Plant sound-light synergistic growth promotion control device

Abstract

The application discloses a plant sound-light synergistic growth promotion control device, and relates to the technical field of mixed intelligent automatic control.The device comprises a spectrum driving module, a photoperiod controller, an acoustic execution array, a sensing and collecting module, and a computing and control module electrically connected with the aforementioned modules; the computing and control module is configured to execute a synergistic growth promotion control process.The application realizes phase coupling of acoustic stimulation and spectrum regulation in the same system, establishes a low-peak multi-tone detection sequence based on blue noise Poisson disk sampling and Schroeder multi-tone phase distribution, generates a directional sound field by jointly solving a multi-channel cross-spectrum square matrix and a pointing vector table, and forms a phase resetting sequence with a minimum duration by combining a phase response curve mapping table matched with crop growth rhythm, so that acoustic stimulation and spectrum formula are synchronized in time domain and phase.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of mixed intelligent automatic control, in particular to a plant sound-light synergistic growth promotion control device in the process of plant seedling cultivation. BACKGROUND

[0002] The development of plant factories and facility agriculture makes artificial light sources, environmental regulation and sensing networks become the core technical links in crop production. At present, in the promotion of plant growth and the regulation of light physiology, the light regulation system with light spectrum formula control as the core is mainly used. The existing light spectrum regulation technology usually adjusts the proportion of red, blue, far-red or green light bands, combines light period and light intensity scheduling, and realizes the regulation of plant morphological construction, photosynthetic efficiency and secondary metabolic products. In some studies, dynamic light spectrum control is introduced, that is, according to light illumination sensing and crop phenotype feedback, the light spectrum formula is dynamically adjusted to improve energy efficiency. However, this method only acts on the light dimension, and does not utilize the mechanical wave response of plants in the non-visual frequency band, so it is difficult to further improve the dry matter accumulation rate under the same energy consumption. On the other hand, as a non-contact control means, sound stimulation has been proved to be able to affect the opening and closing of plant ion channels, membrane potential and gene expression. For example, sound signals in a specific frequency range can promote stomatal opening, accelerate transpiration and photosynthetic rate. However, traditional sound stimulation devices mostly use single frequency or random sweep frequency, which is difficult to stably control the sound field distribution formed in the canopy space, and is easily disturbed by reflection and standing wave, resulting in too high or too low local sound pressure, so that the response of plants is not uniform. In addition, the existing sound stimulation system usually operates independently of the light regulation system, and does not establish the phase relationship between the sound signal and the light period, so it cannot form a stable sound-light synergistic effect.

[0003] In a complex facility environment, noise sources, reflecting surfaces and pipeline structures will change the sound propagation path, so that simple amplitude control cannot maintain the dose consistency at the canopy level. In order to improve this problem, some studies have tried to use a multi-channel array to play sound signals to enhance directivity, but the multi-channel driving is mostly based on fixed weight distribution, and the directivity is not corrected in real time according to the environmental noise and reflection characteristics. In addition, the existing multi-tone signal generation method usually directly superimposes sine waves of different frequencies, resulting in too high peak value in time domain, and the amplifier produces nonlinear distortion when it is overloaded instantaneously, which further affects the accuracy of the stimulation signal. For the light regulation part, most of the traditional systems only control the light period according to the timer or simple logic, lack of phase regulation mechanism synchronized with the crop circadian rhythm, and there is a deviation between the light and the endogenous physiological rhythm of the crops. Even in some high-end systems, physiological models or phenotype feedback are introduced, but the control period is usually in hours, which cannot respond to the physiological fluctuations in minutes.

[0004] In summary, the prior art has the following main problems. First, the sound stimulation and light regulation are not formed in a closed loop, and the two are independently operated in time sequence, which is difficult to lock in phase. Second, the multi-channel acoustic array fails to adaptively correct the direction based on the actual response of the canopy, and the sound energy is unevenly distributed in space, affecting the consistency of growth. Third, the traditional multi-tone signal design has high peak value and strong regularity of frequency point spacing, which is easy to produce standing wave and resonance. Fourth, the light period control lacks a dynamic adjustment mechanism corresponding to the phase response curve of the crop, and cannot achieve optimal light pattern formation under the premise of fixed energy input. Therefore, how to simultaneously realize dynamic cooperative control of sound stimulation and light spectrum formula in a plant factory, establish a sound-light joint stimulation scheme that is consistent in frequency domain, time domain and phase, has become a key technical problem that needs to be solved in the current plant intelligent cultivation system. SUMMARY

[0005] The purpose of the present application is to provide a plant sound-light cooperative growth control device, which realizes phase coupling of acoustic stimulation and light spectrum regulation in the same system, establishes a low-peak multi-tone detection sequence based on blue noise Poisson disc sampling and Schroeder multi-tone phase distribution, generates a directional sound field by joint solving of a multi-channel inter-spectrum matrix and a pointing vector table, and forms a phase reset sequence with the minimum duration by combining a phase response curve mapping table matched with the crop growth rhythm, so as to realize synchronization of sound stimulation and light spectrum formula in time domain and phase. The present application can form a stable and consistent sound-light action area at the canopy reference position, maintain phase alignment of sound-light cooperative output through real-time correction and online confirmation, thereby improving dry matter accumulation efficiency, energy utilization rate and growth consistency, significantly improving crop quality and reducing energy consumption, and having the advantages of high precision, low interference and long-term operation.

[0006] To solve the above technical problems, the present application provides a plant sound-light cooperative growth control device, which comprises a spectrum driving module, a photoperiod controller, an acoustic execution array, a sensing acquisition module, and a computing and control module electrically connected with the above-mentioned modules.

[0007] The computing and control module is configured to execute a cooperative growth control process, which comprises:

[0008] Step 1: generate a target frequency point table based on blue noise Poisson disc sampling in the logarithmic frequency domain, and assign Schroeder multi-tone phase to the generated target frequency point table to synthesize a multi-tone detection sequence with low peak value characteristics, play the multi-tone detection sequence through the acoustic execution array and collect the response through the sensing acquisition module, so as to determine a sound stimulation baseline;

[0009] Step 2: based on the acoustic environment signals of each channel of the array, a multi-channel cross-spectral matrix is constructed, and a pointing vector table is constructed by combining the response collected at the crown reference position, and a set of channel coefficient sets is calculated by solving the multi-channel cross-spectral matrix and the pointing vector table, and a multi-channel directional driving sequence for forming a stable directional sound field at the crown reference position is generated using the calculated channel coefficient set;

[0010] Step 3: based on a phase response curve mapping table matched with the target crop, a set of windows that do not overlap in phase and meet the target cumulative score are selected by calculating the score and efficiency of the candidate window, a phase reset sequence with the smallest cumulative duration is formed, and an acousto-optic coupling plan within the phase reset sequence is generated;

[0011] Step 4: according to the generated multi-channel directional driving sequence and phase reset sequence, the execution of the acoustic execution array and the spectrum driving module is arranged and scheduled, and through online confirmation and correction, the phase-aligned acousto-optic cooperative stimulation at the crown reference position is realized.

[0012] Further, the target frequency point table generated by the calculation and control module in step 1 is specifically configured as follows: a set of frequency points is generated on the logarithmic frequency axis using the dart throwing method of blue noise Poisson disk sampling, wherein candidate frequency points are randomly generated within the range of the logarithmic frequency axis, if the logarithmic distance between the candidate frequency point and any point in the point set is greater than or equal to a minimum logarithmic interval, the candidate frequency point is accepted and added to the point set, and the generation is ended when the activity list is empty and the maximum logarithmic gap between any adjacent points is less than 2 times the minimum logarithmic interval.

[0013] Further, the calculation and control module synthesizes a multi-tone detection sequence with low peak value characteristics in step 1, which specifically includes: calculating a peak evaluation value of the multi-tone detection sequence, if the peak evaluation value is higher than a target upper limit, performing a phase refinement loop, the phase refinement loop includes overall rotation of a working phase table and fine tuning of each phase in the working phase table in order to reduce the peak evaluation value.

[0014] Further, the calculation and control module constructs a multi-channel cross-spectral matrix in step 2, which specifically includes: in the silent environment segment, the sensing and collecting module uses several segments of 2048 samples long with 50% overlap to collect the environment signals of each channel; perform fast Fourier transform on each segment of environment signals, and calculate the average value of the complex product across all segments at each working frequency point for any two channels to form the cross-spectral value of the working frequency point; the cross-spectral values of all channel pairs are organized into a matrix according to the channel index.

[0015] Further, the calculating and controlling module calculating the channel coefficient set in step 2 specifically comprises: performing Cholesky decomposition on the multi-channel cross-spectrum matrix of each working frequency point to obtain a lower triangular matrix; solving a linear equation set through forward substitution and back substitution; calculating a complex inner product of the steering vector and the solving result, and dividing each element of the solving result by the calculated complex inner product at the same time, so that the response of the canopy reference position at the working frequency point is a unit response.

[0016] Further, the calculating and controlling module in step 2 is further configured to perform closed-loop pointing consistency correction, specifically comprising: calculating an amplitude-phase deviation between the actual complex response and the unit response at the canopy reference position for each working frequency point, and repeating the solving process for the frequency point to update the channel coefficient set when the amplitude-phase deviation of any frequency point exceeds a set threshold.

[0017] Further, the calculating and controlling module in step 3 specifically comprises: dividing a complete light period into 1440 equally spaced phase grids before forming the phase reset sequence; for each phase grid and a set of preset durations, calculating a score sum in the phase grid covered by a window as a window score, and dividing the window score by the window duration to obtain an efficiency value; sorting all windows by efficiency value from high to low to form a candidate window sequence.

[0018] Further, the calculating and controlling module in step 3 specifically comprises: starting from the beginning of the candidate window sequence, if the current window does not overlap with the selected window in phase, adding the current window to a selected set and accumulating its score, and terminating the traversal when the cumulative score reaches or exceeds a target cumulative score threshold to form the phase reset sequence.

[0019] Further, the calculating and controlling module in step 4 specifically comprises: the sensing and collecting module collects a monitoring segment of 1 second length at the canopy reference position at 5 second intervals; calculating a complex response at each working frequency point on the monitoring segment, and when the amplitude deviation or phase deviation of the calculated complex response exceeds a preset online confirmation threshold, calling the solving process of step 2 to perform a fast solving on the current frequency point and updating the channel coefficient set.

[0020] Further, the calculating and controlling module is further configured to perform fault safety and recovery, specifically comprising: suspending scheduling and refilling the buffer when it is monitored that the driving buffer of any channel is underloaded; reducing the amplitude coefficient of all channels when it is monitored that the absolute sound level of the canopy reference position exceeds a permission threshold; keeping the current channel coefficient set unchanged to continue execution when the sensing and collecting module does not return valid data in three consecutive monitoring segments.

[0021] The plant sound-light synergistic growth promotion control device has the following beneficial effects:

[0022] The present application realizes dynamic synergy of sound stimulation and spectral modulation in the same control system, so that the plant obtains phase-aligned sound-light action at the canopy level, thereby significantly improving dry matter accumulation efficiency and energy utilization rate. Compared with the traditional single spectral control or independent sound stimulation method, the present application uses the target frequency table generated by blue noise Poisson disc sampling and the Schroeder multi-tone phase distribution method, so that the multi-tone detection sequence has low peak value characteristics in the time domain, reduces the nonlinear distortion caused by energy concentration, and ensures that the sound signal is uniformly and stably distributed in the canopy area. Through joint solving of the multi-channel cross-spectrum matrix and the pointing vector table, a directional sound field with the canopy reference position as the target is established, realizing real-time adaptive compensation of environmental reflection and background noise, so that the sound energy has directionality, consistency and repeatability in spatial distribution.

[0023] Further, the present application introduces a phase response curve mapping table matched with the crop growth rhythm, divides the photoperiod into a high-resolution phase grid, and selects the optimal phase reset sequence through candidate window scoring and efficiency calculation to realize the target cumulative response in the minimum duration. This dynamic regulation based on the phase response curve locks the sound stimulation and spectral formula in the time dimension, can accurately control the acoustic reinforcement window of the crop photosynthetic efficient area, and thereby promotes the synergy of stomatal conductance and photosynthetic electron transfer. The acoustic execution array operates synchronously with the spectral driving module under the multi-channel directional driving sequence, and can maintain stable output of sound-light synergy for a long time through online confirmation and real-time correction. Overall, the present application not only improves the controllability of sound stimulation in space and time, but also reduces redundant input through energy distribution and phase coupling optimization, realizes efficient, uniform and intelligent regulation of the plant growth environment, obtains higher yield and better quality under the same energy consumption, and has good engineering implementability and scalability. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1 The acoustic execution array and plant canopy spatial arrangement principle diagram provided for the embodiments of the present application;

[0025] Figure 2 The sound pressure level distribution diagram of the multi-channel directional sound field in the canopy space provided for the embodiments of the present application;

[0026] Figure 3 The sound pressure level frequency response characteristic diagram of the canopy reference position provided for the embodiments of the present application;

[0027] Figure 4 The plant growth promotion effect comparison curve diagram provided for the embodiments of the present application;

[0028] Figure 5 A sound-light synergistic promotion timing control chart provided for the embodiment of the present application;

[0029] Figure 6 A crown sound pressure level response experimental curve schematic diagram provided for the embodiment of the present application. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0031] Embodiment 1, a plant sound-light synergistic promotion control device, comprising: a spectrum driving module, a photoperiod controller, an acoustic execution array, a sensing acquisition module, and a computing and control module electrically connected with the foregoing modules. The computing and control module is configured to execute a synergistic promotion control process to control the operation of each module electrically connected thereto to complete the plant sound-light synergistic promotion control. In terms of physical arrangement, the acoustic execution array and the lighting fixture are coplanarly installed about 300 millimeters above the cultivation layer, the crown reference position sampling head of the sensing acquisition module is about 100 millimeters away from the crop leaf surface, and the array reference position sampling head is located about 80 millimeters outside the extrapolation of the center line in front of the acoustic execution array. The computing and control module adopts an independent case, is built-in with an industrial-grade main control and an isolation power supply, and is connected with each unit through a shielded cable.

[0032] REFERENCE Figure 1, the acoustic executive array and the plant canopy spatial arrangement principle is as follows: in the plant cultivation system, the cultivation layer is arranged at the bottom, bearing the plant growth substrate and root system. The plant canopy is located above the cultivation layer, composed of stems and leaves of multiple plants. The canopy height is determined according to the crop type and growth stage, and the typical range is 300-600 mm. The acoustic executive array is installed about 300 mm above the cultivation layer and is arranged coplanar with the lighting lamp. The array adopts a rectangular arrangement, and the commonly used configuration is a 32-unit array, arranged in 8 rows and 4 columns or 4 rows and 8 columns. The center distance between each unit is set to 40-60 mm, ensuring that the array aperture can form a directional beam and will not cause grating lobes due to excessive unit spacing. A detachable sound-absorbing frame and diffusion grille are installed in front of the array to reduce the edge diffraction effect and cavity reflection interference. The canopy reference position sampling head is set about 100 mm above the plant leaf surface, which represents the target area where the sound energy actually acts. The sampling head is equipped with an omnidirectional microphone for real-time monitoring of the sound dose reaching the canopy, including amplitude and phase information. The position selection of the sampling head needs to consider the leaf height variation and measurement stability. The array reference position sampling head is set about 80 mm in front of the acoustic executive array along the center line of the array. The sampling head is used to estimate the instantaneous state of the sound emitting end and provide reference data for the construction of the multi-channel cross-spectrum matrix. The two reference position sampling heads use the same sampling rate and clock source to ensure the time consistency of the sound field characteristic evaluation. The whole spatial arrangement follows the principle of optimal sound propagation path, so that the sound energy accurately converges to the canopy reference position after propagating through the free field from the array, and the full-link monitoring of the transmitting end and the receiving end is realized through double-point measurement.

[0033] The spectral drive module uses a multi-channel constant current drive architecture, each channel independently controls a waveband of semiconductor light source, typical configuration is blue light, red light and far red three channels, can be extended to near ultraviolet and green light a total of five channels. Each channel is composed of digital-to-analog converter, isolated gate driver and constant current control board, the control resolution is not less than 16 bits. In order to ensure that the spectral formula remains consistent between different assembly batches, a one-to-one factory calibration process is established. The specific steps are as follows: using the integrating sphere and the traceable spectral irradiance standard source, the irradiation output of each channel at different drive code values is recorded as a lookup table, and the code value step is recommended to be 256 levels; At the end of the production line, the lookup table and the channel serial number are written into the non-volatile memory together. The calculation and control module runs according to the target of each waveband given in the spectral formula, queries the lookup table and uses linear interpolation method to obtain the channel code value. In order to eliminate low-frequency flicker, the channel current uses a fixed carrier frequency pulse width modulation and constant current hybrid method: the pulse width modulation carrier frequency is set to 25000 Hz, and the constant current fine adjustment step is not less than 0.5% of the full scale range of current. When the spectral formula is switched, the ramp transition is performed, and the default transition time is 200 milliseconds, which avoids interference with leaf imaging and acoustic measurement. Considering that the lamp reflector cavity will cause local hot spots, the spectral drive module adds a temperature compensation channel before each channel, the temperature sampling period is set to 1 second, and for every 10 degrees Celsius increase in channel temperature rise, the corresponding code value is automatically reduced by 2% to 3%, and the results are reviewed through the spectral probe of the sensing acquisition module. The direct benefit of this is to maintain the stability of photon output in long-period operation and avoid spectral formula drift during phase reset sequence execution.

[0034] The specific implementation process of the photoperiod controller includes time base establishment, phase grid and phase reset sequence driving. The time base is composed of temperature compensated crystal and network time synchronization. The clock drift is not more than 200 milliseconds within 24 hours, and the network can be corrected once every 6 hours. The light-on time is taken as the phase zero point, and the complete photoperiod is divided into 1440 equal phase grids, each corresponding to 1 minute. The phase response curve mapping table is composed of 1440 scores, corresponding to the target crop and its growth stage one by one. The photoperiod controller completes the daily plan compilation 10 minutes before the daily light-on: first, read the phase response curve mapping table and the execution summary recorded in the last period; then generate a candidate window list according to the candidate window duration set, and the score and efficiency calculation is completed by the calculation and control module; the photoperiod controller receives the phase reset sequence and generates a timestamped action queue, including triggering a multi-tone detection sequence or a sound stimulus baseline at a certain minute, switching the spectral formula at a certain minute, and restoring the regular spectral formula at the window end minute. To ensure strict alignment with the multi-channel directional driving sequence, the photoperiod controller issues a pre-event 30 seconds before the start of each window, and the calculation and control module pre-fills the buffer area accordingly. The reason for using minute-level phase grid is that the coupling response of plant physiology to sound stimulus and spectral formula is most sensitive to minute level, and too coarse grid will reduce the controllability of the phase reset sequence, and too fine grid will significantly increase the scheduling overhead while the growth promotion benefit is limited.

[0035] The acoustic executive array is responsible for the sound energy delivery of the multi-tone detection sequence, the sound stimulus baseline and the multi-channel directional driving sequence. To balance coverage and directionality, it is recommended to configure 16-unit, 32-unit or 64-unit acoustic executive arrays on each cultivation shelf layer, commonly used as a 32-unit rectangular array with a center distance of 40-60 mm. Each unit is composed of a numbered transducer, a dedicated power amplifier and a pre-bandpass filter, and the unit rated sound pressure reaches 85 decibels or above at 1 meter. A detachable sound-absorbing frame and a diffusion grille are installed in front of the array to reduce edge diffraction and cavity reflection. The generation of the multi-tone detection sequence and the sound stimulus baseline follows the target frequency table and the working phase table. The target frequency table is generated by the calculation and control module based on the blue noise Poisson disk sampling dart throwing method. Specifically to the synthesis process, the calculation and control module creates an array in the audio buffer with a length equal to the product of the sampling rate and the sequence duration, typically 48000 samples per second and 2.0 seconds. For each sample, the sine value is accumulated frequency by frequency according to the target frequency table, and the sine angle is determined by the frequency point and the working phase table. After the accumulation is completed, the array is scanned to obtain the maximum absolute value and the root mean square value, and the peak value is evaluated by the ratio of the two. If the peak evaluation value is higher than the target upper limit, the phase refinement loop is executed: first, rotate the working phase table by 5 degrees as a whole, regenerate the array and evaluate it; if the evaluation value decreases, keep the rotation. Then make a 3-degree positive fine-tuning to each frequency point of the working phase table and evaluate it, if it continues to decrease, keep it, otherwise make a 6-degree reverse fine-tuning and evaluate it, and if it still does not decrease, restore the original value. This loop is executed for 10-15 rounds until the peak evaluation value falls within the target range. Normalize the array after peak control to 70-80% of full scale according to the root mean square value, and the multi-tone detection sequence is obtained. The determination of the sound stimulus baseline is completed by one-time playback and two-point measurement: the array reference position and the canopy reference position record the response synchronously, calculate the amplitude and phase of each frequency point, sort the amplitude at the canopy reference position, select a fixed number of working frequency point sequences, such as 24 or 32, and save the phase as the sound stimulus baseline for subsequent long-term operation. The advantage of this is to obtain a working frequency point sequence matched with the actual cavity in the shortest process on site, ensuring that the subsequent directional sound field construction and phase reset sequence execution are based on the effective acoustic dose of the canopy.

[0036] When a stable directional sound field needs to be formed, the acoustic execution array generates a multi-channel directional drive sequence according to a multi-channel cross-spectrum matrix and a pointing vector table. The construction process of the multi-channel cross-spectrum matrix is as follows: in a silent environment section, the sensing and collecting module collects the environment signal of each channel, and uses a window with a length of 2048 samples and an overlap of 50% for segmentation; a fast Fourier transform is performed on each segment; at each working frequency point, the complex product of the frequency domain data of any two channels is calculated and averaged across all segments to obtain a cross-spectrum value; and the cross-spectrum value is combined into a matrix according to the channel index. The construction process of the pointing vector table is as follows: a multi-tone detection sequence is played sequentially by channel, and the response at the canopy reference position is recorded to obtain a complex response of the channel at each working frequency point, which is arranged into a column vector according to the channel number. The calculation and control module takes the multi-channel cross-spectrum matrix and the pointing vector table as input, and obtains a channel coefficient set by matrix decomposition and back substitution, and normalizes the amplitude with the unit response as a constraint. The reason for using this process is that the environmental noise and reflection show stable statistical characteristics in the frequency domain, and the influence of incidental interference can be reduced by cross-spectrum averaging; taking the unit response as the target can directly aim the energy of the directional sound field at the canopy reference position, avoiding the diffusion of energy in non-target areas. The final multi-channel directional drive sequence is composed of sample arrays of each channel, and the generation method is to superimpose the phase advance and amplitude correction of each working frequency point into the sinusoidal superposition of the channel, which is consistent with the synthesis method of the multi-tone detection sequence in implementation, thereby ensuring real-time and maintainability.

[0037] Figure 2A schematic diagram showing the sound pressure level distribution formed in the canopy space after employing the multi-channel directional drive sequence. The diagram establishes a two-dimensional coordinate system, with the horizontal axis representing the horizontal position, ranging from -400 mm to +400 mm, with the zero point located directly below the center line of the array; the vertical axis represents the vertical position, ranging from 0 mm to 400 mm, with the zero point located at the surface of the cultivation layer, with the upward direction being positive. At the position of 0 mm to 15 mm in the vertical position, the physical position of the acoustic execution array is marked, with the array having a horizontal span of 500 mm and the center located at the horizontal coordinate origin. Through the solution of the multi-channel cross-spectral matrix and the steering vector table, a set of channel coefficient sets is obtained, so that each channel forms a coherent superposition when playing the multi-channel directional drive sequence, thereby generating a sound field with directional characteristics in space. Several sound pressure level contours are plotted in the diagram. The innermost contour corresponds to a sound pressure level of 90 decibels, and the contour is in the shape of an ellipse, with the center located at the horizontal coordinate 0 mm and the vertical coordinate about 350 mm, the horizontal semi-axis about 120 mm, and the vertical semi-axis about 80 mm. In order from outside, there are 85 decibel contours, 80 decibel contours, and 75 decibel contours, and the major and minor axes of the ellipses gradually expand as the sound pressure level decreases, with the horizontal semi-axis of the 85 decibel contour about 200 mm and the vertical semi-axis about 140 mm, and the horizontal semi-axis of the 80 decibel contour about 280 mm and the vertical semi-axis about 200 mm. The 75 decibel contour is represented by a dashed line, indicating that it is the outer boundary of the effective action area. At the horizontal coordinate 0 mm and the vertical coordinate 450 mm, the canopy reference position is marked, which is represented by a circle symbol with a cross. The annotation shows that the sound pressure level at this position is 87 decibels. The canopy reference position falls exactly between the 85 decibel and 90 decibel contours, and is located in the center region of the main lobe of the sound field, indicating that by solving the channel coefficient set and applying the unit response normalization constraint, the energy main lobe is successfully aligned with the canopy reference position. In the area with vertical coordinates from 430 mm to 530 mm and horizontal coordinates from -400 mm to +400 mm, the effective action area of the plant canopy is marked with a dashed rectangular frame. This area is highly consistent with the 85 decibel contour, ensuring that the acoustic dose received by each point in the canopy is uniform and reaches the expected intensity. The annotation in the diagram explains that the directional sound field distribution is achieved by solving the channel coefficients of the multi-channel cross-spectral matrix and the steering vector table, thereby directing the energy main lobe to the canopy reference position and avoiding the diffusion of sound energy in non-target areas, improving the delivery efficiency and spatial consistency of the sound stimulation baseline.

[0038] The sensing acquisition module includes three types of detection units: acoustic, optical and environmental. All samples use a unified clock source and are aligned with the time base of the computing and control module. The acoustic detection unit uses two types of sampling heads: array reference position sampling head for estimating the instantaneous state of the sound emitting end, and crown layer reference position sampling head for evaluating the actual dose reaching the crown layer. Each sampling head is connected to an analog-to-digital converter channel with a resolution of 24 bits and a sampling rate of 48000 samples per second or 96000 samples per second. The pre-amplifier has programmable gain. The real-time processing procedure is as follows: every 5 seconds, a 1-second length monitoring segment is intercepted, the same window function as the cross-spectral estimation is used, the complex response at each operating frequency point is calculated, and the amplitude deviation and phase deviation are output. The optical detection unit is composed of a spectral probe and an illuminance probe. The spectral probe is used to confirm that the energy distribution of the target waveband has reached the expected value when the spectral formula is switched, with a sampling period of 10 seconds. The illuminance probe is used to quickly detect illumination mutations and trigger abnormal records. The environmental detection unit includes temperature, relative humidity and carbon dioxide concentration sensors, with a sampling period of 30 seconds, which is used to explain the phenotypic proxy fluctuations caused by leaf temperature fluctuations and transpiration changes, and to avoid misjudging environmental disturbances as acoustic or optical channel abnormalities. In order to reduce the influence of fan and pump noise on acoustic measurement, the sensing acquisition module maintains low-duty-cycle silent sampling during non-window periods, and performs static noise floor update at fixed time every day to ensure that the multi-channel cross-spectral matrix reflects the current state when solving the next time.

[0039] The computing and control module implements the core process orchestration and all signal generation, estimation and correction. Its software hierarchy includes task scheduling layer, signal synthesis layer, frequency domain estimation layer and safety management layer. The task scheduling layer uses millisecond-level time base to drive the action queue, ensuring that the baseline of sound stimulation and the switching of spectral recipe are triggered at the beginning of the phase reset sequence window; at the end of the window, the normal state is restored. The signal synthesis layer provides two types of output: multi-tone detection sequence for step 1 and multi-channel directional driving sequence for step 2. The generation of the multi-tone detection sequence strictly follows the target frequency table and the working phase table and contains the peak control process; the multi-channel directional driving sequence superimposes the phase advance and amplitude correction given by the channel coefficient set on this basis. The frequency domain estimation layer is responsible for generating the multi-channel cross-spectrum matrix, the pointing vector table and calculating the complex response of the crown layer reference position, all processing is based on the fast Fourier transform as the core operator, the window length, overlap rate and frequency index are consistent with the above. The safety management layer provides three types of protection: when the driving buffer of any channel is underloaded, the scheduling is suspended and refilled; when the absolute sound level of the crown layer reference position reaches the permitted threshold, the amplitude of all channels is uniformly reduced and the event is recorded; when the sensor acquisition module does not return valid data for three consecutive monitoring segments, the current channel coefficient set is kept unchanged until the end of the window and an alarm is issued. This layer is also responsible for generating a daily execution summary, which contains the window start and end minutes, cumulative playing time, peak evaluation value, root mean square value, phase deviation statistics and spectral recipe switching timestamp, for production and operation and compliance retention.

[0040] In actual operation, the complete process is executed in the following order. About 20 minutes before the light is turned on every day, the computing and control module triggers the update of the multi-tone detection sequence, generates the target frequency table according to the blue noise Poisson disk sampling, and performs Schroeder multi-tone phase distribution and peak control to obtain a new multi-tone detection sequence. Then a play is performed, the sensor acquisition module synchronously acquires at the array reference position and the crown layer reference position, selects the working frequency sequence and writes it into the sound stimulation baseline. About 10 minutes after the light is turned on, the photoperiod controller issues the action queue of the daily phase reset sequence and the spectral recipe to the task scheduling layer. 30 seconds before the window start minute, the task scheduling layer completes the preloading of the multi-channel directional driving sequence and the preheating of the spectral recipe; at the window start minute, the acoustic execution array and the spectral driving module act synchronously, and the crown layer reference position starts to output monitoring segments at 5-second intervals. If the amplitude deviation or the phase deviation in the monitoring segment exceeds the online confirmation threshold, the fast solving process of the multi-channel cross-spectrum matrix and the pointing vector table is called, and only the channel coefficient set for the working frequency that exceeds the deviation is updated to realize the closed-loop pointing consistency correction with minimum disturbance. When the window end minute comes, a 200-millisecond fade-out is performed and the normal spectral recipe is restored. All windows of the phase reset sequence are completed one after another in a photoperiod, and the execution summary is generated after the light is turned off on the same day, which is used for planning compilation in the next photoperiod.

[0041] To support different scales and crop types, several optional embodiments are given. The first optional embodiment adopts a 64-unit acoustic execution array, the center distance is reduced to 40 mm, and the main lobe is narrower after the array aperture is increased, which is suitable for fruit and vegetable cultivation shelves with a canopy height of more than 500 mm. The working frequency point sequence can be expanded to 48 to cover a wider target frequency band. The second optional embodiment adds near-ultraviolet and green light channels in the spectrum driving module. By superimposing short near-ultraviolet pulses on a specific window of the phase reset sequence through the spectrum formula, it is used to suppress excessive elongation. The spectrum probe completes the confirmation within 10 seconds after switching. The third optional embodiment adds a miniature leaf temperature infrared probe and a weighing tray in the sensing and collecting module, with a sampling period of 10 seconds and 60 seconds, respectively. The complex response of the leaf temperature and weight change and the canopy reference position are recorded in parallel, which is used to explain the situation that the amplitude deviation is temporarily increased but the actual growth state is good in some cases, to avoid unnecessary channel coefficient set update and energy waste. The fourth optional embodiment increases the sampling rate to 96,000 samples per second and the window length to 4,096 samples, which is suitable for formula library scenarios that need to use more working frequency points at the high frequency end, which can effectively improve the resolution of the cross-spectrum estimate.

[0042] The above implementation details embody a deep coupling with the phase response curve mapping table. Setting the light-on time as the phase zero point can establish a stable reference with the crop circadian rhythm. The phase reset sequence selects the window score and efficiency through the minute-level phase grid, ensuring that the target cumulative score is reached in the shortest cumulative duration, thereby achieving a repeatable growth promotion effect with less acoustic energy. Setting the canopy reference position as the common target position of acoustic delivery and online confirmation can maximize the avoidance of error propagation caused by cavity reflection and structural obstruction. Generating a target frequency point table through blue noise Poisson disk sampling, and then supplementing with Schroeder multi-tone phase allocation and peak control, can get a more uniform time energy distribution under the same playback amplitude, reducing mechanical vibration and accessory resonance triggered by transient peaks. The combination of the multi-channel cross-spectrum matrix and the pointing vector table solves the uncertainty caused by random noise in the field. Through piecewise averaging and unit response constraints, the energy main lobe is stabilized and pressed towards the canopy reference position. Online confirmation triggers fast solving only when the amplitude deviation or phase deviation exceeds the threshold, avoiding resource occupation and energy consumption rising caused by continuous solving, while ensuring stability when the environmental conditions fluctuate temporarily.

[0043] Embodiment 2: The computing and control module is configured to perform a cooperative growth promotion control process, including:

[0044] Step 1: generate a target frequency table based on blue-noise Poisson disk sampling in the logarithmic frequency domain, and assign Schroeder multitone phases to the generated target frequency table to synthesize a multitone probe sequence with low peak characteristics, play the multitone probe sequence through an acoustic executive array and collect the response through a sensing and collecting module to determine a sound stimulus baseline;

[0045] Step 2: construct a multichannel cross-spectral matrix based on the environmental signals of each channel of the acoustic executive array, and construct a pointing vector table combined with the response collected at the canopy reference position, solve the multichannel cross-spectral matrix and the pointing vector table to calculate a set of channel coefficient sets, and generate a multichannel directional driving sequence for forming a stable directional sound field at the canopy reference position using the calculated channel coefficient sets;

[0046] Step 3: based on a phase response curve mapping table matched with the target crop, select a set of windows that are phase- non-overlapping and meet the target cumulative score by calculating the score and efficiency of the candidate windows, form a phase reset sequence with the minimum cumulative duration, and generate a sound-light coupling plan in the phase reset sequence;

[0047] Step 4: according to the generated multichannel directional driving sequence and phase reset sequence, arrange and schedule the execution of the acoustic executive array and the spectrum driving module, and through online confirmation and correction, realize phase-aligned sound-light cooperative stimulation at the canopy reference position.

[0048] In step 1, the computing and control module first generates a target frequency point table in the logarithmic frequency domain. The implementation method adopts the dart throwing method of blue noise Poisson disk sampling. The lower limit is set to 100 Hz and the upper limit is set to 8000 Hz on the logarithmic frequency axis, and the minimum logarithmic interval is set to 0.035 octaves. A candidate point is randomly generated in the logarithmic frequency range at the beginning, and if the logarithmic distance between the point and any selected point in the point set is not less than the minimum logarithmic interval, the point is written into the point set and the active list. Then an active point is randomly taken out from the active list, and in the annular band with the active point as the center and the minimum logarithmic interval to twice the minimum logarithmic interval as the radius, up to 30 candidate points are uniformly generated. Each time a candidate point is generated, the logarithmic distance between the candidate point and all points in the point set is calculated, and if all distances are not less than the minimum logarithmic interval, the candidate point is added to the point set and the active list. When an active point fails to generate an acceptable candidate point after 30 attempts, the active point is removed from the active list. When the active list is empty, check the maximum logarithmic gap between any two adjacent points, and if the gap is greater than or equal to twice the minimum logarithmic interval, a point is added at the logarithmic midpoint of the gap and written back to the active list to continue the above process; when the maximum gap is less than twice the minimum logarithmic interval, the generation is terminated. Convert the point set back to linear frequency and sort it from low to high to get the target frequency point table. The size of the target frequency point table usually falls between 24 and 48, and a size less than 24 is easy to cause energy to concentrate on a few frequency points and generate a large peak in the time domain, and a size greater than 48 will significantly increase the computational overhead of a single drive and amplify the difference between the array reference position and the canopy reference position.

[0049] After the target frequency list is determined, Schroeder phase is assigned to each frequency. The specific method is to create a phase list with the same length as the target frequency list, set the phase of the first frequency to 0 degrees, and sequentially increase the phase of the second frequency by a fixed angle step of 137.507764 degrees. When the phase exceeds 360 degrees, it is rolled back to 360 degrees. This allocation method can naturally disperse the peak value after time domain superposition. After the phase allocation is completed, the multi-tone detection sequence is synthesized. The sampling rate of the audio buffer is set to 48000 samples per second, and the duration is set to 2.0 seconds. For each sample in the buffer, sequentially traverse the target frequency list, calculate the sine value of each frequency at that sample time, and superimpose them while using the phase in the phase list as the initial phase of each frequency. After the entire buffer is generated, the maximum absolute value and the root mean square value are calculated to obtain a peak evaluation value. If the peak evaluation value is higher than the preset upper limit (for example, 12), a phase refinement loop is executed: first, rotate the phase list by 5 degrees and re-synthesize and evaluate, if the peak evaluation value decreases, keep the rotation; then, perform a 3-degree positive fine-tuning for each frequency in the phase list, if the peak evaluation value decreases, keep the fine-tuning, otherwise perform a 6-degree negative fine-tuning, if it still does not decrease, restore the original value. The above refinement loop is executed for 10 to 15 rounds until the peak evaluation value falls within the target range. The root mean square value of the sequence is normalized to 70 to 80 of the full scale to ensure sufficient dynamic margin for subsequent driving. The reason for using the above synthesis and refinement method is that the time domain peak value of the multi-tone detection sequence directly determines the instantaneous load of the power amplifier and the mechanical response of the array structure, reducing the passive vibration of non-target structures and the fatigue of connecting parts, while helping to maintain the stability of the crown layer reference position.

[0050] After the multi-tone detection sequence is completed, play the multi-tone detection sequence once and simultaneously collect responses at the array reference position and the crown layer reference position. The collection duration and sampling rate of each position are consistent with the playback. Calculate the amplitude and phase of each frequency for both responses. At the crown layer reference position, sort the frequencies by amplitude from high to low, select a fixed number of frequencies to form a working frequency sequence, commonly 24 or 32, and keep the corresponding phase. Write the working frequency sequence, corresponding phase, and multi-tone detection sequence after peak control into storage as the baseline for subsequent sound stimulation. By selecting the working frequency sequence at the crown layer reference position and sorting it by amplitude, local attenuation points caused by spatial standing waves and cavity reflections can be actively avoided, and limited sound energy can be concentrated in the position where the actual effect is the strongest. Compared with the method of only looking at the array reference position, this method can maintain higher consistency when the cultivation space undergoes minor structural changes (such as pipe displacement or tray replacement).

[0051] In step 2, the computing and control module needs to generate a multi-channel cross-spectral matrix and a table of steering vectors, and then solve a set of channel coefficients to synthesize a multi-channel directional driving sequence for forming a stable directional sound field at the crown reference position. The generation process of the multi-channel cross-spectral matrix is as follows. First, collect the ambient signals of each channel in the silent environment segment, segment the signals using a window with a length of 2048 samples and an overlap of 50, and use a Hanning window. Perform a fast Fourier transform on each segment and take the frequency index closest to the working frequency sequence. At each working frequency, calculate the complex product of the frequency domain data of any two channels and average across all segments to obtain the cross-spectrum value of the channel pair at that frequency. Fill the cross-spectrum values of all channel pairs into the matrix according to the channel index, and independently form a multi-channel cross-spectral matrix for each working frequency. By collecting and averaging across segments in the silent environment segment, the statistical properties of background noise and reflection paths can be stably estimated, thereby suppressing the impact of incidental mechanical impact or short-time wind noise on the results during subsequent solving.

[0052] As Figure 3The figure shows the comparison of the sound pressure level frequency response characteristics of the crown reference position before and after the adoption of the multi-channel directional driving sequence in the present application. The horizontal axis is frequency, in hertz (Hz), using logarithmic scale distribution, ranging from 100 Hz to 8000 Hz, with scale points marked at 100, 200, 500, 1000, 2000, 5000, and 8000 Hz. The vertical axis is sound pressure level, in decibels sound pressure level (dB SPL), ranging from 50 dB to 100 dB, with 5 dB intervals evenly scaled. The dashed curve in the figure represents the sound pressure level frequency response measured at the crown reference position without implementing the multi-channel directional driving sequence. The curve shows a large fluctuation feature, with the sound pressure level fluctuating sharply between 62 dB and 86 dB in the entire frequency range, with a fluctuation amplitude of ±8 dB. This uneven frequency response is caused by reflections, standing waves between the acoustic executive array and the cultivation space cavity, and phase interference between channels. At some frequency points, sound pressure peaks are formed, while at other frequency points, obvious attenuation valleys appear, making the acoustic energy distribution extremely uneven in the frequency domain. The solid curve in the figure represents the sound pressure level frequency response measured at the crown reference position after implementing the multi-channel directional driving sequence of the present application. The curve is significantly flat, with the overall sound pressure level stably maintained at around 82 dB, and the fluctuation amplitude in the full-band range is controlled within ±2 dB. The black solid circles marked on the curve represent the selected frequency points in the working frequency point sequence, including 150, 280, 520, 980, 1850, 3500, and 6500 Hz. The sound pressure levels of these working frequency points are highly consistent, all close to the target sound pressure level of 82 dB, indicated by the dashed horizontal reference line. Through the joint solution of the multi-channel inter-spectrum matrix and the pointing vector table, the present application calculates the coefficient set of each channel, so that the complex response of each working frequency point at the crown reference position is normalized to a unit response. This technical means fundamentally eliminates the effects of cavity reflections and channel interference, accurately aligning the main lobe of the sound field energy to the crown reference position. Comparing the frequency response characteristics before and after direction, it can be seen that the present application significantly improves the uniformity of the acoustic dose, optimizing from the pre-direction fluctuation of ±8 dB to within ±2 dB after direction, ensuring stable and consistent acoustic stimulation intensity at the target position for each working frequency point, providing a reliable acoustic basis for subsequent sound-light synergistic growth control.

[0053] The generation flow of the steering vector table is as follows: sequentially single-channel drive the acoustic executive array, open only one channel and play a multi-tone detection sequence each time, and synchronously collect the response at the canopy reference position. The complex response of the channel is calculated for each working frequency point, and arranged as a column vector of the frequency point according to the channel number. After traversing all channels, the steering vector table covering all working frequency points is obtained. The single-channel sequential playback mode is adopted instead of the simultaneous multi-channel inversion mode, so that the individual differences and installation deviations of the array can be directly reflected in the steering vector table on site, and the cumulative error introduced by manufacturing tolerance or loosening of fixing parts is avoided.

[0054] After obtaining the multi-channel cross-spectrum matrix and the steering vector table, the calculation and control module independently solves the channel coefficient set for each working frequency point. The specific steps are as follows: Cholesky decomposition is performed on the multi-channel cross-spectrum matrix of the frequency point to obtain a lower triangular matrix, and the forward substitution and back substitution are used to sequentially solve the linear equation set, so that the solution vector and the steering vector are consistent in the equation. Then, the complex inner product of the steering vector and the solution result is calculated, and each element of the solution result is divided by the result of the inner product at the same time, so that the response of the canopy reference position at the working frequency point is the unit response. After performing the above steps on all working frequency points, the complete channel coefficient set is obtained. The direct benefit of using the unit response normalization is to strictly align the energy main lobe to the canopy reference position, and to calibrate the amplitude of the multi-channel directional driving sequence to an engineering scale that is easy to review. Subsequent online confirmation only needs to compare the deviation between the complex response and the unit response.

[0055] Reference Figure 5 and Figure 6 , Figure 5The sound-light synergistic promotion timing control chart fully shows the sound-light synergistic control strategy and timing arrangement of the application in a 24-hour light period. The horizontal coordinate represents the time axis, ranging from 0:00 to 24:00, marked at 4-hour intervals, covering the complete day-night cycle; the vertical coordinate is divided into three independent control channels, from top to bottom, the light intensity channel, the spectral formula channel and the acoustic stimulation channel, each channel is arranged vertically to facilitate observation of the synergistic control relationship. The light intensity channel adopts a step function form, maintaining a low level state from 0:00 to 4:00, indicating the dark period; a jump to a high level occurs at 4:00, marking the establishment of the light-on time, i.e. the phase zero point; the high level state lasts until 20:00, corresponding to a 16-hour light period; after 20:00, it falls back to a low level, entering an 8-hour dark period. The spectral formula channel shows the conventional spectral formula with a solid line and the formula switching transition section with a dashed line. The figure shows that at the start time of the acoustic stimulation window, the spectral formula is switched from the conventional mode to the specific formula, and the switching process uses a 200-millisecond slope transition, and the conventional formula is restored at the end of the window. The acoustic stimulation channel shows the activated stimulation window with a black rectangular block, and the figure shows four typical windows: the first window is located 10 minutes after the light is turned on (4:10), with a duration of 30 minutes; the second window is located at 10:00, with a duration of 40 minutes; the third window is located at 14:00, with a duration of 30 minutes; the fourth window is located at 17:00, with a duration of 20 minutes. The selection of each window follows the score and efficiency calculation results of the phase response curve mapping table to ensure that the target cumulative score is reached with the minimum cumulative duration. The timing control chart clearly reflects the core control strategy of the application: by precisely positioning the acoustic stimulation window in the sensitive period of plant physiological response and synchronously switching the spectral formula to form a sound-light coupling effect, precise regulation of plant growth and development is achieved. The timing arrangement uses a minute-level phase grid to ensure control accuracy; reasonable spacing between windows avoids stimulation superposition effect; the sound-light synchronous switching mechanism ensures the maximization of the synergistic promotion effect.

[0056] Figure 6The figure is a curve graph of the crown layer sound pressure level response experiment, which records the sound pressure level time domain characteristics of the acoustic executive array in the actual operation process and its distribution law at different measurement positions in detail. The abscissa represents time, ranging from 0 to 60 minutes, with 10 minutes as an interval for scale annotation; the ordinate represents the sound pressure level, with units of decibels (dB), and the measurement range is 40 to 80 dB, with 10 dB as an interval for scale division. The figure contains three curves: the solid line represents the sound pressure level response of the crown layer reference position, the dashed line represents the sound pressure level response of the array reference position, and the horizontal dashed line represents the ambient noise baseline. From the time domain response characteristics, it can be observed that in the initial stage of 0 to 10 minutes, the sound pressure levels at the two measurement positions are maintained near the ambient noise level, about 45 dB. At the time of 10 minutes, marked by the vertical dashed line "stimulus start", the acoustic executive array starts the multi-channel directional driving sequence, and the sound pressure levels at the two measurement positions rise rapidly. The sound pressure level at the crown layer reference position rises rapidly to 70 dB after the stimulus starts, and then fluctuates slightly in the range of 68-72 dB in the steady state stage of 10 to 50 minutes, indicating that the directional sound field forms a stable acoustic dose delivery at the target position. The sound pressure level at the array reference position rises to 78 dB and maintains in the range of 76-80 dB in the steady state stage, and its sound pressure level is always about 8-10 dB higher than that at the crown layer reference position, which reflects the geometric diffusion loss and air absorption attenuation in the sound wave propagation process. At the time of 50 minutes, marked by the vertical dashed line "stimulus end", the acoustic stimulus stops, and the sound pressure levels at the two measurement positions quickly fall to the ambient noise baseline level. The entire response curve presents an obvious trapezoidal envelope feature, with the rising edge and falling edge time being less than 1 minute, indicating that the system has good dynamic response characteristics. The sound pressure level fluctuation amplitude in the steady state stage is not more than ±2 dB, which proves the effectiveness of the multi-channel inter-spectrum matrix solution and the channel coefficient set optimization, and realizes the stable directional sound field construction in the complex acoustic environment. The experiment curve fully verifies that the invention can maintain a stable and controllable acoustic dose at the crown layer reference position through the closed-loop pointing consistency correction mechanism, providing a reliable physical basis for plant sound-light synergistic growth.

[0057] After the set of channel coefficients is determined, the multi-channel directional driving sequence is generated. An audio buffer with the same length as the playing time is allocated for each channel. Each sample point in the buffer is traversed and the sinusoidal value of the channel at the sample point is calculated sequentially according to the sequence of working frequencies, where the phase is determined by the phase of the Schroeder multi-tone and the phase of the channel in the set of channel coefficients, and the amplitude is determined by the amplitude of the channel in the set of channel coefficients. The contributions of all working frequencies are superimposed to obtain the driving value of the channel at the sample point. To avoid single-channel peak over-limit, amplitude normalization is performed on the buffer of each channel to limit the peak within the range allowed by the device, and then all channels are played back synchronously. During synchronous playback, the crown reference position is sampled at the same sampling rate to obtain feedback. The actual complex response is calculated at each working frequency and compared with the unit response to obtain the amplitude deviation and the phase deviation. If the deviation of any frequency point exceeds the set threshold (for example, the amplitude deviation exceeds 10 within or the phase deviation exceeds 10 degrees within), only the solution process of the multi-channel cross-spectral matrix is repeated once for the frequency point to obtain the updated set of channel coefficients and immediately replace the channel parameters of the frequency point in the playback. Such a closed-loop pointing consistency correction can maintain the main lobe stable alignment with the crown reference position when the device is slightly displaced, the temperature changes or the pipeline airflow changes, without the need to recalculate the whole frequency points, thereby controlling the calculation delay and energy consumption.

[0058] In step 3, the computing and control module forms a phase resetting sequence based on the phase response curve mapping table and generates an acousto-optic coupling plan within the phase resetting sequence. The implementation starts from the establishment of a time reference. Taking the light-on moment as the phase zero point, the complete light period is divided into 1440 equally spaced phase grids, each corresponding to 1 minute. The phase response curve mapping table is an ordered list of length 1440, each element being the response score of that minute, the score being derived from the standardized processing of empirical data or public data of the target crop and the current growth stage. Then a set of candidate windows is constructed, and the duration set is suggested to use 5 minutes, 10 minutes, 15 minutes, 20 minutes, 30 minutes, 45 minutes, and 60 minutes. For each starting minute and each duration, the sum of scores covered by the window is calculated, which is the score of the window; then the score is divided by the duration to get the efficiency value of the window. Sort all candidate windows by efficiency value from high to low to form a candidate window sequence. After setting the target cumulative score threshold, start from the beginning of the candidate window sequence and retrieve the windows one by one. As long as the current window does not overlap with the selected window in phase, add the window to the selected set and accumulate its score; stop retrieving when the cumulative score reaches or exceeds the target cumulative score threshold. If the cumulative score is still insufficient after traversing to the end of the sequence, expand the duration set by adding 75 minutes and 90 minutes, and recompute the score and efficiency before performing the same selection process. Sort the selected set by starting minute from small to large to get the phase resetting sequence. This greedy sorting and non-overlapping selection ensures that high-score and high-efficiency windows are prioritized into the phase resetting sequence, using shorter cumulative duration to exchange for sufficient cumulative score; this is crucial for long-term energy consumption and equipment thermal load, allowing the acoustic stimulus baseline to function in the truly high-sensitivity time period, while remaining silent in the low-sensitivity time period.

[0059] After the phase resetting sequence is determined, an acousto-optic coupling plan is generated for each window. At the window start minute, trigger the playback of the acoustic stimulus baseline, with a playback duration consistent with the window duration; at the same time, issue a spectral recipe to the lighting side, which remains unchanged throughout the window to form stable acousto-optic coupling. When the window end minute arrives, perform a 200 millisecond fade-out and restore the regular spectral recipe. If the interval between adjacent windows is less than 2 minutes, to avoid leaf surface temperature fluctuations or stomatal conductance abnormally rising due to the superposition effect, insert a 30-second low-intensity acoustic stimulus baseline transition segment between the two windows, with an amplitude reduced to 50 of the regular mode, as a transition. The reason for using minute-level phase grids and window score method is that the stomatal response of plants to acoustic stimulation and the light morphogenesis effect of spectral recipes are most significant at the minute level; minute-level discretization not only facilitates implementation on industrial controllers, but also allows the phase resetting sequence to remain comparable under different schedules.

[0060] In step 4, the computing and control module orchestrates and schedules the multi-channel directional driving sequence and the phase resetting sequence, and ensures the phase-aligned acousto-optic synergistic stimulation at the canopy reference location through online validation and correction. The specific approach is as follows: First, the phase resetting sequence is unfolded into a set of time-stamped actions, including starting the acoustic stimulation baseline at a certain minute, switching the spectral recipe at a certain minute, ending the current window and restoring the regular spectral recipe at a certain minute. The set of actions is combined with the playback control of the multi-channel directional driving sequence into a time axis, and the timing base is set to 1 millisecond. 30 seconds before the start of each window, the corresponding length of the multi-channel directional driving sequence is pre-buffered locally to avoid data stream interruption during the window. For the case where the single window duration exceeds 60 minutes, the window is split into several sub-segments not exceeding 30 minutes, and a 200-millisecond fade-in and fade-out transition is inserted at each sub-segment boundary to prevent the transient peak caused by buffer flushing. Online validation is performed at a fixed rhythm, and it is recommended to take a 1-second monitoring segment every 5 seconds. The complex response is calculated at each operating frequency point, and compared with the unit response to obtain the amplitude deviation and phase deviation. If the deviation exceeds the online validation threshold, a fast solution is performed only for that frequency point, the entry of that frequency point in the channel coefficient set is updated, and the replacement is completed without stopping the playback. The rhythm of the monitoring segment is selected to be once every 5 seconds, in order to capture the low-frequency disturbance caused by the fan speed change, pump start-stop and material flow passing through without significantly increasing the computational load. During the online validation process, if the absolute sound level at the canopy reference location is found to exceed the permitted threshold, the amplitudes of all channels are immediately reduced by 20 and the window is continued to be executed; if there is no valid data returned in three consecutive monitoring segments, the current channel coefficient set is kept unchanged until the end of the window, and the event is recorded in the execution summary. The above strategy not only guarantees the safety boundary, but also avoids the forced interruption of the window due to short-time data missing.

[0061] To adapt to different greenhouse layouts and crop types, several optional implementations are provided. One is to lower the lower bound of the target frequency table to 80 Hz and raise the upper bound to 12000 Hz, while raising the sampling rate to 96000 samples per second and the segment window length to 4096 samples. This way, finer frequency band resolution can be obtained in large cultivation layers, suitable for scenarios with canopy height exceeding 600 mm. Two is to raise the size of the working frequency sequence from 32 to 48 and the length of the multi-tone detection sequence from 2.0 seconds to 3.0 seconds, which can provide more stable statistics for the channel coefficient set in complex cavities. Three is to add a short window with a duration of 15 minutes in the phase reset sequence, arranged at the 10-minute and 60-minute positions before noon after turning on the light, and to raise the blue light component to 120% of the regular value in the spectral recipe of this window for quickly opening the stomata and stabilizing the position of the transpiration peak. Four is to add an auxiliary position for self-checking between the array reference position and the canopy reference position, which performs a self-checking playback without the participation of plants once a day at dawn, checks whether the amplitude relationship between the diagonal and off-diagonal elements of the multi-channel cross-spectral matrix is stable, and if the relative change exceeds 15%, forces the update of the complete channel coefficient set before the first window of the day.

[0062] Figure 4The figure clearly shows the significant promotion effect of the sound-light synergistic promotion control device on the biomass accumulation of plants. The horizontal axis represents the treatment days, ranging from 0 to 30 days, marked at intervals of 5 days; the vertical axis represents the biomass growth rate, expressed in percentage, ranging from 0 to 100%, scaled at intervals of 25%. The figure contains two growth curves: the solid line represents the biomass growth trend of the sound-light synergistic treatment group, and the dashed line represents the biomass growth trend of the control group. From the curve trend, it can be observed that the biomass growth rate of both groups shows an increasing trend with time, but the growth rate of the sound-light synergistic treatment group is significantly higher than that of the control group. Specifically, at the 5th day of treatment, the biomass growth rate of the sound-light synergistic treatment group reached about 12.5%, while the control group was only about 6.25%; at the 15th day of treatment, the biomass growth rate of the sound-light synergistic treatment group reached about 42.5%, while the control group was about 30%; at the end of the 30-day experiment, the biomass growth rate of the sound-light synergistic treatment group reached about 95%, while the control group only reached about 60%. Through comparative analysis, it can be known that during the entire 30-day experimental period, the sound-light synergistic treatment group showed a sustained and stable growth advantage over the control group. It is particularly noteworthy that as the treatment time increases, the difference between the two groups shows a gradually widening trend, indicating that the sound-light synergistic growth effect has a cumulative characteristic. At the end of the experiment, the biomass growth rate of the sound-light synergistic treatment group was 35 percentage points higher than that of the control group, with an increase of 58.3%, fully demonstrating that the device can significantly promote plant growth and development by precisely controlling the synergistic effect of acoustic stimulation and light spectrum, and has important application value.

[0063] The above has introduced the present application in detail. The principles and implementation modes of the present application are described by applying specific examples in this paper. The above examples are only used to help understand the method of the present application and its core idea. It should be noted that for ordinary skilled persons in the technical field, without departing from the principles of the present application, the present application can be improved and modified in several ways, and these improvements and modifications also fall within the protection scope of the claims of the present application.

Claims

1. A plant sound-light synergistic growth promotion control device, characterized in that, The device comprises: a computing and control module, a spectrum driving module, a photoperiod controller, an acoustic execution array, and a sensing acquisition module electrically connected to the computing and control module; the computing and control module is configured to execute a synergistic stimulation control process, comprising: Step 1: generating a target frequency point table based on blue noise Poisson disc sampling in a logarithmic frequency domain, and assigning a Schroeder multitone phase to the generated target frequency point table to synthesize a multitone detection sequence with low peak characteristics, playing the multitone detection sequence through the acoustic execution array and collecting the response by the sensing acquisition module, thereby determining an acoustic stimulation baseline; Step 2: constructing a multichannel cross-spectral matrix based on the ambient signals of each channel of the acoustic execution array, and constructing a pointing vector table in combination with the response collected at the canopy reference position, solving the multichannel cross-spectral matrix and the pointing vector table to calculate a set of channel coefficient sets, and generating a multichannel directional driving sequence for forming a stable directional sound field at the canopy reference position using the calculated channel coefficient sets; Step 3: based on a phase response curve mapping table matched with the target crop, selecting a set of windows that are phase-unoverlapped and meet the target cumulative score by calculating the score and efficiency of the candidate window, forming a phase reset sequence with the minimum cumulative duration, and generating an acoustic-optical coupling plan within the phase reset sequence; Step 4: scheduling and dispatching the execution of the acoustic execution array and the spectrum driving module according to the generated multichannel directional driving sequence and the phase reset sequence, and achieving phase-aligned acoustic-optical synergistic stimulation at the canopy reference position through online confirmation and correction; the computing and control module in Step 1 generates the target frequency point table, which is specifically configured to: generate a set of frequency points on the logarithmic frequency axis using the dart-throwing method of blue noise Poisson disc sampling, wherein candidate frequency points are randomly generated within the range of the logarithmic frequency axis, and if the logarithmic distance between the candidate frequency point and any point in the point set is greater than or equal to a minimum logarithmic interval, the candidate frequency point is accepted and added to the point set, until the active list is empty and the maximum logarithmic gap between any adjacent points is less than 2 times the minimum logarithmic interval, the generation is ended; the computing and control module in Step 1 synthesizes a multitone detection sequence with low peak characteristics, which specifically includes: calculating a peak evaluation value of the multitone detection sequence, and if the peak evaluation value is higher than a target upper limit, performing a phase refinement loop, which includes overall rotation of a working phase table and fine tuning of each phase in the working phase table in sequence to reduce the peak evaluation value.

2. The plant sound-light synergistic growth promotion control device according to claim 1, characterized in that, the computing and control module in Step 2 constructs a multichannel cross-spectral matrix, which specifically includes: in the silent environment segment, the sensing acquisition module uses several segments of 2048 samples with 50% overlap to collect the ambient signals of each channel; performing fast Fourier transform on each segment of ambient signals, and calculating the average value of the complex product across all segments at each working frequency point for any two channels to form the cross-spectral value of the working frequency point; organizing the cross-spectral values of all channel pairs into a matrix according to the channel index.

3. The plant sound-light synergistic growth promoting control device according to claim 2, characterized in that, The calculating and controlling module calculating the channel coefficient set in step 2 specifically comprises: performing Cholesky decomposition on the multi-channel cross-spectrum matrix of each working frequency point to obtain a lower triangular matrix; solving a linear equation set through forward substitution and back substitution; calculating a complex inner product of the steering vector and the solution, and dividing each element of the solution by the calculated complex inner product at the same time, so that the response of the canopy reference position at the working frequency point is a unit response.

4. The plant sound-light synergistic growth promoting control device according to claim 3, characterized in that, The calculating and controlling module in step 2 is further configured to perform closed-loop pointing consistency correction, specifically comprising: calculating an amplitude-phase deviation between the actual complex response and the unit response at the canopy reference position for each working frequency point, and repeating the solution process for this frequency point to update the channel coefficient set when the amplitude-phase deviation of any frequency point exceeds the set threshold.

5. The plant acousto-optic synergistic growth control device of claim 1, wherein, The calculating and controlling module in step 3 specifically comprises: dividing a complete photoperiod into 1440 equally spaced phase grids before forming the phase reset sequence; for each phase grid and a set of preset durations, calculating a score sum in the phase grid covered by a window as a window score, and dividing the window score by the window duration to obtain an efficiency value; sorting all windows by efficiency value from high to low to form a candidate window sequence.

6. The plant sound-light synergistic growth promoting control device according to claim 5, characterized in that, The calculating and controlling module in step 3 specifically comprises: starting from the beginning of the candidate window sequence, if the current window does not overlap with the selected window in phase, adding the current window to a selected set and accumulating its score, and terminating the traversal when the cumulative score reaches or exceeds a target cumulative score threshold to form the phase reset sequence.

7. The plant acousto-optic synergistic growth control device of claim 1, wherein, The calculating and controlling module in step 4 specifically comprises: the sensing and collecting module collects a monitoring segment of 1 second length at the canopy reference position at 5 second intervals; calculating a complex response at each working frequency point on the monitoring segment, and when the amplitude deviation or phase deviation of the calculated complex response exceeds the preset online confirmation threshold, calling the solution process in step 2 to perform a quick solution for the current frequency point and updating the channel coefficient set.

8. The plant acousto-optic synergistic growth control apparatus of claim 1, wherein, The calculating and controlling module is further configured to perform fault safety and recovery, specifically comprising: when it is monitored that the drive buffer of any channel is underloaded, suspending scheduling and refilling the buffer; when it is monitored that the absolute sound level of the canopy reference position exceeds a permission threshold, reducing the amplitude coefficient of all channels; when the sensing and collecting module does not return valid data in three consecutive monitoring segments, keeping the current channel coefficient set unchanged and continuing to execute.

Images