Intelligent scheduling and user participation method and system for rice-rice-oil crop accept farming operation

Abstract

The invention discloses an intelligent scheduling and user participation method and system for rice-oil rotation adopting farming operation, and relates to the technical field of intelligent agriculture and agricultural digital management, and the method comprises the steps: obtaining environment, crop phenotypes and user interaction data, and generating a first feature vector of crop urgency and a second feature vector of user behavior; performing primary screening based on the feature vectors to obtain candidate time windows, and calculating group risk parameters; constructing a safety buffer margin by using the parameter to correct a physical residual time length, and delimiting a safety operation feasible region; constructing an objective function containing mechanical cost and user matching degree potential energy in the domain, and determining a scheduling instruction through optimization; according to the method, random uncertainty characteristics of external group interaction behaviors are quantified and converted into dynamic boundary constraint conditions of an agricultural machinery operation scheduling model, an elastic operation time domain with anti-risk capability is constructed, and safe and efficient scheduling of agricultural machinery resources under multi-dimensional complex constraints is realized.

Description

Technical Field

[0001] This invention relates to the field of smart agriculture and digital agricultural management technology, specifically to a method and system for intelligent scheduling and user participation in rice-oilseed rotation agricultural operations. Background Technology

[0002] With the development of agricultural IoT technology, agricultural operation scheduling systems are gradually evolving from single internal operation management to open interactive management. In new production scenarios such as adoption agriculture and shared farms, agricultural machinery operation scheduling is not only affected by weather conditions and crop growth status, but also introduces a large number of external interactive requests, such as remote supervision, on-site participation, and timed harvesting instructions. This scenario constitutes a typical dynamic resource scheduling system with high-dimensional random constraints.

[0003] However, existing agricultural machinery scheduling systems mostly employ linear programming strategies based on deterministic models. In actual operation, the behavioral data generated by external interaction points (i.e., participating users) exhibits high discreteness, randomness, and non-stationarity (e.g., random cancellation of scheduled time slots, time deviations in arrival at the site). This uncertainty directly impacts the original linear scheduling plan, leading to the following problems:

[0004] In high-intensity rice-rice-oilseed rotation, agronomic operations such as the critical point of rice grain shedding and the point of rapeseed cracking and mold growth are insurmountable hard boundaries. Existing scheduling algorithms lack a quantitative defense mechanism against external random disturbance signals (such as high-probability defaults). Once a high-risk external request occupies the critical operation period but is not executed, the system often cannot redeploy agricultural machinery to carry out backup operations in time, resulting in the physical deadline being breached and causing irreversible crop yield loss.

[0005] Therefore, how to transform the uncertainty of external group behavior into quantifiable mathematical constraints while satisfying the rigid time window constraints of crop growth, and construct an elastic scheduling model with anti-disturbance capabilities to ensure the continuity and safety of agricultural machinery operations, is a key technical problem that urgently needs to be solved in the field of high-throughput agricultural production scheduling. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a method and system for intelligent scheduling and user participation in rice-oilseed rotation agricultural operations.

[0007] To achieve the above objectives, the technical solution of the present invention is as follows:

[0008] In a first aspect, this invention discloses a method for intelligent scheduling and user participation in rice-oilseed rotation agricultural operations, comprising the following steps:

[0009] Acquire environmental monitoring data, crop phenotypic image data, and historical interaction log data of associated users for the target plot;

[0010] Time-series analysis of crop phenotypic image data and environmental monitoring data is performed to generate a first feature vector representing the urgency of crop growth; multi-dimensional feature mining is performed on historical interaction log data to generate a second feature vector containing user behavior reliability features and time preference features.

[0011] Within a preset scheduling period, available time slots are determined based on preset job constraints, and the available time slots are filtered according to the first feature vector and the second feature vector to obtain an initial set of candidate time windows.

[0012] For each candidate time period in the initial candidate time window set, the potential participating user group within that time period is identified based on the second feature vector, and a group risk parameter representing the unreliability of the behavior of the potential participating user group is calculated.

[0013] The first feature vector is analyzed to determine the physical remaining time of agronomic operations, and a safety buffer margin is constructed based on the population risk parameters. The safety cutoff time is determined by calculating the difference between the physical remaining time and the safety buffer margin, and the subset of the initial candidate time window set that is earlier than the safety cutoff time is determined as the feasible region of safe operations.

[0014] Within the feasible domain of safe operation, an objective function is constructed that includes the potential energy of mechanical operation cost and the potential energy of user matching degree. The target agricultural operation scheduling instruction is determined by optimizing the objective function.

[0015] Secondly, this invention discloses an intelligent scheduling and user participation system for rice-oilseed rotation agricultural operations, comprising:

[0016] The data acquisition module is used to acquire environmental monitoring data, crop phenotypic image data, and historical interaction log data of associated users for the target plot.

[0017] The feature analysis module is used to perform time-series analysis on crop phenotypic image data and environmental monitoring data to generate a first feature vector representing the urgency of crop growth; and to perform multi-dimensional feature mining on historical interaction log data to generate a second feature vector containing user behavior reliability features and time preference features.

[0018] The time period initial screening module is used to determine the available time periods within a preset scheduling period based on preset job constraints, and to screen the available time periods according to the first feature vector and the second feature vector to obtain an initial candidate time window set.

[0019] The group risk assessment module is used to identify potential participating user groups within each candidate time period in the initial candidate time window set based on the second feature vector, and to calculate the group risk parameter representing the unreliability of the behavior of potential participating user groups.

[0020] The safety boundary calculation module is used to parse the first feature vector to determine the physical remaining time of agronomic operations and construct a safety buffer margin based on the group risk parameters; the safety cutoff time is determined by calculating the difference between the physical remaining time and the safety buffer margin, and the subset of the initial candidate time window set that is earlier than the safety cutoff time is determined as the feasible region of safe operations;

[0021] The optimized scheduling generation module is used to construct an objective function that includes the potential energy of mechanical operation costs and the potential energy of user matching degree within the feasible domain of safe operation, and to determine the target agricultural operation scheduling instructions by optimizing the objective function.

[0022] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0023] 1. A data-driven dynamic safety buffer mechanism was constructed using multi-dimensional feature mining and information entropy algorithms. Instead of a fixed linear time window, the system treats historical interaction logs as a non-stationary random process, quantifying the uncertainty interference at the input end by calculating the entropy of group behavior. Therefore, the calculated risk parameters are used to dynamically adjust the physical cutoff threshold of agronomic operations, establishing an adaptive fault-tolerant control boundary at the mathematical level. This ensures that even when the input data (user behavior) experiences high variance fluctuations, the scheduling instructions can still converge within the safe time domain of agricultural machinery operations, mitigating operational timeliness failures caused by external disturbances (such as grain scattering or mold) from a control theory perspective.

[0024] 2. An objective function was constructed that includes the potential energy of mechanical operation cost, the potential energy of user matching degree, and the potential energy of fragmentation penalty. By introducing the potential energy of fragmentation penalty, the system can automatically identify and reject inferior solutions that, although satisfying the user's time preference, will cause the agricultural machinery to wait ineffectively between two plots (neither able to move to another plot nor rest). This mechanism, while ensuring the user experience (matching degree potential energy), forces the continuity and intensification of mechanical operation, and significantly reduces the energy loss and idle cost of agricultural machinery relocation.

[0025] 3. By establishing a mapping mechanism between the physical space topology map and the cumulative obstruction heat value, the fragmented penalty potential energy generated during software operation is traced back to the physical plot nodes; the system can automatically identify obstruction nodes that have long caused low scheduling efficiency (such as plots with irregular shapes or impassable roads) and output targeted physical modification suggestions; this feature transforms short-term data flow into a basis for long-term farmland mechanization transformation, realizing the system's improvement from adapting to the environment to optimizing the environment. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 This is an overall block diagram of the method in Embodiment 1 of the present invention;

[0028] Figure 2 This is an overall execution flowchart of the method in Embodiment 1 of the present invention;

[0029] Figure 3 This is an overall block diagram of the system in Embodiment 2 of the present invention. Detailed Implementation

[0030] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0031] Application Overview: In the field of deep integration of modern smart agriculture and the experience economy, especially in the adoption model of high-intensity crop rotation of rice-rice-oilseed triple cropping, efficient scheduling of agricultural operations is considered a key hub for balancing biological growth rhythms and the frequency of social interaction. The core of this process lies in handling the conflict between two types of heterogeneous constraints: one is the deterministic hard constraint determined by the biological growth rhythm of crops, such as the requirement to complete harvesting within a specific physical accumulated temperature window; the other is the discrete, random soft constraint initiated by external interactive terminals (i.e., adoption participants), such as non-stationary interactive signals like reservations, cancellations, and delays. An ideal scheduling system needs to utilize agricultural machinery resources as the physical execution carrier, seamlessly embedding highly random external discrete requests into a continuous, time-sensitive agricultural operation flow, thereby maximizing the overall operational efficiency of the system while ensuring the safety of crop physical output.

[0032] However, existing technologies lack quantitative defense mechanisms against the spatiotemporal conflict between deterministic agricultural time boundaries and random interactive disturbances. Traditional linear scheduling models typically assume stable external inputs and cannot accurately identify resource scheduling failures caused by the temporal uncertainty of input data under high-urgency work windows. Specifically, time-penetration failure manifests as the system reserving work windows, but high-variance external signal fluctuations (such as high-probability defaults) cause agricultural machinery to wait ineffectively, leading to delays in actual work time and missing the optimal weather and agronomic timing. Resource fragmentation failure manifests as the system forcing high-frequency cutting of agricultural machinery work paths to respond to discrete random requests, resulting in a surge in kinetic energy loss during machinery relocation and compression of effective work time. Therefore, a safe risk buffer mapping cannot be established between uncertain variables at the input end and deterministic operating conditions at the execution end, causing the system to misjudge scheduling priorities, thus affecting the robustness of work instructions and the stability of farmland output.

[0033] Without establishing an effective disturbance resistance control model, the intelligent scheduling system will be unable to maintain its ability to safeguard agricultural safety. Specifically, the lack of quantification of input risks will cause discrete random variables to breach the defined agronomic defenses, resulting in crop threshing or mold due to missed physical harvesting windows. Simultaneously, the lack of constraint on mechanical cost potential energy will cause ineffective oscillations of the actuators in physical space, making it impossible to meet the stringent time requirements of the three-crop system. Therefore, this application aims to quantify discrete input risks using information entropy theory, constructing an adaptive and flexible operational boundary to achieve an optimal balance between production efficiency and resource allocation.

[0034] Example 1:

[0035] like Figures 1-2 As shown, the intelligent scheduling and user participation method for rice-oilseed rotation agricultural operations includes the following steps:

[0036] Step S1: Obtain environmental monitoring data, crop phenotypic image data, and historical interaction log data of associated users for the target plot;

[0037] Supported by modern agricultural IoT and multimodal data analysis technology, step S1 in this embodiment specifically involves constructing a high-precision multi-source data sensing and integration mechanism, aiming to provide full and real-time underlying data support for subsequent feature engineering and intelligent scheduling. Specifically, for the "environmental monitoring data of the target plot," the system deploys a micro-weather station and a soil moisture sensor network. Before executing the acquisition action, the system pre-configures a data cleaning and alignment rule base. This rule base is constructed based on the statistical characteristics of historical meteorological big data, using a sliding window algorithm to calculate outlier thresholds for each meteorological parameter to filter out noise caused by sensor malfunctions. In actual operation, the system collects parameters such as air temperature, relative humidity, light intensity, wind speed and direction, rainfall, and soil temperature and humidity of the target plot in real time via the MQTT protocol. To ensure data consistency across time, the system uses the NTP network time protocol to perform nanosecond-level time synchronization on all sensor nodes and uniformly sets the acquisition frequency to once every 10 minutes, thereby forming a high-density environmental monitoring time-series data stream.

[0038] For acquiring crop phenotypic image data, this embodiment did not employ ordinary RGB monitoring, but instead constructed a field visual perception system based on multispectral imaging technology. The system deployed high-definition cameras and multispectral cameras at key observation points in the target plot, focusing on capturing the spectral reflectance and texture features of the crop canopy. During image acquisition, the system incorporated an adaptive exposure algorithm that dynamically adjusted shutter speed and gain according to ambient light intensity, ensuring clear phenotypic images were acquired under various weather conditions. To improve the accuracy of subsequent recognition, the system included a pre-processing pipeline, including a dehazing algorithm (based on dark channel prior), distortion correction, and color balance adjustment. Specifically, for key growth characteristics of different crops in the "rice-rice-oilseed" rotation pattern (such as panicle shape during rice heading and color distribution during rapeseed flowering), the system was configured with a specific region of interest (ROI) extraction strategy, which automatically shielded background interference such as field ridges and ditches, focusing on the phenotypic information of the crops themselves. The acquired image data is tagged with precise timestamps and geographic location labels, and stored in a distributed file system in the form of unstructured data, providing high-quality input samples for subsequent growth stage recognition based on convolutional neural networks.

[0039] This embodiment constructs a user behavior profile data warehouse based on the historical interaction log data of associated users. The data sources of this warehouse cover the entire lifecycle of user operations on the adoption platform APP, including but not limited to activity browsing records, application submission times, actual check-in times for historical activities, cancellation records, and feedback. To uncover deeper user behavior characteristics, the system does not directly store raw logs but adopts an event stream-based storage architecture. Specifically, the system predefines an interaction event model, abstracting each user operation into a structured object containing user ID, event type, occurrence time, duration, and contextual information. When acquiring data, the system uses an ETL (Extract, Transform, Load) process to extract historical logs of associated users from the business database and performs data cleaning and aggregation. For example, for a specific user, the system sorts all their activity participation records from the past year by timeline, forming a complete behavioral trajectory sequence. It is worth noting that, to support the subsequent accurate calculation of user credibility and preferences, the system introduced a data quality verification mechanism during the data acquisition phase. This mechanism eliminates duplicate records and abnormal missing values ​​caused by system failures or network latency, ensuring that each interaction log accurately reflects the user's intentions and behavioral habits. Through the aforementioned deep perception and data integration across the three dimensions of environment, crops, and users, this embodiment successfully constructed a heterogeneous data foundation encompassing the physical world (farmland and crops) and the digital world (user behavior), laying a solid foundation for the subsequent generation of high-dimensional feature vectors and the construction of complex potential field models.

[0040] Step S2: Perform time-series analysis on crop phenotypic image data and environmental monitoring data to generate a first feature vector representing the urgency of crop growth; perform multi-dimensional feature mining on historical interaction log data to generate a second feature vector containing user behavior reliability features and time preference features.

[0041] Specifically, the system first performs time-series analysis on crop phenotypic image data and environmental monitoring data to generate the first feature vector. Before performing this step, the system pre-trains and deploys a convolutional neural network model for the "rice-rice-oilseed" rotation scenario. This model uses a deep residual network (ResNet-50) as its backbone architecture. The input layer is designed to receive normalized multispectral image tensors, the intermediate layers extract texture, color, and morphological features of the crop canopy through multi-scale convolutional kernels, and the output layer maps to a predefined set of growth stage labels (such as "rice-tillering stage" and "oilseed rape-silique development stage") using a Softmax function. In actual operation, the system inputs the image acquired by S1 into the model and outputs the current crop growth stage label in real time.

[0042] Next, the system invokes a pre-set crop accumulated temperature model to invert the physical duration. This model is a knowledge base based on the laws of thermodynamics, storing the effective accumulated temperature thresholds required for different crops at various growth stages. The system combines historical effective accumulated temperature records from environmental monitoring data to calculate the remaining accumulated temperature required for the current crop to reach the preset agronomical limit (such as the closing point of the optimal harvest window), and uses future weather forecast data to convert this into the remaining physical duration. It is assumed that the current growth stage needs to reach a certain total accumulated temperature. The accumulated temperature is The predicted daily average effective accumulated temperature is Then the remaining physical time The calculation formula is: To eliminate the difference in physical dimensions and facilitate subsequent weight calculations, the system further introduces a nonlinear sigmoid decay function. After normalization, the agronomic stress index was calculated. The specific form of this function is: ,in Slope factor This is the critical time threshold for a sudden change in urgency. For example, when the remaining time decreases to the critical value, The urgency index will increase exponentially, approaching 1. Finally, the system combines this normalized urgency index with the original remaining physical time to construct a first feature vector that includes both logical urgency and physical timeliness. .

[0043] Simultaneously, the system performs multi-dimensional feature mining on the historical interaction log data of associated users to generate a second feature vector. This process is divided into two dimensions: behavioral reliability modeling and time preference modeling. In the reliability dimension, the system extracts the registration times of all historical activities of the user. With actual check-in time Calculate the time deviation value sequence And further calculate the variance of the sequence. This is used to characterize the user's punctuality stability. Simultaneously, the system retrieves the user's historical number of defaults. Based on these two indicators, the system constructs an individual reliability score. The calculation formula is set as follows ,in and The system uses preset penalty weighting coefficients to ensure that higher variance or more frequent defaults result in lower reliability scores. In the preference dimension, the system statistically analyzes user activity frequency over historical periods, dividing a week into 168-hour time slices to construct an activity probability graph. The system then vectorizes this graph and generates a time distribution feature sequence using a smoothing algorithm. Each element in the sequence represents the probability of a user's participation during a specific time period.

[0044] Specifically, for cold start scenarios where the historical interaction log data of associated users is empty or the data volume is lower than a preset statistical threshold, this embodiment incorporates a cold start mapping mechanism based on land parcel attribute migration. The core of this mechanism lies in using the similarity of physical land parcels to deduce the behavioral similarity of user groups.

[0045] Specifically, the system pre-constructs a static feature library of all historically mature land parcels. When a cold start is triggered, the system first extracts the static attribute features of the current target land parcel and constructs a feature vector. .in, Normalized geographic coordinates (latitude and longitude) represent spatial proximity; One-hot encoding for crop planting types characterizes the homogeneity of agronomic operations; Traffic accessibility ratings (such as a 0-1 score calculated based on road network density and road grade) directly affect users' willingness to arrive and their punctuality rate.

[0046] Subsequently, the system uses Euclidean distance or cosine similarity algorithms to calculate the similarity distance between the target land parcel vector and all historical land parcel vectors in the feature database. The system is based on The values ​​are sorted from smallest to largest, and the Top-N (e.g., N=5) similar plots are selected as the reference set. The system extracts statistical data of the historical user groups associated with these N similar plots from the database, calculates their average reliability score (as the baseline reliability of the group) and a general time preference template (i.e., the historical user activity heatmap of this type of plot). Finally, the system directly assigns the calculated average reliability score as the user behavior reliability feature of the current user, and assigns the general time preference template as the time preference feature, combining them to generate a second feature vector for cold start. This mechanism ensures that even in scenarios involving newly developed plots or newly registered users, the system can still output statistically significant feature inputs.

[0047] Step S3: Within the preset scheduling period, determine the available time periods based on the preset job constraints, and filter the available time periods according to the first feature vector and the second feature vector to obtain an initial candidate time window set;

[0048] Based on the feature vector generated in step S2, step S3 in this embodiment further performs spatiotemporal preliminary screening based on multidimensional constraints, aiming to lock in an initial set of candidate time windows from the time continuum that both meet the rigid requirements of agronomy and have user participation value.

[0049] Specifically, the system first sets a preset scheduling period (e.g., the next 15 days) and discretizes this period into several fixed-length time slices (e.g., one time slice per hour) on the time axis, constructing a sequence of time slices to be filtered. Based on this, the system performs the first round of hard-constraint filtering according to preset job constraints: the system reads the rainfall probability from the weather forecast data obtained by S1. And the resource status table of the agricultural machinery scheduling system. For any time slice If the predicted probability of rainfall exceeds the safety threshold (e.g.) If the agricultural machinery is marked as under maintenance or occupied, the system will set the availability status of that time slice to False (unavailable).

[0050] After completing the hard constraint filtering, the system performs a second round of soft constraint filtering for all time slices marked as available, based on the first and second feature vectors. The core of this process lies in calculating the basic feasibility score for each time slice, which comprehensively reflects the dual value of agricultural necessity and landscape aesthetics. First, the system parses the first feature vector generated in step S2. Extract the agronomic urgency index from it. To dynamically balance the weights of agricultural operations and user experience, the system constructs a dynamic weighting coefficient. In this embodiment, Defined as a nonlinear function positively correlated with agronomic urgency, for example The physical meaning of this formula is: when the crop is in the middle of its growth cycle (… At a lower level, When the value approaches 0.5, user experience becomes as important as agricultural operations; however, as crops approach their optimal harvest time or key agricultural milestones (…), the value increases. Approaching 1). The value quickly rises to 1, forcibly prioritizing agricultural activities, thus achieving an adaptive response to rigid agricultural schedules.

[0051] Meanwhile, to quantify the experiential value of user participation, the system invokes the visual feature analysis module to process the crop phenotypic image data. Unlike traditional RGB analysis, this embodiment converts the image from the RGB color space to the HSV (hue, saturation, lightness) color space to eliminate interference from light intensity on color recognition. The system presets a target hue range matching the current crop growth stage (e.g., the golden yellow range during rapeseed flowering). (Or the yellowish-brown area of ​​rice at maturity). The system traverses the image pixels and calculates the percentage of pixels falling within the target hue range. And combined with atmospheric visibility data from environmental monitoring data Generate landscape aesthetic score The calculation formula is: ,in The weights are normalized. This score directly reflects whether the current plot of land is at its "peak visual appeal," meaning whether it is suitable for users to visit or take photos.

[0052] Finally, based on the above calculation results, the system performs a weighted summation calculation for each available time slice to obtain a basic feasibility score. The calculation logic is as follows: This formula utilizes dynamic weights. This achieves a fusion of agronomic and aesthetic value. The system sets a preset score threshold (e.g., 0.6) and applies it to all... Time slices exceeding this threshold are filtered out and merged according to temporal continuity to output an initial set of candidate time windows. Each time window in this set is not only physically workable, but also logically satisfies the dual criteria of timely agricultural operations and excellent user experience, providing a high-quality solution space for subsequent more refined risk assessment and potential energy optimization.

[0053] Step S4: For each candidate time period in the initial candidate time window set, identify the potential participating user group within that time period based on the second feature vector, and calculate the group risk parameter representing the unreliability of the behavior of the potential participating user group;

[0054] Based on the initial candidate time window set selected in step S3, step S4 in this embodiment further quantifies the uncertainty risk of group behavior. Traditional scheduling systems often only count how many people have signed up, ignoring whether these people are reliable or whether they are consistently reliable. This step calculates group risk parameters to accurately identify high-risk periods that appear to have high participation but are actually highly prone to default.

[0055] Specifically, the system iterates through each candidate time period in the initial candidate time window set. For a specific time period The system first uses the time distribution feature sequence (i.e., the activity probability map) in the second feature vector generated in step S2. This identifies potential user groups participating during that time period. The system reads all associated users. The sequence corresponds to the time period probability value .like Higher than the preset activity probability threshold (e.g.) If so, then the user is determined to be... This identifies the target users for that specific time period. Through this process, the system constructs a set of target users for that time period. .

[0056] Subsequently, the system extracts the individual reliability score for each user in the set from the second feature vector. To comprehensively assess the group's performance risk, the system introduces information entropy theory to measure the dispersion of group behavior.

[0057] First, the system calculates the arithmetic mean of the reliability scores of all individuals in the target user set. The formula is:

[0058] ;

[0059] This indicator reflects the overall average quality of the group; the higher the average, the lower the theoretical risk.

[0060] However, the average value has the drawback of masking extreme values. To compensate for this, the system further calculates the entropy of group behavior. To calculate the entropy value, the system first processes the continuous reliability scores. (Values ​​range from 0 to 1) are discretized and divided into... A preset interval (e.g., divided into 5 intervals): The system counts the number of users falling into each interval and calculates the probability distribution of each interval. Based on this, the system uses the Shannon entropy formula to calculate the group behavior entropy:

[0061] ;

[0062] In physical terms, The higher the value, the more chaotic the reliability distribution of the group (i.e., a mixture of users with extremely high reliability and extremely low reliability). This internal variability often indicates extremely high management difficulty and uncertainty risk.

[0063] Finally, the system generates the final group risk parameters based on the product of the reciprocal of the average confidence level and the group behavioral entropy. The fusion calculation formula used in this embodiment is:

[0064] ;

[0065] in, A preset adjustment coefficient (e.g., 0.5) is used to balance the influence weights of average level and dispersion.

[0066] Step S5: Analyze the first feature vector to determine the physical remaining time of agronomic operations, and construct a safety buffer margin based on the population risk parameters; determine the safe cutoff time by calculating the difference between the physical remaining time and the safety buffer margin, and determine the subset of the initial candidate time window set that is earlier than the safe cutoff time as the feasible region for safe operations;

[0067] The purpose of this step is to establish a reverse elastic modulation mechanism, that is, to use uncertain human factors to define a definite agricultural time limit in reverse, thereby delineating an absolutely safe operational feasible domain.

[0068] Specifically, the system first performs reverse parsing on the first feature vector generated in step S2 to extract the physical dimension component contained therein—the remaining physical time. This value represents, from a purely agronomic perspective, the crop's performance at the current moment. The remaining time window until the task must be completed (such as the critical point of rice grain falling or the critical point of rapeseed bud splitting). However, under the adoption model, considering the uncertainties that user participation may bring (such as users being late causing agricultural machinery to wait, delays in activity organization, etc.), direct use... Using the deadline as a benchmark poses a significant risk of data penetration. Therefore, the system incorporates a safety buffer margin. .

[0069] To construct an accurate safety buffer margin, the system first obtains real-time weather forecast data from external meteorological service providers via API interfaces and extracts the confidence coefficients from it. (range of values) This coefficient reflects the probability of accuracy in short-term nowcasting. A lower confidence level indicates a greater likelihood of sudden weather changes, and the system should have a larger margin of error. It's worth noting that if the external meteorological service provider does not directly provide a confidence coefficient, the system will access the historical meteorological database, extract comparison records of the service provider's forecast data and actual monitoring data for the target area over the past 30 days, calculate the moving average of the forecast accuracy, and use this as the current confidence coefficient. Furthermore, if the atmospheric pressure change rate is detected to exceed a preset strong convection threshold, the system will forcibly [impose a change in atmospheric pressure]. Reduce to a preset low value (e.g., 0.5) to cope with the risk of sudden extreme weather.

[0070] Subsequently, the system combines the population risk parameters calculated in step S4. And the preset benchmark time for agricultural machinery relocation (Refers to the average rigid time required for agricultural machinery to be transferred from the hangar to the target plot and complete commissioning), the safety buffer margin is calculated according to the following nonlinear model:

[0071] ;

[0072] This formula reveals the defensive scheduling process in this embodiment: when the group risk parameter Increase (user is extremely unreliable) or weather confidence Safety buffer margin when reducing (due to unfavorable weather) It will expand exponentially.

[0073] Based on the above calculation results, the system further executes the safety cutoff time. The determination of this is based on the calculation logic of backtracking the physical limit cutoff point by one buffer cycle, i.e. To illustrate the effect of this mechanism intuitively, let's assume the current moment... The remaining time for physics is 8:00. The baseline time is 50 hours (i.e., the deadline is 10:00 AM the day after tomorrow). The duration is 2 hours. If the risk of participating in the group is extremely low at this time ( And the weather forecast is extremely accurate. If the buffer is only 2 hours, and the safe deadline is locked at 08:00 the day after tomorrow, the system is willing to open most of the time for users to experience the system; conversely, if the group risk is extremely high ( And the weather is uncertain. If this happens, the buffer margin will surge to... The safety deadline has been forced to be moved forward to 03:45 the day after tomorrow. The system is based on this dynamically calculated... The initial candidate time window set selected in step S3 is then filtered a second time to remove all windows with an end time later than [the specified time]. The remaining subset, after this selection process, was formally established as the safe operational feasible region. Any activity scheduled within this region is protected by a sufficient buffer time, even in the worst-case scenario of user default. This allows for the scheduling of pure agricultural machinery operations as a backup, thus completely avoiding the risk of missing the farming season due to user behavior.

[0074] Step S6: Within the feasible domain of safe operation, construct an objective function that includes the potential energy of mechanical operation cost and the potential energy of user matching degree, and determine the target agricultural operation scheduling instruction by optimizing the objective function.

[0075] After accurately defining the feasible region for safe operation, step S6 in this embodiment proceeds to the final decision optimization stage. This step aims to solve a multi-objective global optimization problem under complex working conditions. Its core is to treat the random uncertainty of external group behavior as the boundary constraint of the system, and within this constraint range, find an optimal execution time that can simultaneously minimize the cost of mechanical operation, minimize the deviation of discrete interactive request response, and minimize the fragmented loss of agricultural machinery resources.

[0076] Specifically, the system first interacts with the agricultural machinery dispatch center in real time via an API interface to obtain service status data of currently available agricultural machinery. This data includes not only the unique identifier ID of each agricultural machine and its current real-time geographical coordinates, but also... It also records in detail the time occupancy of their scheduled tasks. The time occupancy table shows the busy periods of the agricultural machinery in the future in the form of a timeline, providing a basis for subsequent calculation of idle windows.

[0077] After acquiring the basic data, the system targets each discrete time slice within the safe operation feasible region determined in step S5. We will calculate the three major potential energy components one by one. First is the potential energy of mechanical operation costs. It consists of transfer costs and idle costs. The system calculates the coordinates of the target plot. The system calculates the Euclidean distance from the current coordinates of the agricultural machinery and combines it with the average speed of the machinery to convert it into transfer time; simultaneously, the system calculates the current time slice. The time interval between the scheduled completion time of a task on the agricultural machinery and the actual completion time. To standardize the units of measurement, these two physical quantities are normalized and then weighted and summed; the smaller the value, the higher the machinery scheduling efficiency.

[0078] Specifically, to eliminate the difference in physical dimensions (distance and time), this embodiment employs a dimensionless processing method based on a benchmark value. For transfer distance costs, the system sets a limit threshold for the agricultural machinery's operating radius. (e.g., 10km), calculate the normalized distance. Regarding idle time costs, the system sets a maximum allowed idle threshold. (e.g., 4 hours), calculate normalized time. If the actual value exceeds the threshold, the normalization result is forcibly set to 1. The final formula for calculating the potential energy of mechanical operation costs is as follows: ,in This is the corresponding normalized weighting factor.

[0079] Secondly, there is the potential for user matching. The second feature vector generated in system callback step S2 is used to extract the time distribution feature sequence. This sequence objectively reflects the activity patterns of the user group. The system uses a peak search algorithm to identify the time period with the highest probability value in the sequence and defines it as the ideal time period for users. Subsequently, the system calculates the time slice for the current traversal. and absolute time difference between The larger this difference, the further the scheduling time deviates from the user's psychological expectations, and the greater the potential energy. It then exhibits linear or nonlinear growth.

[0080] Finally, and most importantly, is the fragmentation penalty potential energy. In actual scheduling, if the tasks assigned to agricultural machinery result in a very short waiting time (e.g., 30 minutes) between two plots, this time is insufficient to allow the machinery to be moved to other locations for work, and also prevents the driver from getting adequate rest, representing a pure waste of resources. Therefore, the system calculates the current time slice. The waiting time of the agricultural machinery is calculated by the difference between the end time of adjacent tasks in the agricultural machinery's scheduled task list and the end time of the next scheduled task. The system has a preset threshold for time fragmentation. (For example, 1 hour). The system will detect... Is it within the range? Within this range, once the value falls into this range, the system determines that unacceptable fragmentation loss has occurred and immediately outputs a preset maximum penalty value (e.g., ...). As Conversely, if or (Seamless connection), then It is 0.

[0081] Finally, based on a preset weight vector, the system synthesizes the above three components into a total objective function: The system employs a traversal optimization algorithm to quickly calculate the corresponding time slice within the finite discrete set of the safe operation feasible region. The value is selected, and the time slice that minimizes the total potential energy is chosen. The time slice was ultimately locked as the target execution time. Based on this, the system generated a target agricultural operation scheduling instruction that included the operation time, specified agricultural machinery ID, operation type, and number of users participating. This instruction was then sent to the agricultural machinery vehicle terminal and the user's APP via a message queue, completing a full closed loop from data perception to intelligent decision-making.

[0082] In this process, the system introduces a potential energy threshold circuit breaker mechanism. This mechanism filters out the minimum potential energy value. Then, the system will compare it with the preset acceptable potential energy limit. Perform a comparison. If... This indicates that there are severe resource conflicts or extremely low user matching (i.e., no feasible solution) in all time periods within the current safe operational feasible domain. In this case, the system does not generate automatic scheduling instructions, but instead triggers an anomaly alarm, pushing the scheduling task for this plot to the manual decision-making terminal, suggesting that the dispatcher resolve the conflict by manually adjusting the operational constraints or coordinating backup agricultural machinery resources.

[0083] Step S7: After determining the target agricultural operation scheduling instruction, the system enters the execution and notification phase.

[0084] Specifically, the system monitors the current time and the target job start time in real time. The difference. When the time difference After one hour, the system automatically activates the user notification module. At this time, instead of sending a uniform template to all users, the system invokes the experience activity scheduling module to execute personalized matching logic.

[0085] Spatiotemporal reachability calculation: The system obtains the user's current geographic location coordinates (based on LBS service) and the commuting distance to the target location. Then, the time preference feature in the second feature vector generated in step S2 is retrieved again.

[0086] like (e.g., 50km) and the user's idle probability The system determines that the user meets the conditions for offline participation, automatically pushes a job announcement containing an entry point for offline participation reservations, and intelligently recommends activity content based on the current agricultural type (for example: if it is rapeseed flowering season, recommend flower sea photography + drone aerial photography experience; if it is rice harvesting season, recommend parent-child rice harvesting + threshing experience).

[0087] If the above conditions are not met, the system determines that the user is online, automatically generates a unique low-latency live stream link (RTMP stream), and pushes it to the user's terminal, providing a cloud monitoring option.

[0088] By using this intelligent routing based on geolocation and time, the system effectively reduces the nuisance rate of invalid notifications and ensures the accuracy of the pushed content.

[0089] To achieve deep involvement in crops throughout the year, the system operates an interactive feedback module and establishes a long-term growth record mechanism.

[0090] Specifically, the system uses a weekly time window to periodically trigger the generation of weekly growth reports. The system automatically retrieves crop phenotypic image data stored in step S1, and uses image registration technology to align and fuse images collected from the same observation point over the past 7 days, generating a time-lapse of crop growth. Simultaneously, the system combines environmental monitoring data to automatically extract the week's accumulated temperature increase, total rainfall, and key agricultural events (such as fertilization completion and flowering detection), and uses natural language generation (NLG) technology to automatically generate weekly growth reports for the adopted plots.

[0091] The weekly report and video recordings are pushed to relevant users via the app's message center and visualized on the user's digital farm interface. If the system detects a critical transition in crop growth (such as moving from the jointing stage to the heading stage), it will also trigger real-time highlight push notifications to ensure that users do not miss any key milestones in crop growth, thereby achieving a deep experience upgrade from single-event participation to full-lifecycle companionship.

[0092] Furthermore, this embodiment is not limited to the scheduling of single agricultural activities, but also constructs a long-term mechanism traceability and value closed-loop process. To address the frequent problem of agricultural machinery relocation efficiency in rice-rice-oilseed rotation, the system introduces a feedback mechanism based on physical topology.

[0093] First, the system establishes a physical spatial topology map of the area where the target plot is located, based on a GIS geographic information system. Among them, the node set Representing each independent work site spatial node, edge set This represents the agricultural machinery operation paths (field ridges, farm roads) connecting these plots. The system maps the geometric center of each plot to a node in the graph.

[0094] After a complete scheduling cycle (e.g., from planting to harvesting of a crop season), the system initiates mechanization-based source analysis. The system retrieves all fragmented penalty potential energies calculated in step S6 throughout the entire cycle from historical scheduling logs. The system identifies the geographical location or associated plot node corresponding to each trigger of a maximum penalty value (i.e., ineffective waiting of agricultural machinery or fragmented cutting), and accumulates these penalty values ​​at the corresponding spatial nodes. Through this process, the system generates a cumulative obstruction heat map covering the entire topology. Nodes with higher heat values ​​mean that in past crop rotation cycles, irregular shapes, isolated locations, or impassable roads have most frequently caused efficiency gaps in agricultural machinery scheduling.

[0095] Finally, the system sets preset modification thresholds and automatically identifies abnormal nodes with excessive heat values. Based on graph theory algorithms, the system analyzes the connectivity relationships (such as degree and betweenness centrality) of abnormal nodes in the topology graph and generates targeted physical modification suggestions. For example, if a node has an extremely high heat value and low connectivity with surrounding nodes, the system suggests connecting it with a road; if two adjacent small nodes both have high heat values, the system suggests removing field ridges and merging plots. These suggestions are packaged into an engineering report to guide the construction of farmland infrastructure in the next crop cycle, thereby reducing the system's basic potential energy at the physical level.

[0096] In summary, this embodiment constructs a complete technical chain from multi-source data perception to intelligent decision-making closed loop through the above steps, which completely solves the binary paradox between rigid agricultural time constraints and flexible experience requirements in the rice-rice-oil rotation model. Furthermore, at the physical level, it realizes the adaptive evolution of the production environment through mechanized traceability, thus providing a highly robust engineering paradigm for the large-scale implementation of adoption agriculture under the high-intensity crop rotation system.

[0097] Example 2:

[0098] like Figure 3 As shown, the intelligent scheduling and user participation system for rice-oilseed rotation agricultural operations includes:

[0099] The data acquisition module is used to acquire environmental monitoring data, crop phenotypic image data, and historical interaction log data of associated users for the target plot.

[0100] The feature analysis module is used to perform time-series analysis on crop phenotypic image data and environmental monitoring data to generate a first feature vector representing the urgency of crop growth; and to perform multi-dimensional feature mining on historical interaction log data to generate a second feature vector containing user behavior reliability features and time preference features.

[0101] The time period initial screening module is used to determine the available time periods within a preset scheduling period based on preset job constraints, and to screen the available time periods according to the first feature vector and the second feature vector to obtain an initial candidate time window set.

[0102] The group risk assessment module is used to identify potential participating user groups within each candidate time period in the initial candidate time window set based on the second feature vector, and to calculate the group risk parameter representing the unreliability of the behavior of potential participating user groups.

[0103] The safety boundary calculation module is used to parse the first feature vector to determine the physical remaining time of agronomic operations and construct a safety buffer margin based on the group risk parameters; the safety cutoff time is determined by calculating the difference between the physical remaining time and the safety buffer margin, and the subset of the initial candidate time window set that is earlier than the safety cutoff time is determined as the feasible region of safe operations;

[0104] The optimized scheduling generation module is used to construct an objective function that includes the potential energy of mechanical operation costs and the potential energy of user matching degree within the feasible domain of safe operation, and to determine the target agricultural operation scheduling instructions by optimizing the objective function.

[0105] The above description is merely an example and illustration of the structure of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the structure of the invention or exceed the scope defined in the claims, all of which should fall within the protection scope of the present invention.

[0106] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0107] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for intelligent scheduling and user participation in rice-oilseed rotation agricultural operations, characterized in that: Includes the following steps: Acquire environmental monitoring data, crop phenotypic image data, and historical interaction log data of associated users for the target plot; Time-series analysis is performed on the crop phenotypic image data and the environmental monitoring data to generate a first feature vector representing the urgency of crop growth. Multidimensional feature mining is performed on the historical interaction log data to generate a second feature vector containing user behavior reliability features and time preference features; Within a preset scheduling period, available time periods are determined based on preset job constraints, and the available time periods are filtered according to the first feature vector and the second feature vector to obtain an initial set of candidate time windows. For each candidate time period in the initial candidate time window set, the potential participating user group within that time period is identified based on the second feature vector, and a group risk parameter representing the unreliability of the behavior of the potential participating user group is calculated; The first feature vector is parsed to determine the physical remaining time of agronomic operations, and a safety buffer margin is constructed based on the population risk parameter; the safety cutoff time is determined by calculating the difference between the physical remaining time and the safety buffer margin, and the subset of the initial candidate time window set that is earlier than the safety cutoff time is determined as the feasible region for safe operations; Within the feasible domain of safe operation, an objective function is constructed that includes the potential energy of mechanical operation cost and the potential energy of user matching degree. The target agricultural operation scheduling instruction is determined by optimizing the objective function.

2. The method for intelligent scheduling and user participation in rice-oilseed rotation agricultural operations according to claim 1, characterized in that: The process of generating the first feature vector includes: The crop phenotypic image data is identified based on a pre-trained convolutional neural network model, and the current crop growth stage label is output. The preset crop accumulated temperature model is invoked, and the physical remaining time from the preset agronomic limit operation point is calculated based on the historical effective accumulated temperature in the growth stage label and the environmental monitoring data. The physical remaining time is mapped to a normalized value using a nonlinear S-shaped decay function to obtain the agronomic urgency index. The agronomic urgency index is combined with the physical remaining time to form the first feature vector.

3. The method for intelligent scheduling and user participation in rice-oilseed rotation agricultural operations according to claim 1, characterized in that: The process of generating the second feature vector includes: Extract the time deviation between the user's historical activity registration time and the actual check-in time from the historical interaction log data, and calculate the variance of the time deviation value; Obtain the user's historical number of defaults, and generate an individual reliability score based on the weighted sum of the variance and the historical number of defaults; Statistically analyze the distribution of user activity frequency over historical time periods to construct an activity probability map that maps the relationship between time slices and participation probabilities; The activity probability map is vectorized to obtain a time distribution feature sequence, and the individual reliability score is concatenated with the time distribution feature sequence to generate the second feature vector.

4. The method for intelligent scheduling and user participation in rice-oilseed rotation agricultural operations according to claim 1, characterized in that: The available time periods are filtered based on the first feature vector and the second feature vector, including: The agronomic urgency index is obtained by parsing the first feature vector, and a dynamic weight coefficient is constructed based on the agronomic urgency index; the dynamic weight coefficient is positively correlated with the agronomic urgency index. Visual feature analysis is performed on the crop phenotypic image data to extract color feature parameters that characterize the crop's ornamental attributes. Based on the color feature parameters and the environmental monitoring data, a landscape aesthetic score is calculated and generated. Based on the dynamic weighting coefficients, the agronomic urgency index and the landscape aesthetics score are weighted and summed to obtain the basic feasibility score; The available time periods with a basic feasibility score higher than a preset score threshold are determined as the initial candidate time window set.

5. The method for intelligent scheduling and user participation in rice-oilseed rotation agricultural operations according to claim 3, characterized in that: The calculation process for the group risk parameter includes: For each candidate time period, a set of target users with a participation probability higher than a preset probability threshold is selected based on the activity probability map; the individual reliability score of each user in the target user set is extracted from the second feature vector; Calculate the arithmetic mean of the reliability scores of individuals in the target user set; Discretize the distribution of individual reliability scores in the target user set and calculate the information entropy of the distribution as the group behavior entropy; The group risk parameter is generated based on the product of the reciprocal of the arithmetic mean and the group behavior entropy.

6. The method for intelligent scheduling and user participation in rice-oilseed rotation agricultural operations according to claim 2, characterized in that: A safety buffer margin is constructed based on the aforementioned group risk parameters, and the safety cutoff time is determined, including: Obtain the confidence coefficient of external weather forecast data, and calculate the safety buffer margin for agricultural machinery operations based on the population risk parameter and the confidence coefficient. Subtract the safety buffer margin from the remaining physical duration in the first feature vector, and add the current time to obtain the safety cutoff time.

7. The method for intelligent scheduling and user participation in rice-oilseed rotation agricultural operations according to claim 1, characterized in that: The process of constructing the objective function includes: Obtain service status data from the agricultural machinery dispatch center, the service status data including the current geographical coordinates of available agricultural machinery and the time occupancy table of scheduled tasks; For each discrete time slice within the feasible region of safe operation, perform the following potential energy component calculations: The transfer distance cost is calculated based on the Euclidean distance between the target plot's location coordinates and the current geographical location coordinates; the idle time cost is calculated based on the interval between the discrete time slice and the time occupancy table of the pre-defined tasks; the transfer distance cost and the idle time cost are weighted and summed to obtain the mechanical operation cost potential energy. Extract the time distribution feature sequence from the second feature vector, and determine the time period with the highest probability value in the time distribution feature sequence as the user's ideal time period; calculate the absolute time difference between the discrete time slice and the user's ideal time period, and generate the user matching potential based on the absolute time difference; Calculate the difference between the discrete time slice and the end time of adjacent tasks in the time occupancy table of the pre-defined tasks to obtain the waiting time of the agricultural machinery; determine whether the waiting time of the agricultural machinery is less than the preset time fragment threshold and greater than zero; if so, output the preset maximum penalty value as the fragmentation penalty potential energy. The objective function is constructed based on the weighted sum of the mechanical operation cost potential energy, the user matching degree potential energy, and the fragmentation penalty potential energy; Within the feasible domain of safe operation, the objective function is iterated and optimized, and the discrete time slice corresponding to the minimum function value is selected as the execution time in the objective agricultural operation scheduling instruction.

8. The method for intelligent scheduling and user participation in rice-oilseed rotation agricultural operations according to claim 1, characterized in that: Generating a second feature vector that includes user behavior reliability features and time preference features also includes: In response to the detection that the historical interaction log data of the associated user is empty or the data volume is lower than a preset statistical threshold, a cold start mapping mechanism is triggered: Obtain the static attribute features of the target plot, which include at least the plot's geographical coordinates, crop planting type, and transportation accessibility level; Calculate the similarity distance between the static attribute features of the target land parcel and the attribute features of historical mature land parcels in the database, and select a preset number of similar land parcels based on the similarity distance; Extract the average reliability score and general time preference template of the user groups associated with the similar land parcels; The average reliability score is assigned as the user behavior reliability feature, and the general time preference template is assigned as the time preference feature. The two are combined to generate the second feature vector for cold start.

9. The method for intelligent scheduling and user participation in rice-oilseed rotation agricultural operations according to claim 7, characterized in that: The method further includes: Establish a physical space topology map of the target plot, which consists of several spatial nodes and operation paths connecting the nodes; After completing a full scheduling cycle, the value of the fragmented penalty potential energy triggered within the safe operation feasible domain is accumulated, and the accumulated value is mapped to the corresponding spatial node to generate a cumulative blocking thermal value. Identify abnormal nodes in the physical space topology map whose cumulative resistance thermal value is higher than a preset modification threshold; Based on the connection relationship of the abnormal nodes in the physical space topology map, physical modification suggestions are generated for the target plot.

10. A smart scheduling and user participation system for rice-oilseed rotation agricultural operations, characterized in that: include: The data acquisition module is used to acquire environmental monitoring data, crop phenotypic image data, and historical interaction log data of associated users for the target plot. The feature analysis module is used to perform time-series analysis on the crop phenotypic image data and the environmental monitoring data to generate a first feature vector representing the urgency of crop growth; and to perform multi-dimensional feature mining on the historical interaction log data to generate a second feature vector containing user behavior reliability features and time preference features. The time period initial screening module is used to determine the available time periods based on preset job constraints within a preset scheduling cycle, and to screen the available time periods according to the first feature vector and the second feature vector to obtain an initial candidate time window set. The group risk assessment module is used to identify potential participating user groups within each candidate time period in the initial candidate time window set based on the second feature vector, and to calculate a group risk parameter representing the unreliability of the behavior of the potential participating user group. The safety boundary calculation module is used to parse the first feature vector to determine the physical remaining time of agronomic operations and construct a safety buffer margin based on the group risk parameters; determine the safety cutoff time by calculating the difference between the physical remaining time and the safety buffer margin, and determine the subset of the initial candidate time window set that is earlier than the safety cutoff time as the feasible region for safe operations; The optimized scheduling generation module is used to construct an objective function containing the potential energy of mechanical operation costs and the potential energy of user matching degree within the safe operation feasible domain, and to determine the target agricultural operation scheduling instruction by optimizing the objective function.

Images