Intelligent composting method and system based on tobacco state recognition

Abstract

The invention relates to the technical field of intelligent composting, in particular to an intelligent composting method and system based on tobacco state recognition, and the method comprises the steps: collecting a tobacco sample in a tobacco planting area, and carrying out the image recognition of the tobacco sample to obtain a tobacco sample surface filtering image; performing feature analysis on the surface filtering image of the tobacco sample, and classifying a target tobacco planting area based on environmental parameters around the tobacco sample obtained through real-time monitoring; and performing tobacco anomaly root tracing in the abnormal tobacco planting sub-region, generating a tobacco anomaly solution based on a tobacco anomaly root tracing result, and constructing a compost proportion intelligent analysis model. According to the method, the compost proportion is dynamically adjusted and the fertilizer use is optimized by combining the environment data monitored in real time and the health condition of the tobacco through the intelligent compost proportion analysis model, so that not only is the fertilizer use efficiency improved, but also a customized compost scheme can be provided according to the characteristics of the tobacco in different areas, and the soil quality and the growth condition of the tobacco are improved.

Description

Technical Field

[0001] This application relates to the field of intelligent composting technology, and in particular to an intelligent composting method and system based on tobacco state recognition. Background Technology

[0002] During tobacco cultivation, tobacco plants can deteriorate due to environmental changes. Tobacco is a crop that requires high soil fertility; changes in soil fertility directly affect the growth and development of tobacco.

[0003] When soil fertility declines, tobacco cannot obtain sufficient nutrients, which hinders its growth and reduces its health. Soil fertility is highly affected by environmental factors, and water is an indispensable factor for tobacco growth. Tobacco has different water requirements at different growth stages. Insufficient or excessive soil moisture will affect the growth and development of tobacco. Water requirements vary with different temperatures and humidity levels, and excessively acidic air in the ecological environment can also easily lead to the growth of bacteria in tobacco. Summary of the Invention

[0004] The main purpose of this application is to provide an intelligent composting method and system based on tobacco state recognition to overcome the shortcomings of the prior art.

[0005] To achieve the above objectives, this application provides the following technical solution: Intelligent composting methods based on tobacco state recognition include: Tobacco samples were collected from tobacco growing areas, and image recognition was performed on the tobacco samples to obtain filtered images of the tobacco sample surfaces. Feature analysis is performed on the filtered surface image of the tobacco sample, and the target tobacco planting area is classified based on the environmental parameters around the tobacco sample obtained from real-time monitoring. In the abnormal tobacco planting sub-region, the root cause of tobacco abnormality is traced, and a tobacco abnormality solution is generated based on the root cause tracing results. A composting ratio intelligent analysis model is constructed.

[0006] Preferably, tobacco samples are collected from tobacco-growing areas, and image recognition is performed on the tobacco samples to obtain filtered images of the tobacco sample surface, including: The tobacco planting area that needs to be intelligently composted is identified as the target tobacco planting area, wherein the tobacco in the target tobacco planting area that needs to be composted is the target tobacco. The target tobacco planting area is divided into sub-regions to obtain target tobacco planting areas. The target tobacco is sampled and processed in different target tobacco planting sub-regions to obtain tobacco samples. Acquire an image of the surface of a tobacco sample, and perform grayscale processing on the image of the tobacco sample surface to obtain a grayscale image of the tobacco sample surface; The median filtering algorithm is used to calculate the pixel values ​​of different pixels in the grayscale image of the tobacco sample surface. A filtering window is constructed and controlled to slide on the grayscale image of the tobacco sample surface. During the sliding of the filtering window, the median value of all pixels is calculated in real time within the filtering window until the filtering window has been traversed and the filtered image of the tobacco sample surface is obtained.

[0007] Preferably, feature analysis is performed on the filtered surface image of the tobacco sample, and the target tobacco planting area is classified based on the environmental parameters around the tobacco sample obtained from real-time monitoring, including: Based on the edge detection algorithm, the foreground image and background image are separated in the surface image of the tobacco sample to obtain the foreground image of the tobacco sample, wherein the foreground image of the tobacco sample is the image of the tobacco sample region; Obtain the color histogram of the tobacco sample region image, analyze the color histogram to obtain the color feature information of the tobacco sample, and extract the shape feature information of the tobacco sample from the tobacco sample region image; Real-time monitoring of environmental parameters around the target tobacco, and retrieval of standard color feature information thresholds and standard shape feature information thresholds of the target tobacco under the environmental parameters from the tobacco feature database; The Euclidean distance between the color feature information of the tobacco sample and the standard color feature information threshold is calculated and labeled as a type of Euclidean distance. A preset standard value for the type of Euclidean distance is set. If the type of Euclidean distance is greater than the standard value, the corresponding tobacco sample is labeled as a tobacco sample with abnormal color.

[0008] Preferably, the method further includes performing feature analysis on the filtered surface image of the tobacco sample, classifying the target tobacco planting area based on environmental parameters around the tobacco sample obtained through real-time monitoring, and also includes: The Euclidean distance between the shape feature information of the tobacco sample and the standard shape feature information threshold is calculated and calibrated as the second type of Euclidean distance. A standard value for the second type of Euclidean distance is preset. If the second type of Euclidean distance is greater than the standard value for the second type of Euclidean distance, the corresponding tobacco sample is calibrated as a tobacco sample with abnormal shape. The target tobacco planting sub-regions corresponding to tobacco samples with abnormal color or shape are marked as abnormal tobacco planting sub-regions, while the target tobacco planting sub-regions corresponding to tobacco samples without abnormal color or shape are marked as normal tobacco planting sub-regions.

[0009] Preferably, the root cause of tobacco anomalies is traced in the abnormal tobacco planting sub-region, and a tobacco anomaly solution is generated based on the root cause tracing results. An intelligent analysis model for compost ratios is constructed, including: Obtain a tobacco color feature information comparison database and a tobacco shape feature information comparison database, wherein the tobacco color feature information comparison database includes different tobacco disease types corresponding to different tobacco colors, and the tobacco shape feature information comparison database includes different tobacco maturity levels corresponding to different tobacco shapes; In the abnormal tobacco planting sub-region, based on the tobacco samples with abnormal color and tobacco samples with abnormal shape, the target tobacco with abnormal color or abnormal shape is marked as a target tobacco with abnormal state. Obtain the compost status of different abnormal tobacco planting sub-regions, wherein the compost status includes historical fertilizer usage and historical fertilizer type; obtain the planting duration of different abnormal tobacco planting sub-regions.

[0010] Preferably, the process includes tracing the root causes of tobacco anomalies in the abnormal tobacco planting sub-regions, generating tobacco anomaly solutions based on the tracing results, and constructing an intelligent analysis model for compost ratios. The process also includes: Obtain the planting duration-composting status matching table. Based on the planting duration-composting status matching table, determine whether there is an abnormal tobacco planting sub-region where the composting status and planting duration do not match. If so, a tobacco anomaly solution is generated, wherein the tobacco anomaly solution is to stop composting the corresponding abnormal tobacco planting sub-area until the composting status of the abnormal tobacco planting sub-area matches the planting time. After matching the composting status with the planting duration in the abnormal tobacco planting sub-area, a matching observation time is preset. After the matching observation time, it is determined whether there are still abnormal tobacco planting sub-areas. If abnormal tobacco planting sub-regions still exist, tobacco anomaly tracing and optimization processing will be carried out on the abnormal tobacco planting sub-regions until all target tobacco planting sub-regions are normal tobacco planting sub-regions.

[0011] Preferably, if abnormal tobacco planting sub-regions still exist, then the abnormal tobacco planting sub-regions are subjected to tobacco anomaly tracing and tobacco anomaly optimization processing until all target tobacco planting sub-regions are normal tobacco planting sub-regions, including: In the abnormal tobacco planting sub-region, if there is a target tobacco with an abnormal state, the color feature information and shape feature information of the target tobacco with an abnormal state are imported into the tobacco color feature information comparison database and the tobacco shape feature information comparison database, respectively, to match the tobacco disease type and maturity, and obtain the tobacco state parameters; Obtain the disease type and maximum maturity of the target tobacco plants in the abnormal state, and analyze the tobacco state parameters; If, among the target tobacco plants in abnormal condition, there are tobacco diseases that require discarding, or tobacco plants with a maturity level greater than the maximum maturity level, then the corresponding target tobacco plants in abnormal condition will be marked as discarded target tobacco plants. In the abnormal tobacco planting sub-region, all discarded target tobacco plants will be harvested and discarded, and the remaining target tobacco plants in abnormal condition will be marked as repairable target tobacco plants. Obtain the tobacco state parameters of the repairable target tobacco, calculate the correlation value between the tobacco state parameters and the surrounding environmental parameters of the target tobacco, label it as the first correlation value, and preset the first correlation threshold. If the first correlation value remains within the first correlation threshold, then retrieve the appropriate growth environment parameters of the target tobacco in the big data network, and retrieve the scheme to adjust the environmental parameters of the target tobacco to the appropriate growth environment parameters, and output the second type of tobacco anomaly solution. If the first correlation value does not remain within the first correlation threshold, an intelligent analysis model for composting ratio is constructed, and intelligent composting analysis is performed in the abnormal tobacco planting sub-region. Based on the results of the intelligent composting analysis, repair treatment is carried out on the repairable target tobacco.

[0012] Preferably, an intelligent composting ratio analysis model is constructed to perform intelligent composting analysis within the abnormal tobacco planting sub-region. Based on the results of the intelligent composting analysis, remedial treatment is carried out on the remediatable target tobacco, including: The fertilizer type with the highest repair efficiency for repairable target tobacco is retrieved from the big data network and identified as the target fertilizer type; a compost ratio database for repairable target tobacco is obtained, wherein the compost ratio database for repairable target tobacco records the appropriate compost ratio and fertilizer usage for abnormal tobacco planting sub-areas under different environmental parameters, different shape feature information and color feature information of the target tobacco. An SVM classifier is introduced, and a compost ratio database of repairable target tobacco is imported into the SVM classifier. A compost ratio intelligent analysis model is constructed and run. After the compost ratio intelligent analysis model is run, it is used to generate the target compost ratio and target fertilizer usage amount of the abnormal tobacco planting sub-area based on the shape feature information and color feature information of the repairable target tobacco and the environmental parameters around the target tobacco. Fertilizers of the target fertilizer type are labeled as repairable fertilizers. The total amount of repairable fertilizer is obtained. If the total amount of repairable fertilizer is not less than the target fertilizer usage amount of all abnormal tobacco planting sub-regions, then composting treatment is carried out on all abnormal tobacco planting sub-regions according to the corresponding target composting ratio until there are no abnormal tobacco planting sub-regions in the target tobacco planting area. If the total amount of repairable fertilizer is less than the total target fertilizer usage of all abnormal tobacco planting sub-regions, then the number of repairable target tobacco plants in different abnormal tobacco planting sub-regions is calculated, and composting is prioritized for abnormal tobacco planting sub-regions with a larger number of repairable target tobacco plants according to the corresponding target composting ratio.

[0013] Preferably, an SVM classifier is introduced, into which a compost ratio database of repairable target tobacco is imported, and a smart compost ratio analysis model is constructed and run, including: Data on carbon-nitrogen ratio, moisture content, and pH value at different composting stages were collected to construct a basic compost feature library. The data in the composting basic feature library were scaled; the standardized data were divided into an input feature sample set and a maturity label sample set; an SVM classifier was trained to generate an initial classification model with the input feature sample set as the independent variable and the maturity label sample set as the dependent variable. The compost ratio database of repairable target tobacco is imported into the initial classification model, and a candidate ratio set is generated by feature vector matching; the compost ratio intelligent analysis model is used to perform a maturity correlation analysis on the candidate ratio set and output a predicted maturity index. An evaluation function is constructed with the highest predicted maturity index and the lowest raw material consumption as the optimization objectives; the candidate ratio set is iteratively optimized; the composting raw material ratio adjustment scheme that satisfies the optimal solution of the evaluation function is output; when the difference between the actual compost maturity and the predicted maturity index exceeds the set range, the penalty coefficient of the SVM classifier is dynamically calibrated based on the actual composting data.

[0014] To achieve the above objectives, this application also provides the following technical solution: an intelligent composting system based on tobacco state recognition, applicable to the aforementioned intelligent composting method based on tobacco state recognition, comprising: An image recognition unit is used to collect tobacco samples from tobacco planting areas, perform image recognition on the tobacco samples, and obtain a filtered image of the tobacco sample surface. The feature analysis unit is used to perform feature analysis on the filtered surface image of the tobacco sample and classify the target tobacco planting area based on the environmental parameters around the tobacco sample obtained by real-time monitoring. The fertilizer formulation unit is used to trace the root causes of tobacco anomalies in abnormal tobacco planting sub-regions, generate tobacco anomaly solutions based on the root cause tracing results, and construct an intelligent analysis model for compost formulation.

[0015] (1) This application uses image recognition technology to filter and analyze the features of tobacco surface, which can accurately identify the state of tobacco, such as abnormalities in color and shape. Problem tobacco can be detected in time during the planting process, and early intervention and precise treatment can be carried out to avoid unnecessary waste. Moreover, through the intelligent analysis model of compost ratio, the compost ratio can be dynamically adjusted by combining real-time monitored environmental data and the health status of tobacco, and the fertilizer use can be optimized. This not only improves the fertilizer use efficiency, but also provides customized composting solutions according to the characteristics of tobacco in different regions, thereby improving soil quality and tobacco growth conditions. (2) When abnormal tobacco is discovered, this application can conduct in-depth analysis of its root causes, and provide targeted solutions by combining factors such as the color, shape, and historical fertilizer use of the tobacco. This helps to accurately solve the problem rather than taking blind and uniform treatment measures. Furthermore, the use of composting raw materials can be optimized through intelligent analysis models such as SVM classifiers to avoid overuse or waste of fertilizer and achieve rational use of resources. At the same time, by monitoring the maturity of composting, the best composting effect can be ensured, thereby improving soil fertility and the growth quality of tobacco. Attached Figure Description

[0016] Figure 1 This is a schematic flowchart illustrating the steps of an embodiment of the intelligent composting method based on tobacco state recognition according to this application. Figure 2 This is a schematic diagram of the system architecture of an embodiment of the intelligent composting system based on tobacco state recognition according to this application.

[0017] Figure reference numerals: 1. Image recognition unit; 2. Feature analysis unit; 3. Fertilizer formulation unit. Detailed Implementation

[0018] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings.

[0019] like Figure 1 As shown, the intelligent composting method based on tobacco state recognition proposed in this invention includes: S1. Collect tobacco samples from the tobacco planting area, perform image recognition on the tobacco samples, and obtain a filtered image of the tobacco sample surface; S2. Perform feature analysis on the filtered surface image of the tobacco sample, and classify the target tobacco planting area based on the environmental parameters around the tobacco sample obtained from real-time monitoring. S3. In the abnormal tobacco planting sub-region, trace the root cause of tobacco abnormalities, generate tobacco abnormality solutions based on the root cause tracing results, and construct an intelligent analysis model for compost ratio.

[0020] In this invention, samples need to be collected from tobacco growing areas. These samples may come from different growth stages or regions and are used for subsequent analysis and processing. The collected tobacco samples undergo image processing and recognition to obtain detailed features of the tobacco surface. Image recognition technology can extract various information from the tobacco surface, such as leaf morphology, color, texture, and possible traces of pests and diseases. Filtering is a method of image processing, usually used to remove noise or extract certain features. In the image recognition of tobacco samples, filtered images usually refer to images of the tobacco surface that have been processed to remove unnecessary interference factors (such as changes in lighting or background noise), retaining only the filtered images. This process yields valuable information related to tobacco health and disease. Filtered surface images of tobacco samples obtained through image recognition will be further analyzed for features. Common image features may include the color distribution of tobacco leaves, leaf texture, and signs of pests and diseases. These features are crucial for subsequent analysis. In addition to image information, environmental parameters surrounding the tobacco growing area also need to be collected. These parameters typically include temperature, humidity, soil pH, meteorological conditions (such as rainfall and wind speed), and light intensity. These environmental factors affect the growth and health of tobacco, therefore, it is necessary to combine these parameters to conduct a more accurate analysis and classification of the tobacco growing area. Based on the image characteristics of tobacco samples and surrounding environmental parameters, the system can classify different tobacco planting areas. The classification goal is to categorize planting areas into different types based on factors such as the health status and environmental adaptability of the samples. For example, it may be divided into "healthy areas," "potential problem areas," and "abnormal areas." In healthy areas, tobacco grows well under suitable environmental conditions. In potential problem areas, there may be some adverse factors (such as minor diseases or environmental stress). In abnormal areas, tobacco growth is severely affected, and there may be diseases, pests, or other serious problems. When an abnormality is found in a tobacco planting area, it is necessary to trace its root cause, that is, analyze and track the reasons for the tobacco growth problems. This tracing usually requires combining data such as the image characteristics of tobacco samples, environmental parameters, and planting management history. For example, abnormal leaf color may indicate disease, and abnormal growth status may be related to soil quality or climate change. Through tracing, the root cause of the problem can be determined, such as whether it is due to unsuitable environmental parameters, pest infestation, improper cultivation management, or other factors. Once the root cause of anomalies in a tobacco-growing area is identified, the system can automatically generate targeted solutions based on the source tracing results. These solutions may include: applying certain fertilizers or pesticides to address pests, diseases, or nutrient deficiencies; adjusting planting density, irrigation methods, or fertilization plans to improve growing conditions; optimizing the soil or environment (such as adjusting pH levels, increasing organic matter, etc.); composting is an effective method to improve soil quality and tobacco growth; through composting, soil structure can be improved, soil nutrient content increased, and a better growing environment provided for tobacco; compost ratio refers to the proportion of different organic materials (such as straw, livestock manure, etc.) in the compost; the intelligent analysis model can optimize the compost material ratio based on factors such as soil quality and environmental conditions in the tobacco-growing area to achieve the best soil improvement effect; for example, some soils may lack nitrogen, requiring the addition of nitrogen-rich materials to the compost; while other soils may require more organic materials to improve aeration and water retention; this intelligent model can combine environmental parameters and source tracing results to provide a personalized compost ratio scheme for each tobacco-growing area to ensure optimal tobacco growing conditions.

[0021] In an optional embodiment, tobacco samples are collected from a tobacco-growing area, and image recognition is performed on the tobacco samples to obtain a filtered image of the tobacco sample surface, including: The tobacco planting areas that require intelligent composting are identified as target tobacco planting areas, and the tobacco in the target tobacco planting areas that requires composting treatment is the target tobacco. The target tobacco planting area is divided into sub-regions to obtain target tobacco planting areas. Tobacco samples are then collected from different target tobacco planting sub-regions to obtain tobacco samples. Acquire an image of the tobacco sample surface, and perform grayscale processing on the tobacco sample surface image to obtain a grayscale image of the tobacco sample surface; The median filtering algorithm is used to calculate the pixel values ​​of different pixels in the grayscale image of the tobacco sample surface. A filtering window is constructed and controlled to slide on the grayscale image of the tobacco sample surface. During the sliding of the filtering window, the median value of all pixels is calculated in real time within the filtering window until the filtering window has been traversed and the filtered image of the tobacco sample surface is obtained.

[0022] It should be noted that the target tobacco planting area is the location where tobacco is grown and cultivated, including soil and tobacco plants. The purpose of monitoring environmental parameters within the target tobacco planting area is to identify problems arising during cultivation, as tobacco is a plant susceptible to environmental influences such as disease and over-ripening. Dividing the target tobacco planting area into zones is necessary because the area may be large, and zoning allows for better image recognition, and not all areas contain abnormal tobacco. Due to the potentially large quantity of tobacco, sampling inspection is required. Image recognition first involves grayscale conversion, resulting in a clearer image. Subsequently, image filtering is performed because the acquired image may contain significant noise. Median filtering yields a filtered image of the tobacco sample surface.

[0023] In an optional embodiment, feature analysis is performed on the filtered surface image of the tobacco sample, and the target tobacco planting area is classified based on the environmental parameters surrounding the tobacco sample obtained through real-time monitoring, including: Based on the edge detection algorithm, the foreground image and background image are separated in the surface image of the tobacco sample to obtain the foreground image of the tobacco sample, wherein the foreground image of the tobacco sample is the image of the tobacco sample region; Obtain the color histogram of the tobacco sample region image, analyze the color histogram to obtain the color feature information of the tobacco sample, and extract the shape feature information of the tobacco sample from the tobacco sample region image; Real-time monitoring of environmental parameters around the target tobacco, and retrieval of standard color feature information thresholds and standard shape feature information thresholds of the target tobacco under environmental parameters from the tobacco feature database; The Euclidean distance between the color feature information of the tobacco sample and the standard color feature information threshold is calculated and labeled as a Class I Euclidean distance. A Class I Euclidean distance standard value is preset. If the Class I Euclidean distance is greater than the Class I Euclidean distance standard value, the corresponding tobacco sample is labeled as a tobacco sample with abnormal color.

[0024] It should be noted that the purpose of introducing the edge detection algorithm is to extract the tobacco-related images from the surface image of the tobacco sample, that is, to separate the foreground image and the background image to obtain the tobacco sample area image; color analysis is performed on the tobacco sample within the tobacco sample area image, because the color of diseased tobacco or tobacco with maturity outside the preset range may be different from that of tobacco in a normal state; diseased tobacco usually has yellowing leaves, drooping leaves, whitening of the main vein, and loss of pubescence, while overly mature tobacco leaves may also have maturity spots.

[0025] In an optional embodiment, feature analysis is performed on the filtered surface image of the tobacco sample, and the target tobacco planting area is classified based on the environmental parameters around the tobacco sample obtained from real-time monitoring. The method further includes: Calculate the Euclidean distance between the shape feature information of the tobacco sample and the standard shape feature information threshold, and label it as the second type of Euclidean distance. Preset the standard value of the second type of Euclidean distance. If the second type of Euclidean distance is greater than the standard value of the second type of Euclidean distance, the corresponding tobacco sample is labeled as a tobacco sample with abnormal shape. The target tobacco planting sub-regions corresponding to tobacco samples with abnormal color or shape are marked as abnormal tobacco planting sub-regions, while the target tobacco planting sub-regions corresponding to tobacco samples without abnormal color or shape are marked as normal tobacco planting sub-regions.

[0026] It should be noted that by analyzing color and shape feature information, it can be determined whether the Euclidean distance between the tobacco samples and the corresponding standard values ​​is greater than the standard Euclidean distance value. The larger the Euclidean distance, the lower the data similarity. After obtaining tobacco samples with abnormal color and abnormal shape, the target tobacco planting sub-region containing the tobacco samples with abnormal color or abnormal shape is marked as the abnormal tobacco planting sub-region. Tobacco samples with abnormal color and abnormal shape can coexist in the same sub-region.

[0027] In an optional embodiment, the root cause of tobacco anomalies is traced in the abnormal tobacco planting sub-region, a tobacco anomaly solution is generated based on the root cause tracing results, and an intelligent analysis model for compost ratio is constructed, including: Obtain a tobacco color feature information comparison database and a tobacco shape feature information comparison database. The tobacco color feature information comparison database includes different tobacco disease types corresponding to different tobacco colors, and the tobacco shape feature information comparison database includes different tobacco maturity levels corresponding to different tobacco shapes. In the abnormal tobacco planting sub-region, based on tobacco samples with abnormal color and tobacco samples with abnormal shape, target tobacco with abnormal color or shape is identified as target tobacco with abnormal state. Obtain the compost status of different abnormal tobacco planting sub-regions, including historical fertilizer usage and historical fertilizer type; obtain the planting duration of different abnormal tobacco planting sub-regions.

[0028] It should be noted that the disease type and maturity of tobacco vary under different color and shape characteristics, which can be determined by comparing tobacco color and shape characteristic databases. After identifying the target tobacco with abnormal conditions, it is necessary to process it to ensure that all sub-regions are normal tobacco planting sub-regions. The compost status is directly proportional to and corresponds to the planting time. Within a certain planting time, the amount of fertilizer used is fixed. If the amount and type of fertilizer used within the certain planting time are higher than the standard values, the tobacco is prone to excessive absorption of nutrients from the fertilizer, leading to over-maturity or even disease.

[0029] In an optional embodiment, the method further includes tracing the root causes of tobacco anomalies in the abnormal tobacco planting sub-region, generating tobacco anomaly solutions based on the tracing results, and constructing an intelligent analysis model for compost ratios. Obtain the planting duration-composting status matching table. Based on the planting duration-composting status matching table, determine whether there are any abnormal tobacco planting sub-regions where the composting status and planting duration do not match. If so, a tobacco anomaly solution is generated, in which the tobacco anomaly solution is to stop composting the corresponding abnormal tobacco planting sub-area until the composting status of the abnormal tobacco planting sub-area matches the planting time. After matching the composting status and planting duration in the abnormal tobacco planting sub-regions, a matching observation time is preset. After the matching observation time, it is determined whether there are still abnormal tobacco planting sub-regions. If abnormal tobacco planting sub-regions still exist, tobacco anomaly tracing and optimization processing will be carried out on the abnormal tobacco planting sub-regions until all target tobacco planting sub-regions are normal tobacco planting sub-regions.

[0030] It should be noted that the planting time-composting status matching table can be used to determine whether the current tobacco composting status matches the planting time. If they do not match, composting should be stopped and the planting time should be controlled to match the composting status, thus outputting a tobacco anomaly solution.

[0031] In an optional embodiment, if abnormal tobacco planting sub-regions still exist, tobacco anomaly tracing and optimization processing are performed on these sub-regions until all target tobacco planting sub-regions are normal tobacco planting sub-regions, including: In the abnormal tobacco planting sub-region, if there are target tobacco plants with abnormal status, the color feature information and shape feature information of the target tobacco plants with abnormal status are imported into the tobacco color feature information comparison database and the tobacco shape feature information comparison database respectively to match the tobacco disease type and maturity, and obtain the tobacco status parameters. Obtain the disease type and maximum maturity of the target tobacco plants in abnormal condition, and analyze the tobacco state parameters; If, among the target tobacco plants in abnormal condition, there are tobacco diseases that require discarding, or tobacco plants with a maturity level greater than the maximum maturity level, then the corresponding target tobacco plants in abnormal condition will be marked as discarded target tobacco plants. In the abnormal tobacco planting sub-region, all discarded target tobacco plants will be harvested and discarded, and the remaining target tobacco plants in abnormal condition will be marked as repairable target tobacco plants. Obtain tobacco state parameters of the repairable target tobacco, calculate the correlation value between the tobacco state parameters and the surrounding environmental parameters of the target tobacco, label it as the first correlation value, and preset the first correlation threshold. If the first correlation value remains within the first correlation threshold, then retrieve the appropriate growth environment parameters of the target tobacco in the big data network, and retrieve the scheme to adjust the environmental parameters of the target tobacco to the appropriate growth environment parameters, and output the second type of tobacco anomaly solution. If the first correlation value does not remain within the first correlation threshold, an intelligent analysis model for composting ratio is constructed, and intelligent composting analysis is performed in the abnormal tobacco planting sub-region. Based on the results of the intelligent composting analysis, repair treatment is carried out on the repairable target tobacco.

[0032] It should be noted that if the tobacco planting duration is matched with the composting status, and abnormal tobacco planting sub-areas still exist after a certain period of time (i.e., after the matching observation period), meaning that abnormal target tobacco still exists in these sub-areas, then other methods are needed to repair the tobacco. First, based on a comparison database of tobacco color and shape characteristics, the tobacco state parameters of the abnormal target tobacco are determined, and it is judged whether the abnormal target tobacco can be repaired under the current tobacco state parameters. If it cannot be repaired, it can be directly harvested and discarded. This prevents irreparable tobacco from rotting and affecting the ecological environment of the tobacco planting area. The determination of whether the abnormal target tobacco is... For repairable tobacco, we can determine whether its disease type is the type of tobacco that needs to be discarded, or whether the tobacco maturity is greater than the maximum maturity. If either condition is met, it can be discarded. To obtain solutions for the remaining repairable target tobacco, we need to first determine whether environmental factors have affected the tobacco maturity to the point of being too high or whether disease has occurred. We can use the grey relational analysis method to calculate the grey relational value between the tobacco state parameters of the repairable tobacco and the current surrounding environmental parameters. The larger the grey relational value, the higher the proportion of environmental impact. We need to search for solutions to control the environment based on big data networks, that is, solutions for the second type of tobacco anomalies, and output the solution so that there are no abnormal tobacco planting sub-regions in the target tobacco planting area.

[0033] In an optional embodiment, a smart composting ratio analysis model is constructed to perform smart composting analysis in sub-regions of abnormal tobacco planting. Based on the results of the smart composting analysis, remedial treatment is carried out on remediatable target tobacco, including: Search the big data network for the fertilizer type with the highest repair efficiency for repairable target tobacco and mark it as the target fertilizer type; obtain the compost ratio database of repairable target tobacco, which records the appropriate compost ratio and fertilizer application amount for abnormal tobacco planting sub-areas under different environmental parameters, different shape characteristics and color characteristics of target tobacco; An SVM classifier is introduced, and a compost ratio database of repairable target tobacco is imported into the SVM classifier. A compost ratio intelligent analysis model is constructed and run. After the compost ratio intelligent analysis model is run, it is used to generate the target compost ratio and target fertilizer usage amount of the abnormal tobacco planting sub-area based on the shape feature information and color feature information of the repairable target tobacco and the environmental parameters around the target tobacco. Fertilizers of the target fertilizer type are labeled as repairable fertilizers. The total amount of repairable fertilizer is obtained. If the total amount of repairable fertilizer is not less than the target fertilizer usage amount of all abnormal tobacco planting sub-regions, then composting treatment is carried out on all abnormal tobacco planting sub-regions according to the corresponding target composting ratio until there are no abnormal tobacco planting sub-regions in the target tobacco planting area. If the total amount of repairable fertilizer is less than the total target fertilizer usage of all abnormal tobacco planting sub-regions, then the number of repairable target tobacco plants in different abnormal tobacco planting sub-regions is calculated, and composting is prioritized for abnormal tobacco planting sub-regions with a larger number of repairable target tobacco plants according to the corresponding target composting ratio.

[0034] It should be noted that fertilizers can act as compounds to repair tobacco. For example, when tobacco develops diseases, specific compounds are needed to act on the tobacco to induce chemical reactions at the diseased sites, similar to oxidation, to remove the diseases. These specific compounds can also act as fertilizers, providing nutrients to the tobacco and repairing the diseases. The type of fertilizer with the highest repair efficiency, i.e., the target fertilizer type, should be selected. Since there are many types of fertilizers, composting ratios are necessary during the composting process. These ratios vary depending on the tobacco's state parameters and changes in the surrounding environmental parameters. An intelligent composting ratio analysis model can automatically analyze the composting ratio and automatically determine the amount of fertilizer used based on the different surrounding environmental and state parameters of the target tobacco. The acquired model, the SVM classifier, is the foundation of the intelligent composting ratio analysis model, capable of analyzing and classifying data. After generating the target composting ratio and target fertilizer usage for abnormal tobacco planting sub-regions, it determines whether the total amount of all usable fertilizers can completely cover the target fertilizer usage for all regions. If so, composting is directly applied to all abnormal tobacco planting sub-regions. During composting, the fertilizer usage for different sub-regions is controlled to be equal to the target fertilizer usage, and the fertilizer ratio is equal to the target composting ratio, until no abnormal tobacco planting sub-regions remain. If complete coverage is not possible, composting should be prioritized for sub-regions with a larger quantity of target tobacco that can be repaired, to prevent excessive tobacco in abnormal conditions from damaging the soil.

[0035] In an optional embodiment, an SVM classifier is introduced, into which a compost ratio database of repairable target tobacco is imported, and a smart compost ratio analysis model is built and run, including: Data on carbon-nitrogen ratio, moisture content, and pH value at different composting stages were collected to construct a basic compost feature library. The data in the composting basic feature library were scaled; the standardized data were divided into an input feature sample set and a maturity label sample set; an SVM classifier was trained to generate an initial classification model with the input feature sample set as the independent variable and the maturity label sample set as the dependent variable. The compost ratio database of repairable target tobacco is imported into the initial classification model, and a candidate ratio set is generated by feature vector matching; the compost ratio intelligent analysis model is used to perform a maturity correlation analysis on the candidate ratio set and output a predicted maturity index. An evaluation function is constructed with the optimization objectives of maximizing the predicted maturity index and minimizing raw material consumption. The candidate ratio set is iteratively optimized. The composting raw material ratio adjustment scheme that satisfies the optimal solution of the evaluation function is output. When the difference between the actual compost maturity and the predicted maturity index exceeds the set range, the penalty coefficient of the SVM classifier is dynamically calibrated based on the actual composting data.

[0036] It should be noted that different stages of composting (such as the initial fermentation stage and the later maturation stage) exhibit different physicochemical characteristics. A basic feature library for composting is constructed by collecting data on carbon-to-nitrogen ratio, moisture content, and pH value at each stage. The carbon-to-nitrogen ratio is an important chemical indicator in composting, reflecting the rate of organic matter decomposition and nitrogen source utilization. Moisture content is a key factor in microbial activity during composting; excessively high or low moisture levels will affect composting efficiency. pH value reflects the acidic or alkaline environment during composting; a suitable pH value can promote microbial activity. Different physicochemical data may have different units and magnitudes, therefore, data standardization is necessary. Standardization (such as Z-score standardization or Min-Max standardization) can eliminate scale differences between features, making the model easier to process. The input feature sample set contains various features collected during composting (such as carbon-to-nitrogen ratio, moisture content, and pH value). The maturity label sample set contains maturity labels for compost samples (maturity is a key indicator of whether composting is complete, usually categorized as incomplete, partially mature, or fully mature). During training, the SVM classifier learns how to determine the maturity (e.g., fully mature, partially mature) of compost based on features such as the carbon-to-nitrogen ratio, moisture content, and pH value. The compost mix database contains different compost ratios (i.e., different raw material mixing proportions), each with corresponding physicochemical characteristics. Using feature vector matching, actual compost sample data is matched against features in the mix database to identify candidate mix sets. These candidate mix sets are possible raw material combinations selected based on parameters during the composting process (e.g., carbon-to-nitrogen ratio, moisture content, etc.). The intelligent compost mix analysis model further refines these candidate mixes. The model predicts the maturity of compost. Based on the characteristics of the compost ratio and combined with a trained SVM model, it outputs a predicted maturity index to indicate the composting progress under different ratios. A comprehensive evaluation function is constructed as the optimization objective. This evaluation function is mainly based on two objectives: the highest predicted maturity index means that the selected ratio can achieve the optimal maturity; the lowest raw material consumption means that the optimized ratio can reduce raw material waste and lower costs while meeting the maturity requirements. Iterative optimization algorithms (such as genetic algorithms, particle swarm optimization, etc.) are used to find the optimal solution in the candidate ratio set, that is, to select a ratio that maximizes the maturity index and minimizes raw material consumption. When the difference between the actual compost maturity and the predicted maturity index exceeds the set tolerance range, the penalty coefficient of the SVM classifier needs to be dynamically calibrated. The penalty coefficient in SVM is used to control the tolerance for classification errors. Calibrating the penalty coefficient can improve the model's adaptability to actual compost data. The purpose of dynamic calibration based on actual compost data is to enable the classifier to adjust according to changes in the actual composting process, thereby improving prediction accuracy.

[0037] like Figure 2 As shown, the intelligent composting system based on tobacco state recognition proposed in this invention is applicable to the aforementioned intelligent composting method based on tobacco state recognition, and includes: Image recognition unit 1 is used to collect tobacco samples from tobacco planting areas, perform image recognition on the tobacco samples, and obtain a filtered image of the tobacco sample surface. Feature analysis unit 2 is used to perform feature analysis on the filtered surface image of the tobacco sample and classify the target tobacco planting area based on the environmental parameters around the tobacco sample obtained from real-time monitoring. Fertilizer formulation unit 3 is used to trace the root causes of tobacco anomalies in the abnormal tobacco planting sub-region, generate tobacco anomaly solutions based on the root cause tracing results, and construct an intelligent analysis model for compost formulation.

[0038] The specific embodiments of the invention have been described in detail above, but they are only examples, and this application is not limited to the specific embodiments described above. For those skilled in the art, any equivalent modifications or substitutions to the invention are also within the scope of this application. Therefore, all equivalent changes, modifications, and improvements made without departing from the spirit and principles of this application should be covered within the scope of this application.

Claims

1. A method of intelligent composting based on tobacco condition recognition, characterized in that, The method comprises the following steps: Collecting tobacco samples in a tobacco planting area, performing image recognition on the tobacco samples, and obtaining a tobacco sample surface filtering image; Performing feature analysis on the tobacco sample filtering surface image, classifying the target tobacco planting area based on the environmental parameters of the surrounding area of the tobacco sample obtained through real-time monitoring; In the abnormal tobacco planting sub-area, the root cause of the tobacco abnormality is traced, and a tobacco abnormality solution is generated based on the result of the tobacco abnormality root cause tracing, and a compost ratio intelligent analysis model is constructed.

2. The intelligent composting method based on tobacco condition recognition as claimed in claim 1, wherein, Collecting tobacco samples in a tobacco planting area, performing image recognition on the tobacco samples, and obtaining a tobacco sample surface filtering image, comprising: Determine the tobacco planting area that needs to be composted as the target tobacco planting area, wherein the target tobacco in the target tobacco planting area that needs to be composted is the target tobacco; Divide the target tobacco planting area into target tobacco planting sub-areas, and sample the target tobacco in different target tobacco planting sub-areas to obtain tobacco samples; Obtain a tobacco sample surface image, perform grayscale processing on the tobacco sample surface image, and obtain a tobacco sample surface grayscale image; Based on the median filtering algorithm, the pixel values of different pixel points in the tobacco sample surface grayscale image are calculated, a filtering window is constructed, the filtering window is controlled to slide on the tobacco sample surface grayscale image, and the median value of the pixel values of all pixel points in the filtering window is calculated in real time during the sliding process of the filtering window until the filtering window is traversed and slides, and a tobacco sample surface filtering image is obtained.

3. The intelligent composting method based on tobacco condition recognition as claimed in claim 2, wherein, Perform feature analysis on the tobacco sample filtering surface image, classify the target tobacco planting area based on the environmental parameters of the surrounding area of the tobacco sample obtained through real-time monitoring, comprising: Separate the foreground image and the background image of the tobacco sample surface image based on the edge detection algorithm to obtain a tobacco sample foreground image, wherein the tobacco sample foreground image is a tobacco sample region image; Obtain the color histogram of the tobacco sample region image, analyze the color histogram, obtain the color feature information of the tobacco sample, and extract the shape feature information of the tobacco sample in the tobacco sample region image; Real-time monitoring of the environmental parameters of the target tobacco surrounding area, retrieving the standard color feature information threshold and the standard shape feature information threshold of the target tobacco under the environmental parameters from the tobacco feature library; Calculate the Euclidean distance between the color feature information of the tobacco sample and the standard color feature information threshold, and label it as a class of Euclidean distance. A class of Euclidean distance standard value is preset. If the class of Euclidean distance is greater than the class of Euclidean distance standard value, the corresponding tobacco sample is labeled as a color abnormal tobacco sample.

4. The intelligent composting method based on tobacco condition recognition as claimed in claim 3, wherein, The feature analysis on the tobacco sample filtering surface image and the classification of the target tobacco planting area based on the environmental parameters of the surrounding area of the tobacco sample obtained through real-time monitoring also include: Calculate the Euclidean distance between the shape feature information of the tobacco sample and the standard shape feature information threshold, and label it as a two-class Euclidean distance. A two-class Euclidean distance standard value is preset. If the two-class Euclidean distance is greater than the two-class Euclidean distance standard value, the corresponding tobacco sample is labeled as a shape abnormal tobacco sample; If the extracted tobacco sample does not exist color abnormal tobacco sample or shape abnormal tobacco sample, the corresponding target tobacco sub-planting area is labeled as a normal tobacco sub-planting area.

5. The intelligent composting method based on tobacco condition identification according to claim 4, characterized in that, In the abnormal tobacco sub-planting area, tobacco abnormal root source tracing is performed, and a tobacco abnormal solution is generated based on the tobacco abnormal root source tracing result. A compost ratio intelligent analysis model is constructed, including: Obtain a tobacco color feature information comparison database and a tobacco shape feature information comparison database. The tobacco color feature information comparison database includes different tobacco disease types corresponding to different tobacco colors. The tobacco shape feature information comparison database includes different tobacco maturity corresponding to different tobacco shapes. In the abnormal tobacco sub-planting area, based on the color abnormal tobacco sample and the shape abnormal tobacco sample, the target tobacco with color abnormality or shape abnormality is labeled as a state abnormal target tobacco. Obtain the compost state of different abnormal tobacco sub-planting areas, wherein the compost state includes the historical fertilizer usage and the historical fertilizer type; and obtain the planting duration of different abnormal tobacco sub-planting areas.

6. The intelligent composting method based on tobacco condition identification according to claim 5, characterized in that, In the abnormal tobacco sub-planting area, tobacco abnormal root source tracing is performed, and a tobacco abnormal solution is generated based on the tobacco abnormal root source tracing result. A compost ratio intelligent analysis model is constructed, further including: Obtain a planting duration-compost state matching table, and based on the planting duration-compost state matching table, determine whether the compost state of an abnormal tobacco sub-planting area matches the planting duration in all abnormal tobacco sub-planting areas; If yes, a first-class tobacco abnormal solution is generated, wherein the first-class tobacco abnormal solution is to stop composting in the corresponding abnormal tobacco sub-planting area until the compost state of the abnormal tobacco sub-planting area matches the planting duration; After the compost state of the abnormal tobacco sub-planting area matches the planting duration, a matching observation time is preset. After the matching observation time, it is determined whether there is still an abnormal tobacco sub-planting area; If there is still an abnormal tobacco sub-planting area, tobacco abnormal tracing and tobacco abnormal optimization processing are performed on the abnormal tobacco sub-planting area until all target tobacco sub-planting areas are normal tobacco sub-planting areas.

7. The intelligent composting method based on tobacco condition identification according to claim 6, characterized in that, If there is still an abnormal tobacco sub-planting area, tobacco abnormal tracing and tobacco abnormal optimization processing are performed on the abnormal tobacco sub-planting area until all target tobacco sub-planting areas are normal tobacco sub-planting areas, including: In the abnormal tobacco planting sub-area, if there is a state abnormal target tobacco, the color feature information and the shape feature information of the state abnormal target tobacco are introduced into the tobacco color feature information comparison database and the tobacco shape feature information comparison database respectively to match the disease type and the maturity of the tobacco, and the tobacco state parameters are obtained; The disease type and the highest maturity of the state abnormal target tobacco are obtained, and the tobacco state parameters are analyzed; If the disease type of the state abnormal target tobacco is the disease type of the abandoned tobacco, or the maturity of the state abnormal target tobacco is higher than the highest maturity, the corresponding state abnormal target tobacco is marked as an abandoned target tobacco, and all the abandoned target tobaccos in the abnormal tobacco planting sub-area are picked up and discarded, and the remaining state abnormal target tobaccos are marked as repairable target tobaccos; The tobacco state parameters of the repairable target tobaccos are obtained, the correlation value between the tobacco state parameters and the target tobacco surrounding environment parameters is calculated, and the first correlation value is marked, and a first correlation threshold is preset; If the first correlation value is maintained within the first correlation threshold, the suitable growth surrounding environment parameters of the target tobacco are searched in the big data network, and the scheme of adjusting the target tobacco surrounding environment parameters to the suitable growth surrounding environment parameters is searched, and a second tobacco abnormality solution is output; If the first correlation value is not maintained within the first correlation threshold, a compost ratio intelligent analysis model is constructed, intelligent compost analysis is performed in the abnormal tobacco planting sub-area, and the repairable target tobaccos are repaired based on the intelligent compost analysis result.

8. The intelligent composting method based on tobacco condition identification according to claim 7, characterized in that, The compost ratio intelligent analysis model is constructed, the intelligent compost analysis is performed in the abnormal tobacco planting sub-area, and the repairable target tobaccos are repaired based on the intelligent compost analysis result, including: The fertilizer type with the highest repair efficiency for the repairable target tobacco is searched in the big data network, and the target fertilizer type is marked; a compost ratio database of the repairable target tobacco is obtained, wherein the compost ratio database of the repairable target tobacco records the suitable compost ratio and the fertilizer usage amount of the abnormal tobacco planting sub-area under different target tobacco surrounding environment parameters, different shape feature information and color feature information; The SVM classifier is introduced, the compost ratio database of the repairable target tobacco is introduced into the SVM classifier, the compost ratio intelligent analysis model is constructed and run, wherein the compost ratio intelligent analysis model is used to generate the target compost ratio and the target fertilizer usage amount of the abnormal tobacco planting sub-area according to the shape feature information and the color feature information of the repairable target tobacco and the target tobacco surrounding environment parameters after running; The fertilizer with the target fertilizer type is marked as a repairable fertilizer, the total amount of the repairable fertilizer is obtained, if the total amount of the repairable fertilizer is not less than the target fertilizer usage amount of all the abnormal tobacco planting sub-areas, the compost treatment is performed on all the abnormal tobacco planting sub-areas according to the corresponding target compost ratio until there is no abnormal tobacco planting sub-area in the target tobacco planting area; If the total amount of repairable fertilizer is less than the target fertilizer usage amount of all abnormal tobacco planting sub-regions, the number of repairable target tobaccos in different abnormal tobacco planting sub-regions is calculated, and the abnormal tobacco planting sub-regions with more repairable target tobaccos are preferentially composted according to the corresponding target composting ratio.

9. The intelligent composting method based on tobacco condition identification as claimed in claim 8, wherein, An SVM classifier is introduced, a composting ratio database of repairable target tobaccos is imported into the SVM classifier, and a composting ratio intelligent analysis model is constructed and run, including: Collecting carbon-nitrogen ratio, moisture content, and pH value data at different composting stages to construct a composting basic feature library; Scale unification is performed on the data in the composting basic feature library; the standardized data is divided into an input feature sample set and a maturity label sample set; the input feature sample set is used as the independent variable, and the maturity label sample set is used as the dependent variable, to train the SVM classifier to generate an initial classification model; The composting ratio database of repairable target tobaccos is imported into the initial classification model, and a candidate ratio set is generated through feature vector matching; the candidate ratio set is subjected to maturity correlation analysis by the composting ratio intelligent analysis model, and a predicted maturity index is output; An evaluation function is constructed with the highest predicted maturity index and the lowest raw material consumption as optimization objectives; the candidate ratio set is subjected to iterative optimization; a composting raw material ratio adjustment scheme that meets the optimal solution of the evaluation function is output; when the difference between the actual composting maturity and the predicted maturity index exceeds a set interval, the penalty coefficient of the SVM classifier is dynamically calibrated based on the actual composting data.

10. An intelligent composting system based on tobacco condition recognition, which is suitable for the intelligent composting method based on tobacco condition recognition according to any one of claims 1-9, characterized in that, Including: An image recognition unit is configured to collect tobacco samples in a tobacco planting area, perform image recognition on the tobacco samples, and obtain a filtered image of the surface of the tobacco samples; A feature analysis unit is configured to perform feature analysis on the filtered surface image of the tobacco samples, and classify target tobacco planting areas based on environmental parameters around the tobacco samples obtained through real-time monitoring; A fertilizer ratio unit is configured to trace the root cause of tobacco abnormalities in abnormal tobacco planting sub-regions, generate a tobacco abnormality solution based on the root cause tracing result, and construct a composting ratio intelligent analysis model.

Images