Paddy field water body data on-line measurement method and system and medium
By removing noise and screening the Raman spectral data in rice field water, an XGBoost measurement model was constructed, which solved the problem of low signal-to-noise ratio of Raman spectral measurement in rice field water, and achieved higher measurement accuracy and robustness.
Patent Information
- Application Number
- CN202411682671.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-22
- Publication Date
- 2025-03-04
AI Technical Summary
In rice fields, when measuring Raman spectroscopy, due to unknown interference from inorganic and organic matter, the signal-to-noise ratio is low, making it difficult to accurately measure important parameters such as pH value and dissolved oxygen.
By removing the key features in Raman spectral data, using SHAP values to screen out significant key features, constructing an XGBoost measurement model, reducing model construction time and improving measurement accuracy.
It effectively reduces the model construction time, improves the accuracy of target data measurement, and optimizes the hyperparameters of the XGBoost model through the Optuna algorithm, achieving better robustness and measurement effects.
Smart Images
Figure CN119557595A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of water body data monitoring, and in particular to an online measurement method, system and medium for paddy field water body data. Background Art
[0002] Paddy field water data monitoring plays a vital role in agricultural production. It is not only related to the growth and development of rice, but also directly affects the balance and stability of the paddy field ecosystem. For example, pH and dissolved oxygen (DO) measurements, pH is a key indicator to measure the acidity and alkalinity of paddy field water, while dissolved oxygen is an important indicator to measure the health status and self-purification capacity of paddy field water. The pH value and dissolved oxygen content of paddy field water directly affect the survival status of aquatic organisms, and thus affect the balance of the paddy field ecosystem.
[0003] Raman spectroscopy is an effective analytical tool for determining the properties of a sample and assessing the quality of its components based on molecular vibration and rotation information, capable of measuring multiple parameters simultaneously with high sensitivity and rapid speed. Importantly, it does not require additional reagents or sample preparation, making it a convenient and environmentally friendly option for measuring data in paddy water.
[0004] However, paddy field water contains many unknown inorganic and organic substances, and the Raman signal of the substance to be measured is indeed much smaller than the Raman signal of these unknown substances and the fluorescence interference excited by them and other unknown interferences from the environment. Therefore, when establishing a Raman measurement model, it is necessary to use a variety of denoising methods, feature extraction methods, and regression methods for permutation testing in the hope of obtaining the best combination of processing methods for the sample to be measured. This is a process that requires skills and energy, and its accuracy cannot be guaranteed. Summary of the invention
[0005] To solve the above problems, the present invention provides an online measurement method for paddy field water body data. The method removes noise from Raman key features and screens out significant key features based on SHAP values. The extracted significant key feature variables are significantly correlated with the target to be measured. There is no need to use numerous denoising methods, feature extraction methods, and regression methods for permutation testing, which effectively reduces the model construction time and improves the accuracy of target data measurement.
[0006] To achieve the above object, the present invention provides the following technical solutions.
[0007] The present invention provides a method for online measurement of paddy field water body data, comprising the following steps:
[0008] Determine the type of water body data to be tested; obtain the water body Raman spectrum data set and the corresponding wavenumber sequence;
[0009] Determine the substance molecules related to the type of water data to be tested, take the characteristics of the scattered frequency shifts generated by the substance molecules to the incident light as key features, and determine the positions of the key features in the Raman spectrum data set based on the wave number sequence to construct a key feature sequence;
[0010] Identify the local inflection points of the Raman data sequence and construct a key candidate feature sequence; by calculating the average coefficient of variation of the distance between the key candidate feature sequence and the key feature, screen out the key candidate features with an average coefficient of variation greater than 1, complete noise removal, and obtain a new key feature sequence;
[0011] Construct an XGBoost measurement model, calculate the SHAP value based on the specific sample to obtain the predicted value of the water body data type measured by the XGBoost measurement model, perform label significance analysis on each screened key feature according to the SHAP value, and obtain several significant key features;
[0012] The significant key features of the Raman spectral data of the water body to be tested are extracted as input, and the prediction results of the water body data are obtained through the XGBoost measurement model.
[0013] Preferably, the method further includes optimizing the hyperparameters of the XGBoost measurement model by using the Optuna algorithm to obtain an optimized XGBoost measurement model; inputting Raman spectrum data of the water body to be measured, and obtaining a prediction result of the water body data measurement target by using the optimized XGBoost measurement model.
[0014] Preferably, the construction of the key feature sequence comprises the following steps:
[0015] Obtain a water Raman spectral data set D and a corresponding wave number sequence W, where the Raman data set D contains n data, expressed as D = [D1, D2, ..., D n ]; the wave number sequence W is a real-valued sequence collected at intervals within a fixed wave number, expressed as W = [W1, W2, ..., W n ];
[0016] Determine the substance molecules related to the type of water body data to be tested, calculate the two-dimensional chemical structure of the substance molecules, and form the substance molecular configuration as initial geometric data;
[0017] The molecular configuration of the substance is optimized by density functional theory method and basis set, the overestimation of the calculated harmonic number is corrected, and the optimized geometric parameters of the substance molecule are obtained;
[0018] Generate theoretical Raman spectra based on optimized geometric parameters and identify key features;
[0019] Construct key feature sequence based on theoretical key features:
[0020]
[0021] Among them, K a+m is the strength of the key feature, W a+m is the position of the key feature in D, 1≤a≤n, 1≤a+j-1≤n, 0≤m≤j-1.
[0022] Preferably, the screening of the key features comprises the following steps:
[0023] Screening and obtaining local inflection points of the Raman spectral sequence as key candidate features:
[0024]
[0025] Among them, KC b+q is the strength of the key candidate feature, WC b+q is the position of the key candidate feature in D, 1≤b≤n, 1≤b+p-1≤n, 0≤q≤p-1.
[0026] Calculate the distance between a key candidate feature i and all key data in all samples:
[0027]
[0028] Determine the average distance:
[0029]
[0030] Where S is the number of samples, b≤i≤b+p-1.
[0031] Determine candidate key features KC b+q The coefficient of variation is:
[0032]
[0033] Eliminate the key candidate features with an average coefficient of variation greater than 1 and obtain f key features
[0034]
[0035] Preferably, the XGBoost measurement model is based on Tree SHAP analysis to obtain a SHAP value.
[0036] Preferably, before measuring with the XGBoost measurement model, the method further comprises removing the baseline of the Raman spectrum data by polynomial fitting to obtain the denoised Raman spectrum data for model prediction.
[0037] The present invention also provides an online measurement system for paddy field water body data, the system comprising:
[0038] processor;
[0039] a memory having stored thereon a computer program executable on the processor;
[0040] Wherein, when the computer program is executed by the processor, the steps of the online measurement method of paddy field water body data are implemented.
[0041] The present invention also provides a computer-readable storage medium, on which a data processing program is stored. When the data processing program is executed by a processor, the steps of the online measurement method for paddy field water body data are implemented.
[0042] Beneficial effects of the present invention:
[0043] The present invention proposes an online measurement method, system and medium for paddy field water body data. The invention removes the noise of Raman key features and screens out significant key features based on SHAP values. The extracted significant key feature variables are significantly associated with the target to be measured, and there is no need to use a large number of denoising methods, feature extraction methods, and regression methods for permutation testing, which effectively reduces the model construction time and improves the accuracy of target data measurement. The invention uses the Optuna algorithm to optimize the XGBoost measurement model, and obtains a measurement effect that is comparable to or even better than the classic model. The constructed model has very good robustness. BRIEF DESCRIPTION OF THE DRAWINGS
[0044] Figure 1 is a flow chart of a method for online measurement of paddy field water body data according to an embodiment of the present invention;
[0045] Figure 2 is a spectrum diagram of 40 key candidate feature variables of an embodiment of the present invention;
[0046] Figure 3 is a SHAP value bee swarm diagram of an embodiment of the present invention, wherein, Figure 3 (a)- Figure 3 (b) is the SHAP value bee swarm diagram of the DO measurement model. Figure 3 (c)- Figure 3 (d) Beeswarm diagram of the SHAP value of the pH measurement model. DETAILED DESCRIPTION
[0047] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.
[0048] Example
[0049] Paddy field water contains many unknown inorganic and organic substances. The Raman signal of the substance to be measured itself is indeed much smaller than the Raman signal of these unknown substances and the fluorescence interference excited by them and other unknown interferences from the environment. Therefore, when establishing a Raman measurement model, it is necessary to use a variety of denoising methods, feature extraction methods, and regression methods to perform permutation tests in the hope of obtaining the best combination of processing methods for the sample to be measured. This is a process that requires skills and energy, and its accuracy cannot be guaranteed. For this reason, the process of an online measurement method for paddy field water body data proposed in the present invention is as follows: Figure 1 As shown, the following steps are included:
[0050] S1: Determine the type of water body data to be tested; obtain the water body Raman spectrum data set and the corresponding wavenumber sequence.
[0051] S2: Determine the material molecules related to the type of water data to be tested, take the characteristics of the scattered frequency shift generated by the material molecules to the incident light as the key features, and determine the position of the key features in the Raman spectrum data set based on the wavenumber sequence to construct the key feature sequence.
[0052] S3: Determine the local inflection points in the Raman spectrum sequence and construct a key candidate feature sequence; by calculating the average coefficient of variation of each key candidate feature in the key candidate feature sequence, screen out the key candidate features with an average coefficient of variation greater than 1 to obtain the screened key features.
[0053] S4: Construct an XGBoost measurement model, calculate the SHAP value based on the specific sample to obtain the predicted value of the water body data type measured by the XGBoost measurement model, perform label significance analysis on each screened key feature according to the SHAP value, and obtain several significant key features.
[0054] S5: Extract the significant key features of the Raman spectral data of the water body to be tested as input, optimize the hyperparameters of the XGBoost measurement model through the Optuna algorithm, and obtain the optimized XGBoost measurement model; input the Raman spectral data of the water body to be tested, and obtain the prediction result of the water body data measurement target through the optimized XGBoost measurement model.
[0055] Specifically, constructing a key feature sequence, i.e., key feature identification, includes the following steps:
[0056] Obtain a water Raman spectral data set D and a corresponding wave number sequence W, where the Raman data set D contains n data, expressed as D = [D1, D2, ..., D n ]; the wave number sequence W is a real-valued sequence collected at intervals within a fixed wave number, expressed as W = [W1, W2, ..., W n ];
[0057] In the Raman data set, the scattered frequency shift sequence generated by the molecules of the substance to be tested to the incident light is defined as the key feature sequence. The present invention constructs the two-dimensional chemical structure and spatial structure of the molecule, further optimizes the molecular spatial geometric parameters, uses the optimized spatial geometric parameters to generate a theoretical Raman spectrum, and calibrates it with the actual spectrum to complete the key feature identification.
[0058] Specifically, determine the substance molecules related to the type of water data to be tested, calculate the two-dimensional chemical structure of the substance molecules, and form the substance molecular configuration as the initial geometric data; optimize the substance molecular configuration through the density functional theory method and basis set, correct the overestimation of the calculated harmonic number, and obtain the optimized geometric parameters of the substance molecules; generate theoretical Raman spectra based on the optimized geometric parameters to determine the key features; construct the key feature sequence according to the theoretical key features: Among them, K a+m is the strength of the key feature, W a+m is the position of the key feature in D, 1≤a≤n, 1≤a+j-1≤n, 0≤m≤j-1.
[0059] The types of water body data to be tested selected in the experiment in this embodiment are dissolved oxygen and pH value. Nitrogen fertilizer is used in large quantities in paddy fields, which usually exists in the form of ammonia nitrogen. The presence of dissolved oxygen will promote the oxidation of ammonia nitrogen into nitrite (NO2-) and nitrate (NO3-). It can be seen that the dissolved oxygen content in the paddy field solution is associated with the nitrogen content. And the pH value of the solution has an important influence on the form and concentration of nitrate. Under acidic conditions (low pH), nitrate mainly exists in the form of ammonium ions (NH4+). As the pH value of the solution increases, ammonia ions will gradually convert into ammonia (NH3), which is manifested as a decrease in the concentration of nitrate. Therefore, there is a correlation between the nitrogen content and oxygen content in the solution, as well as the same pH value. In this embodiment, the Raman characteristic peak of nitrogen is used for analysis as one of the key features of DO content and pH value measurement, as an example.
[0060] First, the calculations were performed using Gaussian09 quantum chemistry software, which is commonly used to calculate energy, molecular orbitals, and molecular vibration distributions in molecules. The two-dimensional chemical structures of oxygen (O2) and nitrogen (N2) molecules were constructed using ChemDraw software, and then the two molecular configurations formed using Chem3D software were used as the initial geometric data of O2 and N2, which were optimized using the molecular modeling software Gaussian09. In these calculations, the density functional theory method B3LYP and the 6-31G (d) basis set were used. The overestimation of the calculated harmonic number was corrected by the correction program.
[0061] Furthermore, the theoretical Raman spectra were generated using Gaussian09 and Origin software using the optimized spatial geometric parameters of O2 and N2 molecules. The theoretical Raman peaks of OO and NN molecules are located at 1564.5 cm -1 and 2341.3cm -1 The experimentally observed vibration peak position is 1554cm -1 and 2330cm -1 Compared with the theoretical value, it is shifted to the left by about 10cm -1 The overall observed displacement is consistent. Through the above analysis, it is first clear that 1554cm -1 and 2330cm -1 Some of the key features.
[0062] Furthermore, the local inflection points in the key feature sequence are removed to construct the key candidate feature sequence Among them, KC b+q is the strength of the key candidate feature, WC b+q is the position of the key candidate feature in D, 1≤b≤n, 1≤b+p-1≤n, 0≤q≤p-1.
[0063] For data sequences with large noise, filtering of key candidate features can solve the problem of too many feature points. Noise has random characteristics, and the probability of noise points stably appearing at the same position in the Raman sequence is low. Calculate the average coefficient of variation of the distance between a key candidate feature and all key data in all samples. When the coefficient of variation is greater than 1, it means that the candidate feature position is highly discrete and random, that is, it is judged as a noise point and is removed. Assuming there are S samples, calculate the distance between a key candidate feature i and all key data in all samples:
[0064]
[0065] Determine the average distance:
[0066]
[0067] Where S is the number of samples, b≤i≤b+p-1;
[0068] Determine candidate key features KC b+q The coefficient of variation is:
[0069]
[0070] Eliminate the key candidate features with an average coefficient of variation greater than 1 and obtain f key features
[0071]
[0072] The present invention aims at the data sequence with large noise, and the filtering of key candidate features can solve the problem of too many feature points. Noise has random characteristics, and the probability of noise points stably appearing at the same position in the Raman sequence is not high. The average coefficient of variation of the distance from a key candidate feature to all key data in all samples is calculated. When the coefficient of variation is greater than 1, it means that the discrete degree of the candidate feature position is large and the randomness is strong, that is, it is judged as a noise point and is eliminated.
[0073] In this embodiment, by calculating the coefficient of variation of the average distance between the key candidate feature and the four key features determined previously, taking the value 1 as the threshold, and eliminating the candidate points with large position dispersion, 40 key candidate feature variables were obtained. Next, the polynomial fitting method was performed to remove the baseline, and the obtained Figure 2 The spectrum graph shown in Figure 2. Using all 40 key candidate feature variables, an XGBoost measurement model was constructed.
[0074] SHAP analysis based on algorithmic models is a common method for feature importance prediction. For the predicted value of a specific sample, the SHAP value is the contribution of each feature in the sample to the predicted value. SHAP methods include Kernel SHAP, Deep SHAP, and Tree SHAP. Among them, Tree SHAP has a fast calculation speed and good interpretation effect on black box models. The present invention uses TreeSHAP to interpret the XGBoost tree model. The SHAP model interpreter is used to analyze the impact of key candidate features on the model output and rank the importance, such as Figure 3 (a)- Figure 3 (d) Figure 3 (a)- Figure 3 (b) is the SHAP value bee swarm diagram of the DO measurement model. Figure 3 (c)- Figure 3 (d) Beeswarm diagram of the SHAP value of the pH measurement model.
[0075] SHAP Swarm Figure 3 The distribution of the impact of 40 key candidate features on DO and pH model outputs is shown. Each point represents a sample, the number of samples is stacked vertically, and light and dark distributions correspond to high and low values of the samples. Figure 3In (a), for variable 15 (OO), the greater the OO concentration, the greater the SHAP value, indicating that the OO concentration is positively correlated with the DO concentration; the greater the SHAP value of variable 27 (NN), the smaller the DO value, indicating that the negative impact on the DO concentration is greater. The dissolved oxygen content and nitrogen content in the paddy field solution are related, but the relationship is more complicated. Generally speaking, nitrogen exists in different forms in paddy fields, mainly ammonia nitrogen and nitrate nitrogen. First, higher ammonia nitrogen concentrations usually increase the activity of microorganisms, which consume dissolved oxygen for ammonia oxidation, thereby reducing the dissolved oxygen level in the water. Secondly, nitrate nitrogen can increase the nitrogen load in the water body through nitrification, which may cause changes in the dissolved oxygen level. Usually, this change is manifested as the nitrification process of nitrogen requires oxygen consumption. Finally, the water ecosystem of the paddy field will also be affected by the nitrogen content. Excessive nitrogen will lead to the proliferation of algae, and the growth and death process of algae will also affect the dissolved oxygen level. In short, an increase in nitrogen content in the paddy field usually reduces the dissolved oxygen level, verifying the correctness of the logical relationship between the NN variable and the target measurement in the SHAP value diagram. In addition, as the pH value changes, the solubility of dissolved oxygen will also change. Specifically, under acidic conditions, the concentration of hydrogen ions in water molecules is higher, which can affect the combination of oxygen molecules and water. That is, at lower pH values, the solubility of oxygen usually increases, and at higher pH values, the solubility of oxygen usually decreases. The positive and negative correlations between variables 31 (HS) and 16 (OH) and DO content observed in the experiment are consistent with the above physical principles. For the pH measurement model, Figure 3 In (c), the four key features 16, 31, 15, and 27 show positive, negative, positive, and negative effects, respectively.
[0076] Since there may be correlations between candidate key features, resulting in information redundancy, we tried to combine the input data sets from high to low according to the feature importance, and determined the significant key features for modeling based on the evaluation indicators of the training set and the test set. As shown in Table 1, for the DO and pH models, 25 and 34 significant key features were finally determined for model training, respectively, which indeed achieved better performance than the training model of all 40 candidate key features. The performance of the XGBoost model of the key candidate feature variables is shown in Table 1.
[0077] Table 1 XGBoost model performance of key candidate feature variables
[0078]
[0079] In order to further improve the model's measurement accuracy, robustness, and generalization ability on independent data sets, the present invention uses the Optuna algorithm to optimize the hyperparameters of the XGBoost model. Optuna uses the Bayesian optimization algorithm to search the hyperparameter space, which has the advantages of concise code, easy operation, and highly modularized writing. It can make up for the defects of low computational efficiency, high cost, and easy generation of local optimal values in parameter adjustment methods such as grid search and random search. The Optuna algorithm is used to optimize the parameters of the XGBoost model on the training set, and the fitness judgment standard is set to the standard deviation obtained by the 10-fold cross validation of the model in each iteration. Substitute the optimized hyperparameters into the model to obtain the optimized XGBoost model.
[0080] This example compares the optimized XGBoost model with a combination of multiple preprocessing methods, feature extraction methods, and regression methods, as shown in Table 2. As shown in Table 2, the Optuna-XGBoost model based on the denoising and feature extraction method achieves comparable or even better measurement results than the classic model.
[0081] Table 2 Model comparison
[0082]
[0083] This embodiment also provides a computer-readable storage medium, which stores a computer program, which can be used to execute the above Figure 1 Provided is an online measurement method for paddy field water data.
[0084] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).
[0085] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. A method for online measurement of paddy field water data, characterized in that: The following steps are involved: Determine the type of water body data to be tested; Obtain water Raman spectrum data set and corresponding wavenumber series; Determine the substance molecules related to the type of water data to be tested, take the characteristics of the scattered frequency shifts generated by the substance molecules to the incident light as key features, and determine the positions of the key features in the Raman spectrum data set based on the wave number sequence to construct a key feature sequence; Identify the local inflection points of the Raman data sequence and construct a key candidate feature sequence; by calculating the average coefficient of variation of the distance between the key candidate feature sequence and the key feature, screen out the key candidate features with an average coefficient of variation greater than 1, complete noise removal, and obtain a new key feature sequence; Construct an XGBoost measurement model, calculate the SHAP value based on the specific sample to obtain the predicted value of the water body data type measured by the XGBoost measurement model, perform label significance analysis on each screened key feature according to the SHAP value, and obtain several significant key features; The significant key features of the Raman spectral data of the water body to be tested are extracted as input, and the prediction results of the water body data are obtained through the XGBoost measurement model.
2. The method for online measurement of paddy field water body data according to claim 1, characterized in that: It also includes optimizing the hyperparameters of the XGBoost measurement model through the Optuna algorithm to obtain the optimized XGBoost measurement model; inputting the Raman spectrum data of the water body to be measured, and obtaining the prediction results of the water body data measurement target through the optimized XGBoost measurement model.
3. The method for online measurement of paddy field water body data according to claim 1, characterized in that: The key feature sequence is constructed, comprising the following steps: Obtain a water Raman spectral data set D and a corresponding wave number sequence W, where the Raman data set D contains n data, expressed as D = [D1, D2, ..., D n ]; the wave number sequence W is a real-valued sequence collected at intervals within a fixed wave number, expressed as W = [W1, W2, ..., W n ]; Determine the substance molecules related to the type of water body data to be tested, calculate the two-dimensional chemical structure of the substance molecules, and form the substance molecular configuration as initial geometric data; The molecular configuration of the substance is optimized by density functional theory method and basis set, the overestimation of the calculated harmonic number is corrected, and the optimized geometric parameters of the substance molecule are obtained; Generate theoretical Raman spectra based on optimized geometric parameters and identify key features; Construct key feature sequence based on theoretical key features: Among them, K a+m is the strength of the key feature, W a+m is the position of the key feature in D, 1≤a≤n, 1≤a+j-1≤n, 0≤m≤j-1.
4. The method for online measurement of paddy field water body data according to claim 3, characterized in that: The screening of the key features comprises the following steps: The local inflection points of the Raman spectrum sequence are screened and obtained as key candidate features, which are expressed as: Among them, KC b+q is the strength of the key candidate feature, WC b+q is the position of the key candidate feature in D, 1≤b≤n, 1≤b+p-1≤n, 0≤q≤p-1; Calculate the distance between a key candidate feature i and all key data in all samples: Determine the average distance: Where S is the number of samples, b≤i≤b+p-1; Determine candidate key features KC b+q The coefficient of variation is: Eliminate the key candidate features with an average coefficient of variation greater than 1 and obtain f key features 5. The method for online measurement of paddy field water data according to claim 1, characterized in that: The XGBoost measurement model is based on Tree SHAP analysis to obtain SHAP values.
6. The method for online measurement of paddy field water body data according to claim 1, characterized in that: It also includes removing the baseline of the Raman spectral data by polynomial fitting before measuring with the XGBoost measurement model, and obtaining the denoised Raman spectral data for model prediction.
7. A rice field water body data online measurement system, characterized in that: The system comprises: processor; a memory having stored thereon a computer program executable on the processor; Wherein, when the computer program is executed by the processor, the steps of the online measurement method for paddy field water body data as described in any one of claims 1 to 6 are implemented.
8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a data processing program, and when the data processing program is executed by a processor, the steps of the online measurement method for paddy field water body data according to any one of claims 1 to 6 are implemented.