iteration. k , starting from the initial iteration, through the state update equation corresponding to the prediction error obtained by the Kalman model of the grain storage duration information using the Kalman filtering algorithm, the state update equation corresponding to the Kalman gain, the time update equation corresponding to the state vector, and the time update equation corresponding to the prediction error, obtain the state vector SC of the current iteration. k ;
[0078] Update the current iteration number, and then continue to obtain the state vector of the current iteration until the prediction error corrected by the Kalman gain in the state update equation corresponding to the prediction error meets the iteration stop condition; usually, the iteration stop condition is that the prediction error corrected by the Kalman gain reaches convergence, or the prediction error corrected by the Kalman gain is less than the set error threshold, to ensure that the predicted storage duration of the grains meets the required accuracy; since how to specifically set the corresponding iteration stop condition in the Kalman filtering algorithm actually belongs to the prior art, it will not be elaborated here.
[0079] After the iteration stops (assuming that the iteration number k = kmax at this time), according to the state vector SC obtained from the final iteration. kmax And the Kalman filtering state space model AGE including the state vector and storage duration information of the current iteration. k = C·SC k + VN k , that is, substitute the state vector obtained from the final iteration into this Kalman filtering state space model to obtain the calculated AGE. kmax , as the predicted storage duration of the grains.
[0080] Embodiment of the grain storage duration prediction system
[0081] This embodiment provides a technical solution of a grain storage duration prediction system, which includes a processor. There are executable program instructions stored in the processor, and the executable program instructions are used to be executed to implement the grain storage duration prediction method in the grain storage duration prediction method embodiment described above.
[0082] Since the specific principle and working mode of the grain storage duration prediction system in this embodiment have been described in detail in the above-mentioned embodiment of the grain storage duration prediction method, they will not be elaborated here.
[0083] It should be understood that the above specific embodiments of the present invention are only used for exemplary illustration or explanation of the principle of the present invention, and do not constitute a limitation on the present invention.
Claims
1. A method for predicting the storage duration of grain kernels, characterized in that, Including: Taking the biophoton radiation data of the grain kernels to be predicted as the storage duration information for each iteration, starting from the first iteration, the predicted error, the state update equation corresponding to the Kalman gain, and the state vector obtained by using the Kalman filter algorithm through the Kalman model of the grain kernel storage duration information, and the time update equation corresponding to the predicted error are used to obtain the state vector of the current iteration; Updating the current iteration number, and then continuing to obtain the state vector of the current iteration until the predicted error corrected by the Kalman gain in the state update equation corresponding to the predicted error satisfies the iteration stop condition; According to the state vector obtained from the last iteration and the Kalman filter state space model including the state vector and storage duration information of the current iteration, the predicted storage duration of the grain kernels is obtained; The Kalman model is determined by identifying the coefficients of the Kalman filter state space model that respectively include the state vectors of the current iteration and the previous iteration and the state vector and storage duration information of the current iteration by using the known storage duration of the grain kernel samples and the biophoton radiation data as the state vector.
2. The method for predicting the storage duration of grain kernels according to claim 1, wherein Starting from the first iteration, the method of obtaining the state vector of the current iteration through the predicted error, the state update equation corresponding to the Kalman gain, and the state vector and the time update equation corresponding to the predicted error includes: Starting from the first iteration, at each iteration, the predicted error before being corrected by the Kalman gain of the current iteration is obtained through the time update equation corresponding to the predicted error; The parameters in the time update equation corresponding to the predicted error include the predicted error before being corrected by the Kalman gain of the current iteration, the noise information of the previous iteration, and the predicted error corrected by the Kalman gain of the previous iteration; The Kalman gain of the current iteration is obtained through the state update equation corresponding to the Kalman gain; The parameters in the state update equation corresponding to the Kalman gain include the Kalman gain of the current iteration, the noise information, and the predicted error before being corrected by the Kalman gain of the current iteration; The state vector of the current iteration is obtained through the time update equation corresponding to the state vector; The parameters in the time update equation corresponding to the state vector include the state vector of the current iteration, the Kalman gain, the storage duration information, and the state vector of the previous iteration; The predicted error corrected by the Kalman gain of the current iteration is obtained through the state update equation corresponding to the predicted error; The parameters of the state update equation corresponding to the predicted error include the Kalman gain of the current iteration and the predicted errors before and after being corrected by the Kalman gain of the current iteration.
3. The method for predicting the storage duration of grain kernels according to claim 2, wherein The method of obtaining the state vector of the current iteration through the time update equation corresponding to the state vector includes: Substituting the state vector of the previous iteration, the storage duration information of the current iteration, and the obtained Kalman gain of the current iteration into the time update equation corresponding to the state vector to obtain the state vector of the current iteration; In the case where the current iteration is the first iteration, substitute the set initial value of the state vector as the state vector of the previous iteration into the time update equation corresponding to the state vector; in the case where the current iteration is not the first iteration, directly substitute the state vector of the previous iteration into the time update equation corresponding to the state vector.
4. The method for predicting the storage duration of grain kernels according to claim 2, wherein The ways to obtain the Kalman gain of the current iteration through the state update equation corresponding to the Kalman gain include: Substitute the Kalman gain of the current iteration, the noise information, and the predicted error before being corrected by the Kalman gain of the current iteration into the state update equation corresponding to the Kalman gain to obtain the Kalman gain of the current iteration; In the case where the current iteration is the first iteration, substitute the set noise information value as the noise information of the current iteration into the state update equation corresponding to the Kalman gain; in the case where the current iteration is not the first iteration, directly substitute the noise information of the current iteration into the state update equation corresponding to the Kalman gain.
5. The method for predicting the storage duration of grain kernels according to claim 2, wherein The Kalman filter state space model including the state vectors of the current iteration and the previous iteration also includes a zero-mean normal input white noise sequence; the Kalman filter state space model including the state vector of the current iteration and the storage duration information also includes a zero-mean additive Gaussian white noise; The noise information of the previous iteration in the time update equation corresponding to the predicted error is the noise variance matrix of a zero-mean normal input white noise sequence; The noise information of the current iteration in the state update equation corresponding to the Kalman gain is the noise variance matrix of a zero-mean additive Gaussian white noise.
6. The method for predicting the storage duration of grain kernels according to any one of claims 1-5, characterized in that The identification of the coefficients of the Kalman filter state space model including the state vectors of the current iteration and the previous iteration and including the state vector of the current iteration and the storage duration information is carried out by using the least squares linear model identification method.
7. A prediction system for the storage duration of grain kernels, comprising a processor, characterized in that, The processor stores executable program instructions, and the executable program instructions are used to be executed to implement the grain kernel storage duration prediction method according to any one of claims 1-6.