Agricultural meteorological data intelligent acquisition system based on big data

Abstract

The invention discloses an agricultural meteorological data intelligent acquisition system based on big data, relates to the technical field of agricultural meteorological data processing, and adopts an image sensor to intelligently identify a crop growth stage, dynamically adjusts meteorological data sampling frequency based on a sensitivity matrix, and realizes optimal balance between monitoring precision and energy consumption. Weight coefficients of meteorological factors are continuously corrected through a closed-loop feedback mechanism, so that a sampling strategy can adapt to growth requirements of different crops, and meanwhile, the execution control module ensures the reliability of the system through threshold judgment and multi-node data verification. Through combination of growth stage identification and dynamic frequency calculation, accurate monitoring is realized; a feedback regulation mechanism enables the system to have continuous optimization capability and can adapt to a complex agricultural environment; the multi-module collaborative design gives consideration to the data acquisition efficiency and the system stability.

Description

Technical Field

[0001] The present invention relates to the technical field of agricultural meteorological data processing, and in particular to an intelligent agricultural meteorological data collection system based on big data. Background Art

[0002] In agricultural production, meteorological data (such as temperature, humidity, and light) have an important impact on crop growth, pest and disease control, and disaster warning. Traditional meteorological data acquisition systems usually use a fixed sampling frequency and cannot be dynamically adjusted according to the crop growth stage or meteorological anomalies, resulting in the following problems: Low data acquisition efficiency: The fixed sampling frequency may result in key meteorological events (such as frost and drought) not being captured in a timely manner, or data redundancy in non-critical periods, wasting storage and computing resources. Lack of growth stage adaptability: Different crop growth stages (such as germination, flowering, and maturity) have different sensitivities to meteorological factors, but existing systems cannot automatically identify phenological periods and adjust monitoring strategies.

[0003] In recent years, with the development of image recognition technology, some studies have attempted to determine crop growth status based on phenotypic characteristics, but the dynamic matching of monitoring parameters with growth stages has yet to be effectively addressed. Furthermore, while the application of big data technology has enhanced meteorological data analysis capabilities, a significant technological gap remains in the intelligent front-end data collection. Summary of the Invention

[0004] In view of the shortcomings of the existing technology, the present invention provides an intelligent collection system for agricultural meteorological data based on big data.

[0005] In order to achieve the above object, the technical solution of the present invention is as follows:

[0006] An intelligent agricultural meteorological data collection system based on big data, including

[0007] The phenological phase recognition module is used to obtain crop canopy image data through an image sensor and determine the current growth stage data based on a preset color ratio threshold;

[0008] Meteorological data acquisition module, used to obtain temperature, humidity and light intensity data at the sampling frequency fed back by the dynamic frequency calculation module;

[0009] Sensitivity matrix storage module, which stores the mapping relationship table between growth stage data and meteorological factor weight coefficients;

[0010] A dynamic frequency calculation module is used to query the mapping relationship table according to the current growth stage data to obtain the maximum weight coefficient, and calculate the target sampling frequency data in combination with the real-time fluctuation data of meteorological factors;

[0011] A feedback adjustment module is used to compare the historical meteorological event capture result data with the expected capture rate data to generate weight coefficient correction instruction data;

[0012] The execution control module is used to control the meteorological data acquisition module to switch the sampling frequency and send data verification requests to adjacent nodes when the target sampling frequency data exceeds the preset threshold; and to receive the correction instruction data output by the feedback adjustment module to update the mapping relationship table.

[0013] Compared with the prior art, the present invention has the following beneficial effects:

[0014] 1. Through the synergistic effect of the phenological period recognition module and the sensitivity matrix, differentiated collection of key meteorological factors for different growth stages is achieved, crop growth stages are automatically identified, and meteorological data collection strategies are dynamically adjusted;

[0015] 2. Based on big data analysis, a mapping relationship between growth stages and meteorological factor sensitivity is established, the sampling frequency is optimized, a feedback adjustment mechanism is introduced, and the weight coefficient is corrected according to the historical disaster capture rate to improve monitoring accuracy;

[0016] 3. Through the combination of big data analysis and intelligent control technology, the accuracy, efficiency and reliability of agricultural meteorological monitoring have been significantly improved, providing better technical solutions for smart agriculture. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The disclosure of the present invention is described with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of the present invention. In the drawings, the same reference numerals are used to refer to the same components. Among them:

[0018] Figure 1 This is a schematic diagram of module functions of the present invention;

[0019] Figure 2 This is a working diagram of the dynamic frequency calculation module of the present invention;

[0020] Figure 3 This is a working diagram of the feedback regulation module of the present invention;

[0021] Figure 4 Schematic diagram of the process of dynamically adjusting the sampling frequency of the execution control module of the present invention;

[0022] Figure 5 This is a working diagram of the data compensation module of the present invention;

[0023] Figure 6 This is a working diagram of the abnormal learning unit of the present invention. DETAILED DESCRIPTION

[0024] It is easy to understand that according to the technical solution of the present invention, without changing the essential spirit of the present invention, a person skilled in the art can propose a variety of interchangeable structural modes and implementation modes. Therefore, the following specific embodiments and drawings are only exemplary descriptions of the technical solution of the present invention and should not be regarded as the entire invention or as a limitation or restriction of the technical solution of the present invention.

[0025] Application Overview:

[0026] Agrometeorological monitoring is a key technical support for the development of modern agriculture. The precise collection and analysis of crop growth environment parameters directly affects the scientific nature and timeliness of agricultural production decisions. Traditional agrometeorological monitoring systems mainly use a fixed-frequency data collection mode, which has the following technical defects:

[0027] 1. Lack of targeted sampling strategies: Existing systems usually adopt a unified sampling frequency and cannot dynamically adjust according to the differences in sensitivity of crops to meteorological factors at different growth stages, resulting in insufficient data collection during critical growth periods or waste of resources during non-critical periods.

[0028] 2. Insufficient environmental adaptability: Conventional systems find it difficult to respond to sudden changes in meteorological conditions in real time. Monitoring blind spots often occur when extreme weather events occur, affecting the timeliness of disaster warnings.

[0029] In response to the above-mentioned defects in the existing technology, the basic concept of this application is to perform real-time analysis of crop canopy images through a phenological period recognition module, and accurately judge the current growth stage based on a preset color ratio threshold; the meteorological data acquisition module obtains key meteorological parameters such as temperature, humidity and light intensity at a dynamically changing sampling frequency according to system instructions, ensuring the targeted and timely nature of data acquisition.

[0030] The core of the system lies in the mapping table between growth stages and meteorological factor weight coefficients established in the sensitivity matrix storage module. This design enables the system to understand the differences in sensitivity to various meteorological factors at different growth stages. The dynamic frequency calculation module queries this mapping table to obtain the maximum weight coefficient. It then performs intelligent calculations based on real-time fluctuations in meteorological factors to output the optimal target sampling frequency. The feedback adjustment module continuously monitors the deviation between historical meteorological event capture results and expected values ​​and generates weight coefficient correction instructions to optimize system performance.

[0031] After introducing the basic concept of the present invention, embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0032] Example

[0033] like Figure 1 As shown in the figure, an intelligent agricultural meteorological data collection system based on big data includes:

[0034] The phenological period recognition module is used to obtain crop canopy image data through an image sensor and determine the current growth stage data based on a preset color ratio threshold.

[0035] The process of judging the current growth stage data based on the preset color ratio threshold is as follows:

[0036] Extract R channel pixel ratio data of crop canopy image data. When the ratio data exceeds a preset threshold corresponding to the current growth stage, generate phenological period switching instruction data and update the current growth stage data.

[0037] Specifically, the phenological period recognition module uses image sensors (such as RGB cameras, multispectral cameras) to collect crop canopy images in real time, and automatically determines the current growth stage of the crop (such as germination period, growth period, flowering period, maturity period, etc.) based on color feature analysis (R channel pixel ratio).

[0038] The logic of its implementation is as follows:

[0039] The color characteristics of different growth stages are obviously different, for example, red / white is the main color during the flowering stage, and yellow / brown is the main color during the mature stage;

[0040] The pixel ratio threshold of the R / G / B channels can be used to quantitatively characterize the changes in the growth stage;

[0041] Dynamically update growth stage data to provide a basis for subsequent adjustments to meteorological data collection strategies.

[0042] The implementation process is roughly as follows:

[0043] The image sensor captures wheat canopy images at regular intervals (e.g., at 10 a.m. every day) and performs preprocessing such as denoising, cropping, and light normalization on the images to ensure the accuracy of color analysis.

[0044] Extract the RGB three-channel data of the image, count the pixel ratio of the R channel (red), and determine whether the pixel ratio of the R channel (red) exceeds the preset threshold;

[0045] The amount of preset thresholds can be adjusted according to the crop type, the following table shows an example of one of them:

[0046] Growth stage R channel ratio threshold Typical color characteristics budding stage ＜15％ Mainly green (chlorophyll-dominated) Growth period 15％～30％ dark green Flowering period ＞40％ Red / white ear Maturity 30％～50％ The ear turns yellow / brown

[0047] According to the above table, when the R channel ratio is detected to be greater than 40% for three consecutive days and the current stage is the "growth stage", it is determined to have entered the "flowering stage";

[0048] Generate phenological period switching instruction data and update the system global growth stage label.

[0049] Example: Winter wheat flowering period monitoring;

[0050] Initial state: The system defaults to the growth stage of "growth period" and the average proportion of R channel is 25%;

[0051] Monitoring process:

[0052] Day 1: The R percentage suddenly increased to 45% (because the ear began to show color), but it did not last, so the stage was not switched for the time being;

[0053] Day 3: If the R ratio continues to be greater than 42%, the system will determine that the disease has entered the "flowering stage";

[0054] Linkage effect:

[0055] The dynamic frequency calculation module automatically increases the sampling frequency of temperature and humidity (flowering period is sensitive to frost);

[0056] The feedback adjustment module records the abnormal meteorological events (such as low temperature) in this stage for subsequent weight coefficient optimization.

[0057] The meteorological data acquisition module is used to obtain temperature, humidity and light intensity data at the sampling frequency fed back by the dynamic frequency calculation module.

[0058] The meteorological data acquisition module is the system's data input unit. Its primary function is to collect key meteorological data such as temperature, humidity, and light intensity in real time at a dynamically adjusted sampling frequency and transmit this data to subsequent processing modules. It can collaborate with multiple sensors (temperature, humidity, and light sensors), supporting flexible switching from low to high frequencies and sampling at different frequencies.

[0059] Example: Frost warning for strawberry greenhouses;

[0060] Workflow diagram:

[0061] By default, the basic sampling frequency (such as 1 time / 10 minutes) is used, the growth stage is "growth period", and the data is recorded: temperature 15°C, humidity 65%, and light 2000 lux.

[0062] The dynamic frequency calculation module detects a sudden drop in temperature to 2°C (the coefficient of variation exceeds the threshold) and automatically increases the sampling frequency to 1 time / 1 minute; the real-time data stream triggers a frost alarm, notifying farmers to start heating equipment.

[0063] The sensitivity matrix storage module stores the mapping relationship table between growth stage data and meteorological factor weight coefficients.

[0064] The sensitivity matrix storage module stores the sensitivity weights of different crop varieties to meteorological factors (temperature, humidity, and light) at various growth stages in a structured form. The matrix is ​​stored independently by crop type (such as rice, wheat, and corn), and the weight coefficients are generated by inverting historical disaster data. It also supports real-time correction of the feedback adjustment module and uses a key-value database (such as Redis) or a relational database (such as MySQL) to achieve millisecond-level response.

[0065] The weight coefficient is calculated through the inversion of historical disaster statistics, and the formula is as follows:

[0066]

[0067] Taking the sensitivity matrix of winter wheat as an example, the weights of meteorological factors at different stages are:

[0068] Germination: ("temp": 0.6, "hum": 0.3, "light": 0.1);

[0069] Tiller: ("temp": 0.5, "hum": 0.4, "light": 0.1);

[0070] bloom: ("temp": 0.8, "hum": 0.1, "light": 0.1);

[0071] Maturity: (“temp”: 0.3, “hum”: 0.5, “light”: 0.2).

[0072] For example, suppose there are 20 disasters during the rice seedling stage, including:

[0073] Low temperature results in 15 events, high temperature results in 2 events → Temperature weight = (15 + 2) / 20 × 0.9 (attenuation coefficient) = 0.76;

[0074] Drought causes 3 times → humidity weight = 3 / 20×0.9 = 0.14;

[0075] Insufficient light results in 0 times → Lighting weight = 0.1 (bottom line value).

[0076] Example: High temperature warning during corn tasting period.

[0077] The phenological phase recognition module determines that the current stage is the "tasseling" stage of corn;

[0078] The sensitivity matrix storage module returns weights based on the "tasseling" stage of corn: ("temp": 0.7, "hum": 0.2, "light": 0.1).

[0079] The dynamic frequency calculation module is used to query the mapping relationship table according to the current growth stage data to obtain the maximum weight coefficient, and calculate the target sampling frequency data in combination with the real-time fluctuation data of meteorological factors.

[0080] like Figure 2 As shown in the figure, this module represents the system's primary computational processing. It dynamically adjusts the sampling frequency by quantitatively analyzing crop sensitivity and weather volatility. The module simultaneously considers the impact of both the crop growth stage (static weight) and real-time weather fluctuations (dynamic risk) on the sampling frequency. It automatically reduces the sampling frequency after verifying the risk has been resolved, thus avoiding resource waste and enabling rapid response to sudden weather events.

[0081] The dynamic frequency calculation module includes:

[0082] The fluctuation analysis unit is used to calculate the coefficient of variation data of the target meteorological factor in the last 24 hours.

[0083] The calculation process is as follows:

[0084] Input the sampling data sequence of the target meteorological factor (such as temperature) in the last 24 hours X={x1,x2,...,x n};

[0085] The calculation formula is: coefficient of variation

[0086] Among them, σ is the standard deviation and μ is the mean;

[0087] In arid areas with large temperature differences between day and night, the daily temperature CV can reach 15%, while in greenhouse environments the CV is usually <5%.

[0088] an amplification calculation unit, for multiplying the maximum weight coefficient by the square root of the coefficient of variation to generate dynamic amplification factor data;

[0089] The calculation formula is: Dynamic magnification

[0090] The square root operation of the coefficient of variation is used to prevent high-frequency sampling overload, and when CV is less than 5%, K=1 is forced to avoid meaningless amplification.

[0091] Calculation example:

[0092] Flowering temperature weight W max =0.8,

[0093] The frequency synthesis unit is used to multiply the basic sampling frequency data with the dynamic amplification factor data and output the target sampling frequency data to the execution control module.

[0094] The calculation formula is: target sampling frequency ftarget =f0×(1+K×S scale );

[0095] Among them, f0 is set as the basic frequency, which is set to 1 time / 10 minutes; S scale is a configurable scaling factor.

[0096] Continued example: f target =6×(1+0.25×10)=1 time / 2.4 minutes.

[0097] The ramp-down control module is used to gradually reduce the target sampling frequency data according to a preset attenuation coefficient until the basic sampling frequency data is restored when no meteorological abnormal data is detected within N consecutive sampling periods, where N is a preset positive integer.

[0098] The decay rule is set as:

[0099] Frequency attenuation after every N normal sampling times Δf=β×f current ;

[0100] Where β is the configurable attenuation coefficient, which is set to 0.2; f current is the real-time frequency, when f current The decay stops when ≤f0.

[0101] The Fluctuation Analysis Unit is also used to:

[0102] Detect the sudden change gradient of the coefficient of variation data. When the gradient value exceeds the preset warning line, send emergency calibration request data to the feedback adjustment module. The feedback adjustment module responds to the request and immediately starts the weight coefficient correction process.

[0103] The calculation method is:

[0104] The sliding window calculates the instantaneous rate of change of CV (t is the gradient at the moment):

[0105] Instantaneous rate of change

[0106] Default warning line G threshold If it is set to 5% / hour, for example, if the CV suddenly rises from 8% to 15%, it means that the instantaneous rate of change exceeds the default warning line;

[0107] The feedback adjustment module responds to the request of the emergency calibration request data and starts the weight coefficient correction process as follows:

[0108] The meteorological type corresponding to the coefficient of variation data is marked as an emergency factor, the sampling frequency is forced to be increased to the highest frequency, and the real-time correction thread of the weight coefficient is started.

[0109] Example: Event Characteristics:

[0110] Light CV gradient from 2% / h→12% / h (exceeding the threshold);

[0111] The temperature CV is increased synchronously (but the gradient does not exceed the limit).

[0112] The response result is:

[0113] Mark light as an emergency factor;

[0114] Temporarily override the original weight and increase the lighting sampling frequency to 1 time / 30 seconds;

[0115] The feedback adjustment module completes the correction of the light weight coefficient +0.15 within 2 minutes.

[0116] Complete example of the dynamic frequency calculation module workflow:

[0117] The scene is a vegetable base;

[0118] Baseline configuration:

[0119] f0=1 time / 15 minutes, S scale =8, N=3, β=0.15.

[0120] The effect comparison results are shown in the following table:

[0121] index Fixed frequency mode Dynamic Frequency Mode Average number of sampling times per day 96 times 127 times Disaster capture rate 68％ 92％ Sensor battery life 30 days 27 days

[0122] The feedback adjustment module is used to compare the historical meteorological event capture result data with the expected capture rate data to generate weight coefficient correction instruction data.

[0123] like Figure 3 As shown in Figure 1, this module primarily quantitatively evaluates the deviation between actual monitoring results and theoretical expectations, dynamically correcting the weight coefficients in the sensitivity matrix and achieving continuous evolution of system performance. The module relies on establishing a complete "monitoring-assessment-correction" processing chain. In actual deployments, single deviations trigger fine-tuning, while continuous deviations trigger version upgrades. By setting these two sets of parameters, it filters out false corrections caused by random fluctuations.

[0124] The specific process of comparing the historical meteorological event capture result data with the expected capture rate data is as follows:

[0125] Statistics are collected on the number of meteorological anomalies actually captured in the high-frequency sampling mode in the most recent M growth cycles, and the percentage deviation from the theoretical capture number is calculated. When the percentage exceeds the preset tolerance, the corresponding weight coefficient data in the mapping relationship table is adjusted proportionally in the direction of the deviation; where M is a preset positive integer, and the theoretical capture number is the number of disasters that should be captured in this growth stage predicted based on historical data.

[0126] The data on the number of meteorological anomaly events actually captured are the actual captured data, which are meteorological anomaly events confirmed during the high-frequency sampling period. For example, a temperature > 35°C for 2 hours is recorded as one high temperature event;

[0127] The theoretical capture number is a Poisson distribution prediction based on historical data. For example, the theoretical number of frosts during the heading period of a certain variety of wheat is 2.3 times per season.

[0128] The formula for calculating the deviation percentage is:

[0129] Among them, N actual is the actual captured data, and is the theoretical capture number.

[0130] Example: 3 actual frost captures, theoretical 2 → δ = (3-2) / 2×100% = +50%;

[0131] The preset tolerance is ±20% → a correction is triggered (δ>+20%).

[0132] The weight coefficient adjustment process is:

[0133] The proportional correction formula is:

[0134] W new =W old ×(1+λ·δ), where λ is a configurable damping coefficient used to prevent over-adjustment and its default value is 0.5.

[0135] Correction direction:

[0136] δ>0: Too much actual capture → current weight is too high → lower the coefficient;

[0137] δ<0: Actual capture is insufficient → current weight is too low → increase the coefficient.

[0138] Example:

[0139] The original value of humidity weight during the flowering period of a rice field is 0.4, and the actual drought event capture rate is low (δ = -30%), then:

[0140] New weight = 0.4 × (1 + 0.5 × 0.3) = 0.46.

[0141] When the direction of the weight coefficient adjustment data is consistent for K consecutive times in the same growth stage, the version iteration instruction data of the sensitivity matrix is ​​triggered, where K is a preset threshold and K≥3.

[0142] Verify the continuous consistency in advance, record the direction of the latest K corrections (↑ / ↓), and when K consecutive corrections are in the same direction, it is determined to be a systematic deviation, triggering a matrix version upgrade.

[0143] An example is shown in the following table: K = 3;

[0144] Correction batch Deviation δ Adjust direction Consistency Count 1 -25％ ↑ 1 2 -18％ ↑ 2 3 -30％ ↑ 3 → Trigger iteration

[0145] A complete example of the feedback adjustment module workflow:

[0146] The scenario is winter wheat planting monitoring;

[0147] Problem found: The actual capture rate of late spring frost during the jointing period from 2021 to 2023 was only 40%, significantly lower than the theoretical value of 65%.

[0148] Feedback regulation process:

[0149] Statistical data for M = 2 growing seasons, calculated δ = -38.5% (exceeds the tolerance);

[0150] Temperature weight from 0.6→0.73 (λ=0.5);

[0151] After three consecutive quarters of same-direction revisions (K=3), the release iteration matrix is:

[0152] The temperature weight at the jointing stage is increased to 0.8;

[0153] Added a new temperature drop gradient monitoring item.

[0154] Effect: The capture rate of late spring cold weather in 2024 increased to 78%, and the false alarm rate decreased by 12%.

[0155] The interaction between the feedback regulation module and the meteorological data acquisition module includes:

[0156] The sensor energy consumption data during high-frequency sampling is obtained. When the ratio of the energy consumption increment per unit time to the capture rate increase exceeds the preset economic coefficient, the sampling frequency degradation recommendation data is generated for the execution control module to make a decision.

[0157] The core of this process is to achieve energy consumption optimization by quantitatively evaluating the cost-effectiveness of "increased energy consumption - increased capture rate"; avoid energy waste caused by oversampling (especially for solar-powered equipment); and intelligently reduce unnecessary sampling while ensuring the accuracy of key meteorological monitoring, ultimately forming a complete control chain of "energy consumption monitoring → economic analysis → strategy adjustment → effect confirmation".

[0158] The specific process is:

[0159] The sensor energy consumption data in high-frequency sampling mode is obtained through real-time monitoring of the current sensor, and the difference ΔE between the sensor energy consumption at the current frequency and the lowest frequency is calculated.

[0160] The calculation formula for the increase in capture rate is: where N high is the number of captures at the current frequency, and is the number of captures at the lowest frequency.

[0161] The calculation formula of the economic coefficient is: Among them, θ eco It is the preset economic coefficient, with a typical value of 0.5, indicating that the energy consumption increase must be ≤0.5mA for every 1% increase in capture rate.

[0162] Example:

[0163] ΔE=120mAh / day,CR gain =30%→R eco =4>θ eco → Trigger downgrade recommendation.

[0164] When the execution control module responds to the downgrade suggestion data: it prioritizes reducing the sampling frequency data of non-critical meteorological factors, retains the high-frequency sampling channels of critical meteorological factors, and sends downgrade execution confirmation data to the feedback adjustment module; the critical meteorological factors are factors ranked in the top P% of the sensitivity matrix selection weight, where P is a real number between 0 and 100.

[0165] The key meteorological factors are set as the factors with the top 20% weight in the sensitivity matrix, such as the temperature during the flowering period.

[0166] The process of non-critical factor degradation is:

[0167] The sampling frequency is reduced to the minimum frequency, and the power supply of redundant sensors is turned off, such as the CO2 sensor during the non-photosynthetic period.

[0168] The key factor is to maintain the current high-frequency sampling and retain one spare verification channel to ensure synchronous verification of adjacent node data.

[0169] In practical applications, the economic coefficient R eco Make settings to achieve the effect of graded response, as shown in the following table:

[0170] Downgrade <![CDATA[Trigger condition (R eco Value range)]]> Response Action Level 1 (mild) <![CDATA[0.5<R eco ≤1.0]]> The frequency of key factors decreased by 50% Level 2 (moderate) <![CDATA[1.0<R eco ≤2.0]]> Turn off non-critical sensors Level 3 (Radical) <![CDATA[R eco >2.0]]> Keep only single factor high frequency sampling

[0171] Example: The scenario is frost monitoring in a vineyard;

[0172] Initial state:

[0173] Full sensor high-frequency sampling (temperature / humidity / light / wind speed), average daily energy consumption: 680mAh, frost capture rate 92%.

[0174] Problem Discovery: R eco =1.8, the humidity sensor contributes 60% of the energy consumption but the frost correlation is only 10%.

[0175] Downgrade execution:

[0176] Turn off the humidity sensor, change the light sampling from 1 / 2min to 1 / 10min, and maintain high-frequency monitoring of temperature and wind speed.

[0177] result:

[0178] Energy consumption is reduced to 420mAh (saving 38%), and the frost capture rate is only reduced by 2 percentage points (90%). The system confirms the downgrade is successful and updates the economic coefficient threshold to θ eco =0.7.

[0179] The execution control module is used to control the meteorological data acquisition module to switch the sampling frequency and send data verification requests to adjacent nodes when the target sampling frequency data exceeds the preset threshold; and to receive the correction instruction data output by the feedback adjustment module to update the mapping relationship table.

[0180] like Figure 4 As shown in FIG, the process of dynamically adjusting the sampling frequency of this module is shown.

[0181] The execution control module is the main response structure of the system, which is mainly used to dynamically control the execution of sampling frequency, ensure data reliability through multi-node cross-validation, and update monitoring strategy parameters in real time.

[0182] The execution control module is also used to:

[0183] Receive meteorological verification data returned by adjacent nodes, calculate the median deviation value of meteorological data between the current node and the adjacent nodes, and trigger sensor fault alarm data when the deviation value exceeds the preset tolerance.

[0184] The median deviation value process is:

[0185] Collect the data of the current node and ≥2 adjacent nodes during the same period, sort them, take the median as the benchmark value, and then calculate the absolute deviation:

[0186] Median deviation value: Where X median is the median data, X local The data of the current node.

[0187] The preset tolerance is adjusted by presets. Different preset values ​​are set according to different data types. Specific examples are shown in the following table:

[0188] Meteorological factors General tolerance Emergency Mode Tolerance temperature 5％ 10％ humidity 8％ 15％ illumination 15％ 25％

[0189] The process of triggering sensor fault alarm data is:

[0190] Monitor data and hardware status respectively, and classify alarms. The specific classifications are shown in the following table:

[0191] level Trigger Conditions Disposal method Level 3 Single deviation exceeds limit Recording logs Level 2 Deviation exceeds the limit for 3 consecutive times Email Alert Level 1 Abnormal hardware indicators SMS + automatic switching of backup sensors

[0192] Example: The scenario is greenhouse cluster temperature monitoring;

[0193] Initial state:

[0194] 20 nodes form a Mesh network, with a basic sampling frequency of 1 time / 5 minutes;

[0195] Event trigger:

[0196] Node A detects a sudden temperature rise (30°C → 38°C);

[0197] Execution Control Module:

[0198] Immediately increase the sampling rate of node A to 1 time / 30 seconds, request the adjacent nodes B / C to verify the data, and find that the median of the data of node B is 32℃

[0199] Diagnosis:

[0200] Direct sunlight on the temperature sensor at node A causes abnormal readings, triggering a secondary alarm and switching to the backup PT100 sensor.

[0201] Subsequent processing:

[0202] Automatically correct the A node data record and update the mapping table to add the "sun protection factor" compensation item.

[0203] Example 2: This example adds a data compensation module based on Example 1. The rest of the content is consistent with Example 1. The specific content is as follows:

[0204] Also included is a data compensation module for:

[0205] When it is detected that the current sampling frequency data is higher than the lowest sampling frequency, the historical meteorological trend data of the adjacent nodes are obtained, and the linear interpolation algorithm is used to generate the compensated meteorological data of the missing time points.

[0206] This module is mainly used to ensure data integrity and can solve two data loss problems caused by high-frequency sampling in actual use:

[0207] Active Missing: Intermittent data gaps caused by active frequency reduction due to energy consumption management;

[0208] Passive loss: Data loss due to sensor failure or communication interruption.

[0209] The continuity and availability of data curves are ensured through adjacent node data collaboration and intelligent interpolation algorithms.

[0210] The working process of this module is as follows Figure 5 shown.

[0211] In the selection of adjacent nodes, they are classified in the order of spatially closest nodes, environmentally similar nodes, and historically relevant nodes. The spatially closest nodes represent nodes with a GPS distance ≤ 100 m, environmentally similar nodes represent nodes with the same crop variety + soil type, and historically relevant nodes represent nodes with a correlation coefficient greater than 0.9 for the past 7 days.

[0212] For the missing time point t, given the data of the adjacent nodes at time t1 and t2 (t1<t<t2):

[0213] X t is the data corresponding to the missing time point t.

[0214] The coefficient of variation (CV) of the data of the nodes involved in the interpolation is obtained as the credibility assessment of the interpolated values, and CV < 15% is manually set as compensation for high-quality data.

[0215] Example: The scenario is drought monitoring in cotton fields;

[0216] Problem scenario: The master node loses communication for 6 hours due to a sandstorm, and the sampling frequency drops from once every 10 minutes to once every 2 hours.

[0217] Compensation Execution:

[0218] Three adjacent nodes (distance 80-120 m) were selected and linear interpolation was performed to generate soil moisture data;

[0219] After testing, CV = 12% (high quality compensation);

[0220] Effect verification:

[0221] The error compared with the actual data restored subsequently was less than 8%, successfully capturing the key drought turning point.

[0222] Example 3: This example adds an abnormal learning unit and a compensation verification unit on the basis of Example 1. The rest of the content is consistent with Example 1. The specific content is as follows:

[0223] The feedback regulation module includes an anomaly learning unit for:

[0224] Mark the correction instruction data rejected by manual operation, extract the environmental feature data corresponding to the decision moment to build a rejected case library, and automatically skip the correction process when the matching degree between the new feedback data and the case library features exceeds the preset critical value.

[0225] As an auxiliary unit of the feedback regulation module, this functional unit mainly converts human intervention behaviors into machine-understandable rules, thereby avoiding invalid or harmful automated corrections, converting agronomists' on-site judgments into system knowledge, reducing the frequency of human intervention and improving decision-making quality.

[0226] like Figure 6 The figure shows the actual workflow of this module. It is triggered when a correction instruction generated by the system is manually rejected through the console, and a snapshot of the environment status 15 minutes before the rejection is recorded.

[0227] After building the rejected case library, the similarity is calculated when receiving new feedback data, and when the similarity is greater than a preset critical value, the correction is skipped and the rejected case library is updated.

[0228] The preset critical value is dynamically adjusted, and the adjustment formula is:

[0229] where threshold new is the new preset critical value, threshold base is the old preset threshold, and accuracy is the accuracy of the last 10 skip decisions. The base value of the preset threshold is 0.75, and the dynamic range is ±0.05.

[0230] Example: Vegetable greenhouse;

[0231] Problem scenario:

[0232] The system frequently suggested increasing the temperature weight (due to interference with the sensor due to reflections from greenhouse film), and the agronomist rejected the correction three times in a row;

[0233] Learning process:

[0234] Extract common features: light > 80klux and humidity > 85%;

[0235] Establish feature rules: skip temperature correction when the above conditions are met at the same time;

[0236] Implementation effect:

[0237] Ineffective corrections were reduced by 67% and the frequency of manual intervention decreased by 41%.

[0238] The feedback regulation module includes a compensation verification unit for:

[0239] Receive the corrected weight coefficient data output by the feedback adjustment module, load the test data set at the beginning of the next growth stage, calculate the percentage increase in the meteorological event capture rate before and after the correction, and roll back to the previous version of the weight coefficient data when the increase percentage is lower than the preset threshold.

[0240] This module also serves as an auxiliary unit of the feedback adjustment module. Its main function is to ensure that each weight coefficient correction can bring actual benefits through a strict A / B testing mechanism, prevent erroneous corrections that may reduce system efficiency, use data to prove the effectiveness of the adjustment, and establish a traceable weight coefficient iteration history.

[0241] The meteorological stable period is selected as the time to start the verification trigger. The test data consists of historical benchmark data, real-time monitoring data and edge cases. The historical benchmark data is the data of the same period of the previous three years, the real-time monitoring data is the newly collected data in the current growth stage, and the edge cases are special meteorological events manually marked.

[0242] The process of calculating the percentage increase in meteorological event capture rate before and after correction mainly uses two-dimensional evaluation indicators; the two-dimensional evaluation indicators are:

[0243] Key event capture rate:

[0244] False alarm rate change: FR change =FR new -FR old ;

[0245] Among them, CR new is the capture rate of the new version, CR old is the capture rate of the old version, FR new is the false positive rate of the new version, FR old This is the false positive rate for older versions.

[0246] Must also meet: CR improve ≥Threshold (default 15%), FR change ≤ tolerance value (default +5%);

[0247] Example: wheat scab warning;

[0248] Corrections:

[0249] The humidity weight during flowering period was increased from 0.35 to 0.42;

[0250] Verification process:

[0251] Load the test dataset (including 20 historical scab events);

[0252] The new version caught 19 times (95%), and the old version caught 16 times (80%);

[0253] False alarm rate from 12% to 14% (within the tolerance range);

[0254] result:

[0255] CR improve=18.75%>threshold;

[0256] The new version is officially released.

[0257] The technical scope of the present invention is not limited to the contents of the above description. Those skilled in the art can make various deformations and modifications to the above embodiments without departing from the technical idea of ​​the present invention, and these deformations and modifications should all fall within the protection scope of the present invention.

Claims

1. An intelligent agricultural meteorological data collection system based on big data, characterized by: include The phenological phase recognition module is used to obtain crop canopy image data through an image sensor and determine the current growth stage data based on a preset color ratio threshold; Meteorological data acquisition module, used to obtain temperature, humidity and light intensity data at the sampling frequency fed back by the dynamic frequency calculation module; Sensitivity matrix storage module, which stores the mapping relationship table between growth stage data and meteorological factor weight coefficients; A dynamic frequency calculation module is used to query the mapping relationship table according to the current growth stage data to obtain the maximum weight coefficient, and calculate the target sampling frequency data in combination with the real-time fluctuation data of meteorological factors; A feedback adjustment module is used to compare the historical meteorological event capture result data with the expected capture rate data to generate weight coefficient correction instruction data; The execution control module is used to control the meteorological data acquisition module to switch the sampling frequency and send a data verification request to the adjacent node when the target sampling frequency data exceeds the preset threshold; And receive the correction instruction data output by the feedback adjustment module to update the mapping relationship table.

2. The intelligent agricultural meteorological data collection system based on big data according to claim 1, characterized in that: The specific process of comparing the historical meteorological event capture result data with the expected capture rate data is as follows: Count the number of meteorological anomalies actually captured in the high-frequency sampling mode during the most recent M growth cycles, and calculate the percentage deviation from the theoretical capture number. When the percentage exceeds a preset tolerance, adjust the corresponding weight coefficient data in the mapping relationship table proportionally in the direction of the deviation; where M is a preset positive integer, and the theoretical capture number is the number of disasters that should be captured in this growth stage predicted based on historical data; And when the direction of the weight coefficient adjustment data is consistent for K consecutive times in the same growth stage, the version iteration instruction data of the sensitivity matrix is ​​triggered, where K is a preset threshold and K≥3.

3. The intelligent agricultural meteorological data collection system based on big data according to claim 1, characterized in that: The interaction between the feedback adjustment module and the meteorological data acquisition module includes: The sensor energy consumption data during the high-frequency sampling period is obtained. When the ratio of the energy consumption increment per unit time to the capture rate increase exceeds the preset economic coefficient, the sampling frequency downgrade recommendation data is generated for the execution control module to make a decision; When the execution control module responds to the downgrade suggestion data: it prioritizes reducing the sampling frequency data of non-critical meteorological factors, retains the high-frequency sampling channels of critical meteorological factors, and sends downgrade execution confirmation data to the feedback adjustment module; the critical meteorological factors are factors ranked in the top P% of the sensitivity matrix, where P is a real number between 0 and 100.

4. The intelligent agricultural meteorological data collection system based on big data according to claim 1, characterized in that: The dynamic frequency calculation module includes: Fluctuation analysis unit, used to calculate the coefficient of variation data of the target meteorological factor in the last 24 hours; an amplification calculation unit, for multiplying the maximum weight coefficient by the square root of the coefficient of variation to generate dynamic amplification factor data; A frequency synthesis unit, configured to multiply the basic sampling frequency data by the dynamic amplification factor data, and output the target sampling frequency data to the execution control module; The ramp-down control module is used to gradually reduce the target sampling frequency data according to the preset attenuation coefficient until it is restored to the lowest sampling frequency when no meteorological abnormal data is detected within N consecutive sampling periods, where N is a preset positive integer.

5. The intelligent agricultural meteorological data collection system based on big data according to claim 4 is characterized in that: The fluctuation analysis unit is further used for: Detect the sudden change gradient of the coefficient of variation data. When the gradient value exceeds the preset warning line, send emergency calibration request data to the feedback adjustment module. The feedback adjustment module responds to the request and immediately starts the weight coefficient correction process.

6. The intelligent agricultural meteorological data collection system based on big data according to claim 1, characterized in that: The process of determining the current growth stage data according to the preset color ratio threshold is as follows: Extract R channel pixel ratio data of crop canopy image data. When the ratio data exceeds a preset threshold corresponding to the current growth stage, generate phenological period switching instruction data and update the current growth stage data.

7. The agricultural meteorological data intelligent collection system based on big data according to claim 1, characterized in that: The execution control module is also used for: Receive meteorological verification data returned by adjacent nodes, calculate the median deviation value of meteorological data between the current node and the adjacent nodes, and trigger sensor fault alarm data when the deviation value exceeds the preset tolerance.

8. The agricultural meteorological data intelligent collection system based on big data according to claim 1, characterized in that: Also included is a data compensation module for: When it is detected that the current sampling frequency data is higher than the lowest sampling frequency, the historical meteorological trend data of the adjacent nodes are obtained, and the linear interpolation algorithm is used to generate the compensated meteorological data of the missing time points.

9. The intelligent agricultural meteorological data collection system based on big data according to claim 1, characterized in that: The feedback adjustment module includes an abnormality learning unit, which is used to: Mark the correction instruction data rejected by manual operation, extract the environmental feature data corresponding to the decision moment to build a rejected case library, and automatically skip the correction process when the matching degree between the new feedback data and the case library features exceeds the preset critical value.

10. The agricultural meteorological data intelligent collection system based on big data according to claim 1, characterized in that: The feedback adjustment module includes a compensation verification unit, which is used to: Receive the corrected weight coefficient data output by the feedback adjustment module, load the test data set at the beginning of the next growth stage, calculate the percentage increase in the meteorological event capture rate before and after the correction, and roll back to the previous version of the weight coefficient data when the increase percentage is lower than the preset threshold.

Images