A method for judging the stage of tobacco leaf baking process

Abstract

The present invention proposes a method for judging the stage of tobacco leaf baking process, which belongs to the technical field of tobacco leaf baking. The method for judging the stage of tobacco leaf baking process, before actual baking, inputs preset baking parameters of the entire baking process through the control end of the baking room; in the entire baking process of actual baking, according to the calibrated data collection time t of the baking process stage, collects the hyperspectral information of tobacco leaf state index X; and establishes characteristic spectral data of tobacco leaf state index X at the collection time t through the hyperspectral information collected at the time t; compares the characteristic spectral data of tobacco leaf state index X at the actual baking time t with the pre-calibrated hyperspectral information; and judges the stage of the current tobacco leaf baking process through the comparison result. The automation of the stage judgment of the tobacco leaf baking control process can be realized, and the efficiency of judging the stage of the baking process can be improved.

Description

Technical Field

[0001] The invention belongs to the technical field of tobacco leaf baking, and in particular relates to a method for judging the stage of tobacco leaf baking process based on hyperspectral technology. Background Art

[0002] Tobacco leaf baking is an important step in the cigarette production process, and the quality of baking directly determines the quality of tobacco leaves. In China, intensive baking rooms are used for baking, and each baking process takes about 120-170 hours. The entire tobacco leaf baking process is divided into three stages: yellowing period, color fixing period and dry tendon period. The baking conditions and tobacco leaf states corresponding to each stage are different. In the actual production process, baking personnel visually inspect the state of tobacco leaves through the observation window of the intensive baking room to determine the stage of the tobacco leaf baking process and then set the baking conditions. This method of visually judging the baking stage of tobacco leaves relies on the personal experience of the baking personnel. It is not only inefficient, but also has large errors in the judgment results, which is not conducive to the homogenization of tobacco leaf baking.

[0003] In response to the above problems, the invention patent with patent publication number CN111466605A proposes a solution to use machine vision to replace human eyes. It proposes a method, system and device for judging the yellowing and browning degree of tobacco leaves by extracting the RGB value of tobacco leaf images. However, the purpose of the device is to judge the yellowing and browning degree of tobacco leaves, not to judge the baking process of tobacco leaves. More importantly, in addition to the color of tobacco leaves, their morphology, moisture content and main vein state will change during the baking process. It is inaccurate to rely on the color of tobacco leaves to judge the state of tobacco leaves. The RGB image information in the above-mentioned prior art is a kind of image information based on the spectral characteristics close to human visible light. Therefore, the prior art only plays an effect of replacing human eyes for observation. The RGB image information obtained still belongs to the technical scope of multi-spectrum, and is limited by the wide band characteristics of multi-spectrum.

[0004] The invention patent with the patent publication number CN109164009A discloses a real-time online monitoring and recording system for tobacco leaf baking. The system monitors and records the moisture content of tobacco leaves by measuring the weight change of tobacco leaves, monitors the color of tobacco leaves and judges the tobacco leaf baking process through near-infrared spectroscopy, and finally feeds back the moisture content and color monitoring results to the baking operator through a mobile client. The invention patent has the following inevitable defects: 1. It can be clearly seen from the invention content of the invention that the system is designed for a baking oven, which is used to implement tobacco leaf baking experiments, and is essentially different from the structure, tobacco loading and control system of a large intensive baking room used in actual production. The use of this prior art has considerable limitations and is not suitable for actual production. 2. The system uses tobacco leaf color spectrum data obtained by a near-infrared camera to judge the tobacco leaf baking process. The near-infrared spectral range is an extension of the visible light spectral range of humans, but does not include the visible light range. Therefore, near-infrared belongs to one of the multi-spectral spectra, and is also limited by the characteristics of the wide band of the multi-spectral spectrum. For example, in the prior art, the moisture data of tobacco leaves is monitored by weight method, and cannot be obtained through the spectral information of near-infrared images. However, the state changes of tobacco leaves during baking are multi-dimensional, including color, shape, moisture content, main vein state, etc. Judging the tobacco leaf baking process only from the color will often result in large deviations. 3. The final result of the system's judgment on the state of tobacco leaves is not clear, and it is impossible to judge the stage of the baking process according to the three-stage method currently used in production (yellowing period, color fixing period and dry tendon period). Summary of the invention

[0005] The purpose of the present invention is to provide a method for judging the stage of tobacco leaf baking process, to realize the automation of stage judgment of tobacco leaf baking control process, to improve the efficiency of judging the stage of baking process, and to make it suitable for large-scale intensive baking room structure, tobacco loading capacity and control requirements used in actual production.

[0006] The second purpose of the present invention is to provide a method for determining the stage of tobacco leaf baking process. The tobacco leaf state during the baking process, especially the tobacco leaf state indicators, are comprehensively analyzed in multiple dimensions to obtain a clear judgment result of the stage of the tobacco leaf baking process, so as to reduce the judgment error and further improve the homogenization of tobacco leaf baking.

[0007] The third purpose of the present invention is to provide a method for judging the stage of tobacco leaf baking process. The tobacco leaf baking process stage is monitored by using hyperspectral technology, and the image and spectral characteristics of the narrow band and high dimension of the hyperspectral technology are used to obtain the hyperspectral information of tobacco leaf state indicators at each stage of the baking process. The information is rich and the authenticity is high, which is used as the basis for judging the stage of the leaf baking process, thereby achieving the purpose of improving the accuracy of judging the tobacco leaf state.

[0008] The fourth object of the present invention is to provide a method for determining the stage of tobacco leaf baking process. By processing the rich hyperspectral images and spectral data collected during the tobacco leaf baking process, the comprehensiveness and accuracy of the tobacco leaf state spectral data used in determining the stage of the tobacco leaf baking process can be improved.

[0009] The object of the present invention is achieved in that:

[0010] To achieve the above-mentioned purpose, the present invention proposes a method for judging the stage of tobacco leaf baking process, wherein before actual baking, preset baking parameters of the entire baking process are input through a baking room control terminal; wherein, during the entire baking process of actual baking, high-spectral information of a tobacco leaf state index X is collected according to a calibrated data collection time t of the baking process stage; and characteristic spectral data of the tobacco leaf state index X at the collection time t is established based on the high-spectral information collected at the time t; the characteristic spectral data of the tobacco leaf state index X at the actual baking time t is compared with the pre-calibrated high-spectral information; and the stage of the current tobacco leaf baking process is judged by the comparison result.

[0011] The method for determining the stages of the tobacco leaf baking process as described above, wherein the tobacco leaf baking process stages include a yellowing period, a color fixing period and a dry tendon period; each of the yellowing period, the color fixing period and the dry tendon period can also be divided into three periods: an early period, a middle period and a final period.

[0012] The method for determining the stage of tobacco leaf baking process as described above, wherein the pre-calibrated baking process data collection time t includes the pre-calibrated information collection time t and the baking process stage at the time t; and the information collection time t is segmented according to the stage of the baking process.

[0013] The method for judging the stage of tobacco leaf baking process as described above, wherein the time t for information collection is segmented, and two levels of segmentation are adopted; the first level segmentation is to segment the yellowing stage, the color fixing stage and the dry tendon stage; the second level segmentation is to segment each process stage of the yellowing stage, the color fixing stage and the dry tendon stage into the early stage, the middle stage and the final stage.

[0014] In the above-mentioned method for determining the stage of tobacco leaf baking process, the data collection time t can be set according to the equal time interval of data collection; or can be set according to the segmented time of the baking process stage.

[0015] As described above, in the method for determining the stage of the tobacco leaf baking process, the data collection time t of the actual baking process is the same as the data collection time t of the pre-calibrated baking process; and a time tag m defining the data collection time t is added to the collected hyperspectral data.

[0016] In the method for determining the stage of tobacco leaf baking process as described above, the collected hyperspectral information is the hyperspectral information of tobacco leaf state indicators at the data collection time t during tobacco leaf baking.

[0017] The method for judging the stage of tobacco leaf baking process as described above, wherein the tobacco leaf state index is composed of seven tobacco leaf state indexes, namely, tobacco leaf moisture content, leaf color, main vein color, tobacco leaf color uniformity, main vein morphology, tobacco leaf morphology, and main vein moisture content.

[0018] As described above, the method for determining the stage of the tobacco leaf baking process, wherein the determination of each baking process stage selects at least one tobacco leaf state indicator X or a combination of more than one of the seven tobacco leaf state indicators.

[0019] As described above, in the method for determining the stage of the tobacco leaf baking process, the characteristic spectral data of the tobacco leaf state index collected at time t is composed of a hyperspectral characteristic matrix Xt of the tobacco leaf state index X constructed based on the original hyperspectral information of the tobacco leaf state index X.

[0020] In the above-mentioned method for determining the stage of tobacco leaf baking process, the hyperspectral feature matrix Xt of the tobacco leaf state index X at time t is constructed by the following method:

[0021] S1: Start the actual baking. During the baking process, collect the original hyperspectral information of tobacco leaves and construct the hyperspectral information matrix of actual baked tobacco leaves. The hyperspectral information matrix Xt at time t is:

[0022] Xt=(x1 … x n )

[0023] Where: t is the time of hyperspectral information collection, unit: min;

[0024] n is the dimension of hyperspectral information data;

[0025] S2: After data processing of matrix Xt, matrix Yt is formed:

[0026] Yt=(y1 … y n )

[0027] Where: t is the time of hyperspectral information collection, unit: min;

[0028] n is the dimension of hyperspectral information data;

[0029] S3: Segment the matrix Yt according to the respective characteristic bands of the pre-calibrated tobacco state index X to establish a tobacco characteristic index matrix Yxt corresponding to the tobacco characteristic index X:

[0030] Where t is the time of hyperspectral information collection, unit min;

[0031] xTobacco leaf condition index X.

[0032] The method for judging the stage of tobacco leaf baking process as described above, wherein the tobacco leaf characteristic index matrix Yxt is optimally divided into 7 tobacco leaf characteristic index matrices Yxt based on tobacco leaf moisture content, leaf color, main vein color, tobacco leaf color uniformity, main vein morphology, tobacco leaf morphology, and main vein moisture content.

[0033] The method for determining the stage of the tobacco leaf baking process as described above, wherein the data processing method described in step S2 is any one of the data processing methods such as correction, averaging, smoothing, differentiation, normalization, dimensionality reduction, etc., or a combination of two or more methods.

[0034] In the method for determining the stage of tobacco leaf baking process as described above, the data processing method in step S2 is preferably the same data processing method used when calibrating the spectral data characteristics of the corresponding tobacco leaf status indicators.

[0035] The method for judging the stage of tobacco leaf baking process as described above, wherein the characteristic spectral data of the tobacco leaf state index X at the actual baking time t is compared with the pre-calibrated high-spectral information, is to compare the characteristic spectral matrix Yxt of each tobacco leaf state index with the characteristic matrix Ct of the corresponding calibrated tobacco leaf state index.

[0036] The method for determining the stage of the tobacco leaf baking process as described above, wherein the characteristic matrix Yxt of the tobacco leaf state index is compared with the characteristic matrix Ct of the calibrated corresponding tobacco leaf state index, can be implemented by the following method:

[0037] S21: Start the actual baking. According to the original hyperspectral information of tobacco leaves collected at time t during the baking process, construct the hyperspectral information matrix Xt of the actual baking tobacco leaf state index X:

[0038] Xt=(x1 … x n )

[0039] S22: Process the hyperspectral information matrix Xt and construct the spectral feature matrix Yxt of each tobacco leaf status index X:

[0040] Yxt=(y1 … y n );

[0041] S23: Compare the characteristic matrix Yxt of each tobacco leaf state index X with the characteristic band matrix Ct of the calibrated tobacco leaf state index X.

[0042] In the above-mentioned method for determining the stage of tobacco leaf baking process, the comparison in step S23 can be further performed by the following method:

[0043] S321: Calculate the Euclidean distance between the characteristic matrix Yxt of the tobacco leaf state index X and the corresponding data of the spectral characteristic matrix Cxt of the calibrated tobacco leaf state index X;

[0044] S322: performing data processing on the calculated Euclidean distance values ​​of the corresponding points to obtain a distance matrix Zx of the tobacco leaf state index X;

[0045] S323: dividing the distance matrix Zx into distance matrix Zxt according to the stages of the baking process; wherein the distance matrix Zxt is taken as including all the distance matrix data at time t in different stages of the baking process;

[0046] S324: Calculate the average value of the data in the distance matrix Zxt of each baking process stage.

[0047] S325: Calculate the sum of the data average values ​​of the distance matrix Zxt of at least two tobacco leaf state indicators in the baking process stage;

[0048] S326: Determine the stage of the baking process or the period of the stage according to the sum of the data average values ​​of the distance matrix Zxt of the tobacco leaf state index.

[0049] The method for judging the stage of tobacco leaf baking process as described above, wherein the stage standard of the baking process is judged based on the sum of the average values ​​of the data of the distance matrix Zxt, and the early, middle and late periods of each stage are determined in descending order using the sum of the average values ​​of the data of the distance matrix Zxt; when the spectral characteristic data of the tobacco leaf state index when the distance matrix Zxt data is the smallest is closest to the calibrated spectral characteristics of the tobacco leaf state, it is judged that the stage of the tobacco leaf process is ended.

[0050] The method for determining the stage of tobacco leaf baking process as described above, wherein the determination of the stage of tobacco leaf baking process is performed in the order of yellowing stage, color fixing stage and dry tendon stage.

[0051] As described above, the method for judging the stage of tobacco leaf baking process can further judge the period of each stage, namely, the yellowing stage, the color fixing stage and the dry tendon stage, by selecting the data of the corresponding tobacco leaf state index X according to the baking process stage to be judged.

[0052] As described above, in the method for judging the stage of tobacco leaf baking process, for judging the yellowing stage, color fixing stage and dry tendon stage of tobacco leaf baking process, the tobacco leaf state index X is preferably selected as at least the moisture content and color of tobacco leaves.

[0053] As described above, in the method for judging the stage of tobacco leaf baking process, the tobacco leaf state index X is preferably selected as the main vein color and tobacco leaf color uniformity for judging the yellowing period of tobacco leaf baking process.

[0054] As described above, in the method for judging the stage of tobacco leaf baking process, the tobacco leaf state index X is preferably selected as the color of tobacco leaf main vein, the shape of tobacco leaf main vein and the shape of leaf blade for judging the color fixing period of tobacco leaf baking process.

[0055] As described above, in the method for judging the stage of tobacco leaf baking process, for judging the dry tendon stage of tobacco leaf baking process, the tobacco leaf state index X is preferably selected as the moisture content of the tobacco leaf main vein.

[0056] The method for judging the stages of tobacco leaf baking process as described above, wherein, after judging each process stage of yellowing stage, color fixing stage and dry tendon stage, the early, middle and final periods of each process stage can be further judged.

[0057] The method for judging the stages of the tobacco leaf baking process as described above, wherein the early, middle and final periods of each process stage, namely the yellowing period, the color fixing period and the dry tendon period, are judged; the basis for the judgment is the judgment data of all data collection moments t in each period of the early, middle and final periods of each process stage.

[0058] The technical effects of the present invention are remarkable:

[0059] The present invention provides a method for judging the stage of tobacco leaf baking process, which continuously collects tobacco leaf state data during the whole process of tobacco leaf baking, compares and processes the collected original hyperspectral data in real time, obtains comparative data of each stage of tobacco leaf baking process in real time, and realizes automation of stage judgment of tobacco leaf baking control process through model decision-making, thereby improving the efficiency of judging the stage of baking process. Thus, the defect of low efficiency in judging tobacco leaf baking stage by visual inspection combined with personal experience of baking personnel in the prior art is overcome. Experiments have proved that the present invention can meet the automatic control requirements of large-scale intensive baking room structure, tobacco loading capacity and control system used in actual production.

[0060] In addition, in the present invention, the data accuracy of the tobacco leaf status at each stage of the tobacco leaf baking process is greatly improved by collecting rich hyperspectral images and spectral data of the tobacco leaf baking process. The tobacco leaf status during the baking process, especially the tobacco leaf status indicators, is comprehensively analyzed in multiple dimensions to obtain a clear judgment result of the tobacco leaf baking process stage, thereby reducing the judgment error and improving the homogenization of tobacco leaf baking. This overcomes the defect of the prior art that the judgment result of the tobacco leaf baking stage is judged by visual inspection and personal experience, which has a large error and is not conducive to the improvement of the homogenization of tobacco leaf baking.

[0061] Furthermore, the present invention uses hyperspectral technology to monitor the tobacco leaf baking process. By using the narrow band, high-dimensional image and spectral characteristics of hyperspectral technology, the hyperspectral information of the tobacco leaf status at each stage of the baking process is obtained, which has the characteristics of rich information and high authenticity. Then, by processing the rich hyperspectral images and spectral data collected during the tobacco leaf baking process, the data accuracy of the tobacco leaf status at each stage of the tobacco leaf baking process is greatly improved.

[0062] At the same time, the method for judging the stage of tobacco leaf baking process of the present invention. Through the rich hyperspectral images and spectral data processing methods collected during the tobacco leaf baking process, the three stages used in production practice can be judged simultaneously during the tobacco leaf baking process, and the accuracy of all judgment bases is improved at the same time, without combining other judgment methods. Thereby overcoming the defects of the limitation of the range of tobacco leaf status indicators used in judging the baking process stages in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 A flow chart of a method for constructing a hyperspectral original information matrix of actual flue-cured tobacco leaves of the present invention;

[0064] Figure 2 A flow chart of a method for comparing actual and calibrated tobacco leaf state characteristic spectra of the present invention;

[0065] Figure 3 Flow chart of a specific embodiment of the method for comparing actual and calibrated tobacco leaf state characteristic spectra of the present invention.

[0066] The accompanying drawings described herein are only for explanation purposes and are not intended to limit the scope of the present invention in any way. In addition, the shapes and proportional dimensions of the various components in the figures are only schematic, used to help understand the present invention, and are not specifically limited to the shapes and proportional dimensions of the various components of the present invention. Those skilled in the art can select various possible shapes and proportional dimensions to implement the present invention according to the teachings of the present invention. DETAILED DESCRIPTION

[0067] Example 1: The tobacco leaves in this example are Cui Bi No. 1 middle tobacco leaves produced in Sanming, Fujian.

[0068] The present invention provides a method for judging the stage of tobacco leaf baking process. Before actual baking, preset baking parameters of the entire baking process are input through a baking room control terminal. During the entire baking process of actual baking, the hyperspectral information of tobacco leaf state index X is collected according to the calibrated data collection time t of the baking process stage; and characteristic spectrum data of tobacco leaf state index X at the collection time t is established based on the hyperspectral information collected at the collection time t. The characteristic spectrum data of the tobacco leaf state index X at the actual baking time t is compared with the pre-calibrated hyperspectral information, and the stage of the current tobacco leaf baking process is judged by the comparison result.

[0069] In the present invention, tobacco leaf state data during tobacco leaf baking is continuously collected during the whole process of tobacco leaf baking, and the collected original hyperspectral data is compared and processed in real time, and the comparative data of each stage in the tobacco leaf baking process is obtained in real time. Through model decision-making, the stage judgment of the tobacco leaf baking control process is automated, and the efficiency of judging the stage of the baking process is improved. Thus, the defect of the prior art that the tobacco leaf baking stage is judged by visual inspection combined with the personal experience of the baking personnel is overcome, which is inefficient.

[0070] In the present invention, the stage of tobacco leaf baking process is monitored by using hyperspectral technology to obtain rich hyperspectral information of tobacco leaf state at each stage of the baking process. Therefore, the hyperspectral information of tobacco leaf state has the characteristics of high authenticity, which is used as the basis for judging the stage of leaf baking process, greatly improving the accuracy of judging tobacco leaf state.

[0071] In particular, the present invention can improve the data accuracy of tobacco leaf status at each stage of the tobacco leaf baking process by processing the rich hyperspectral images and spectral data collected during the tobacco leaf baking process. By comparing the obtained data of tobacco leaf status at each stage of the baking process with the pre-calibrated data of tobacco leaf status at each stage, the accuracy of judgment at each stage of the tobacco leaf baking process is effectively improved.

[0072] In the present invention, the tobacco leaf baking process includes at least a yellowing period, a color fixing period and a dry tendon period to meet the actual production needs of tobacco leaf baking. In a preferred embodiment of the present invention, each of the yellowing period, the color fixing period and the dry tendon period can be divided into an early stage, a middle stage and a late stage, so that a more accurate judgment of the yellowing period, the color fixing period and the dry tendon period can be achieved.

[0073] It should be noted that the calibration method of pre-calibrated hyperspectral information proposed in the present invention was applied for separately on the same day as the present case.

[0074] In the first embodiment of the present invention, the pre-calibrated baking process data collection time t includes the pre-calibrated information collection time t and the baking process stage at the time t; and the information collection time t is segmented according to the baking process stage.

[0075] A preferred implementation of this embodiment 1 is that two levels of segmentation are used for segmenting the time t of information collection. The first level segmentation is segmenting the yellowing stage, the color fixing stage and the tendon drying stage. The second level segmentation is segmenting each process stage of the yellowing stage, the color fixing stage and the tendon drying stage into the early stage, the middle stage and the final stage.

[0076] Specifically in the present embodiment 1, the collection time t of the tobacco leaf hyperspectral information is set at an interval of 1 hour. This baking takes 160 hours, and the baking stages of this batch of tobacco leaves are segmented as follows: the 1st to 100th hour is the yellowing stage, the 101st to 135th hour is the color fixing stage, and the 136th to 160th hour is the dry tendon stage. Among them: the 0th to 33th hour is the early yellowing stage, the 34th to 66th hour is the middle yellowing stage, and the 67th to 100th hour is the late yellowing stage. The 101st to 112th hour is the early color fixing stage, among which: the 113th to 124th hour is the middle color fixing stage, and the 125th to 135th hour is the late color fixing stage. The 136th to 144th hour is the early dry tendon stage, among which: the 145th to 153th hour is the middle dry tendon stage, and the 153th to 160th hour is the late dry tendon stage.

[0077] In the above-mentioned possible implementation of this embodiment 1, the data collection time t can be set according to the equal time interval of data collection of 1 hour. This implementation can improve the data collection density.

[0078] In another implementation of this embodiment 1, the segmented time can be set according to the calibrated baking process stage. This implementation reduces the data processing and computing volume by reducing the data collection density, and is suitable for devices with insufficient computing power.

[0079] In this embodiment 1, the data collection time t of the actual baking process is the same as the data collection time t of the pre-calibrated baking process, and a time tag m defining the data collection time t is added to the collected hyperspectral data. Specifically, in this embodiment, a time tag m is added to each collected hyperspectral information, which is the 1st hour, the 2nd hour, ... the 160th hour.

[0080] In the present embodiment 1, the collected hyperspectral information is the hyperspectral information of the tobacco leaf state index at the data collection time t during tobacco leaf baking. The tobacco leaf state index is composed of 7 tobacco leaf state indexes, namely, tobacco leaf moisture content, leaf color, main vein color, tobacco leaf color uniformity, main vein morphology, tobacco leaf morphology, and main vein moisture content. Long-term and extensive production practice has proved that the 7 tobacco leaf state indicators in the invention are sufficient to express the state of tobacco leaves under baking. Therefore, it is not necessary to use all the huge information collected by hyperspectral technology, which is conducive to reducing the pressure and intensity of computer processing and calculation of information data, thereby speeding up the speed of data feedback.

[0081] A preferred implementation of this embodiment 1 is that, for the determination of each stage of the baking process, at least one tobacco leaf state indicator X or a combination of more than one of the seven tobacco leaf state indicators is selected.

[0082] Specifically in this embodiment 1: for the determination of the yellowing stage, color fixing stage and dry rib stage during the baking process, the tobacco leaf state index X is selected as the tobacco leaf moisture content X1 and the leaf color X2 as the determination basis. Experiments have shown that

[0083] In this embodiment, the characteristic spectrum data of the tobacco leaf state index collected at time t is composed of a hyperspectral feature matrix Xt of the tobacco leaf state index X constructed based on the original hyperspectral information of the tobacco leaf state index X.

[0084] In one specific embodiment, the hyperspectral feature matrix Xt of the tobacco leaf state index X at time t is constructed using the following method: Figure 1 As shown:

[0085] S1: Start the actual baking. During the baking process, collect the original hyperspectral information of tobacco leaves and construct the hyperspectral information matrix of actual baked tobacco leaves. The hyperspectral information matrix Xt at the acquisition time t is:

[0086] Xt=(x1 … x n )

[0087] Where: t is the time of hyperspectral information collection, unit: min;

[0088] n is the dimension of hyperspectral information data;

[0089] S2: After data processing of matrix Xt, matrix Yt is formed:

[0090] Yt=(y1 … y n )

[0091] Where: t is the time of hyperspectral information collection, unit: min;

[0092] n is the dimension of hyperspectral information data;

[0093] S3: Segment the matrix Yt according to the respective characteristic bands of the pre-calibrated tobacco state index X to establish a tobacco characteristic index matrix Yxt corresponding to the tobacco characteristic index X:

[0094] Where t is the time of hyperspectral information collection, unit min;

[0095] xTobacco leaf condition index X.

[0096] Specifically in this embodiment 1, according to the original hyperspectral information of the tobacco leaf moisture content index X1 and the leaf color index X2 at the acquisition time t, characteristic bands are extracted after data processing, wherein 2 characteristic bands are screened out for tobacco leaf moisture content and 5 characteristic bands are screened out for tobacco leaf color. The tobacco leaf characteristic index band matrices of the tobacco leaf moisture content index X1 and the leaf color index X2 are obtained respectively:

[0097] Where n = 2; m = 1 to 160

[0098] Where n = 5; m = 1 to 160

[0099] In the present invention, the tobacco leaf characteristic index matrix Yxt is optimally divided into 7 tobacco leaf index characteristic matrices Yxt according to tobacco leaf moisture content, leaf color, main vein color, tobacco leaf color uniformity, main vein morphology, tobacco leaf morphology, and main vein moisture content.

[0100] Specifically in this embodiment 1, the same method as above is used to finally form 7 characteristic band matrices, namely, tobacco leaf moisture content index X1, leaf color index X2, main vein color index X3, tobacco leaf color uniformity index X4, main vein morphology index X5, tobacco leaf morphology index X6, and main vein moisture content index X7.

[0101] The data processing method described in step S2 in this embodiment 1 is any one of the data processing methods such as correction, averaging, smoothing, differentiation, normalization, dimension reduction, etc., or a combination of two or more methods.

[0102] In a preferred implementation of this embodiment 1, the data processing method in step S2 is the same data processing method used when calibrating the spectral data characteristics of the corresponding tobacco leaf state indicators.

[0103] In this embodiment 1, the characteristic spectral data of the tobacco leaf state index X at the actual baking time t is compared with the pre-calibrated hyperspectral information, that is, the characteristic spectral matrix Yxt of each tobacco leaf state index X is compared with the spectral characteristic matrix C of the corresponding calibrated tobacco leaf state index X.

[0104] In one embodiment of the present invention, the characteristic matrix Yxt of the tobacco leaf state index is compared with the spectral characteristic matrix C of the calibrated corresponding tobacco leaf state index, and the following method can be used, specifically as follows: Figure 2 As shown:

[0105] S21: Start the actual baking. According to the original hyperspectral information of tobacco leaves collected at time t during the baking process, construct the hyperspectral information matrix Xt of the actual baking tobacco leaf state index X:

[0106] Xt=(x1 … x n )

[0107] S22: Process the hyperspectral information matrix Xt and construct the spectral feature matrix Yxt of each tobacco leaf status index X:

[0108] Yxt=(y1 … y n );

[0109] S23: Compare the characteristic matrix Yxt of each tobacco leaf state index X with the characteristic band matrix Cxt of the calibrated tobacco leaf state index X.

[0110] The comparison in step S23 can be further performed by the following method: Figure 3 As shown:

[0111] S321: Calculate the Euclidean distance between the characteristic matrix Yxt of the tobacco leaf state index X and the corresponding data of the spectral characteristic matrix Cxt of the calibrated tobacco leaf state index X;

[0112] S322: performing data processing on the calculated Euclidean distance values ​​of the corresponding points to obtain a distance matrix Zx of the tobacco leaf state index X;

[0113] S323: dividing the distance matrix Zx into distance matrix Zxt according to the stages of the baking process; wherein the distance matrix Zxt is taken as including all the distance matrix data at time t in different stages of the baking process;

[0114] S324: Calculate the average value of the data in the distance matrix Zxt of each baking process stage.

[0115] S325: Calculate the sum of the data average values ​​of the distance matrix Zxt of at least two tobacco leaf state indicators in the baking process stage;

[0116] S326: Determine the stage of the baking process according to the sum of the data average values ​​of the distance matrix Zxt of the tobacco leaf state index.

[0117] Specifically, in this embodiment 1, the distance matrix Z of the tobacco leaf moisture index X1 and the leaf color index X2 obtained in steps S321 to S323 is X1 and Z X2 They are

[0118]

[0119]

[0120] Among them, the distance matrix Z of tobacco leaf moisture index X1 and leaf color index X2 X1 and Z X2 The data includes all the distance matrix data at time t during the yellowing stage of the baking process.

[0121] Specifically, in this embodiment 1, according to the method of steps S324 to S326, the matrix Z is respectively X1 and Z X2 The distance matrix Z of the leaf color index X2 is obtained by averaging the data of columns 1-100, 101-135, and 136-160. X1 and Z X2 The mean values ​​of the column data obtained at each stage are listed as follows:

[0122] Table 1

[0123] matrix m＝1-100 m＝101-135 m＝136-160 <![CDATA[Z X1 ]]> 0.245 0.419 0.304 <![CDATA[Z X2 ]]> 0.364 0.664 0.718 Mean Sum 0.609 1.083 1.022

[0124] In the present invention, the stage standard for judging the baking process based on the sum of the data average values ​​of the distance matrix Zx1t is to determine the early, middle and final stages of each stage in descending order using the sum of the data average values ​​of the distance matrix Zxt; the spectral characteristic data of the tobacco leaf state index at the minimum is closest to the calibrated tobacco leaf state spectral characteristics, and it is judged that the stage of the tobacco leaf process has ended.

[0125] Specifically in this embodiment 1, the sum of the means of the above list data columns 1-100 is the smallest, indicating that the spectral characteristics of the tobacco leaves at this time are closest to the spectral characteristics of the tobacco leaves yellowing period m=1~100h, and it is determined that the tobacco leaves are roasted in the yellowing period at this time.

[0126] In the present invention, the tobacco leaf baking process stage is judged according to the order of yellowing stage, color fixing stage and dry tendon stage.

[0127] In a feasible implementation mode of the present invention, the data of the corresponding tobacco leaf state index X can be further selected based on the judged yellowing period, color fixing period and dry tendon period of the baking process, and the early, middle and final stages of each of the yellowing period, color fixing period and dry tendon period can be judged.

[0128] In a feasible embodiment of the present invention, for judging the early, middle and final stages of the yellowing period during the tobacco leaf baking process, the tobacco leaf state index X is preferably selected as the tobacco leaf moisture content and the tobacco leaf color.

[0129] Specifically, in this embodiment 1, after determining that the tobacco leaves are in the yellowing stage, the spectrum matrix Z of the main vein color index X3 is used. X3 and tobacco leaf color uniformity spectrum matrix Z X4 , and further judge whether the tobacco leaves are in the early, middle and late stages of the yellowing period. X1 and Z X2 Using the same method, the mean values ​​of the column data obtained are listed as follows:

[0130] Table 2

[0131] matrix m＝1-33 m＝34-66 m＝67-100 <![CDATA[Z3]]> 0.651 0.347 0.288 <![CDATA[Z4]]> 0.415 0.354 0.114 Mean Sum 1.066 0.701 0.402

[0132] The data in the table show that the sum of the label m data at each moment when m=1-100 in the yellowing period can be obtained, and the sum of the data is the smallest when m=67-100, so it is determined that m=67-100 is at the end of the yellowing period.

[0133] Further, at the end of the yellowing period when m=67-100, the data of all the label m at each moment are summed up, and the results are as shown in Table 3:

[0134] Table 3

[0135] Time label m Row data sum 67 0.148 68 0.483 69 0.375 … … 99 0.119 100 0.014

[0136] The data in the list show that when m=100, the total data is the smallest, indicating that the spectral characteristics of the tobacco leaves at this time are closest to the spectral characteristics of the calibrated end time of the tobacco leaves yellowing period at 100h. It is determined that the tobacco leaves should be roasted at the end of the yellowing period at this time.

[0137] In the present invention, the judgment of the early, middle and late stages of the color fixing period during the tobacco leaf baking process adopts the same method as the yellowing period. The state index X adopted is preferably selected from the color of the tobacco leaf main vein, the shape of the tobacco leaf main vein and the shape of the leaf.

[0138] The determination of the early, middle and final stages of the color fixing period during the tobacco leaf baking process is performed in the same manner as the yellowing period. The tobacco leaf state index X is preferably selected as the moisture content of the tobacco leaf main vein.

[0139] The preferred selection scheme for the tobacco leaf state index X during the tobacco leaf baking process of the present invention is described in the following table, and each specific implementation is no longer described one by one (see Table 4):

[0140] Table 4

[0141]

[0142]

[0143] In the present invention, the judgment of the early, middle and final stages of each process stage of the yellowing stage, the color fixing stage and the tendon drying stage is based on the judgment data of all data collection time t in each sub-stage of the early, middle and final stages of each process stage. Since the applicant filed another application on the same day for the calibration method of the basis for the above judgment, it will not be repeated in this case.

[0144] Since the present invention processes the rich hyperspectral images and spectral data collected during the tobacco leaf baking process, the data accuracy of the tobacco leaf status at each stage of the tobacco leaf baking process is greatly improved. By comparing the obtained data of the tobacco leaf status at each stage of the baking process with the pre-calibrated data of the tobacco leaf status at each stage, the accuracy of the judgment at each stage of the tobacco leaf baking process is effectively improved.

[0145] In addition, in the present invention, tobacco leaf state data during tobacco leaf baking is continuously collected in real time throughout the whole process of tobacco leaf baking, and the collected original hyperspectral data can be processed in real time, and compared with the evaluation standards in real time and directly, so as to obtain the comparative data of each stage in the tobacco leaf baking process in real time, and realize the automation of stage judgment of tobacco leaf baking control process through model decision-making. The method of the present invention is conducive to realizing AI control of stage judgment of tobacco leaf baking control process through the accumulation of a large amount of data in production practice.

[0146] Example 2

[0147] This Example 2 uses the same tobacco leaf type as Example 1. The method used is basically the same as that of Example 1, with the first difference being that the tobacco leaf moisture content X1 is eliminated from the two characteristic indicators of tobacco leaf moisture content X1 and leaf color X2 selected in Example 1 for judging the tobacco leaf baking stage, and only the leaf color is used to judge the baking stage. The second difference is that when judging the early, middle and late stages of the yellowing period, the main vein color indicator is eliminated, that is, only color uniformity is used for period judgment. The final results of the experiment show that when the above two indicators are eliminated for judgment, the accuracy of stage judgment and period judgment will decrease (see Table 5).

[0148] Table 5

[0149] test Stage judgment accuracy / % Period judgment accuracy / % Example 1 98.3 98.9 Example 2 84.1 76.5

[0150] The test data of this embodiment 2 show that when judging the stages of the baking process or the period of each process, appropriately selecting the type of tobacco leaf state index X is beneficial to the accuracy of stage judgment and period judgment.

[0151] Example 3

[0152] This Example 2 uses the same tobacco leaf type as Example 1. The method used is basically the same as that of Example 1, except that four tobacco leaf state indicators X, namely, tobacco leaf moisture content X1, leaf color X2, main vein color X3 and color uniformity X4, are used as the judgment indicators for the baking stage and baking period. The final results of the experiment show that the judgment accuracy can be slightly improved by using the method of Example 2. However, the difference in accuracy with Example 1 is not significant, and the method of Example 3 greatly increases the computing pressure of the computer.

[0153] Table 6

[0154] test Stage judgment accuracy / % Period judgment accuracy / % Total calculation time / s Example 1 98.3 98.9 128 Example 3 98.5 99.3 315

[0155] The test data of Example 3 show that the increase in the types of tobacco leaf state indicators X has a relatively large impact on the data calculation time when the stages or periods of the baking process are being judged. Appropriately selecting the number of types of tobacco leaf state indicators X is conducive to improving the feedback speed of data for stage judgment and period judgment.

[0156] The detailed explanation of the above-mentioned embodiments is only intended to explain the present invention so as to facilitate a better understanding of the present invention. However, these descriptions cannot be interpreted as limitations on the present invention for any reason. In particular, the various features described in different embodiments may also be arbitrarily combined with each other to form other embodiments. Unless there is a clear description to the contrary, these features should be understood to be applicable to any embodiment and are not limited to the described embodiments.

Claims

1. A method for determining the stage of tobacco leaf baking process, wherein before actual baking, preset baking parameters of the entire baking process are input through a baking room control terminal; characterized in that: In the whole baking process of actual baking, the hyperspectral information of tobacco leaf state index X is collected according to the data collection time t of the calibrated baking process stage; and characteristic spectrum data of tobacco leaf state index X at the collection time t is established based on the hyperspectral information collected at the time t; the characteristic spectrum data of tobacco leaf state index X at the actual baking time t is compared with the pre-calibrated hyperspectral information; and the stage of the current tobacco leaf baking process is determined by the comparison result; The tobacco leaf condition index is composed of seven tobacco leaf condition indexes, namely, tobacco leaf moisture content, leaf color, main vein color, tobacco leaf color uniformity, main vein morphology, tobacco leaf morphology, and main vein moisture content; For determining each stage of the baking process, at least two combinations of the seven tobacco leaf status indicators are selected.

2. The method for determining the stage of tobacco leaf baking process according to claim 1, characterized in that: The tobacco leaf baking process stages include yellowing stage, color fixing stage and dry tendon stage; each stage of the yellowing stage, color fixing stage and dry tendon stage is further divided into three periods: early stage, middle stage and final stage.

3. The method for determining the stage of tobacco leaf baking process according to claim 1, characterized in that: The pre-calibrated baking process data collection time t includes the pre-calibrated information collection time t and the baking process stage at the time t; and the information collection time t is segmented according to the stage of the baking process.

4. The method for determining the stage of tobacco leaf baking process according to claim 3, characterized in that: The time t of information collection is segmented in two levels; the first level segmentation is to segment the yellowing period, the color fixing period and the tendon drying period; the second level segmentation is to segment each process stage of the yellowing period, the color fixing period and the tendon drying period into the early, middle and final periods.

5. The method for determining the stage of tobacco leaf baking process according to claim 1 or 4, characterized in that: The data collection time t can be set according to the equal time intervals of data collection; or according to the segmented time of the baking process stage.

6. The method for determining the stage of tobacco leaf baking process according to claim 1, characterized in that: The data collection time t of the actual baking process is the same as the data collection time t of the pre-calibrated baking process; and a time tag m defining the data collection time t is added to the collected hyperspectral data.

7. The method for determining the stage of tobacco leaf baking process according to claim 1, characterized in that: The collected hyperspectral information is the hyperspectral information of tobacco leaf status indicators at the data collection time t during tobacco leaf baking.

8. The method for determining the stage of tobacco leaf baking process according to claim 1, characterized in that: The characteristic spectrum data of the tobacco leaf state index collected at the time t is composed of a hyperspectral characteristic matrix Xt of the tobacco leaf state index X constructed based on the original hyperspectral information of the tobacco leaf state index X.

9. The method for determining the stage of tobacco leaf baking process according to claim 8, characterized in that: The hyperspectral feature matrix Xt of the tobacco leaf state index X at time t is constructed using the following method: S1: Start the actual baking. During the baking process, collect the original hyperspectral information of tobacco leaves and construct the hyperspectral information matrix of actual baked tobacco leaves. The hyperspectral information matrix Xt at time t is: Xt= Where: t is the time of hyperspectral information collection, unit min; n is the dimension of hyperspectral information data; S2: After data processing of matrix Xt, matrix Yt is formed: Yt= Where: t is the time of hyperspectral information collection, unit min; n is the dimension of hyperspectral information data; S3: Segment the matrix Yt according to the respective characteristic bands of the pre-calibrated tobacco state index X to establish a tobacco characteristic index matrix Yxt corresponding to the tobacco characteristic index X: Where t is the time of hyperspectral information collection, unit min; x Tobacco leaf condition index X.

10. The method for determining the stage of tobacco leaf baking process according to claim 6 or 9, characterized in that: The tobacco leaf characteristic index matrix Yxt is divided into 7 tobacco leaf characteristic index matrices Yxt according to tobacco leaf moisture content, leaf color, main vein color, tobacco leaf color uniformity, main vein morphology, tobacco leaf morphology, and main vein moisture content.

11. The method for determining the stage of tobacco leaf baking process according to claim 8, characterized in that: The data processing method described in step S2 is any one of the correction, averaging, smoothing, differentiation, normalization, and dimension reduction data processing methods, or a combination of two or more methods.

12. The method for determining the stage of tobacco leaf baking process according to claim 9, characterized in that: The data processing method in step S2 is the same data processing method used when calibrating the spectral data characteristics of the corresponding tobacco leaf state indicators.

13. The method for determining the stage of tobacco leaf baking process according to claim 8, characterized in that: The method of comparing the characteristic spectral data of the tobacco leaf state index X at the actual baking time t with the pre-calibrated hyperspectral information is to compare the characteristic spectral matrix Yxt of each tobacco leaf state index with the characteristic matrix Ct of the corresponding calibrated tobacco leaf state index.

14. The method for determining the stage of tobacco leaf baking process according to claim 13, characterized in that: The characteristic matrix Yxt of the tobacco leaf state index is compared with the characteristic matrix Ct of the calibrated corresponding tobacco leaf state index, and the following method can be used: S21: Start the actual baking. According to the original hyperspectral information of tobacco leaves collected at time t during the baking process, construct the hyperspectral information matrix Xt of the actual baking tobacco leaf state index X: Xt= ; S22: Process the hyperspectral information matrix Xt and construct the spectral feature matrix Yxt of each tobacco leaf status index X: Yxt= ; S23: Compare the characteristic matrix Yxt of each tobacco leaf state index X with the characteristic band matrix Ct of the calibrated tobacco leaf state index X.

15. The method for determining the stage of tobacco leaf baking process according to claim 14, characterized in that: The comparison in step S23 may be further performed by the following method: S321: Calculate the Euclidean distance between the characteristic matrix Yxt of the tobacco leaf state index X and the corresponding data of the spectral characteristic matrix Cxt of the calibrated tobacco leaf state index X; S322: performing data processing on the calculated Euclidean distance values ​​of the corresponding points to obtain a distance matrix Zx of the tobacco leaf state index X; S323: dividing the distance matrix Zx into distance matrix Zxt according to the stages of the baking process; wherein the distance matrix Zxt is taken as including all the distance matrix data at time t in different stages of the baking process; S324: Calculate the average value of the data in the distance matrix Zxt of each baking process stage; S325: Calculate the sum of the data average values ​​of the distance matrix Zxt of at least two tobacco leaf state indicators in the baking process stage; S326: Determine the stage of the baking process or the period of the stage according to the sum of the data average values ​​of the distance matrix Zxt of the tobacco leaf state index.

16. The method for determining the stage of tobacco leaf baking process according to claim 15, characterized in that: The standard for judging the stage of the baking process based on the sum of the data averages of the distance matrix Zxt is to determine the early, middle and final periods of each stage in order from large to small using the sum of the data averages of the distance matrix Zxt; when the spectral characteristic data of the tobacco leaf state index when the distance matrix Zxt data is the smallest is closest to the calibrated tobacco leaf state spectral characteristics, it is judged that the stage of the tobacco leaf process is over.

17. The method for determining the stage of tobacco leaf baking process according to claim 1 or 4, characterized in that: The determination of the tobacco leaf baking process stage is carried out according to the order of yellowing stage, color fixing stage and dry tendon stage.

18. The method for determining the stage of tobacco leaf baking process according to claim 1 or 17, characterized in that: According to the baking process stage to be judged, the data of the corresponding tobacco leaf state index X is selected to further judge the period of each stage, namely the yellowing stage, the color fixing stage and the dry tendon stage.

19. The method for determining the stage of tobacco leaf baking process according to claim 18, characterized in that: For judging the yellowing stage, color fixing stage and dry tendon stage in the tobacco leaf baking process, the tobacco leaf state index X is selected to be at least the tobacco leaf moisture content and the tobacco leaf color.

20. The method for determining the stage of tobacco leaf baking process according to claim 18, characterized in that: To judge the yellowing period of tobacco leaves during the baking process, the tobacco leaf status index X is selected as the main vein color and the uniformity of tobacco leaf color.

21. The method for determining the stage of tobacco leaf baking process according to claim 18, characterized in that: For judging the color-fixing period of the tobacco leaf baking process, the tobacco leaf state index X is selected as the color of the tobacco leaf main vein, the shape of the tobacco leaf main vein and the shape of the leaf.

22. The method for determining the stage of tobacco leaf baking process according to claim 18, characterized in that: The early, middle and final periods of each process stage of the yellowing period, color fixing period and tendon drying period are judged; the basis for the judgment is the judgment data of all data collection moments t in each period of the early, middle and final periods of each process stage.

Images