Wild-imitating three-dimensional intelligent cultivation system and method for ganoderma lucidum based on Internet of Things

Abstract

The invention discloses a ganoderma lucidum wild-imitating three-dimensional intelligent cultivation system and method based on the Internet of Things, and relates to the technical field of agricultural planting. The system comprises a three-dimensional cultivation frame module, wherein the three-dimensional cultivation frame module adopts a multi-layer adjustable structure to realize efficient space utilization; the Internet of Things sensor network is distributed on each cultivation layer to monitor parameters such as COconcentration, trace element content, hypha activity and the like in real time; the wild environment simulation module restores wild growth conditions through an adjustable spectrum LED and a temperature and humidity control unit; the intelligent decision and control center processes data based on an AI algorithm and outputs a control instruction; and the user interaction interface realizes state monitoring. The method comprises the steps of building a growth situation evaluation index system, collecting environment and image data, generating a regulation and control strategy through AI analysis, automatically adjusting environment parameters and continuously optimizing. According to the scheme, through closed-loop regulation and control of environmental parameters and biological states, the ganoderma lucidum yield and the active ingredient content are effectively increased, and meanwhile the land resource utilization efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of agricultural planting technology, specifically to an intelligent, three-dimensional, simulated wild cultivation system and method for Ganoderma lucidum based on the Internet of Things. Background Technology

[0002] Reishi mushroom, a traditional and precious Chinese medicinal herb, holds an important position in the field of traditional Chinese medicine, with its market demand continuously growing and its application scope expanding. In modern agricultural technology, reishi cultivation has gradually evolved from traditional forest understory planting to greenhouse cultivation, using artificially constructed planting environments to control certain environmental parameters. Current technologies include some planting bases using IoT sensor networks to monitor basic environmental parameters such as temperature and humidity, combined with automated equipment for environmental regulation; simultaneously, three-dimensional cultivation technology has been initially applied in the field of edible fungi cultivation, increasing the planting capacity per unit area through multi-layered structures. These technical solutions typically include environmental monitoring modules, basic control equipment, and simple feedback control systems, enabling periodic data collection and threshold-triggered adjustments of the planting environment, forming a basic environmental management process.

[0003] However, in the existing technology, the regulation of environmental parameters and the monitoring of biological growth status in the cultivation of Ganoderma lucidum are always disconnected, making it impossible to achieve dynamic environmental optimization based on biological activity feedback. This makes it difficult to accurately match the physiological needs of Ganoderma lucidum at each growth stage during the cultivation process. Summary of the Invention

[0004] This invention provides an IoT-based intelligent three-dimensional cultivation system and method for simulated wild Ganoderma lucidum, which can solve the technical problem of the disconnect between environmental control and biological growth state in traditional Ganoderma lucidum cultivation. To achieve the above objective, this invention provides the following technical solution:

[0005] This invention provides an IoT-based simulated wild three-dimensional intelligent cultivation system for Ganoderma lucidum, comprising:

[0006] Three-dimensional cultivation rack module: It adopts a multi-layer adjustable structure, with each layer equipped with an independent substrate cultivation trough and environmental control device;

[0007] Internet of Things (IoT) sensor network: distributed across each cultivation layer, used to monitor CO2 concentration, trace element content, mycelial activity, temperature and humidity, and light intensity parameters in real time;

[0008] Wild-like environment simulation module: including an adjustable spectrum LED lighting unit, a temperature and humidity control unit, and a matrix control unit;

[0009] Intelligent Decision and Control Center: Processes sensor data based on AI algorithms and outputs control commands to the execution devices;

[0010] User interface: used for system status monitoring and parameter settings.

[0011] In one alternative embodiment, the three-dimensional cultivation rack module adopts a layered and rotatable design, which can automatically adjust the angle according to the growth stage of Ganoderma lucidum to ensure uniform light exposure.

[0012] In one alternative embodiment, the Internet of Things sensor network includes a mycelial activity monitoring sensor that assesses the mycelial growth status in real time using near-infrared spectroscopy.

[0013] In one alternative embodiment, the AI ​​algorithm is an optimization model based on artificial neural networks and genetic algorithms, which can dynamically adjust environmental parameters according to historical data and real-time monitoring values.

[0014] In one optional embodiment, the system supports remote monitoring and management, and uses a cloud server to perform data storage and analysis functions.

[0015] A second aspect of the present invention provides a method for intelligent, three-dimensional, simulated wild cultivation of Ganoderma lucidum based on the Internet of Things, comprising the following steps:

[0016] A growth status assessment index system for Ganoderma lucidum was constructed, including mycelial expansion rate, cap morphology, cap area, and the proportion of fruiting bodies.

[0017] Data and image data of the growth environment of Ganoderma lucidum are collected through an Internet of Things (IoT) sensor network;

[0018] Data is analyzed using AI algorithms to generate environmental control strategies;

[0019] The system automatically adjusts light, temperature, humidity, CO2 concentration, and nutrient solution supply according to the strategy.

[0020] Continuously monitor and optimize system parameters to achieve high-quality and high-yield production of Ganoderma lucidum.

[0021] Compared with the prior art, the beneficial effects of the present invention are:

[0022] 1) This invention provides an IoT-based simulated wild three-dimensional intelligent cultivation system and method for Ganoderma lucidum. The solution achieves efficient space utilization through a multi-layer adjustable structure of the three-dimensional cultivation rack module. Each layer of independent substrate cultivation trough and environmental control device ensures differentiated management of different growth levels, thereby avoiding resource waste and increasing yield per unit area.

[0023] 2) Based on the Internet of Things sensor network distributed in each cultivation layer, CO2 concentration, trace element content, mycelial activity, temperature, humidity and light intensity parameters are monitored in real time. This design improves the limitations of traditional single parameter monitoring, comprehensively covers key factors for Ganoderma lucidum growth, and significantly improves environmental response speed and monitoring dimensions.

[0024] 3) Utilizing the adjustable-spectrum LED lighting unit, temperature and humidity control unit, and matrix regulation unit included in the simulated wild environment module, the system highly replicates the wild growth conditions of Ganoderma lucidum, effectively promoting the biosynthesis of active ingredients such as polysaccharides and triterpenes. Through an intelligent decision-making and control center that processes sensor data using AI algorithms and outputs control commands to the execution equipment, this mechanism replaces manual experience to achieve dynamic environmental regulation, enhancing the system's adaptability and ensuring that Ganoderma lucidum is in the optimal environmental state at each growth stage.

[0025] 4) The user interface is used for system status monitoring and parameter setting, enabling operators to monitor the growth status in real time and make necessary interventions. This solution constructs a closed-loop control process of "perception-analysis-decision-execution-feedback". By constructing a Ganoderma lucidum growth status evaluation index system including mycelial expansion speed, cap morphology, cap area and fruiting body ratio, it achieves multi-dimensional quantitative evaluation of the growth status.

[0026] 5) Environmental and image data collected by IoT sensor networks provide comprehensive input for regulation; AI algorithms analyze the data to generate environmental regulation strategies, ensuring that regulation is forward-looking and accurate; the system automatically adjusts light, temperature, humidity, CO2 concentration and nutrient solution supply according to the strategy to achieve refined management of environmental parameters; the system continuously monitors and optimizes system parameters to form a dynamic optimization cycle, ultimately achieving the goal of synergistic improvement of high quality and high yield of Ganoderma lucidum.

[0027] 6) This invention solves the core problem of the disconnect between environmental control and biological growth by deeply integrating three-dimensional planting, IoT sensing and AI decision-making, and shows significant advantages in improving the quality stability of Ganoderma lucidum, increasing yield per unit area and optimizing resource utilization efficiency. Detailed Implementation

[0028] The technical solutions in the embodiments of the present invention will be clearly and completely described below. 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.

[0029] Example 1:

[0030] Traditional Ganoderma lucidum cultivation suffers from problems such as unstable quality, low yield, poor land utilization, and insufficient precision in environmental control, making it difficult to achieve both high yield and high quality. Specifically: greenhouse cultivation, due to its highly homogenized environment, cannot simulate the light gradient, microclimate differences, and substrate heterogeneity under wild conditions, leading to inhibited synthesis of active ingredients such as polysaccharides and triterpenes; forest cultivation, while approximating the wild state, is affected by uncontrollable natural factors such as season, rainfall, and pests, resulting in large fluctuations in fruiting body formation rate and extremely low yield per unit area; planar artificial cultivation is limited by its two-dimensional spatial layout, resulting in significant waste of land resources, and each spawn bag is in a similar environment, making it impossible to differentiate temperature, humidity, CO2, and spectral composition according to growth stages, leading to redundant consumption of nutrients and energy. These problems collectively result in poor consistency of Ganoderma lucidum products, inconsistent content of effective components, and insufficient stability in large-scale production, hindering its high-quality application in the modern traditional Chinese medicine industry.

[0031] The present invention proposes the following:

[0032] A smart, three-dimensional, simulated wild cultivation system for Ganoderma lucidum based on the Internet of Things includes:

[0033] Three-dimensional cultivation rack module: It adopts a multi-layer adjustable structure, with each layer equipped with an independent substrate cultivation trough and environmental control device;

[0034] Internet of Things (IoT) sensor network: distributed across each cultivation layer, used to monitor CO2 concentration, trace element content, mycelial activity, temperature and humidity, and light intensity parameters in real time;

[0035] Wild-like environment simulation module: including an adjustable spectrum LED lighting unit, a temperature and humidity control unit, and a matrix control unit;

[0036] Intelligent Decision and Control Center: Processes sensor data based on AI algorithms and outputs control commands to the execution devices;

[0037] User interface: used for system status monitoring and parameter settings.

[0038] This embodiment provides a three-dimensional intelligent cultivation system for Ganoderma lucidum that integrates spatial reconstruction, multi-source sensing, environmental simulation, intelligent inference, and human-machine collaboration. Its core lies in achieving three-layer coupling of physical space, information space, and decision-making space through a modular architecture: the three-dimensional cultivation rack module constructs a vertical physical carrier; the Internet of Things sensor network forms a full-dimensional sensing neural network; the simulated wild environment module acts as a precise execution effector; the intelligent decision-making and control center undertakes central computing and strategy generation functions; and the user interface constitutes the operational entry point for the human-machine closed loop. This system does not rely on a single empirical model or fixed program, but is driven by real-time dynamic data to reproduce the spatiotemporal heterogeneous environment required for the wild growth of Ganoderma lucidum within a limited space, thereby supporting the orderly expression of its physiological rhythms and secondary metabolism.

[0039] The three-dimensional cultivation rack module adopts a multi-layer adjustable structure, meaning it has a rigid support frame with adjustable number of layers (typically 4–6 layers), adjustable layer height (the distance between adjacent layers can be adjusted electrically or manually within the range of 30–80 cm), and adaptable load-bearing capacity (single layer load-bearing capacity ≥50 kg). Each layer of the independently set substrate cultivation tank is a rectangular or arc-shaped tank made of food-grade polypropylene (PP) or modified polycarbonate (PC). The inner wall has a microporous hydrophobic coating to balance air permeability and water retention. The bottom of the tank integrates a permeable screen and a guide channel to facilitate the recovery of excess nutrient solution. The environmental control device is an embedded micro-actuator, including a micro-atomizing nozzle (orifice diameter ≤50 μm), a micro CO2 micro-valve (diameter Φ2 mm), and a micro temperature control patch (PTC ceramic heating / TEC semiconductor cooling composite structure), all of which are linearly arranged along the long side of the cultivation tank and form an integrated assembly relationship with the tank body. This structure can be replaced with a cantilever telescopic tray structure, or a pneumatic lifting linkage mechanism can be used to replace the electric screw adjustment. Alternatively, the cultivation trough can be replaced with a detachable honeycomb bag holder to accommodate different sizes of bags.

[0040] The IoT sensor network is distributed across each cultivation layer, with sensor nodes deployed in a "one node per tank" or "two tanks sharing one node" configuration. The node shells are made of IP67-rated, moisture- and bacteria-resistant ABS engineering plastic, and contain multi-module sensor chips. The monitored CO2 concentration parameter is acquired using non-dispersive infrared (NDIR) technology, with a range of 0–5000 ppm and a resolution of 1 ppm. Trace element content parameters specifically refer to the ion activity values ​​of nitrogen (N), phosphorus (P), potassium (K), selenium (Se), and zinc (Zn) in the matrix, simultaneously detected by an ion-selective electrode (ISE) array. Mycelial activity parameters are non-destructive biological indicators, acquired using near-infrared spectroscopy (NIRS, wavelength range 900–1700 nm), which collects the mycelial layer reflectance spectrum. After preprocessing, the absorbance ratio of characteristic bands (e.g., 1200 nm / 1450 nm) is extracted. The activity criterion is (nm). Temperature, humidity, and light intensity parameters are acquired by a digital temperature and humidity sensor (SHT45) and a silicon photodiode light sensor (TSL2591), respectively, with the sampling frequency programmable from 1 to 60 minutes. This network can be replaced with LoRaWAN wireless networking instead of the default Zigbee protocol. The sensors can also be directly attached to the surface of the mushroom bags using a flexible electronic skin patch structure, or periodically inspected using a drone equipped with a multispectral camera to supplement blind spots in static deployment.

[0041] The simulated wild environment module includes an adjustable-spectrum LED lighting unit, a temperature and humidity control unit, and a matrix control unit. The adjustable-spectrum LED lighting unit consists of four types of LED chips—red (600-650 nm), blue (400-450 nm), far-red (720-750 nm), and white (correlated color temperature 2700-6500 K)—proportionally packaged on the same PCB substrate. The spectral ratio supports 0.1% step adjustment, with a light intensity output range of 0–2000 Lux and a response time ≤100 ms. The temperature and humidity control unit includes an ultrasonic humidifier (atomized particle size ≤5 μm), a Peltier (TEC) heat exchange module (temperature control range 15–35°C, accuracy ±1°C), and a variable-frequency axial flow fan (airflow adjustable from 0-100 m³ / h). These three components work together to maintain a stable microenvironment within the layer. The matrix control unit is a slow-release water-fertilizer coupling device, containing an osmotic pressure-driven nutrient solution micropump (flow rate 0.1-5... The module features adjustable flow rates (mL / min) and a pH / EC dual-parameter feedback control module, which delivers nutrient solution directionally to the deep substrate via capillary tubes. This module can be replaced with a UV-Vis broadband LED array instead of the four-band combination. The temperature and humidity unit can also adopt an evaporative cooling + heat pump composite system to improve energy efficiency. The substrate control unit can also integrate microbial agent slow-release capsules, which release antagonistic strains to inhibit miscellaneous bacteria through temperature / humidity triggering.

[0042] The intelligent decision-making and control center processes sensor data based on AI algorithms and outputs control commands to the execution devices. Its hardware platform is an edge computing terminal (typical configuration: ARM Cortex-A72 quad-core processor + 2 GB LPDDR4 memory + eMMC 16 GB storage), running a lightweight AI inference engine. The AI ​​algorithm takes sensor time-series data as input to construct a multi-dimensional state vector (including 12-dimensional features such as CO2 trend slope, mycelial activity fluctuation entropy, and temperature and humidity coupling deviation). After the artificial neural network (ANN) completes state recognition, the genetic algorithm (GA) searches for the optimal parameter combination within a preset constraint space (such as light intensity ≤1500 Lux, humidity ≤90% RH) and outputs low-level control signals such as PWM duty cycle, valve opening, and fan speed. The control commands are distributed to the execution devices at each layer via Modbus RTU or CAN bus. The command update cycle is synchronized with or doubled with the sensor sampling cycle. The center can be replaced by a cloud-based collaborative architecture, where feature extraction and rapid response are completed at the edge, while complex strategy optimization is handled by a cloud server GPU cluster. The AI ​​model can also use graph neural networks (GNNs) to model the interlayer environmental coupling relationship, or introduce Bayesian optimization to replace genetic algorithms to improve convergence efficiency.

[0043] The user interface is used for system status monitoring and parameter setting. Its software architecture supports both B / S and C / S modes: the web client is developed based on the Vue.js framework, providing a 3D visualized cultivation rack topology map (real-time rendering of temperature and humidity cloud maps and mycelial activity thermograms for each layer), historical data curve overlay analysis (supporting multi-parameter cross-comparison), and anomaly warning pop-ups (e.g., automatically highlighting a decrease in mycelial activity >15% for 3 consecutive hours); the mobile app is compatible with Android / iOS systems, integrating offline mode and a voice command parsing module (e.g., "Increase the red light ratio of the second layer to 75%"); all parameter settings undergo double verification (local logic verification + cloud rule base matching), and modifications automatically generate operation logs and trigger a system re-evaluation process. This interface can be replaced with an AR glasses projection-style interaction, locating each cultivation layer through spatial anchor points, and allowing gesture operations to retrieve corresponding parameters; it can also be expanded into a multi-tenant management interface, supporting tiered authorization access for three roles: planting base administrators, agricultural experts, and quality supervisors.

[0044] The modules described above are not isolated: the three-dimensional cultivation rack module provides a spatial deployment benchmark and physical isolation boundary for the sensor network, ensuring that environmental regulation between different layers does not interfere with each other; the data acquired by the sensor network constitutes the basis for the regulation of the simulated wild environment module, and its monitoring granularity (such as mycelial activity at the single-trough level) directly determines the spatial accuracy of light / temperature and humidity regulation; the execution effect of the simulated wild environment module, in turn, affects the sensor readings, forming a closed-loop feedback chain; the intelligent decision and control center, as a data hub, not only analyzes sensor inputs but also orchestrates and executes command outputs, while mapping process data to the user interface to achieve cognitive alignment; the user interface provides human intervention capabilities, offering a coverage mechanism when AI strategies fail or special agricultural operations (such as pre-harvest stress induction). The modules achieve loosely coupled integration through unified timestamps, standardized data protocols (JSONSchema defining field semantics), and event-driven buses (such as MQTT Topic hierarchical subscription).

[0045] Through the above technical solution, this invention achieves closed-loop control of the entire growth process of Ganoderma lucidum within a limited indoor space, characterized by spatial differentiation, temporal dynamics, and refined parameters. For example, in the early stage of mycelial culture, the system receives low light intensity (200 Lux) and stable temperature and humidity (25±1°C, 75±5% RH) through the lower cultivation tank, combined with a high-nitrogen substrate supply, to promote rapid mycelial spread. When the sensor network detects that the mycelial activity index of a certain layer exceeds the threshold (>0.85) and the CO2 accumulation rate increases, the intelligent decision center immediately triggers an environmental switch for that layer: the LED lighting unit increases the proportion of red light to 65%, the temperature and humidity control unit raises the humidity to 85% RH and starts micro-ventilation to reduce CO2 to 1000 ppm, and the substrate control unit simultaneously slows down the nitrogen supply while increasing the pulse injection of trace elements; the user interface simultaneously pushes a "Primordial Induction Period Start" prompt for technicians to confirm or fine-tune. The reason this process can solve the problems of lagging environmental control, mutual interference between levels, and slow human response in traditional cultivation is that the three-dimensional cultivation rack module provides the physical control basis, the Internet of Things sensor network provides the accuracy of state judgment, the simulated wild environment module provides the execution capability boundary, the intelligent decision and control center provides the timeliness of strategy generation, and the user interface provides the flexibility of human-machine collaboration. All five are indispensable and work together to promote the smooth transformation of Ganoderma lucidum from vegetative growth to reproductive growth, ultimately achieving the technical effect of synergistic improvement in high quality and high yield.

[0046] Example 2:

[0047] Based on the above embodiments, this embodiment further provides:

[0048] The three-dimensional cultivation rack module adopts a layered and rotatable design, which can automatically adjust the angle according to the growth stage of Ganoderma lucidum to ensure uniform light exposure.

[0049] The "layered rotatable design" refers to the fact that each cultivation layer of the three-dimensional cultivation rack is equipped with an independent rotation mechanism. This mechanism consists of a drive motor, a reducer, a rotation support shaft, and an angle feedback sensor. The rotational movements of each layer are decoupled and do not interfere with each other, allowing for differentiated angle control based on the growth stage of the Ganoderma lucidum at different layers. The rotation support shaft is arranged perpendicular to the bottom of the cultivation trough, with its axis coinciding with the center of the layer, ensuring the stability of the substrate and the center of gravity of the spawn bag during rotation and preventing slippage or overturning. The drive motor is a stepper motor or servo motor with high-precision positioning capabilities and a minimum rotation resolution of 0.5°. The angle feedback sensor is an absolute photoelectric encoder or a magnetic angle sensor, providing real-time feedback to the intelligent decision-making and control center on the actual tilt angle of the current layer, forming the basis of closed-loop control. This design is not limited to synchronous rotation of the entire layer; it can also support independent rotation of a single cultivation trough. For example, in a 4-layer architecture, if the 2nd and 3rd layers are in the fruiting body differentiation stage, while the 1st layer is still in the mycelial spreading stage, the system will only drive the 2nd and 3rd layers to rotate, while the 1st layer remains horizontally stationary, thereby achieving on-demand, zoned, and precise light adaptation.

[0050] "Based on the growth stages of Ganoderma lucidum" means that the system divides the growth process into five stages based on a preset model of the entire growth period of Ganoderma lucidum: mycelial germination stage, mycelial propagation stage, primordium induction stage, fruiting body differentiation stage, and fruiting body maturity stage. Each stage corresponds to different phototropic response intensities and photoperiod requirements: for example, during the mycelial propagation stage, the mycelium is not sensitive to light and mainly avoids light, with rotation paused or maintaining a 0° horizontal posture; while from the fruiting body differentiation stage to the maturity stage, it exhibits significant positive phototropism and needs to periodically rotate so that the edges of the cap are successively illuminated by the main light source to avoid morphological defects such as elongated stipe and skewed cap caused by continuous light exposure on one side. This stage model is dynamically corrected by a fusion of three criteria: time dimension (number of days), physiological indicators (such as the output value of the mycelial activity sensor reaching the threshold, and image recognition determining the appearance of primordia), and environmental parameters (temperature and humidity jump points), rather than being called in a fixed time sequence.

[0051] "Automatic angle adjustment" refers to the intelligent decision-making and control center receiving multi-source data from the IoT sensor network, calling the built-in stage identification module to determine the current growth stage of the Ganoderma lucidum, and looking up a table to obtain the recommended rotation strategy for that stage—including the starting angle, target angle, rotation rate (0.1°-2° / min), single rotation amplitude (5°-30°), rotation cycle (6-24 h / time), and whether to enable the reciprocating oscillation mode; then, a PWM control signal is generated to drive the rotation actuator of the corresponding layer; during the rotation, the angle feedback sensor continuously transmits the measured value, and the control center executes a PID correction algorithm to ensure that the actual angle deviation is ≤±0.8°; this automatic adjustment process does not require manual intervention and supports remote manual overwrite command input.

[0052] "Ensuring uniform illumination" does not mean that the illuminance at each point is absolutely consistent, but rather that within a set time window (such as 24 hours), the difference in the cumulative received photon flux density (PPFD) of each region on the cap projection surface is controlled within ±15%. This effect is achieved through the synergy of spatial rotation and temporal scheduling: on the one hand, rotation changes the angle between the cap normal and the incident light vector, causing the originally backlit area to periodically turn into the frontlit area; on the other hand, combined with the dynamic spectral adjustment of the LED lighting unit (such as increasing the proportion of red light during the fruiting body stage), the biological utilization efficiency of unit light energy is improved. The two work together to reduce the spatial non-uniformity index (SNI) of light distribution on the cap surface from ≥40% under static structure to ≤12%, which is significantly better than traditional fixed three-dimensional frames.

[0053] A clear control link relationship is formed between the components: the light intensity sensor and mycelial activity monitoring sensor in the Internet of Things sensor network provide the basis for stage judgment; the intelligent decision and control center completes stage identification and strategy generation; the rotary actuator receives instructions and acts; the angle feedback sensor constitutes a closed-loop verification link; the user interface can display the real-time tilt angle of each layer, historical rotation trajectory and the next planned action time, and supports abnormal status alarms (such as motor stall, angle deviation).

[0054] Through the above technical solutions, this invention achieves the following: Based on the multi-layer adjustable three-dimensional cultivation rack defined in this invention, a layered independent rotatable mechanism is introduced, allowing each cultivation layer to autonomously, precisely, and dynamically adjust its spatial orientation according to the actual development process of the Ganoderma lucidum it carries; because the rotation of each layer is driven by the growth stage rather than a fixed program, the ineffective energy consumption of "rotating for the sake of rotation" is avoided; because the rotation angle and rate are programmable and adjustable, it adapts to the differentiated light response characteristics of different Ganoderma lucidum varieties (such as Ganoderma lucidum preferring strong phototropism, while Ganoderma lucidum is relatively shade-tolerant); because closed-loop feedback control is adopted, the repeatability accuracy and long-term stability of angle adjustment are guaranteed; finally, because the light received by the cap tends to be balanced in all directions, morphological deviations are effectively suppressed, the roundness and marketability of the fruiting bodies are improved, and the synthesis and accumulation of secondary metabolites (such as ganoderic acids and polysaccharides) are promoted. Experimental data show that, compared with the control group without rotation function, this embodiment increased the first-grade product rate (cap diameter error ≤ ±8%, no lopsided crown, no deformity) from 63.2% to 89.7% within the same cultivation cycle, verifying the effectiveness of this technology in solving the quality deterioration problem caused by uneven light distribution in Ganoderma lucidum three-dimensional cultivation.

[0055] Example 3:

[0056] Based on the above embodiments, this embodiment further provides:

[0057] The Internet of Things (IoT) sensor network includes mycelial activity monitoring sensors that use near-infrared spectroscopy to assess mycelial growth status in real time.

[0058] This technical solution focuses on building a direct, dynamic, and non-invasive ability to perceive the physiological state of the organism itself—namely, the mycelium—during Ganoderma lucidum cultivation. Traditional Ganoderma lucidum cultivation relies on visual observation of lagging indicators such as mycelial spread speed, color changes, or substrate surface film formation. This approach cannot quantify mycelial metabolic activity, biomass accumulation rate, and early stress response, leading to significant time lags in environmental regulation. For example, when the mycelium has actually entered the nutrient depletion period but still appears vigorous, the system continues to supply a high-nitrogen substrate, which can induce contamination by other microorganisms. Furthermore, 24-48 hours before primordia differentiation begins, mycelial metabolic activity experiences a characteristic surge. Without accurate identification of this period, light, temperature, and humidity parameters cannot be switched to the fruiting body induction mode in time, directly affecting the uniformity of fruiting and the efficiency of effective component synthesis. The technical features defined in this embodiment address the aforementioned problem of inaccurate regulation caused by the "invisibility of biological processes," providing an in-situ physiological sensing pathway that can be embedded in existing IoT architectures and work in conjunction with environmental sensors.

[0059] The mycelial activity monitoring sensor refers to a type of optical sensing unit specifically designed for the quantitative analysis of the physiological state of Ganoderma lucidum mycelia. Its core is a near-infrared (780-2500 nm) light source-detector integrated module, including an emitter (such as a halogen tungsten lamp or broadband LED), a collimating optical system, a sample interaction cavity (adapted for embedded installation on the sidewall or top of the substrate cultivation trough), a reflected / transmitted light collection optical path, and a high-sensitivity InGaAs linear array detector. This sensor is independent of sampling, does not require damage to the mycelium pack, and can operate stably for extended periods in the high-humidity (≥75% RH) and low-light (≤1000 Lux) microenvironment of Ganoderma lucidum cultivation. Its structure supports multiple installation methods: it can be fixed to the inner side of each beam of a three-dimensional cultivation rack, illuminating the surface of the mycelium pack below at a 45° angle; alternatively, a retractable probe design can be used, driven by a robotic arm to periodically insert vertically into the substrate surface 2-3 times. Local scanning is performed at a distance of cm. In terms of materials, the optical window is made of quartz glass (resistant to moisture and heat, and resistant to organic solvent corrosion), and the outer shell is made of food-grade 304 stainless steel or PFA-coated aluminum alloy, meeting the cleanliness requirements of GAP planting environments. As an optional embodiment, this sensor can also be replaced with an alternative based on the principle of fluorescence lifetime imaging (FLIM), which indirectly reflects the redox state and energy metabolism level by exciting the intrinsic NADH fluorescence of hyphae and analyzing its decay time constant, thereby achieving multimodal cross-validation of hyphal activity.

[0060] Near-infrared spectroscopy is a technique that utilizes the absorption characteristics of overtone and combination vibrations in the near-infrared region to establish a mathematical mapping relationship between spectral response and mycelial physiological parameters. In practice, after acquiring raw spectral data, standard normal variable transformation (SNV) and first-order derivative preprocessing are performed to eliminate baseline drift and scattering interference. Then, using partial least squares regression (PLSR) or convolutional neural network (CNN) models, the intensity and peak shape changes of spectral characteristic peaks (such as the OH stretching vibration at 1450 nm and the combined CH and OH vibration at 1940 nm) are converted into four key activity indicators: mycelial biomass (g / L), relative water content (%), respiratory quotient (RQ), and β-glucan synthase activity index. Model training data comes from concurrent offline biochemical assays (such as the anthrone method for polysaccharide determination, the BCA method for protein determination, and the MTT colorimetric method for dehydrogenase activity determination). The calibration set covers the mycelial stage (0-15 days), primordia formation stage (16-22 days), and fruiting body expansion stage (23-45 days). d) Throughout the entire growth cycle, ensure the model's generalization ability across growth stages; the sampling frequency is synchronized with the IoT main network, set to 10 minutes / time by default, but can also be dynamically adjusted according to the growth stage; during the critical window of primordium induction (days 16-18), the frequency is increased to 2 minutes / time to capture the inflection point of active mutation. As an optional embodiment, this analysis technique can also employ a transfer learning strategy to adapt the already modeled Ganoderma lucidum spectral model through a small number of fine-tuned samples of target varieties (such as Ganoderma sinense), reducing the deployment threshold for new varieties.

[0061] The two aforementioned technical features constitute a closed-loop sensing chain: the mycelial activity monitoring sensor serves as the hardware carrier, providing the raw spectral signal input; near-infrared spectroscopy analysis technology is the algorithmic core, completing the interpretation from physical signals to biological meaning. Their synergistic effect is manifested in the following ways: the spatial arrangement density of the sensors (one per layer) ensures the independent characterization of mycelial state in each layer of the three-dimensional cultivation rack, avoiding data aliasing caused by inter-layer obstruction or airflow disturbance; and the "growth stage weighting factor" embedded in the spectral analysis model assigns differentiated interpretation logic to the same spectral feature at different developmental stages. For example, an increase in the intensity of the 1940 nm peak indicates sufficient nutrient reserves during the mycelial stage, while suggesting active cell wall remodeling during the primordium stage, thus supporting the AI ​​decision center in outputting differentiated control commands.

[0062] Through the above technical solution, real-time, non-destructive, and quantitative tracking of the physiological processes inside Ganoderma lucidum mycelium is achieved. By introducing mycelial activity as a core biological response variable, the system no longer relies solely on empirical regulation based on external environmental parameters (such as temperature and CO2 concentration), but can instead identify the actual metabolic needs of the mycelium: when spectral analysis shows that the mycelial expansion rate reaches a threshold (>0.8 mm / d) and the moisture content is stable at 62%±3%, the intelligent decision center automatically triggers a nutrient solution pulse supply for 15 minutes, with a dosage calculated at 5 mL per mycelial bag; when the respiratory quotient (RQ) is detected to be below 0.75 for three consecutive samples, indicating suppressed carbon metabolism, the system immediately reduces the CO2 concentration to 800 ppm and increases the ventilation frequency to avoid anaerobic stress; when the primordium induction period activity index shows a step increase of more than 20%, the system increases the red light ratio from 50% to 70% 12 hours in advance, simultaneously increasing the humidity from 75% RH to 85% RH. These clearly causal regulatory actions stem directly from the combined application of mycelial activity monitoring sensors and near-infrared spectroscopy analysis technology. This solves the problems of regulatory lag, misjudgment, and resource misallocation caused by the lack of mycelial state perception in the background technology, significantly improving the matching accuracy between environmental parameters and biological needs. It provides an irreplaceable data foundation for subsequent AI algorithms to generate high-confidence regulatory strategies.

[0063] Example 4:

[0064] Based on the above embodiments, this embodiment further provides:

[0065] The AI ​​algorithm is an optimization model based on artificial neural networks and genetic algorithms, which can dynamically adjust environmental parameters based on historical data and real-time monitoring values.

[0066] The technical problem addressed by this embodiment is that the single-rule control, fixed threshold feedback, or empirical adjustment methods used in traditional Ganoderma lucidum cultivation systems are difficult to adapt to the differentiated, nonlinear, and time-varying environmental parameters required by Ganoderma lucidum throughout its entire growth period (mycelial growth, primordium differentiation, and fruiting body development). The rigidity of the control strategy leads to improper light ratios, temperature and humidity fluctuations exceeding the physiological tolerance range, and CO2 supply lagging behind changes in metabolic rate, which in turn causes problems such as decreased mycelial vitality, low primordium formation rate, deformed fruiting bodies, or inhibited synthesis of effective components, ultimately restricting yield stability and quality consistency.

[0067] AI algorithms refer to the core computational logic module executed by the intelligent decision-making and control center. Their function is to map multi-source heterogeneous sensor inputs (including CO2 concentration, trace element content, mycelial activity, temperature, humidity, and light intensity) into an executable set of environmental control instructions. They do not rely on pre-set program scripts but achieve closed-loop adaptive decision-making through data-driven processes. AI algorithms are not general-purpose machine learning frameworks but rather a hybrid optimization architecture specifically tailored for modeling the physiological responses of Ganoderma lucidum.

[0068] Artificial Neural Network (ANN) serves as the feature learning and trend prediction unit, employing a three-layer feedforward structure (input layer with 12 nodes, corresponding to 6 types of sensors × 2 time steps; hidden layer: 16 ReLU activated neurons; output layer: 5 nodes, corresponding to target temperature, target humidity, target light intensity, target red-blue light ratio, and target CO2 concentration, respectively). It receives a continuous 10-minute sampling sequence in a sliding time window manner (i.e., each round of input contains a 60-dimensional vector), and outputs recommended settings for various environmental parameters within the next 5 minutes. Its training data comes from historical sensor data of Ganoderma lucidum at different growth stages and corresponding phenotypic observation labels (such as cap unfolding rate and fruiting body thickness). Weight updates use the Adam optimizer, with the learning rate initially set to 0.001 and decaying with each training round. The ANN does not directly participate in parameter optimization but instead constructs a nonlinear mapping relationship of "input state → ideal output," providing a high-quality initial solution space constraint for the genetic algorithm.

[0069] The Genetic Algorithm (GA) serves as the global search and policy evolution unit, performing multi-objective parameter combination optimization within the neighborhood of the recommended values ​​output by the ANN. Its population size is set to 50, and the encoding method is real-number encoding, with each individual being a 5-dimensional vector ([T, RH, Lux, R / B, CO2]). The fitness function comprehensively and weightedly evaluates three indicators: mycelial activity index (the ratio of biomass to redox state obtained from near-infrared spectroscopy), fruiting body morphological integrity score (the roundness of the cap and the clarity of its edges based on image recognition), and unit energy consumption output ratio (the dry weight yield corresponding to each kilowatt-hour of electricity). The selection operation adopts the tournament method, with a crossover probability of 0.85 and a mutation probability of 0.15. The asynchronous long-press parameter dimensions are adaptively scaled (e.g., temperature variation ±0.3°C, CO2 variation ±20 ppm). Each GA iteration calls the ANN prediction model to quickly evaluate new individuals, avoiding high-cost physical experimental verification, thus completing the generation of a single regulatory strategy in milliseconds.

[0070] The optimization model based on artificial neural networks and genetic algorithms emphasizes a collaborative mechanism of functional decoupling and process coupling between the two: ANN undertakes the cognitive function—extracting the growth response patterns of Ganoderma lucidum from massive historical data and establishing an environment-phenotype association model; GA undertakes the "decision-making" function—within the reasonable feasible domain defined by ANN, guided by multi-objective fitness, it searches for the Pareto optimal parameter combination under the current real-time operating conditions; the two avoid the defects of traditional single models, such as being prone to getting trapped in local optima, weak generalization ability, and high response latency, through a two-level mechanism of "ANN pre-screening + GA fine search"; this hybrid model is deployed on an edge computing gateway (such as NVIDIA Jetson AGX Orin), supports online incremental learning, and after each adjustment is executed, the system automatically collects the actual environmental response curve and subsequent 24-hour phenotypic change data, and sends them back to the ANN weight fine-tuning module, so that the model can continuously adapt to the characteristics of specific strains and local climate disturbances.

[0071] Historical data refers to the structured time-series dataset accumulated and stored in the local database or cloud server during system operation. It contains complete records for at least three consecutive cultivation cycles. Each record includes: timestamp, raw sensor readings at each layer, action logs of the executed equipment (such as LED switching times and humidifier start-stop durations), manually labeled growth stage tags (divided according to "GB / T 29371.2-2012 Ganoderma lucidum production technical specifications"), and physicochemical indicators (polysaccharide and triterpene content) detected after harvesting. This dataset is used for ANN offline pre-training and GA initial population generation after standardization (Z-score normalization).

[0072] Real-time monitoring values ​​refer to the latest raw sampling data acquired by the IoT sensor network within the current control cycle, without smoothing filtering. The sampling frequency is 10 minutes / time. After the data undergoes lightweight verification at the edge (removing abrupt outliers and interpolating data from short-term communication interruptions), it is directly input into the ANN inference engine. This data stream has strong timeliness, ensuring that GA optimization is always anchored to the real dynamic operating conditions, rather than static historical averages.

[0073] Dynamic adjustment of environmental parameters manifests as closed-loop control behavior: the system triggers a complete decision-making cycle every 10 minutes; real-time values ​​are collected → the ANN generates baseline suggestions → the GA searches for the optimal solution within a ±15% disturbance range → commands are output to actuators such as LED drivers, temperature control relays, and CO2 solenoid valves → sampling is performed again 10 minutes after execution, forming a minimum control cycle of "sensing-decision-execution-feedback"; this dynamism is not only reflected in the successive correction of parameter values, but also in the phased transitions of the control strategy. For example, when the ANN identifies a decrease in the mycelial activity index after three consecutive samplings and a blue shift in the near-infrared absorption peak (indicating a shift in metabolism towards secondary metabolism), the GA automatically increases the red light proportion to 65%-75% and simultaneously raises the target humidity value to 82%-86% RH. This strategy switching requires no manual intervention and does not rely on pre-programmed stage switching logic.

[0074] The various technical features form a tight causal link: the high-confidence trend prediction provided by ANN significantly compresses the search space dimension and iteration number of GA; the Pareto optimal solution output by GA verifies and strengthens the mapping accuracy of ANN; historical data ensures the completeness of the model's prior knowledge, and real-time monitoring values ​​give the model robustness in dealing with sudden disturbances (such as the instantaneous drop in LED illuminance caused by power grid voltage fluctuations); the two together support the core behavior of "dynamic adjustment"; it is neither a simple proportional adjustment nor an open-loop timing control, but a closed-loop optimization based on physiological feedback with memory and evolution.

[0075] Through the above technical solutions, precise, personalized, and adaptive control of the Ganoderma lucidum growth environment was achieved. For example, during the fruiting body expansion period, when real-time monitoring showed that the CO2 concentration in a certain layer suddenly dropped to 650 ppm (below the optimal range of 800–1200 ppm), and near-infrared spectroscopy showed that the mycelial activity index decreased by 12% simultaneously, the ANN immediately predicted that this CO2 deficiency would lead to a 18% reduction in the dry matter accumulation rate of the fruiting body. The GA then searched for the optimal CO2 supplementation strategy under the constraints of temperature 27.5–28.5°C, humidity 84–87% RH, red light ratio 70–75%, and light intensity 950–1050 Lux: starting the CO2 generator to supply gas at a constant flow of 0.8 L / min for 9 minutes to raise the concentration back to 980 ppm, while fine-tuning the LED blue light power by 5% to balance the photosynthesis / respiration ratio. After this strategy was implemented, the subsequent three samplings showed that the mycelial activity index recovered to 96% of the initial value, and the daily weight gain rate of the fruiting body recovered to 112% of the average value of the control group. This process fully demonstrates that relying solely on either ANN or GA algorithms cannot simultaneously achieve both response speed and global optimality. Only by integrating the two can the complex technical problems pointed out in the background technology, such as "rigid regulation, high energy consumption, and unsuitable growth," be solved. This will improve the level of intelligence and physiological adaptability of environmental regulation without increasing hardware investment, providing an algorithmic foundation for the stable acquisition of high-quality Ganoderma lucidum fruiting bodies.

[0076] Example 5:

[0077] Based on the above embodiments, this embodiment further provides:

[0078] The system supports remote monitoring and management, and uses cloud servers to perform data storage and analysis.

[0079] This embodiment focuses on the technical features of "remote monitoring and management" and "data storage and analysis supported by cloud servers." Its core lies in constructing a cross-regional, highly reliable, and scalable cloud-based collaborative architecture to overcome the inherent limitations of localized control systems in terms of operational response timeliness, multi-source data fusion capabilities, long-term trend modeling accuracy, and system disaster recovery resilience. This technical solution does not simply upload local data to a server; rather, it integrates multiple layers of technologies, including protocol adaptation, edge preprocessing, time-series database modeling, and lightweight AI service deployment, enabling the system to possess platform-level management capabilities for large-scale Ganoderma lucidum production bases.

[0080] "Remote monitoring and management" refers to the ability for users to perceive the status of the entire cultivation process, intervene in anomalies, and optimize strategies without being physically present on-site. Specifically, the system is configured with a lightweight MQTT / CoAP communication protocol stack at the edge. Data and image frames collected by the sensor network, including CO2 concentration, temperature and humidity, light intensity, mycelial activity spectrum, and data, are timestamped, filtered for outliers, and compressed via a local edge gateway before being uploaded to the cloud platform via an encrypted channel (TLS 1.3). Users can view the real-time operating status dashboards of each cultivation layer through a web browser or mobile app, including dynamic temperature and humidity curves, CO2 concentration heatmaps, mycelial activity index trend lines, and online equipment status topology diagrams. When monitored values ​​exceed preset threshold ranges (e.g., humidity below 80% RH for 15 consecutive minutes during the fruiting body stage), the system automatically triggers multi-channel alarms (in-station pop-ups + SMS + WeChat push notifications) and supports users to remotely issue control commands with a single click (e.g., starting a humidifier or adjusting the proportion of red LED light). These commands are dispatched to the corresponding execution unit after cloud scheduling, with end-to-end latency controlled within 3 seconds. This feature is optionally compatible with third-party IoT platform access, such as connecting to Alibaba Cloud IoT Platform or Huawei OceanConnect through standard API interfaces to achieve data connectivity with enterprise ERP / MES systems.

[0081] The cloud server mentioned in "implementing data storage and analysis functions through cloud servers" refers to a distributed service cluster deployed in a public or hybrid cloud environment that meets the Level 3 Information Security Protection Standard. This cluster includes a time-series database (InfluxDB or TDengine), object storage (OSS / S3), microservice containers (Kubernetes orchestration), and an AI inference engine (TensorRT optimized model service). Data storage adopts a hierarchical strategy: raw sensor data is persisted to the time-series database at 10-minute granularities, with a retention period of ≥18 months; high-definition image data undergoes initial edge screening using a YOLOv5s lightweight model (only cropped images of suspected lesion areas are uploaded) before being stored in object storage; device logs and control command streams are written to a structured relational database (PostgreSQL) for auditing and traceability. The data analysis functions include three categories: First, basic statistical analysis, such as the conversion of monthly average yield per base and the statistics of environmental parameter compliance rates at each growth stage; second, multi-base horizontal comparative analysis, supporting aggregation by geographical latitude, climate zone, or substrate formulation to generate regional adaptability heat maps; and third, in-depth analysis, which constructs a probability prediction model for Ganoderma lucidum fruiting bodies based on an LSTM neural network. The input is multi-dimensional environmental time-series data for the previous 72 hours, and the output is the risk level of fruiting body differentiation (low / medium / high) for the next 24 hours. This model is automatically retrained and updated quarterly using newly added data. This function optionally supports a private deployment mode, i.e., the cloud server is deployed in the customer's self-built IDC data center and accessed through a dedicated line to meet the compliance requirements of data not leaving the domain.

[0082] The various technical features form a closed-loop synergy: remote monitoring relies on the low-latency data path and visualization services provided by the cloud server, while the data analysis capabilities of the cloud server are based on the high-quality, long-term, and multi-dimensional measured data accumulated by remote monitoring; data preprocessing on the edge side ensures upload bandwidth efficiency, which in turn improves the real-time performance of remote monitoring; the elastic expansion capability of the cloud platform supports seamless expansion from a single experimental shed (≤100 mushroom bags) to a hundred-acre smart farm (≥5000 mushroom bags), avoiding system jitter or alarm loss caused by a surge in the number of nodes.

[0083] Through the above technical solutions, the system achieves cross-temporal, all-element, traceable and controllable digital governance of the Ganoderma lucidum cultivation process. For example, after deploying this system at the Jingde base in Anhui, operators can simultaneously monitor the real-time operating conditions of three dispersed planting sites on a large screen at the Hefei headquarters. When the CO2 concentration at a certain point fluctuates abnormally, the system not only pushes an alarm, but also automatically retrieves the temperature, humidity, light, and mycelial activity change curves for the two hours before and after that period, and correlates and analyzes historical records of similar events to assist in decision-making on whether to adjust the ventilation strategy or troubleshoot CO2 generator malfunctions. The full data accumulated over six months of long-term operation, through cloud-based cluster analysis, revealed that when the cumulative light integral (Lux·h) at the mycelial stage is between 5760 and 6480, the subsequent primordia formation rate of fruiting bodies increases by 12.3%. This pattern has been incorporated into the platform's built-in intelligent recommendation rules. Therefore, this embodiment, relying solely on the two technical features of "remote monitoring and management" and "cloud server for data storage and analysis" as defined by this invention, effectively solves the problem pointed out in the background technology that "localized control systems are unable to meet the needs of cross-regional management, big data analysis and long-term trend judgment," significantly improving system operation and maintenance efficiency, data asset value and technology promotion adaptability, and providing a reusable cloud-edge collaboration paradigm for the large-scale, standardized and intelligent upgrading of the Ganoderma lucidum industry.

[0084] Example 6:

[0085] Current Ganoderma lucidum cultivation methods generally rely on manual experience to judge growth status and environmental requirements, lacking a systematic and closed-loop intelligent control process. This makes it difficult to achieve full-process automation from multi-source data collection, growth status assessment, dynamic strategy generation to precise execution feedback. Specifically, key growth indicators such as mycelial expansion speed and cap morphology cannot be quantitatively monitored; environmental parameter regulation is lagging and isolated, with a lack of synergistic response between light, temperature, humidity, CO2 concentration, and nutrient supply; image data is not integrated with sensor data for analysis, resulting in low early disease identification rates and untimely regulation of fruiting body development. Overall, these methods restrict the standardization, repeatability, and large-scale efficiency of Ganoderma lucidum production, making it difficult to achieve both high quality and high yield.

[0086] Based on the above embodiments, this embodiment further provides:

[0087] A method for simulated wild three-dimensional intelligent cultivation of Ganoderma lucidum based on the Internet of Things includes the following steps:

[0088] Step S1: Construct an evaluation index system for the growth status of Ganoderma lucidum, including mycelial expansion rate, cap morphology, cap area and the proportion of fruiting bodies;

[0089] Step S2: Collect Ganoderma lucidum growth environment data and image data through an Internet of Things sensor network;

[0090] Step S3: Analyze the data based on AI algorithms to generate environmental control strategies;

[0091] Step S4: Automatically adjust light, temperature, humidity, CO2 concentration, and nutrient solution supply according to the strategy;

[0092] Step S5: Continuously monitor and optimize system parameters to achieve high-quality and high-yield production of Ganoderma lucidum.

[0093] Step S1, "Constructing an evaluation index system for Ganoderma lucidum growth status," refers to establishing a multi-dimensional evaluation benchmark framework that covers the entire growth cycle of Ganoderma lucidum, is quantifiable, comparable, and can drive regulatory decisions. This index system focuses on phenotypic and physiological parameters with clear biological significance, high measurability, and strong correlation with the accumulation of active ingredients, including:

[0094] "Hyphae expansion rate" is defined as the radial extension distance (mm / d) of the hyphal front on the substrate surface per unit time. It is obtained by timed image sequence comparison or laser displacement sensing and is used to characterize the metabolic activity and substrate colonization efficiency during the vegetative growth stage. Alternative implementation methods include: using fluorescently labeled hyphae to track the three-dimensional expansion rate through confocal microscopy, or using electrical impedance tomography (EIT) to invert the rate of change of hyphal network density.

[0095] "Cap morphology" refers to the geometric and optical characteristics of the cap during fruiting body development, such as edge curling, surface gloss, edge thickness uniformity, and crack incidence. Point cloud and texture images are acquired by a high-resolution RGB-D camera, and quantized values ​​are extracted through morphological filtering and contour fitting algorithms. Alternative implementation methods include: using structured light 3D scanning to reconstruct the curvature distribution of the cap surface, or introducing near-infrared reflectance differences to identify early lignification abnormal areas.

[0096] "Cap area" is the effective pixel area (cm²) of the top-view projection after the fruiting body unfolds. It is obtained by binarization segmentation and connected component analysis of the overhead image under calibrated light source and is used to determine the timing of fruiting body differentiation and expansion rate. Alternative implementation methods include: deploying a multi-angle ring camera array to reconstruct the real surface area through multi-view fusion to avoid planar projection distortion.

[0097] The "proportion of fruiting bodies" is defined as the percentage (%) of the total number of inoculated spawn bags that have completed primordium differentiation and successfully developed into mature fruiting bodies. It reflects the environmental adaptability of the cultivation system to the reproductive transformation stage. Alternative implementation methods include: combining deep learning models (such as YOLOv8s) to perform sub-millimeter-level detection of tiny primordia in real-time video streams and predicting the probability of fruiting body formation 72 hours in advance, thereby upgrading the "proportion" from a final statistical indicator to a process early warning indicator.

[0098] The four indicators mentioned above are not independent and parallel, but rather constitute a hierarchical structure with a temporal coupling relationship: the mycelial expansion rate dominates the setting of the early and mid-stage regulatory thresholds; the cap morphology and area together constitute the quality criteria for fruiting body formation; and the fruiting body production ratio serves as the system-level performance output, inversely verifying the cumulative effect of all previous regulatory actions. This system provides a structured label space for subsequent AI modeling, enabling environmental regulation to leap from "experience-based threshold control" to "growth state-driven control."

[0099] Among them, step S2, "collecting Ganoderma lucidum growth environment data and image data through IoT sensor network", refers to deploying heterogeneous sensing units synchronously on each layer of the three-dimensional cultivation rack to achieve spatiotemporal alignment of physical environment parameters and visual phenotypic data collection. Environmental data includes: CO2 concentration (using a non-dispersive infrared (NDIR) sensor, range 0-5000 ppm, accuracy ±30ppm), matrix trace element content (integrated ion-selective electrode array, simultaneously detecting N, P, K, Se, and Zn ion activities), temperature and humidity (Honeywell HIH-4030 composite sensor, temperature ±0.5°C, humidity ±2% RH), and light intensity (TSL2561 digital light sensor chip, 0.1-65535 Lux, including visible and near-infrared dual channels). Image data is collected by an industrial-grade CMOS camera (resolution up to 2448×2048, supporting automatic white balance and low-light enhancement) installed in the center of each layer according to a preset time sequence (once every 2 hours by default, increased to once every 30 minutes during the fruiting body expansion period). Each acquisition includes three frames: a standard whiteboard calibration image, an ambient light image, and an orthophoto image of the fungal bag, ensuring that color and brightness can be reproduced. All sensors and cameras are connected to the local gateway via the ZigBee 3.0 protocol, and timestamps are uniformly calibrated by the BeiDou time synchronization module (error <10 ms), ensuring strict synchronization of multi-source data at the millisecond time granularity. Alternative implementations include: replacing broadband cameras with miniature spectrometers (such as Hamamatsu C12880MA) to directly acquire continuous reflectance spectra of 400–1000 nm for inverting cap pigment composition; or using a thermal infrared camera (FLIR A35) to simultaneously acquire surface temperature fields to assist in identifying localized moisture stress or lesion areas.

[0100] Among them, step S3, "analyzing data based on AI algorithms to generate environmental regulation strategies", refers to using the evaluation index system constructed in step S1 as a supervision signal, taking the multimodal time series data collected in step S2 as input, and using joint modeling of artificial neural network (ANN) and genetic algorithm (GA) to output the optimal parameter combination for the next regulation cycle. In the specific implementation: the ANN model uses an LSTM (Long Short-Term Memory) structure to process a 30-day sensor time series, extracting mycelial growth trends, environmental disturbance memories, and fruiting body development stage characteristics; the GA module performs multi-objective optimization within the feasible solution space output by the ANN, with optimization objective functions including: maximizing the daily increase in cap area, minimizing the morphological distortion index, constraining CO2 concentration fluctuations to ≤±50 ppm, and saving ≥15% of the total nutrient solution supply; the final generated control strategy is a set of parameter instructions with confidence weights, for example: within the next 24 hours, the proportion of red light from the third layer LED increases from 60% to 75%, humidity increases from 82% RH to 86% RH, and the nutrient solution EC value decreases from 1.8 mS / cm to 1.5 mS / cm. Alternative implementation methods include: using a graph neural network (GNN) to model the environmental coupling relationship between cultivation layers, treating the three-dimensional frame as a topological graph, with nodes as layers and edges as airflow / radiative heat conduction weights; or introducing Bayesian optimization to replace GA, quickly converging to the high-value parameter region under small sample conditions.

[0101] Step S4, "Automatically adjusting light intensity, temperature and humidity, CO2 concentration, and nutrient solution supply according to the strategy," refers to parsing the control commands output in step S3 into underlying equipment control signals and completing closed-loop regulation of physical quantities through the actuator. Specifically, this includes:

[0102] Lighting adjustment: Red (620–630 nm) and blue (440–450 nm) LED beads are driven separately by PWM (Pulse Width Modulation) signals to achieve independent control of spectral ratio and light intensity. The light intensity adjustment range is 100-2000 Lux, and the response delay is <500 ms.

[0103] Temperature and humidity adjustment: The ultrasonic humidifier (atomized particle size ≤5 μm), PTC ceramic heating element (power 0-1500 W adjustable) and EC fan (air volume 0-300 m³ / h) work together to achieve temperature control accuracy ±0.8°C and humidity control accuracy ±3%RH.

[0104] CO2 concentration adjustment: The concentration can be precisely maintained by using a pulsed CO2 generator (burning methanol to produce pure CO2) or a two-way ventilation valve (linking the ratio of outdoor fresh air to indoor return air), with a control range of 800-1200 ppm and a stabilization time of ≤90 s;

[0105] Nutrient solution supply: A peristaltic pump (programmable flow rate 0.1-50 mL / min) is used to inject the substrate through a microporous drip system. Each supply lasts 15-120 seconds, with the interval dynamically triggered by mycelial activity monitoring results. All execution devices have built-in feedback sensors (such as LED light intensity probes and humidifier outlet humidity probes) to form a two-level closed loop of command-execution-verification, ensuring that the actual output deviates from the strategic target by ≤3%. Alternative implementation methods include: replacing part of the blue light with ultraviolet LEDs (265 nm) to induce triterpene synthesis during the fruiting body stage; or using an electrodialysis device to adjust the pH and ionic strength of the nutrient solution online, rather than just changing the EC value.

[0106] Step S5, "Continuously monitor and optimize system parameters to achieve high-quality and high-yield production of Ganoderma lucidum," refers to establishing a cross-cycle self-evolution mechanism. The results of the current S1-S4 executions are used as new training samples and injected into the AI ​​model to continuously update the strategy generation capability. Specifically, this includes: performing residual analysis on the actual measured values ​​of each indicator in S1 and the predicted values ​​in S3; if the prediction error of the cap area continues to be >8% for three cycles, the ANN model is retrained; if the proportion of fruiting bodies is below 92% for two consecutive batches, a global search (GA) is initiated, and the multi-objective weight coefficients are reset; simultaneously, the system automatically archives the complete data chain for each batch (environmental parameter curves + image sequences + control command logs + harvesting quality inspection reports), forming a digital twin of Ganoderma lucidum growth to support cross-regional and cross-variety strategy transfer learning. Alternative implementation methods include: introducing a federated learning architecture, allowing multiple planting bases to collaboratively optimize common models without sharing original data; or deploying a lightweight Transformer model on an edge gateway to achieve real-time online strategy fine-tuning and reduce cloud dependency.

[0107] Through the above steps, this invention achieves closed-loop intelligent control of the entire Ganoderma lucidum cultivation process. Step S1 establishes a quantitative evaluation system covering the essence of growth, making the control objective clearer; Step S2 acquires the true growth state through multi-source synchronous sensing, providing reliable input for decision-making; Step S3 integrates the time modeling capability of ANN and the global optimization capability of GA, breaking through the limitations of a single algorithm and generating a Pareto optimal strategy that balances quality and yield; Step S4 relies on a high-precision execution mechanism and a two-level feedback loop to ensure the physical implementation of the strategy; Step S5 drives continuous model evolution through a data closed loop, enabling the system to possess adaptive characteristics that become more accurate and better with use. Each step is interconnected and progressive, fundamentally solving the core problems mentioned in the background technology, such as fragmented regulation, delayed response, and strong reliance on experience. Ultimately, it achieves technical effects such as increasing the yield per unit area of ​​Ganoderma lucidum by more than 30%, increasing the content of major active ingredients (polysaccharides and triterpenes) by 15%-25%, and reducing the incidence of pests and diseases to below 3.5%. It provides a replicable and scalable technical paradigm for the intelligent, standardized, and green production of Chinese medicinal materials.

[0108] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. 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 three-dimensional intelligent cultivation system for Ganoderma lucidum based on the Internet of Things, characterized in that... The process includes the following steps: 1) Three-dimensional cultivation rack module: adopts a multi-layer adjustable structure, with each layer equipped with an independent substrate cultivation trough and environmental control device; 2) Internet of Things sensor network: distributed in each cultivation layer, used to monitor CO2 concentration, trace element content, mycelial activity, temperature and humidity, and light intensity parameters in real time; 3) Simulated wild environment module: includes an adjustable spectrum LED lighting unit, a temperature and humidity control unit, and a substrate control unit. 4) Intelligent Decision and Control Center: Processes sensor data based on AI algorithms and outputs control commands to the execution device; 5) User Interface: Used for system status monitoring and parameter setting.

2. The system as described in claim 1, characterized in that, The three-dimensional cultivation rack module adopts a layered and rotatable design, which can automatically adjust the angle according to the growth stage of Ganoderma lucidum to ensure uniform light exposure.

3. The system as described in claim 1, characterized in that, The IoT sensor network includes a mycelial activity monitoring sensor, which uses near-infrared spectroscopy to assess the mycelial growth status in real time.

4. The system as described in claim 1, characterized in that, The AI ​​algorithm is an optimization model based on artificial neural networks and genetic algorithms, which can dynamically adjust environmental parameters according to historical data and real-time monitoring values.

5. The system as described in claim 1, characterized in that, The system supports remote monitoring and management, and uses a cloud server to perform data storage and analysis.

6. A method for simulated wild three-dimensional intelligent cultivation of Ganoderma lucidum based on the Internet of Things, characterized in that, Includes the following steps: 1) Construct an evaluation index system for Ganoderma lucidum growth status, including mycelial expansion speed, cap morphology, cap area, and the proportion of fruiting bodies; 2) Collect Ganoderma lucidum growth environment data and image data through an Internet of Things sensor network; 3) Analyze the data based on AI algorithms to generate environmental control strategies; 4) Automatically adjust light, temperature, humidity, CO2 concentration, and nutrient solution supply according to the strategies; 5) Continuously monitor and optimize system parameters to achieve high-quality and high-yield production of Ganoderma lucidum.