Typical user selection method and system for tobacco yield prediction

Abstract

The invention provides a typical user selection method and system for tobacco yield prediction, and belongs to the field of tobacco pre-production, and the method comprises the steps: obtaining a remote sensing image of a research region; calculating the weight of each factor by adopting an expert scoring method and a variable coefficient method to obtain a factor proportion result; making a scoring standard of each factor in the research region; based on the factor proportion result and the scoring standard, establishing a planting level model, and substituting the tobacco field information of each tobacco grower in the research area for calculation to obtain a planting level score of each tobacco grower; the NDVI of the tobacco field is calculated based on the remote sensing image, and the current-season tobacco growth vigor of each tobacco grower is obtained after the average value is taken; the tobacco growers are classified according to the planting level score and the current-season tobacco growth vigor, and a tobacco grower typical household list is obtained. According to the scheme, the planting level model and the tobacco field mature period NDVI are combined, the current-season planting level and the tobacco leaf quality of tobacco growers can be analyzed and evaluated more comprehensively, a tobacco station is helped to select typical households with a more scientific and accurate method, and the tobacco leaf purchasing and allocating efficiency is improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of tobacco yield prediction, and particularly relates to a typical household selection method and system for tobacco leaf yield prediction. BACKGROUND

[0002] The tobacco industry in China follows the Tobacco Monopoly Law of the People's Republic of China and adopts the mode of planned planting and planned purchase. Each year, the tobacco company and its commissioned units purchase all the tobacco leaves produced according to the planting area stipulated in the tobacco leaf purchase contract. However, because the climate and agricultural level each year are not the same, the yield and quality of the tobacco leaves produced from the same planting area also have certain fluctuations, so the yield and quality after reprocessing need to be predicted each year. The planning department of the State Council allocates the plan according to the prediction results of each province, autonomous region and municipality, so as to reasonably complete the overall tobacco leaf purchase. Therefore, yield prediction and quality prediction (referred to as "pre-production") are an important part of the tobacco purchase work, and provide an important decision basis for tobacco purchase and purchase allocation.

[0003] At present, the tobacco pre-production mainly adopts the method of selecting typical households from tobacco farmers. Specifically, the tobacco technicians of each tobacco station determine the typical households that can represent the local tobacco planting conditions this year according to the understanding of individual tobacco farmers and the planting conditions obtained by visiting and investigating the tobacco fields during the planting process, and infer the local tobacco yield and quality conditions this year based on the yield and quality of the typical households after harvesting and reprocessing. Generally, a tobacco technician selects three tobacco farmers, which represent good, medium and poor planting conditions. It should be noted that the specific number of typical households selected may vary according to the area of the tobacco production area and the size of the tobacco station, but the principle of selecting typical households is the same. After the typical households are selected, the yield of different grades of tobacco leaves of the typical households, the proportion of each grade of upper, middle and lower tobacco leaves are counted after the tobacco leaves are harvested and reprocessed, the yield and quality of the tobacco planting this year in the local area are predicted from the results of the typical households, and the data are uniformly reported to the higher level, and finally provide data for the decision of the tobacco purchase and allocation plan of this year.

[0004] Because plants have strong reflection in the near-infrared band and high reflectivity, and have strong absorption in the red band and low reflectivity, the normalized difference vegetation index (NDVI) can quantitatively reflect the growth conditions of vegetation by calculating the difference between the near-infrared band and the red band. In typical spectral vegetation indices, NDVI is one of the indices most suitable for monitoring crop growth dynamics, and can effectively reflect the growth and health of crops.

[0005] The current tobacco yield prediction method mainly depends on the experience and subjectivity of the tobacco technicians of the local tobacco station, and cannot exclude the abnormal situation that the typical household may have a fluctuation in the planting level this year or the tobacco field of the typical household suffers from disasters, so the current selection of the typical household is prone to deviation, thereby causing a large deviation in the final prediction of the yield and quality or leading to multiple prediction results.

[0006] The yield of tobacco planting is closely related to the operation of each planting link, and the existence of diseases and meteorological disasters may greatly affect the planting situation, and the tobacco technicians mainly rely on visits to understand the situation of the tobacco field, and the visual and experience-based identification method is difficult to comprehensively and quantitatively reflect the planting situation of the season.

[0007] The existing application discloses a selection and analysis method and device for a typical representative area of a corn planting environment (patent application number: CN201610221767.6), which selects a typical representative area of a corn planting environment, and cannot be applied to the selection of a typical household for tobacco yield prediction, and the application of the typical representative area and the typical household is not the same. SUMMARY

[0008] The present application provides a typical household selection method and system for tobacco yield prediction, aiming to solve the problems that the determination of the existing typical household relies on human experience and subjectivity, is prone to deviation, and the tobacco technicians rely on visual and experience-based identification to understand the situation of the tobacco field, and it is difficult to comprehensively and quantitatively reflect the planting situation of the season.

[0009] In order to achieve the above purpose, the application adopts the following technical scheme, comprising:

[0010] Determine a research area, and obtain a remote sensing image of the research area;

[0011] According to the experience of provincial and municipal experts, factors affecting the planting level in the research area are screened, and the weights of the factors are calculated by using the expert scoring method and the coefficient of variation method, to obtain a factor proportion result;

[0012] According to the experience of municipal or county experts, a scoring standard of each factor in the research area is formulated;

[0013] Based on the factor proportion result and the scoring standard, a planting level model is established, and the tobacco field information of each tobacco farmer in the research area is substituted into the planting level model to calculate the planting level score of each tobacco farmer;

[0014] Based on the remote sensing image, the NDVI of the tobacco field corresponding to each tobacco farmer in the research area is calculated, and the average value is taken to obtain the tobacco growth of each tobacco farmer in the season;

[0015] The tobacco farmers in the study area are classified according to the planting level score and the tobacco growth in the current season, and the list of typical households of different categories is obtained.

[0016] As preferred, the expression for calculating the weight of each factor is wherein X is the weight, X a is the weight of factor a, and a, b,.., n are different factors respectively, CV is the coefficient of variation, represents the mean of the expert score of factor n, and σn represents the standard deviation of the expert score of factor n.

[0017] As preferred, the expression of the planting level model is Y=A n *X1+B n *X2+...+N n *X n , wherein Y is the planting level score, A n ~N n are the scores obtained by the tobacco farmer according to the scoring standard in the factor, X1~X n are equivalent to X a ~X n , and a, b,.., n are the weights of different factors respectively.

[0018] As preferred, the factors affecting the planting level include but are not limited to the years of tobacco planting experience, the education level, whether the tobacco farmer has received planting training, the degree of mechanization, the average yield per mu of land in the past three years, the average yield per mu of land in the past three years, the proportion of high-quality tobacco, medium-quality tobacco, and low-quality tobacco in the upper part of the tobacco leaves when purchased last year, the proportion of high-quality tobacco, medium-quality tobacco, and low-quality tobacco in the middle part of the tobacco leaves when purchased last year, the proportion of high-quality tobacco, medium-quality tobacco, and low-quality tobacco in the lower part of the tobacco leaves when purchased last year, and the subjective score of the planting level in previous years.

[0019] As preferred, the scoring standard of the same factor is different according to different study areas.

[0020] As preferred, the tobacco farmers in the study area are classified according to the planting level score and the tobacco growth in the current season, and the list of typical households of different categories is obtained, specifically as follows:

[0021] The planting level scores of the tobacco farmers are sorted from high to low and divided into three categories: good planting level, medium planting level, and poor planting level;

[0022] The tobacco growth in the current season of the tobacco farmers is sorted from high to low and divided into three categories: good tobacco growth in the current season, medium tobacco growth in the current season, and poor tobacco growth in the current season;

[0023] The tobacco farmers with good planting level and good seasonal tobacco growth, the tobacco farmers with medium planting level and medium seasonal tobacco growth, and the tobacco farmers with poor planting level and poor seasonal tobacco growth are combined to obtain the typical household list of the good, medium and poor categories of tobacco farmers.

[0024] Preferably, the division criteria of dividing into three categories of good planting level, medium planting level and poor planting level or dividing into three categories of good seasonal tobacco growth, medium seasonal tobacco growth and poor seasonal tobacco growth are equally allocated.

[0025] A typical household selection system for tobacco yield prediction comprises:

[0026] An image acquisition module is configured to determine a research area and acquire remote sensing images of the research area.

[0027] A factor weight calculation module is configured to filter factors affecting planting level in the research area according to the experience of provincial and municipal experts, and calculate the weights of the factors by using expert scoring method and coefficient of variation method to obtain a factor proportion result.

[0028] A scoring standard establishment module is configured to establish scoring standards of the factors in the research area according to the experience of municipal or county-level experts.

[0029] A planting level score calculation module is configured to establish a planting level model based on the factor proportion result and the scoring standards, and calculate the planting level scores of the tobacco farmers in the research area by substituting the tobacco field information of the tobacco farmers into the planting level model.

[0030] A seasonal tobacco growth calculation module is configured to calculate the NDVI of the tobacco fields corresponding to the tobacco farmers in the research area based on the remote sensing images, and obtain the seasonal tobacco growth of the tobacco farmers by taking the mean value.

[0031] A typical household list generation module is configured to classify the tobacco farmers in the research area based on the planting level scores and the seasonal tobacco growth to obtain the typical household list of the tobacco farmers in different categories.

[0032] An electronic device comprises a memory and a processor, the memory is configured to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement the typical household selection method for tobacco yield prediction according to any one of the above.

[0033] A computer readable storage medium storing a computer program, the computer program is executed by a computer to implement the typical household selection method for tobacco yield prediction according to any one of the above.

[0034] The present application has the following beneficial effects:

[0035] (1) By establishing a planting level model, this scheme can more objectively describe the planting level of tobacco farmers in different regions, avoid the bias problem that may exist due to excessive reliance on the subjective judgment of tobacco technicians in traditional methods, and improve the accuracy of selecting typical households.

[0036] (2) The NDVI index at the maturity of tobacco fields can effectively determine the actual planting situation in the region during the season. The NDVI index measured in practice can effectively reflect the abnormal planting level caused by special circumstances such as weather, pests and diseases, or fluctuations in planting level. By identifying and classifying typical households through the NDVI index, the planting situation in the season can be fully and quantitatively reflected, avoiding the bias of relying on human experience and subjective experience in identification methods.

[0037] (3) This scheme combines the planting level model with the NDVI index of tobacco field maturity, which integrates expert experience, tobacco technicians’ understanding of tobacco farmers’ planting level, and the actual growth of tobacco leaves in the current season. It can more comprehensively analyze and evaluate the planting level of tobacco farmers in the current season and the quality of tobacco leaves, avoiding the evaluation limitations brought about by a single indicator. It can help tobacco stations select typical households in a more scientific and accurate way, and improve the efficiency of tobacco leaf procurement and allocation and the quality of tobacco leaves.

[0038] (4) This scheme constructs a planting level model by combining expert scoring and the coefficient of variation to summarize the planting level of tobacco farmers. Then, it combines remote sensing technology to directly measure and compare the NDVI index of the planting situation in the current year. Typical households with expected tobacco production are selected in a scientific and quantifiable way, providing more objective and scientific decision-making basis for the selection of typical households to tobacco industries at all levels. In addition, more planting data are accumulated through this method to optimize the planting level model and provide a more accurate model for the expected tobacco production in the following year. Attached Figure Description

[0039] Figure 1 This is a flowchart of a typical household selection method for predicting tobacco yield in this invention.

[0040] Figure 2 Mind map of a specific implementation scheme in Embodiment 1 of the present invention

[0041] Figure 3 This is a schematic diagram of the structure of a typical user selection system in this invention. Detailed Implementation

[0042] Example 1

[0043] like Figure 1 As shown, a method for selecting typical households for tobacco yield forecasting includes the following steps:

[0044] S11. Determine the study area and acquire remote sensing images of the study area;

[0045] S12. Based on the experience of provincial and municipal experts, the factors affecting planting levels in the study area were screened, and the weights of each factor were calculated using the expert scoring method and the coefficient of variation method to obtain the factor proportion results.

[0046] S13. Develop scoring criteria for each factor in the research area based on the experience of municipal or district / county level experts;

[0047] S14. Based on the factor proportion results and scoring criteria, establish a planting level model, substitute the tobacco field information of each tobacco farmer in the study area into the planting level model for calculation, and obtain the planting level score of each tobacco farmer.

[0048] S15. Based on remote sensing images, calculate the NDVI of the tobacco fields corresponding to each tobacco farmer in the research area, and take the average value to obtain the tobacco growth of each tobacco farmer in the current season.

[0049] S16. Based on the planting level score and the current tobacco growth, tobacco farmers in the study area were classified to obtain a list of typical tobacco farmers in different categories.

[0050] This embodiment constructs a planting level model by combining expert scoring and a weighted assessment of the coefficient of variation to summarize the planting level of tobacco farmers. It then combines remote sensing technology to directly measure and compare the NDVI index of the current year's planting situation. This scientific and quantifiable method selects typical households with high expected tobacco yields, providing a more objective and scientific basis for selecting typical households to the tobacco industry at all levels. The mind map of the above scheme is shown below. Figure 2 As shown, the specific solution process is as follows:

[0051] Step 1: Build a user planting level model

[0052] A. Using a combination of expert scoring and the coefficient of variation method, provincial and municipal experts screened the influencing factors of the planting level model and calculated the specific weight proportions to obtain the factor proportion results.

[0053] The expert scoring method involves all experts proposing factors that may affect the judgment of planting level, including but not limited to years of tobacco planting experience, education level, and number of training sessions. Based on the factors proposed by all experts, all experts score the degree of influence of each factor on the planting level. The scoring results are shown in Table 1.

[0054] Table 1

[0055]

[0056] The coefficient of variation is a normalized measure of the dispersion of the probability distribution. It reflects the differences in the impact of various factors on planting levels and serves as the basis for determining the weights of each factor. It is calculated by dividing the mean by the standard deviation, and its formula is as follows:

[0057]

[0058] in, σn represents the mean of the n-factor expert scores, σn represents the standard deviation of the n-factor expert scores, and CVn represents the mean of the n-factor expert scores. n Let be the coefficient of variation for n factors.

[0059] The above expression is based on an example of n factors. The formula for calculating the coefficient of variation for other factors is the same; simply replace the n factor with the corresponding factor and substitute it into the calculation.

[0060] The weight of a factor is determined based on the coefficient of variation. Taking factor 'a' as an example, the calculation formula is as follows:

[0061]

[0062] Among them, X a This refers to the final weight of factor 'a' in the planting level model, i.e., the factor proportion of factor 'a'. Different factors a, b, ..., n are substituted into the above formula to calculate the final weight of each factor. The final weights of all factors are then summed to obtain the factor proportion result.

[0063] The influencing factors (i.e., factors affecting the planting level) in the planting level model that can be used for reference in this embodiment are shown in Table 2:

[0064] Table 2

[0065]

[0066]

[0067] B. The scoring standards for each region are determined by experts at the municipal and district / county levels respectively. The scoring standards available for reference in this embodiment are shown in Table 3:

[0068] Table 3

[0069]

[0070]

[0071] The selection of factors mentioned above utilizes the experience of provincial and municipal experts, while the determination of scoring standards relies on the experience of municipal or district / county-level experts, as factor selection is a relatively objective process. For example, factors such as years of tobacco-growing experience and educational level certainly influence tobacco-growing performance and can be uniformly determined by provincial and municipal experts. However, for specific scoring standards for particular factors, it is more accurate to have experts at a lower administrative level (municipal or district / county level) conduct the evaluation. The specific reasons are as follows:

[0072] For example, the scoring standards for tobacco cultivation experience differ between a city / county / district with decades of tobacco-growing history and one that has only been undertaking tobacco cultivation for a few years. In the former, over ten years is required to be considered an experienced tobacco farmer, while in the latter, three years is sufficient. Similarly, the scoring standards for training differ depending on whether a city / county / district has received training. A city / county / district that prioritizes tobacco and conducts training regularly may have only held sporadic training sessions, leading to different evaluation criteria. Other influencing factors, such as average yield and tobacco leaf quality ratio, also vary by region. Therefore, the scoring standards for the same factor differ depending on the research area and must be determined based on local conditions. For these reasons, it is more accurate for the scoring standards to be determined by experts at the administrative level one or two levels below the administrative level.

[0073] C, The local tobacco station and all tobacco technicians will uniformly score all tobacco farmers in the jurisdiction.

[0074] It is recommended that all tobacco technicians review the standards together to ensure they are suitable for local conditions. When assigning scores, the scores from the tobacco technicians responsible for a particular tobacco farmer should be considered the primary opinion, while also taking into account the suggestions of other tobacco technicians.

[0075] D. The planting level model based on expert scoring method obtains the planting level score of local tobacco farmers, that is, the planting level score of each tobacco farmer in the study area. The expression of the planting level model is as follows.

[0076] Y = A n *X1+B n *X2+...+N n *X n ,

[0077] Where Y represents the planting level score, and A n ~N n These are the scores that the tobacco farmer obtained for this factor according to the scoring criteria, X1 to X2 respectively. n Equivalent to X a ~X n , where represents the weights of different factors.

[0078] The planting level scores of each tobacco farmer were summarized and compiled into a table as shown in Table 4.

[0079] Table 4

[0080]

[0081] Step 2: Use remote sensing technology to obtain the average NDVI index of local tobacco leaves at the maturity period, and obtain the tobacco growth status of the current season.

[0082] NDVI values ​​are most accurate during the mid-season when crops are at their most active growth stage. Therefore, under clear, unclouded conditions during the tobacco ripening period, surface reflectance data from Sentinel-2 radios on any given day within the study area were obtained using Google Earth Engine. Vector files from the same time period (within ±2 hours of noon) were filtered and cropped. Then, NDVI vegetation index was extracted using near-infrared and red bands to obtain the NDVI index for tobacco fields belonging to different tobacco farmers. Finally, the NDVI index for all tobacco fields belonging to the same farmer in the same region on the same day was obtained, and the mean value for all tobacco fields belonging to that farmer was calculated. The average value of the region is used to represent the growth of tobacco in the current season, which is the growth of tobacco in the corresponding tobacco field of the tobacco farmer in the study area.

[0083] Step 3: Based on the planting level model and the current tobacco growth status based on the NDVI index, provide a recommended list of typical local households.

[0084] The Y-value results of tobacco farmers in the study area based on the planting level model were used to obtain a list corresponding to the local planting level from largest to smallest (as shown in Table 5). The top 33% of tobacco farmers were considered to have "good planting level", 33% to 66% were considered to have "medium planting level", and the remaining tobacco farmers were considered to have "poor planting level". The NDVI results were then sorted from largest to smallest (as shown in Table 6) and divided into three equal groups: the top 33% of tobacco farmers were considered to have "good tobacco growth in the current season", 33% to 66% were considered to have "medium tobacco growth in the current season", and 66% to 99% were considered to have "poor tobacco growth in the current season".

[0085] Based on the criteria of simultaneously meeting "good planting level" and "good tobacco growth in the current season", "medium planting level" and "medium tobacco growth in the current season", and "poor planting level" and "poor tobacco growth in the current season", a suggested list of typical tobacco growers with good, medium and poor performance in the local area for this year is provided for local tobacco technicians to select or randomly select.

[0086] This means that typical households at different levels must meet both the same criteria in terms of planting level score and tobacco growth in the current season.

[0087] Table 5

[0088]

[0089] Table 6

[0090]

[0091]

[0092] Example 2

[0093] like Figure 3 As shown, a typical household selection system for tobacco yield forecasting includes:

[0094] Image acquisition module 10: Used to determine the study area and acquire remote sensing images of the study area;

[0095] Factor weight calculation module 20: It is used to screen the factors affecting the planting level in the study area based on the experience of provincial and municipal experts, and to calculate the weight of each factor using the expert scoring method and the coefficient of variation method to obtain the factor proportion results;

[0096] Scoring Criteria Development Module 30: Used to develop scoring criteria for various factors in the research area based on the experience of municipal or district / county level experts;

[0097] Planting level score calculation module 40: It is used to establish a planting level model based on the factor proportion results and scoring criteria, and to calculate the planting level score of each tobacco farmer by substituting the tobacco field information of each tobacco farmer in the study area into the planting level model.

[0098] The seasonal tobacco growth calculation module 50 is used to calculate the NDVI of the tobacco fields corresponding to each tobacco farmer in the research area based on remote sensing images, and then take the average value to obtain the seasonal tobacco growth of each tobacco farmer.

[0099] Module 60, which generates a list of typical tobacco farmers, is used to classify tobacco farmers in the study area by combining planting level scores and the current tobacco growth, and to obtain lists of typical tobacco farmers in different categories.

[0100] One implementation of the above system is as follows: In the image acquisition module 10, the study area is determined and remote sensing images of the study area are acquired. In the factor weight calculation module 20, factors affecting planting level in the study area are screened based on the experience of provincial and municipal experts, and the weights of each factor are calculated using expert scoring and the coefficient of variation method to obtain the factor proportion results. In the scoring standard formulation module 30, scoring standards for each factor in the study area are formulated based on the experience of municipal or district / county experts. In the planting level score calculation module 40, a planting level model is established based on the factor proportion results and the scoring standards. The tobacco field information of each tobacco farmer in the study area is substituted into the planting level model to calculate the planting level score of each tobacco farmer. In the current season tobacco growth calculation module 50, the NDVI of the tobacco field corresponding to each tobacco farmer in the study area is calculated based on the remote sensing images, and the average value is taken to obtain the current season tobacco growth of each tobacco farmer. In the typical tobacco farmer list generation module 60, tobacco farmers in the study area are classified by combining the planting level score and the current season tobacco growth to obtain a list of typical tobacco farmers in different categories.

[0101] Example 3

[0102] Based on the above embodiments, this embodiment provides an electronic device.

[0103] Example 4

[0104] Based on the above embodiments, this embodiment provides a storage medium.

[0105] The above description is only a specific embodiment of the present invention, but the technical features of the present invention are not limited thereto. Any changes or modifications made by those skilled in the art within the scope of the present invention are covered by the patent scope of the present invention.

Claims

1. A method for selecting typical households for tobacco yield prediction, characterized in that, include: Determine the study area and acquire remote sensing images of the study area; Based on the experience of provincial and municipal experts, factors affecting planting levels in the study area were screened, and the weights of each factor were calculated using expert scoring and the coefficient of variation method to obtain the factor proportion results. Scoring criteria for each factor in the research area were developed based on the experience of municipal or district / county level experts. Based on the factor proportion results and scoring criteria, a planting level model was established. The tobacco field information of each tobacco farmer in the study area was substituted into the planting level model for calculation to obtain the planting level score of each tobacco farmer. The NDVI of each tobacco farmer's tobacco field in the research area was calculated based on remote sensing images, and the average value was taken to obtain the tobacco growth of each farmer in the current season. By combining planting level scores and the current tobacco growth, tobacco farmers in the study area were classified into typical households in different categories.

2. The method for selecting typical households for tobacco yield prediction according to claim 1, characterized in that, The expression for calculating the weights of each factor is as follows: Where X is the weight, X a Let be the weight of factor 'a', where a, b, ..., n are different factors, and CV is the coefficient of variation. σn represents the mean of the n-factor expert scores, and σn represents the standard deviation of the n-factor expert scores.

3. The method for selecting typical households for tobacco yield prediction according to claim 2, characterized in that, The expression for the planting level model is Y = A n *X1+B n *X2+...+N n *X n Where Y is the planting level score, and A n ~N n These are the scores that the tobacco farmer obtained for this factor according to the scoring criteria, X1 to X2 respectively. n Equivalent to X a ~X n , where represents the weights of different factors.

4. The method for selecting typical households for tobacco yield prediction according to claim 1, characterized in that, Factors affecting planting levels include, but are not limited to, years of tobacco planting experience, education level, whether or not planting training has been received, level of mechanization, average output value per acre over the past three years, average yield per acre over the past three years, the proportion of high-grade, medium-grade, and low-grade tobacco leaves in the upper part of the plant at the time of last year's purchase, the proportion of high-grade, medium-grade, and low-grade tobacco leaves in the middle part of the plant at the time of last year's purchase, the proportion of high-grade, medium-grade, and low-grade tobacco leaves in the lower part of the plant at the time of last year's purchase, and subjective scoring of planting levels in previous years.

5. The method for selecting typical households for tobacco yield prediction according to claim 4, characterized in that, The scoring criteria for the same factor vary depending on the research region.

6. The method for selecting typical households for tobacco yield prediction according to claim 1, characterized in that, Based on planting level scores and the current tobacco growth, tobacco farmers in the study area were classified, resulting in a list of typical tobacco farmers in different categories, as follows: The planting level scores of each tobacco farmer were ranked from high to low and divided into three categories: good planting level, medium planting level, and poor planting level. The tobacco growth of each tobacco farmer in the current season was ranked from high to low and divided into three categories: good tobacco growth, medium tobacco growth, and poor tobacco growth. By summarizing tobacco farmers who have good planting skills and good tobacco growth in the current season, tobacco farmers who have medium planting skills and medium tobacco growth in the current season, and tobacco farmers who have poor planting skills and poor tobacco growth in the current season, a list of typical tobacco farmers in the categories of good, medium, and poor is obtained.

7. The method for selecting typical households for tobacco yield prediction according to claim 6, characterized in that, The classification criteria for dividing tobacco into three categories—good planting level, medium planting level, and poor planting level—or three categories—good tobacco growth in the current season, medium tobacco growth in the current season, and poor tobacco growth in the current season—are based on equal distribution.

8. A typical household selection system for predicting tobacco yield, characterized in that, include: Image acquisition module: used to determine the study area and acquire remote sensing images of the study area; Factor weight calculation module: It is used to screen factors affecting planting level in the study area based on the experience of provincial and municipal experts, and calculate the weight of each factor using expert scoring method and coefficient of variation method to obtain the factor proportion result; Scoring Criteria Development Module: Used to develop scoring criteria for various factors in the research area based on the experience of municipal or district / county level experts; Planting level score calculation module: It is used to establish a planting level model based on the factor proportion results and scoring criteria, and to calculate the planting level score of each tobacco farmer by substituting the tobacco field information of each tobacco farmer in the study area into the planting level model. The seasonal tobacco growth calculation module is used to calculate the NDVI of the tobacco fields corresponding to each tobacco farmer in the research area based on remote sensing images, and then take the average value to obtain the seasonal tobacco growth of each tobacco farmer. The module for generating a list of typical tobacco farmers is used to classify tobacco farmers in the study area by combining planting level scores and the current tobacco growth, and to obtain lists of typical tobacco farmers in different categories.

9. An electronic device, characterized in that, It includes a memory and a processor, the memory being used to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement a typical household selection method for tobacco yield prediction as described in any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a computer, it implements a typical household selection method for tobacco yield prediction as described in any one of claims 1-7.

Images