A greenhouse tomato weekly yield prediction method based on temperature, humidity and radiation accumulation

Abstract

The application relates to the technical field of agricultural management, and particularly discloses a greenhouse tomato weekly yield prediction method based on temperature, humidity and radiation accumulation. The method comprises the following steps: acquiring inflorescence batch flowering time, productive open flower quantity, synchronous fruit load, temperature, water vapor pressure deficit and time sequence of crown layer absorption photosynthetically active radiation; determining target inflorescence batches and target weekly maturity attribution probability; mapping the time sequence into stage normalized development process to form a temperature, humidity and radiation accumulation spectrum; obtaining reproductive realization parameters according to the accumulation spectrum of the flower organ formation stage and the flowering and fruit setting stage; combining response equivalent production pollination coverage parameters to determine effective fruit number; determining average fruit weight according to the accumulation spectrum of the post-fruit setting stage and the synchronous fruit load; and finally obtaining target harvesting week unit area yield prediction value.

Description

Technical Field

[0001] This invention relates to the field of agricultural management technology, and more specifically, to a method for predicting the weekly yield of greenhouse tomatoes based on temperature, humidity, and accumulated radiation. Background Technology

[0002] Greenhouse tomatoes are characterized by continuous flowering, continuous fruit setting, and staggered harvesting. Their weekly yield is influenced by a variety of factors, including temperature, air humidity, radiation conditions, pollination status, fruit load, and ripening process. Even with the same number of productive open flowers, different inflorescence batches may exhibit significant differences in the number of effective fruits, average fruit weight, and marketable fruit maturity time due to variations in developmental stage, environmental experience, and pollination conditions. Therefore, estimations based solely on historical yields, current flower count, or fixed empirical parameters are insufficient to accurately reflect the marketable fruit yield for a specific harvest week.

[0003] Existing literature (Effects of regulating relative humidity on growth, water status, and yield of greenhouse melons and tomatoes, 2021) studied the effects of relative humidity at different growth stages on photosynthetic characteristics, growth period, and yield of tomatoes, indicating that humidity conditions have a significant impact on the flowering and fruit setting period and fruit enlargement period of tomatoes. However, this literature mainly evaluated physiological indicators and final yield under different humidity treatments, and did not establish a weekly yield prediction method corresponding to specific inflorescence batches for continuously changing temperature, humidity, and radiation conditions.

[0004] The publication CN111988985A, titled "Controlled Agricultural Systems and Methods for Agriculture," proposes detecting plant flowers or buds, using typical flower-fruit conversion rates to predict fruit quantity, and combining this with average fruit weight and typical maturity time to predict yield and harvest time. While this method can update conversion rates using temperature, humidity, light intensity, and pollination conditions, it still primarily relies on flower quantity, typical conversion rates, average fruit weight, and maturity time, making it difficult to fully reflect the actual reproductive status of different inflorescence batches, differences in production pollination, concurrent fruit load, and maturity distribution.

[0005] Furthermore, the actual vibration experienced by the flower organs during the production pollination process is not entirely consistent with the input state of the pollination device. Existing methods are difficult to compare and evaluate diagnostic testing conditions and production pollination conditions. When the detection signal is weak, improper parameter processing may also affect the judgment of fruit setting.

[0006] Therefore, there is a need for a method that can reliably predict the marketable yield of greenhouse tomatoes at the target harvest week by comprehensively considering the continuous environmental experience, pollination status, fruit load, and ripening process of inflorescence batches. Summary of the Invention

[0007] To overcome the aforementioned deficiencies in the prior art, this invention provides a method for predicting weekly yield of greenhouse tomatoes based on temperature, humidity, and irradiance accumulation. By mapping environmental time series to a stage-normalized developmental process, an irradiance accumulation spectrum under temperature and humidity conditions is formed within the sub-intervals of equal developmental processes. Reproductive realization parameters are obtained using the actual vibration response of the same diagnostic flower at two levels. Combined with production pollination coverage parameters, concurrent fruit load, and ripening probability, the problem mentioned in the background art is solved.

[0008] To achieve the above objectives, the present invention provides the following technical solution: A method for predicting weekly yield of greenhouse tomatoes based on temperature, humidity, and accumulated radiation includes the following steps: Step 1: Obtain the flowering time of the inflorescence batch, the number of productive open flowers, the fruit load and temperature at the same time, the water vapor pressure deficit, and the time series of photosynthetically active radiation absorbed by the canopy, and determine the target inflorescence batch and the probability of its maturity at the target week. Step 2: Based on the temperature development rate of the variety and development stage, the time series is mapped to the stage-normalized development process. Within the sub-intervals of the same development process, the cumulative radiation and residence time are calculated according to the temperature and water vapor pressure deficit joint interval to form the temperature and humidity state irradiance accumulation spectrum. Step 3: Based on the accumulation spectrum of the flower formation stage and the flowering and fruit setting stage, obtain the reproductive realization parameters from the reproductive environment response library. The response library stores the accumulation spectrum and reproductive realization parameters of the calibrated inflorescence batch. The parameters are determined by the two levels of fertile pollen release obtained by the first actual vibration response and the second actual vibration response of the same diagnostic flower at the same diagnostic frequency. The anther cone response amplitude of the second actual vibration response is higher than that of the first actual vibration response. Step 4: Determine the equivalent production pollination coverage parameter based on the proportion of productive open flowers that reach the first actual vibration response under actual pollination of the target inflorescence batch, and determine the effective number based on the reproductive realization parameter, coverage parameter and the number of productive open flowers. Step 5: Determine the average fruit weight based on the accumulation spectrum and fruit load during the same period after fruit set. Summarize the effective number, average fruit weight, and target week maturity attribution probability of each target inflorescence batch to obtain the predicted yield per unit area for the target harvest week.

[0009] As a further aspect of the present invention, a continuous maturation process is determined based on the flowering time, temperature time series, and temperature development rate corresponding to the variety and development stage of the inflorescence batch. The target week maturity assignment probability is determined based on the continuous maturation process and the pre-calibrated cumulative distribution function of variety commercial maturity. The inflorescence batch with a target week maturity assignment probability not lower than a preset contribution threshold is determined as the target inflorescence batch.

[0010] As a further aspect of the present invention, each developmental stage is divided into multiple iso-developmental process sub-intervals; when the developmental process interval corresponding to the sampling period spans at least two iso-developmental process sub-intervals, the canopy absorbed photosynthetically active radiation and dwell time of the sampling period are respectively allocated to the corresponding iso-developmental process sub-intervals according to the overlap ratio between the developmental process interval and each iso-developmental process sub-interval; when the developmental process increment is zero, the canopy absorbed photosynthetically active radiation and dwell time of the sampling period are allocated to the iso-developmental process sub-interval where the current developmental process is located.

[0011] As a further aspect of the present invention, the reproductive environment response library is established separately for each tomato variety. The establishment process includes: applying calibration excitation of multiple candidate frequencies to a preset excitation position of a representative diagnostic flower, detecting the actual vibration response of the preset excitation position and the anther cone, and determining the mechanical transmission relationship between the preset excitation position and the anther cone based on the calibration result that the pollen release amount is not higher than the background threshold; adjusting the excitation input to the same representative diagnostic flower according to the mechanical transmission relationship, so that the anther cone reaches the first actual vibration response and the second actual vibration response sequentially at the same diagnostic frequency, and detecting the first fertile pollen release amount and the second fertile pollen release amount in two mutually isolated collection periods.

[0012] As a further aspect of the present invention, the reproductive realization parameters include a fertile release capacity component and a first response release component; the fertile release capacity component is determined based on the sum of the first fertile pollen release amount and the second fertile pollen release amount relative to the reference fertile pollen release amount, which is obtained by pre-calibration of reference diagnostic flowers under the same tomato variety, flowering age range, detection system, and two-level response conditions; the first response release component is determined based on the proportion of the first fertile pollen release amount in the sum of the first fertile pollen release amount and the second fertile pollen release amount; when the sum of the first fertile pollen release amount and the second fertile pollen release amount is lower than a preset effective detection limit, the first response release component is not generated.

[0013] As a further aspect of the present invention, candidate temperature ranges and candidate vapor pressure deficit ranges are first set, and a preset change threshold is determined based on the repeated measurement error. For adjacent candidate temperature ranges or adjacent candidate vapor pressure deficit ranges, when the change amplitude of the fertile release capacity component of the calibrated inflorescence batch under the corresponding environmental treatment, or when the change amplitude of the first response release component is generated in both adjacent environmental treatments, exceeds the preset change threshold and the change direction is consistent in independent repeated experiments, the boundary of the corresponding candidate range is retained; otherwise, the corresponding adjacent candidate ranges are merged.

[0014] As a further aspect of the present invention, when the anther cone vibration frequency of a productive open flower under actual pollination is within the allowable frequency range of the first actual vibration response, and the anther cone response amplitude reaches the target amplitude of the first actual vibration response, and the effective duration reaches the first duration or the equivalent cumulative response amount reaches the calibrated response amount threshold within the effective period of reaching the target amplitude, it is determined that the productive open flower has reached the first actual vibration response; the equivalent production pollination coverage parameter is determined based on the proportion of the number of productive open flowers that have reached the first actual vibration response to the sampled number extracted from the target inflorescence batch.

[0015] As a further aspect of the present invention, the method for predicting the weekly yield of greenhouse tomatoes based on temperature, humidity and irradiance accumulation can be used on various Internet platforms for Internet security management of greenhouse tomatoes, or provide an Internet platform or industrial APP for monitoring the yield of greenhouse tomatoes.

[0016] The technical effects and advantages of the greenhouse tomato weekly yield prediction method based on temperature, humidity and irradiance accumulation of this invention are as follows: This invention uses inflorescence batches as continuous prediction objects, mapping environmental time series to stage-normalized developmental processes, and accumulating radiation and residence time within the combined temperature and vapor pressure deficit interval to reduce information loss caused by natural time misalignment and environmental mean-of-factometry. It determines the mechanical transmission relationship from the preset excitation position to the anther cone through calibration excitation with pollen release not exceeding the background threshold, ensuring that both the two-level diagnostic response and production pollination coverage use the actual vibration response of the anther cone as a unified evaluation benchmark. It obtains reproductive realization parameters through the two-level actual vibration responses of the same diagnostic flower, and determines the effective number by combining the response-equivalent production pollination coverage parameters. When the first response release component is not generated, it uses the effective rate relationship of varieties that do not include this component to avoid mistakenly treating insufficient detection information as zero. It determines the average fruit weight based on the temperature and humidity irradiance accumulation spectrum and the concurrent fruit load during the post-fruit set stage, and then determines the target weekly yield contribution of each target inflorescence batch by combining the target week maturity attribution probability. Therefore, different tomato varieties can be predicted by establishing corresponding calibration libraries. The environmental accumulation, reproductive realization, pollination coverage, effective number, average fruit weight, and target week contribution of each inflorescence batch form a traceable calculation link. Attached Figure Description

[0017] Figure 1 This is a flowchart of a method for predicting weekly greenhouse tomato yield based on temperature, humidity, and irradiance accumulation, according to the present invention.

[0018] Figure 2 This is a schematic diagram of the preset excitation position, the actual vibration response detection position of the anther cone, and the pollen particle detection area for a representative diagnostic flower. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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.

[0020] Example 1 This embodiment was implemented in a 1200m² multi-span glass greenhouse, divided into six environmental monitoring zones, planted with medium-fruit tomato variety A at a planting density of 2.5 plants / m². The target harvest week was June 23-29, 2025, and the prediction results were generated and solidified before 00:00 on June 23, 2025. Before the prediction date, the measured time series was used; from the prediction date to the end of the target harvest week, a predicted time series determined and solidified based on greenhouse environmental control settings, external weather forecasts, greenhouse light transmittance parameters, shading plans, and supplemental lighting plans was used, and was not replaced by subsequent measured values.

[0021] like Figure 1 As shown, this invention determines the target inflorescence batch and the probability of maturity at the target week by acquiring the time series of inflorescence batch flowering time, productive open flower number, concurrent fruit load and temperature, water vapor pressure deficit, and canopy absorption of photosynthetically effective radiation; maps the time series to a stage-normalized developmental process and forms a temperature and humidity-controlled irradiance accumulation spectrum; obtains reproductive realization parameters from the reproductive environment response library to determine the response-equivalent production pollination coverage parameters and effective number; determines the average fruit weight based on the accumulation spectrum of the post-fruit set stage and concurrent fruit load, and summarizes to obtain the predicted yield per unit area for the target harvest week.

[0022] Each environmental monitoring zone collects air temperature, relative humidity, and canopy absorbed photosynthetically active radiation every 5 minutes. The vapor pressure deficit is determined based on the air temperature and relative humidity at the same sampling time. The flowering time of each inflorescence batch, the number of productively open flowers, and the concurrent fruit load are recorded. Flowers that enter a productively open state for two consecutive days are classified as one inflorescence batch. Deformed flowers, damaged flowers, flowers that have ceased production, and representative diagnostic flowers are not included in the number of productively open flowers.

[0023] The developmental process of variety A is divided into four stages: flower formation, flowering and fruit setting, fruit enlargement, and commercial maturity. The post-fruit setting stage begins at the end of the flowering and fruit setting stage and ends at the end of the commercial maturity stage, including the fruit enlargement and commercial maturity stages. The baseline, optimum, upper limit temperatures, and developmental duration for each stage are as follows: 10℃, 24℃, 35℃ and 10.8 equivalent days of optimum temperature; 10℃, 25℃, 36℃ and 5.5 equivalent days of optimum temperature; 8℃, 24℃, 34℃ and 28.0 equivalent days of optimum temperature; 9℃, 24℃, 35℃ and 12.0 equivalent days of optimum temperature.

[0024] The temperature development rate at the optimum temperature is 1 optimum temperature equivalent day / day; when the temperature is not higher than the baseline temperature or not lower than the upper limit temperature, the temperature development rate is zero; between the baseline temperature and the optimum temperature, the temperature development rate increases linearly from zero to 1; between the optimum temperature and the upper limit temperature, the temperature development rate decreases linearly from 1 to zero. The starting point of the flower formation stage is determined by a pre-calibrated flower formation initiation development amount, which in this embodiment is approximately 12 days before effective flowering.

[0025] For each inflorescence batch, the temperature development rate and sampling duration for each sampling period are continuously accumulated and divided by the corresponding stage development amount, so that the stage-normalized development process for each stage increases from 0 to 1. The stage-normalized development process is used to divide the iso-developmental process sub-intervals and form the temperature and humidity irradiation accumulation spectrum.

[0026] The continuous maturation process is a continuous coordinate system distinct from the stage-normalized development process. Starting from the flowering time of each inflorescence batch, the development rate is continuously accumulated according to the temperature at each developmental stage, with the cumulative development value corresponding to the cumulative maturity probability of commercial fruit of variety A reaching 0.500 defined as 1.00. The continuous maturation process is not truncated at 1.00, thus allowing values ​​less than or greater than 1.00.

[0027] Each developmental stage is divided into six iso-developmental process sub-intervals. When a sampling period crosses the boundary of an iso-developmental process sub-interval, the amount of photosynthetically active radiation absorbed by the canopy and the duration of residence are allocated according to the overlap ratio between the developmental process interval of that sampling period and each iso-developmental process sub-interval. When the developmental process increment is zero, the amount of photosynthetically active radiation absorbed by the canopy and the duration of residence of that sampling period are allocated to the iso-developmental process sub-interval where the current developmental process is located.

[0028] For example, the stage-normalized developmental progression in a 5-minute sampling period increased from 0.162 to 0.174, crossing the boundary at the one-sixth mark, with proportions of 0.389 and 0.611 for the two preceding and following isochronous developmental progression sub-intervals, respectively. The average canopy absorbed photosynthetically active radiation during this sampling period was 620 μmol·m⁻¹. -²·s - ¹, corresponding to a radiation dose of 0.186 mol·m - ², respectively distributed 0.072 mol·m - ² and 0.114 mol·m - ²; the stay durations were 1.945 min and 3.055 min, respectively.

[0029] In this embodiment, candidate temperature ranges and candidate vapor pressure deficit ranges are set at intervals of 2℃ and 0.3kPa. When comparing candidate temperature ranges, sub-ranges with the same development stage, equal developmental process, vapor pressure deficit range, and radiation accumulation range are matched; when comparing candidate vapor pressure deficit ranges, sub-ranges with the same development stage, equal developmental process, temperature range, and radiation accumulation range are matched.

[0030] Each adjacent environment processing is configured with three independent replicates. The absolute value of the difference between the replicate means is calculated in all matching comparison layers, and Table 1 records the maximum variation of each candidate boundary in all matching comparison layers. In this embodiment, the combined replicate standard deviation of the two reproductive realization components is 0.026, and 0.08, which is three times the combined replicate standard deviation rounded to the nearest whole number, is used as the preset variation threshold.

[0031] When adjacent environmental processes both generate the first response release component, the change magnitudes of the fertile release capacity component and the first response release component are compared respectively; if the maximum change magnitude of any component exceeds the preset change threshold and the change directions of the three independent repetitions are consistent, the corresponding candidate interval boundary is retained; otherwise, the corresponding adjacent candidate intervals are merged.

[0032] When neither of the adjacent environmental processes generates a first response release component, the first response release component is not compared; only the fertile release capacity component is compared. If the maximum change of the fertile release capacity component exceeds a preset change threshold and the change direction of the three independent repetitions is consistent, the boundary of the corresponding candidate interval is retained; otherwise, the corresponding adjacent candidate intervals are merged.

[0033] When only one environmental process in an adjacent environmental process generates the first response release component, the first response release component is not compared, only the fertile release capacity component is compared; when the maximum change amplitude of the fertile release capacity component exceeds the preset change threshold and the change direction of the three independent repetitions is consistent, the corresponding candidate interval boundary is retained; otherwise, the corresponding adjacent candidate intervals are merged.

[0034] Table 1. Results of all candidate boundary processing

[0035] In Table 1, the three independent replicates corresponding to each retained boundary all maintain the same direction of change, ultimately forming four temperature ranges: 14℃ to 18℃, 18℃ to 22℃, 22℃ to 30℃, and 30℃ to 34℃, as well as four water vapor pressure deficit ranges: 0 kPa to 0.6 kPa, 0.6 kPa to 1.2 kPa, 1.2 kPa to 1.8 kPa, and 1.8 kPa to 2.4 kPa.

[0036] Each interval adopts a left-closed, right-open configuration, with the highest temperature interval and the highest vapor pressure deficit interval including their right endpoints. All sampled values ​​in this embodiment fall within the aforementioned interval ranges. Subsequently, within each developmental stage and iso-developmental process sub-intervals, the canopy absorbed photosynthetically active radiation is accumulated according to the combined temperature and vapor pressure deficit intervals, and the residence time is statistically analyzed to form a temperature and humidity-controlled irradiance accumulation spectrum.

[0037] The flowering and fruit setting stage of inflorescence batch B18 lasted 132 hours, with each developmental stage sub-interval lasting 22 hours. The first sub-interval accumulated 14 hours and 2.20 mol·m⁻² under the conditions of "18℃ to 22℃, 0.6 kPa to 1.2 kPa" and "22℃ to 30℃, 0.6 kPa to 1.2 kPa", respectively. - ² and 8h, 0.90mol·m - ²; the corresponding values ​​for the second subinterval are 10h and 1.65 mol·m. - ² and 12h, 2.15mol·m - ².

[0038] The third sub-interval accumulated 13 hours and 2.35 mol·m⁻¹ under the conditions of "22℃ to 30℃, 0.6 kPa to 1.2 kPa" and "22℃ to 30℃, 1.2 kPa to 1.8 kPa", respectively. - ² and 9h, 1.40 mol·m - ²; the corresponding values ​​for the fourth subinterval are 9h and 1.60 mol·m. - ² and 13h, 2.30mol·m - ²; the corresponding values ​​for the fifth subinterval are 12h and 1.95 mol·m. - ² and 10h, 1.70 mol·m - ²; The corresponding values ​​for the sixth subinterval are 8h and 1.20 mol·m. - ² and 14h, 2.60 mol·m - ². The total radiation during this stage was 22.00 mol·m². - ². Store separately when the same temperature and humidity conditions occur at different developmental locations.

[0039] Reproductive environment response databases were established separately for each tomato variety. The reproductive environment response database for variety A was established from 96 calibrated inflorescence batches, with 18 representative diagnostic flowers selected from each calibrated inflorescence batch.

[0040] like Figure 2 As shown, the preset excitation position is located 8 mm from the calyx on the pedicel, and the middle of the anther cone is the actual vibration response detection position. Calibration excitations of 15 Hz, 30 Hz, 50 Hz, and 100 Hz were sequentially applied to the same representative diagnostic flower, each lasting 0.20 s, with a 3 s interval between adjacent calibration excitations. At each candidate frequency, the actual acceleration amplitude from zero to peak value at the preset excitation position was controlled to be 10.0 m·s². - ².

[0041] Using a 0.20s background detection window, the mean number of background particles was 2.1, and the standard deviation was 0.9. The background threshold of 5 particles was obtained by adding three times the standard deviation to the mean number of background particles and rounding up. If the number of particles detected during the calibration excitation period did not exceed 5, the pollen release was considered to be below the background threshold. This calibration result was used to determine the mechanical transmission relationship from the preset excitation position to the anther cone.

[0042] At 15Hz, 30Hz, 50Hz and 100Hz, the ratios of the actual acceleration amplitude of the anther cone from zero to peak value to the actual acceleration amplitude of the preset excitation position from zero to peak value are 0.74, 1.08, 1.32 and 0.61, respectively, and the signal coherence coefficients are 0.91, 0.95, 0.97 and 0.89, respectively. In this embodiment, 50Hz is selected as the diagnostic frequency.

[0043] Adjusting the excitation input based on the mechanical transmission relationship, the same representative diagnostic flower was made to sequentially achieve the first and second actual vibration responses at a diagnostic frequency of 50 Hz. The zero-to-peak response amplitude of the anther cone in the first actual vibration response was 18 m / s². - ², duration 0.35s; the amplitude of the anther cone zero-to-peak response of the second actual vibration response is 32 m·s². - ², with a duration of 0.35s.

[0044] After the first actual vibration response ends, the first collection channel is closed and the detection area is cleared. After the particle count returns to the background range, the second collection channel is opened to isolate the first collection period from the second collection period. Pollen that continues to be released is collected during the second collection period.

[0045] In the first collection period, 71,515 pollen grains were detected in 18 representative diagnostic flowers from a calibrated inflorescence batch, with a cumulative background particle count of 90 grains, a collection and detection efficiency of 0.82, and a fertile pollen ratio of 0.87, yielding a first fertile pollen release of 4,210 grains / flower. In the second collection period, 22,424 pollen grains were detected, with a cumulative background particle count of 78 grains, a collection and detection efficiency of 0.79, and a fertile pollen ratio of 0.84, yielding a second fertile pollen release of 1,320 grains / flower.

[0046] The reference fertile pollen release was pre-calibrated using 30 reference diagnostic flowers, divided into three independent replicates, each including 10 reference diagnostic flowers. Under the same tomato variety, flower age range, detection system, and two-level response conditions, the combined average of the two levels of fertile pollen release was 6000 seeds / flower, with a standard deviation of 310 seeds / flower. Therefore, the reference fertile pollen release was determined to be 6000 seeds / flower, and the preset effective detection limit was determined to be 600 seeds / flower.

[0047] The sum of the first and second fertile pollen release amounts of the above-mentioned calibrated inflorescence batches is 5530 pollen grains / flower, and the fertile release capacity component is 5530 / 6000=0.922; the sum of the two levels of fertile pollen release amounts is not lower than the preset effective detection limit, and the first response release component is 4210 / 5530=0.761.

[0048] The accumulated irradiance spectra under temperature and humidity conditions for each calibrated inflorescence batch during the flower formation and flowering / fruit setting stages were correlated and stored with the corresponding reproductive realization parameters. The reproductive realization parameters include the fertile release capacity component and the first response release component.

[0049] The radiation intensity and residence time in each temperature and vapor pressure deficit region of the temperature and humidity irradiance accumulation spectrum are expanded according to the unified spectral unit order. Each dimension is standardized using the mean and standard deviation of all valid entries in the reproductive environment response database of the same species; dimensions with a standard deviation of zero are not included in the spectral distance calculation; spectral units where no corresponding temperature and humidity conditions have occurred are represented by both zero radiation intensity and residence time.

[0050] Only valid entries with the same reproductive realization parameter generation status and complete data are compared. If there are entries with zero spectral distance, only all zero-distance entries are used and weighted equally; if there are no zero-distance entries, the five valid entries with the smallest spectral distance are selected, and the weights are determined by normalizing the inverse of the spectral distance.

[0051] When predicting B18, the spectral distances of the five nearest valid entries were 0.18, 0.22, 0.25, 0.29, and 0.33, with corresponding weights of 0.270, 0.221, 0.194, 0.168, and 0.147; their fertility release capacity components were 0.82, 0.86, 0.83, 0.87, and 0.80, and their first response release components were 0.69, 0.73, 0.70, 0.74, and 0.68, respectively. The weighted fertility release capacity component of B18 was 0.836, and the first response release component was 0.708, subsequently adjusted to 0.84 and 0.71, respectively.

[0052] This embodiment uses 72 historical inflorescence batches from four complete production seasons to pre-calibrate the cumulative distribution function of commercial maturity for variety A. When the continuous maturity progression is 0.78, 0.86, 0.98, 1.00, 1.09, and 1.14, the corresponding cumulative maturity probabilities are 0.100, 0.198, 0.450, 0.500, 0.711, and 0.802, respectively. Linear interpolation is used between adjacent calibration points.

[0053] The target week maturity attribution probability is the cumulative maturity probability at the end of the target harvest week minus the cumulative maturity probability at the beginning of the target harvest week. Inflorescence batches with a target week maturity attribution probability of not less than 0.05 are identified as target inflorescence batches; therefore, B17, B18, and B19 are identified as target inflorescence batches.

[0054] Mechanical vibration pollination is used in production, and 60 productive open flowers are randomly selected from each target inflorescence batch according to the proportion of productive open flowers in the six environmental monitoring zones.

[0055] When the anther cone vibration frequency of a productive open flower under actual pollination is between 47 Hz and 53 Hz, and the zero-to-peak response amplitude of the anther cone reaches 18 m·s⁻¹ - ², and the effective duration reaches 0.35s, or the response amplitude reaches 18m·s. - The calibrated equivalent cumulative response reached 6.3 m·s² within the effective time period. - ¹When the productive open flower reaches the first actual vibration response (equivalent cumulative response is when the anther cone vibration frequency is between 47Hz and 53Hz and the response amplitude reaches 18m·s), it is determined that the flower has reached the first actual vibration response (equivalent cumulative response is when the anther cone vibration frequency is between 47Hz and 53Hz and the response amplitude reaches 18m·s). - The cumulative value of the product of the actual response amplitude and the corresponding duration within each effective time period of ²; the calibrated response threshold is 6.3 m·s. - ¹).

[0056] In B17, B18, and B19, 51, 49, and 45 flowers, respectively, reached the first actual vibration response. Based on the proportion of productive open flowers that reached the first actual vibration response to the sample size, the equivalent production pollination coverage parameters for the three target inflorescence batches were determined to be 0.850, 0.817, and 0.750, respectively.

[0057] A relationship between varietal effectiveness rates was established using 64 historical calibrated inflorescence batches. Each entry recorded the fertile release capacity component, the first response release component, the response-equivalent production pollination coverage parameter, and the effectiveness rate calibration value. The effectiveness rate calibration value is the proportion of the number of fruits that ultimately meet the preset commercial standards in the inflorescence batch to the corresponding number of productive open flowers.

[0058] During the prediction, the fertile release capacity component, the first response release component, and the response equivalent production pollination coverage parameter were standardized according to the mean and standard deviation of 64 historical calibrated inflorescence batches. The three-dimensional Euclidean distance was calculated, and the effective rate was obtained according to the aforementioned zero distance treatment rule and distance reciprocal weighting rule.

[0059] The distances to the five nearest items for B18 are 0.22, 0.27, 0.31, 0.36, and 0.41, respectively, with corresponding weights of 0.272, 0.222, 0.193, 0.166, and 0.146. The effective rate values ​​are 0.72, 0.70, 0.74, 0.69, and 0.71, respectively. The weighted effective rate for B18 is 0.713. Using the same method, the effective rates for B17 and B19 are 0.749 and 0.675, respectively.

[0060] The impact of the equivalent production pollination coverage parameter on fruit set is already included in the effective rate obtained through the variety effective rate relationship. Therefore, the effective number is determined by multiplying the productive open flower number and the effective rate, and the equivalent production pollination coverage parameter is no longer multiplied by the effective rate repeatedly.

[0061] A calibration library for average fruit weight was established according to tomato varieties. Each entry includes the cumulative irradiance spectrum under temperature and humidity conditions after fruit set, the fruit load at the same time, and the measured average fruit weight of the corresponding inflorescence batch. The fruit load at the same time was converted into an equivalent actual load according to the relative assimilation requirement coefficient corresponding to the number of fruit batches and developmental age.

[0062] During prediction, the actual accumulated spectrum formed before the prediction time point in the post-fruit set stage, the predicted accumulated spectrum formed after the prediction time point based on the solidified prediction time series, and the fruit load of the same period are combined into a unified vector. The vector is standardized according to the mean and standard deviation of all valid entries in the average fruit weight calibration library. The five valid entries with the smallest distance are selected, and the average fruit weight is determined according to the aforementioned zero-distance processing rule and distance reciprocal weighting rule.

[0063] The distances to the five nearest entries for B18 are 0.20, 0.24, 0.29, 0.34, and 0.40, respectively, with corresponding weights of 0.277, 0.231, 0.191, 0.163, and 0.138. The measured average fruit weights are 0.203, 0.199, 0.205, 0.200, and 0.204 kg / fruit, respectively. The weighted average fruit weight for B18 is 0.2021 kg / fruit. Using the same method, the average fruit weights for B17 and B19 are 0.2057 kg / fruit and 0.1987 kg / fruit, respectively.

[0064] Table 2 Calculation results of target inflorescence batches

[0065] In Table 2, C, R, and U represent the fertile release capacity component, the first response release component, and the response equivalent production pollination coverage parameter, respectively. All values ​​shown in the table have been rounded; the effective number, weekly yield contribution, and summary results are calculated using the unrounded raw intermediate values.

[0066] The effective number, average fruit weight, and target week maturity probability of the three target inflorescence batches were summarized to obtain a predicted yield per unit area of ​​1.7305 kg / m² for the target harvest week; based on a planting area of ​​1200 m², the predicted total weight of marketable fruit for the target harvest week was 2076.6 kg.

[0067] To verify the predictive accuracy of the method of this invention, independent verification was conducted for eight consecutive harvest weeks from June 23 to August 17, 2025, without altering the reproductive environment response library, the variety effectiveness rate relationship, the average fruit weight calibration library, the cumulative distribution function of variety commercial maturity, the temperature and humidity state interval boundaries, the preset change threshold, the number of nearest neighbor entries, and the distance weighting rules. The verification harvest weeks and their corresponding inflorescence batches were not included in the above calibration.

[0068] The forecast results for each harvest week are generated and fixed before the start of the corresponding target harvest week, and are not corrected using measured environmental data after the start of the target harvest week. At harvest, fruits that meet the preset commercial standards are weighed daily, and the weight of commercial fruits in the six environmental monitoring zones is recorded using an electronic scale with a graduation of 0.1 kg. The total weight of commercial fruits for each week is divided by the planting area of ​​1200 m² to obtain the measured weekly yield per unit area.

[0069] The predicted values ​​for the eight harvest weeks were 1.7305, 1.812, 1.924, 2.038, 2.116, 1.985, 1.807 and 1.654 kg / m², respectively, while the corresponding measured values ​​were 1.708, 1.845, 1.886, 2.101, 2.061, 1.944, 1.862 and 1.619 kg / m², respectively.

[0070] Taking the target harvest week from June 23 to 29, 2025 as an example, the predicted total weight of marketable fruit was 2076.6 kg, while the actual total weight of marketable fruit was 2049.6 kg, a difference of 27.0 kg, with a relative error of 1.32%.

[0071] Statistics from the eight harvest weeks above show that the average absolute error between the predicted and measured values ​​was 0.0428 kg / m², the root mean square error was 0.0447 kg / m², the average absolute relative error was 2.25%, the maximum absolute relative error was 3.00%, and the average prediction deviation was 0.0051 kg / m².

[0072] The coefficient of determination was determined by subtracting the ratio of the sum of squares of the predicted residuals to the sum of squares of the deviations of the measured values ​​from their mean by 1, and the resulting coefficient of determination was 0.914.

[0073] Example 2 This embodiment follows the steps in Embodiment 1 for determining the target inflorescence batch, forming the temperature and humidity irradiation accumulation spectrum, determining the average fruit weight, and summarizing the weekly yield, to explain how to determine the effective rate and the number of effective responses when the first response release component obtained from the reproductive environment response library has not been generated.

[0074] The first and second fertile pollen release amounts for inflorescence batch B26 are 180 and 120 pollen grains / flower, respectively, with a total of 300 pollen grains / flower, which is below the preset effective detection limit of 600 pollen grains / flower. The ratio of the sum of the two fertile pollen release amounts to the reference fertile pollen release amount is 300 / 6000 = 0.050. Therefore, the fertile release capacity component is 0.050, and no first response release component is generated.

[0075] Sixty productive open flowers were randomly selected from B26 in strata, of which 47 flowers reached the first actual vibration response. The equivalent production pollination coverage parameter of the response was 47 / 60=0.783.

[0076] Variety A was further evaluated using 21 historically calibrated inflorescence batches where the first response release component was not generated, establishing a relationship between the fertility release capacity component and the pollination coverage parameter of the response equivalent production. During prediction, the fertile release capacity component and the response equivalent production pollination coverage parameter were standardized, and a two-dimensional Euclidean distance was calculated. The five most effective entries with the smallest distances were selected, and their weights were determined by normalizing to the inverse of the distance.

[0077] The effective rate calibrated values ​​of the five most recently effective entries are 0.25, 0.28, 0.27, 0.24 and 0.29, respectively, with corresponding weights of 0.270, 0.222, 0.194, 0.168 and 0.146, respectively. The weighted effective rate is 0.264.

[0078] The productive number of open flowers for B26 is 9.6 flowers / m², resulting in an effective number of 9.6 × 0.264 = 2.53 fruits / m². Throughout this process, the first response release component is kept ungenerated, and it is not used as a numerical zero input for the variety effectiveness rate relationship that includes the first response release component.

[0079] In the candidate vapor pressure deficit interval treatment corresponding to B26, when no first response release component is generated in adjacent environmental treatments, only the fertile release capacity component is compared.

[0080] The fertile release capacity components of the three independent replicates to the left of a candidate boundary are 0.052, 0.049, and 0.051, respectively, and to the right are 0.056, 0.054, and 0.055, respectively. The variation of the mean of the replicates on both sides is 0.004, which does not exceed the preset variation threshold. Therefore, the candidate boundary is merged.

[0081] The fertile release capacity components of the three independent replicates to the left of the other candidate boundary are 0.052, 0.049 and 0.051, respectively, and to the right are 0.149, 0.153 and 0.151, respectively. The variation range of the mean of the replicates on both sides is 0.100, which exceeds the preset variation threshold, and the variation direction of the three independent replicates is consistent. Therefore, this candidate boundary is retained.

[0082] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for predicting weekly yield of greenhouse tomatoes based on temperature, humidity, and irradiance accumulation, characterized in that, include: Step 1: Obtain the flowering time of the inflorescence batch, the number of productive open flowers, the fruit load and temperature at the same time, the water vapor pressure deficit, and the time series of photosynthetically active radiation absorbed by the canopy, and determine the target inflorescence batch and the probability of its maturity at the target week. Step 2: Based on the temperature development rate of the variety and development stage, the time series is mapped to the stage-normalized development process. Within the sub-intervals of the same development process, the cumulative radiation and residence time are calculated according to the temperature and water vapor pressure deficit joint interval to form the temperature and humidity state irradiance accumulation spectrum. Step 3: Based on the accumulation spectrum of the flower formation stage and the flowering and fruit setting stage, obtain the reproductive realization parameters from the reproductive environment response library. The response library stores the accumulation spectrum and reproductive realization parameters of the calibrated inflorescence batch. The parameters are determined by the two levels of fertile pollen release obtained by the first actual vibration response and the second actual vibration response of the same diagnostic flower at the same diagnostic frequency. The anther cone response amplitude of the second actual vibration response is higher than that of the first actual vibration response. Step 4: Determine the equivalent production pollination coverage parameter based on the proportion of productive open flowers that reach the first actual vibration response under actual pollination of the target inflorescence batch, and determine the effective number based on the reproductive realization parameter, coverage parameter and the number of productive open flowers. Step 5: Determine the average fruit weight based on the accumulation spectrum and fruit load during the same period after fruit set. Summarize the effective number, average fruit weight, and target week maturity attribution probability of each target inflorescence batch to obtain the predicted yield per unit area for the target harvest week.

2. The method for predicting weekly yield of greenhouse tomatoes based on temperature, humidity, and irradiance accumulation according to claim 1, characterized in that, The continuous maturation process is determined based on the flowering time, temperature time series, and temperature development rate corresponding to the variety and development stage of the inflorescence batch. The target week maturity assignment probability is determined based on the continuous maturation process and the pre-calibrated cumulative distribution function of variety commercial maturity. The inflorescence batches with a target week maturity assignment probability not lower than a preset contribution threshold are identified as the target inflorescence batches.

3. The method for predicting weekly yield of greenhouse tomatoes based on temperature, humidity, and irradiance accumulation according to claim 1, characterized in that, Each developmental stage is divided into multiple iso-developmental process sub-intervals. When the developmental process interval corresponding to the sampling period spans at least two iso-developmental process sub-intervals, the canopy absorbed photosynthetically active radiation and residence time of the sampling period are allocated to the corresponding iso-developmental process sub-intervals according to the overlap ratio between the developmental process interval and each iso-developmental process sub-interval. When the developmental process increment is zero, the canopy absorbed photosynthetically active radiation and residence time of the sampling period are allocated to the iso-developmental process sub-interval where the current developmental process is located.

4. The method for predicting weekly yield of greenhouse tomatoes based on temperature, humidity, and irradiance accumulation according to claim 1, characterized in that, The reproductive environment response library is established separately for each tomato variety. The establishment process includes: applying calibration excitation of multiple candidate frequencies to a preset excitation position of a representative diagnostic flower, detecting the actual vibration response of the preset excitation position and the anther cone, and determining the mechanical transmission relationship between the preset excitation position and the anther cone based on the calibration result that the pollen release amount is not higher than the background threshold; adjusting the excitation input to the same representative diagnostic flower according to the mechanical transmission relationship, so that the anther cone reaches the first actual vibration response and the second actual vibration response in sequence at the same diagnostic frequency, and detecting the first fertile pollen release amount and the second fertile pollen release amount in two mutually isolated collection periods.

5. The method for predicting weekly yield of greenhouse tomatoes based on temperature, humidity, and irradiance accumulation according to claim 1, characterized in that, The reproductive fulfillment parameters include a fertile release capacity component and a first response release component; The fertile release capacity component is determined based on the sum of the first fertile pollen release and the second fertile pollen release relative to the reference fertile pollen release, which is obtained by pre-calibrating a reference diagnostic flower under the same tomato variety, flower age range, detection system, and two-level response conditions. The first response release component is determined based on the proportion of the first fertile pollen release amount to the sum of the first fertile pollen release amount and the second fertile pollen release amount; when the sum of the first fertile pollen release amount and the second fertile pollen release amount is lower than the preset effective detection limit, the first response release component is not generated.

6. The method for predicting weekly yield of greenhouse tomatoes based on temperature, humidity, and irradiance accumulation according to claim 5, characterized in that, First, candidate temperature ranges and candidate vapor pressure deficit ranges are set, and a preset change threshold is determined based on the repeated measurement error. For adjacent candidate temperature ranges or adjacent candidate vapor pressure deficit ranges, if the change amplitude of the fertile release capacity component of the calibrated inflorescence batch under the corresponding environmental treatment, or the change amplitude of the first response release component when both adjacent environmental treatments generate the first response release component, exceeds the preset change threshold and the change direction is consistent in independent repeated experiments, the boundary of the corresponding candidate range is retained; otherwise, the corresponding adjacent candidate ranges are merged.

7. The method for predicting weekly yield of greenhouse tomatoes based on temperature, humidity, and irradiance accumulation according to claim 1, characterized in that, When the anther cone vibration frequency of a productive open flower under actual pollination is within the allowable frequency range of the first actual vibration response, and the anther cone response amplitude reaches the target amplitude of the first actual vibration response, and the effective duration reaches the first duration or the equivalent cumulative response amount reaches the calibrated response amount threshold within the effective period of reaching the target amplitude, the productive open flower is determined to have reached the first actual vibration response; the equivalent production pollination coverage parameter is determined based on the proportion of the number of productive open flowers that have reached the first actual vibration response to the sampled number, which is extracted stratified from the target inflorescence batch.

8. The method for predicting weekly yield of greenhouse tomatoes based on temperature, humidity, and irradiance accumulation according to claim 5, characterized in that, Once the first response release component has been generated, the effective rate is obtained from the pre-calibrated variety effective rate relationship based on the fertile release capacity component, the first response release component, and the response equivalent production pollination coverage parameters. When the first response release component is not generated, the effective rate is obtained from the variety effective rate relationship that does not include the first response release component; the effective number is determined based on the effective rate and the number of productive open flowers; the variety effective rate relationship is calibrated by the ratio of the number of fruits that finally reach the preset commercial standard in the inflorescence batch to the corresponding number of productive open flowers.

Images