Coupled Internet of Things regulation and control system based on multispectrum and hypha feature library

Abstract

The invention discloses a coupled Internet of Things regulation and control system based on multispectrum and hypha feature library, and relates to the technical field of image processing, and the system comprises the following steps: dividing a photosensitive area of an image sensor into a plurality of channels according to light source spectral features; collecting light intensity distribution data and spectrum distribution data of a plurality of channels, and when it is monitored that the real-time light intensity or the real-time spectrum of the corresponding channel is larger than a judgment threshold value, marking the corresponding channel as a to-be-regulated state; a photosensitive area of the image sensor is divided into a plurality of visible light, ultraviolet and near-infrared independent channels according to spectral characteristics of a light source, each channel is provided with an exclusive gain control module, when light intensity of a single channel is suddenly changed due to extreme illumination such as strong light of thunder and lightning and germicidal lamp faults, only the gain of the channel is adjusted in a targeted manner, and other channels maintain conventional parameters; the problem of overexposure or underexposure of the whole image under traditional single-channel regulation and control is avoided, it is ensured that the image features of the hypha growth area are always clear, and effective data support is provided for subsequent hypha monitoring.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a coupled Internet of Things control system based on multispectral and mycelial feature libraries. Background Technology

[0002] In the large-scale cultivation of *Pleurotus ostreatus*, growers typically rely on image sensors to collect images of mycelium on the substrate surface. By identifying features such as substrate texture and mycelial morphology, they monitor mycelial growth and assist in regulating the greenhouse environment, such as sterilization, ventilation, and dehumidification. However, existing monitoring systems for *Pleurotus ostreatus* greenhouses have technical limitations under extreme light conditions, making it difficult to guarantee monitoring stability and control accuracy.

[0003] When exposed to sudden bursts of intense sunlight through the greenhouse film during lightning strikes, or when germicidal lamps malfunction and produce instantaneous super-bright flashes, the existing system's image sensors, equipped with only a single automatic gain control module, cannot adapt to the spectral characteristics of different light sources. This easily leads to malfunctions in the gain control module, resulting in overexposure (complete white) or underexposure (complete black) in the acquired images. At this time, the effective features originally relied upon by the system algorithm, such as substrate texture and mycelial morphology, are completely lost, making normal mycelial state identification impossible. Due to the lack of differentiated control based on spectral features and a mycelial feature reference mechanism, the existing system defaults to a full-area outbreak of contaminants based on the highest risk logic, triggering all control equipment (sterilization equipment, ventilation equipment, dehumidification equipment) to operate at full capacity. This false triggering behavior not only wastes energy such as electricity and consumables but also triggers a chain of negative consequences: excessive ventilation causes rapid moisture loss from the substrate surface, leading to drying of the culture medium and affecting mycelial water absorption; excessive sterilization destroys the mycelial activity, ultimately causing large-scale mycelial death and reducing the yield and quality of *Mushroom arborescens* cultivation.

[0004] In summary, there is an urgent need for an intelligent system that can adapt to multispectral characteristics, avoid misjudgment under extreme lighting conditions, and ensure the accuracy of mycelial monitoring and regulation. Summary of the Invention

[0005] The purpose of this invention is to solve the problem that when the image sensor of the existing system acquires an image that is instantly overexposed or underexposed, the effective features such as the texture of the material surface and the morphology of mycelium that the system algorithm originally relied on are completely lost, making it impossible to complete the normal mycelium state identification. Therefore, a new IoT control system method based on multispectral and mycelium feature library is proposed.

[0006] To achieve the above objectives, the present invention employs the following technology: a coupled Internet of Things (IoT) control system based on multispectral and hyphal feature libraries, comprising the following steps:

[0007] The photosensitive area of ​​the image sensor is divided into multiple channels based on the spectral characteristics of the light source.

[0008] Collect light intensity distribution data and spectral distribution data from multiple channels. When the real-time light intensity or real-time spectrum of the corresponding channel is detected to be greater than the judgment threshold, the corresponding channel is marked as a state to be controlled.

[0009] A spectral reflectance feature library of *Flammulina velutipes* mycelium was constructed. When the corresponding channel was marked as a state to be adjusted, the standard reflectance data of the corresponding channel in the feature library was retrieved, and the real-time reflectance was compared with the standard value to obtain the feature comparison result.

[0010] Based on the channel's uncontrolled state and feature comparison results, adjust the corresponding channel gain;

[0011] Real-time image data from multiple channels is collected and synthesized to obtain a complete fabric image. The accuracy of the complete fabric image is verified. If it meets the requirements, the mycelial features of the complete fabric image are compared with the feature library standard.

[0012] If local features become blurred due to adjustment of a single channel, adjust the gain of other channels;

[0013] Record the light intensity data, gain parameters, feature comparison results, and synthetic image quality for each adjustment, and input the data into the *Agaricus esculentus* mycelial spectral reflectance feature library.

[0014] Furthermore, the method for acquiring real-time image data from multiple channels and synthesizing a complete fabric image includes:

[0015] Define the physical range of image acquisition for multiple channels; mark the stitching boundary on the images of the corresponding channels, and delineate the texture observation area on both sides of the stitching boundary;

[0016] Within the texture observation area of ​​the corresponding channel, the texture direction of the hyphae is obtained. The observation area is divided into multiple grids using the grid method, and the number of hyphae in each grid is recorded to obtain a reference table of texture direction and hyphae number.

[0017] By comparing the reference tables of texture observation areas in multiple channels, the orientation deviation and density difference are obtained;

[0018] Based on the orientation deviation, the rotation angle of the corresponding channel image is obtained, and based on the density difference, the image contrast of the corresponding channel is obtained;

[0019] A complete fabric image is obtained by stitching together channel images based on rotation angle and image contrast.

[0020] Furthermore, the method for obtaining the texture orientation of hyphae includes:

[0021] A polarizing filter is installed in each channel, and the filter angle is divided into multiple angles to acquire multi-polarization images of the same area.

[0022] By comparing the brightness distribution of the polarization angle image in each channel, the dominant direction of the hyphal texture in the corresponding channel can be obtained;

[0023] By comparing the dominant directions of hyphal textures in multiple channels, the texture direction of the hyphal texture is obtained.

[0024] Furthermore, the method for verifying the accuracy of the complete fabric image includes:

[0025] Multiple mycelia are selected as judgment samples from real-time images across multiple channels.

[0026] Based on the judgment samples, a directional baseline is drawn in the real-time image of the visible light channel of multiple channels, and the angle between the baseline and the horizontal axis of the real-time image is measured, and the coordinates of multiple feature points are marked.

[0027] Align the real-time image of the channel to be stitched with the visible light channel image to obtain the hyphal direction and coordinates of multiple feature points of the mycelium of the judgment sample in the real-time image of the channel to be stitched; and judge the hyphal direction and coordinates of multiple feature points of the mycelium of the judgment sample in the real-time image of the channel to be stitched to obtain the judgment result.

[0028] If the angle between the hyphae direction and the baseline in the real-time image of the channel to be spliced ​​is lower than the preset angle and the coordinate deviation of multiple feature points is lower than the preset length, then the real-time image of the channel to be spliced ​​is accurate, and the complete fabric image meets the accuracy verification.

[0029] Furthermore, the method for dividing the photosensitive area of ​​the image sensor into multiple channels based on the spectral characteristics of the light source includes:

[0030] Full-spectrum data of all light sources in the greenhouse were collected, and the spectral range and intensity peak of natural light, germicidal lamps and supplemental lights were recorded. The reflectance and absorptivity of the antler mushroom mycelium in each spectral band were measured to obtain the dataset.

[0031] Based on the dataset and the location of all light sources in the studio, the wavelengths are divided into visible light channels, ultraviolet light channels, and near-infrared light channels.

[0032] Furthermore, methods for acquiring light intensity distribution data and spectral distribution data from multiple channels include:

[0033] Collect light intensity distribution data and spectral distribution data within a preset time period, and calculate the average light intensity and spectral target proportion of multiple channels;

[0034] Collect the operating parameters of the light source corresponding to multiple channels;

[0035] Based on the average light intensity of multiple channels and the proportion of spectral targets, the judgment thresholds for multiple channels are obtained;

[0036] Collect real-time light intensity and real-time spectral data from multiple channels within a preset time period, and compare the real-time light intensity and real-time spectral data from multiple channels with a judgment threshold.

[0037] If the real-time light intensity or real-time spectrum of the corresponding channel is greater than the judgment threshold and the working parameters are displayed normally, then the corresponding channel is marked as a candidate state to be controlled.

[0038] The real-time light intensity and real-time spectral data of the channel marked as a candidate state to be regulated are compared with the judgment threshold multiple times. If the real-time light intensity and real-time spectral data are all greater than the judgment threshold multiple times, the corresponding channel is marked as a state to be regulated.

[0039] Furthermore, methods for obtaining feature comparison results include:

[0040] Based on the mycelial spectral reflectance feature library of *Pleurotus ostreatus*, the baseline spectral features and core feature types of multiple channels of the mycelial samples to be compared are obtained; the real-time spectral features of the mycelial samples to be compared are obtained.

[0041] Environmental consistency calibration is performed on the real-time spectral features to obtain calibrated real-time spectral features. The calibrated real-time spectral features are then compared with the core feature types to obtain multi-channel comparison results.

[0042] Based on the multi-channel comparison results, the overall consistency is judged; and the feature comparison results are obtained.

[0043] Furthermore, based on the channel's state to be adjusted and the feature comparison results, methods for adjusting channel gain include:

[0044] Obtain the reflectivity deviation type of the channel to be controlled, and adjust the gain of the corresponding channel according to the type of the channel to be controlled and the reflectivity deviation type;

[0045] Select non-critical monitoring areas within the channel to be regulated, adjust the gain of the corresponding channel according to the type of the channel to be regulated and the type of reflectivity deviation, collect reflectivity data of the channel to be regulated after gain adjustment, and obtain the pre-adjustment result;

[0046] Based on the pre-adjustment results, the gain adjustment range is adjusted, and the reflectivity data of the channel to be adjusted is re-acquired and compared with the preset standard range in the feature library to verify whether the reflectivity meets the standard. If the reflectivity meets the standard, the gain adjustment of the channel to be adjusted is completed.

[0047] Furthermore, methods for adjusting the gain of other channels include:

[0048] Obtain the local feature blurring region and associated non-adjusted channel that appears after single-channel adjustment;

[0049] Based on the blurred regions of local features and the associated non-adjusted channels, the sharpness of existing features is analyzed to obtain auxiliary adjusted channels;

[0050] Based on the local feature fuzzy region and the adjacent non-sensitive region of the local feature fuzzy region, the pre-adjustment confirmation direction and pre-adjustment confirmation magnitude of the auxiliary adjustment channel are obtained;

[0051] The pre-adjustment confirmation direction and pre-adjustment confirmation magnitude are applied to the fuzzy correlation region of the auxiliary channel, and a multi-channel composite image of the fuzzy correlation region is acquired; if the local feature fuzzy region in the multi-channel composite image is eliminated and the spectral features of each channel are consistent, then the gain adjustment of other channels ends.

[0052] Furthermore, methods for performing existing feature sharpness analysis include:

[0053] Obtain standard features from the spectral reflectance feature library of *Flammulina velutipes* mycelium, and obtain spectral features of regions corresponding to the blurred regions of local features in the associated unadjusted channels;

[0054] By comparing spectral features with standard features, the differences in the sharpness analysis of existing features are obtained.

[0055] In summary, due to the adoption of the above-mentioned technology-based IoT control system coupled with multispectral and mycelial feature libraries, the beneficial effects of this invention are:

[0056] This invention divides the photosensitive area of ​​an image sensor into multiple independent channels for visible light, ultraviolet light, and near-infrared light based on the spectral characteristics of the light source. Each channel is equipped with a dedicated gain control module. When extreme light conditions such as strong lightning or germicidal lamp malfunction cause a sudden change in the light intensity of a single channel, the gain of that channel is adjusted only, while other channels maintain normal parameters. This avoids the problem of overexposure or underexposure of the overall image under traditional single-channel control, ensuring that the image features of the mycelial growth area are always clear, and providing effective data support for subsequent mycelial monitoring.

[0057] This invention ensures the accuracy of the synthesized substrate surface image, providing a reliable basis for assessing mycelial growth status. During the multi-channel image synthesis process, this invention effectively eliminates orientation deviations and density differences during channel splicing by defining texture observation areas, statistically analyzing mycelial density using a grid method, and comparing mycelial orientation across channels. It also employs an accuracy verification mechanism involving multiple mycelial samples and multiple operator reviews. This avoids problems such as broken mycelial textures and blurred edges in the synthesized image, ensuring that the complete substrate surface image truly reflects the mycelial growth status and provides growers with accurate visual references for assessing mycelial growth and developing management strategies.

[0058] This invention constructs a spectral reflectance feature library of *Agaricus esculentus* mycelia, storing standard spectral features of mycelia under different growth stages and environmental parameters. When a channel is marked as being under control, the system determines whether gain anomalies affect mycelial features by accurately comparing real-time reflectance with standard values, avoiding blind gain adjustments caused by the lack of feature references in traditional systems. At the same time, accurate feature identification can prevent the system from misjudging a global outbreak of miscellaneous bacteria, thereby avoiding full-load idling of control equipment such as sterilization, ventilation, and dehumidification, reducing energy waste, protecting mycelial activity, and preventing large-scale mycelial death due to over-control.

[0059] This invention uses existing feature clarity analysis to screen auxiliary adjustment channels, combines pre-adjustment in non-sensitive areas to verify the adjustment direction and magnitude, and then applies the parameters to the fuzzy correlation area. This eliminates local fuzziness and ensures the continuity of spectral features of each channel, making the mycelial texture cross channels without abrupt changes and the reflectance transition natural. This avoids image quality problems in traditional control and ensures that all material surface image features are uniform and traceable. Attached Figure Description

[0060] Figure 1 The system block diagram of the IoT control system based on multispectral and hyphal feature library of the present invention is shown.

[0061] Figure 2 The schematic diagram of the IoT control system based on multispectral and mycelial feature library of the present invention is shown.

[0062] Figure 3 A diagram of mycelium on the substrate surface of the present invention is shown. Detailed Implementation

[0063] The following will describe clearly and completely the technology of the IoT control system based on the coupling of multispectral and mycelial feature libraries in the embodiments of the present invention with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0064] To more clearly and intuitively demonstrate the practical application effects and advantages of the IoT-based control system coupled with multispectral and mycelial feature libraries, and to verify its feasibility and effectiveness, the invention is further described below with reference to embodiments. Through specific scenario simulations and data calculations, the system's role in actual *Pleurotus ostreatus* cultivation is explained in detail, helping readers better understand the technical details and practical value of the invention. The invention is further described below with reference to embodiments.

[0065] Example 1:

[0066] See Figures 1-3The photosensitive area of ​​the image sensor is divided according to the spectral characteristics of the light source in the greenhouse, and clearly divided into multiple channels, including visible light channel, ultraviolet light channel and near-infrared light channel. Each channel is equipped with a dedicated gain control module and presets the initial gain parameters adapted to the corresponding spectrum to ensure that the initial gain of each channel matches the reflectance characteristics of the antler mushroom mycelium under the spectrum, laying the foundation for subsequent independent adjustment.

[0067] It should be noted that methods for dividing the photosensitive area of ​​the image sensor into multiple channels based on the spectral characteristics of the indoor light source include:

[0068] Full-spectrum data of all light sources in the greenhouse were collected. A high-precision spectrometer was used to record the spectral range and intensity peak of natural light, germicidal lamps and supplementary lights. The reflectance and absorptivity of the antler mushroom mycelium in each spectral band were measured simultaneously to establish a dataset and provide a basis for subsequent channel division.

[0069] Based on the dataset, the wavelengths with stable mycelial reflectance are divided into visible light channels (e.g., 400-760nm), ultraviolet light channels (e.g., 200-400nm), and near-infrared light channels (e.g., 760-1100nm). The start and end wavelengths of each wavelength are defined to obtain the wavelength range of each wavelength, which is different from the conventional fixed wavelength division. Combined with the distribution of light sources in the greenhouse, on the photosensitive surface of the sensor, the area corresponding to the direct natural light area is assigned to the visible light channel, the area near the germicidal lamp is assigned to the ultraviolet light channel, and the area covered by the supplementary light is assigned to the near-infrared light channel, ensuring that each area overlaps with the irradiation range of the corresponding light source.

[0070] Each region is equipped with a dedicated photosensitive element. For example, a silicon-based photosensitive element is used in the visible light channel to match its band sensitivity; a cadmium selenide element is used in the ultraviolet light channel to enhance the ultraviolet response; and an indium gallium arsenide element is used in the near-infrared light channel to improve the infrared capture capability. The element parameters are matched to the band range of each band.

[0071] The photosensitive areas of each channel are spectrally isolated. Nanoscale filter strips are installed between the photosensitive areas of adjacent channels. The filter strips only allow the target wavelength of the corresponding channel to pass through. For example, the filter strip of the ultraviolet channel blocks visible light and near-infrared light. The isolation effect is tested by spectral distribution data to ensure that the optical crosstalk rate between channels is less than 3% of that of the near-infrared channel.

[0072] Establish a channel performance correlation table, which records the band range, photosensitive element model, filter strip parameters and compatible light source of each channel, and correlates the mycelial response characteristics of the channel band in the first step. This is used to quickly match the channel parameters during subsequent sensor maintenance to ensure that each channel meets the spectral characteristics of the light source and the mycelial monitoring requirements in the long term.

[0073] Secondly, an independent spectral monitoring unit is installed for each spectral channel. The unit collects the light intensity and spectral distribution data of the corresponding channel in real time and transmits the data to the central control module every microsecond. When a sudden increase (such as a strong light due to lightning in the ultraviolet channel) or a sudden decrease in light intensity is detected in a certain channel, the channel is immediately marked as a state to be adjusted, providing a basis for precise adjustment.

[0074] A spectral reflectance feature library of *Agaricus esculentus* mycelium was constructed. The library stores the reflectance range of mycelium in each spectral channel under normal growth conditions. When a channel is marked as to be adjusted, the standard reflectance data of that channel in the feature library is retrieved. The real-time reflectance is compared with the standard value to determine whether the gain abnormality affects the identification of mycelial features, thus avoiding blind adjustment.

[0075] It should be noted that the methods for constructing the spectral reflectance feature library of *Agaricus esculentus* mycelium include:

[0076] Six key stages of the entire growth cycle of *Pleurotus ostreatus* were selected, including mycelial germination, mycelial full growth, primordium formation, fruiting body growth, and tide transition. Thirty healthy mycelial samples were selected at each stage, and spectral reflectance data of each sample in the visible, ultraviolet, and near-infrared light channels were collected using a high-precision spectrometer to obtain the original spectral curves, laying the data foundation for the feature library.

[0077] The primary classification labels are divided according to the growth stage, and the secondary classification labels are divided according to the mycelial health status. At the same time, the corresponding greenhouse environmental parameters are recorded, so that each group of spectral data is associated with clear scene information, which is different from the conventional feature library without scene. The mycelial health status includes no contamination, slight contamination, and severe contamination, and the greenhouse environmental parameters include temperature, humidity, and light intensity.

[0078] By subtracting ambient light interference from the collected dark background spectrum and correcting the spectrometer system error using the reflectance of a standard white board, the error of the spectral reflectance data of each channel after calibration is ensured to be less than 2%, providing pure data for the extraction of accurate features. The dark background spectrum is the spectrum inside the greenhouse when there are no mycelia.

[0079] Based on classification labels, the core spectral feature set of healthy hyphae at each growth stage is extracted. That is, the peak wavelength of reflection is identified in the visible light channel, such as 550nm; the absorption valley wavelength is extracted in the ultraviolet light channel, such as 280nm; and the reflectance ratio of characteristic bands is calculated in the near-infrared light channel, such as the reflectance ratio of 850nm to 950nm, to form the core spectral feature set of each stage, thus avoiding the blindness of feature extraction.

[0080] Correlation analysis was conducted between the core spectral feature set and the environmental parameters inside the greenhouse to clarify the fluctuation range of the feature parameters under different environments. For example, when the humidity increases, the reflectance ratio of the near-infrared light channel decreases by 10%, so that the feature library has environmental adaptability rather than fixed feature values.

[0081] Every three months, collect spectral data of a new batch of antler mushrooms and repeat the above process. Compare the new data with the existing feature library. If new spectral features are found, such as the shift in the reflectance peak of new healthy mycelia, update the core spectral feature set for the corresponding stage to ensure that the feature library closely matches the mycelial characteristics of actual cultivation in the long term.

[0082] It should be noted that the methods for obtaining the comparative results of the mycelial spectral characteristics of *Pleurotus ostreatus* include:

[0083] From the existing library of mycelial reflectance features of *Agaricus esculentus*, the baseline spectral features of the mycelial samples to be compared are retrieved. The baseline spectral features include the corresponding growth stage and environmental parameters. The corresponding growth stage is, for example, the period when the mycelium is fully grown. The environmental parameters are, for example, the current temperature and humidity in the greenhouse. The core feature types of multiple channels are identified. The core feature types include reflectance peaks, absorption valleys, and feature band ratios, which serve as reference standards for subsequent comparisons.

[0084] The real-time spectral characteristics of the mycelial samples to be compared are obtained, that is, the position of the reflection peak, the position of the absorption valley, and the position of the characteristic band ratio of each channel are separated from the sample spectral data collected in real time. This ensures that the real-time spectral characteristics are completely consistent with the type of the reference spectral characteristics, and avoids comparison deviation due to mismatch of feature types.

[0085] Environmental consistency calibration is performed on real-time spectral features. Environmental consistency calibration means adjusting the numerical distribution of real-time spectral features based on the environmental parameters associated with the reference spectral features and the actual environment inside the greenhouse. The real-time environment includes real-time light intensity and air humidity. If the current light intensity is weaker than the reference environment, the intensity features of the real-time reflectance peak are appropriately corrected to make the environmental adaptability of the real-time spectral features consistent with the reference spectral features and improve the accuracy of comparison.

[0086] The calibrated real-time spectral features are compared with the core feature types. The visible light channel focuses on comparing the morphological consistency of the reflection peaks, such as whether the peaks are flat or sharp. The ultraviolet light channel compares the positional matching of the absorption valleys. The near-infrared light channel compares the consistency of the trend of the reflection peaks in the characteristic bands, such as whether the ratio of the characteristic bands changes in the same direction. The consistency of each type of feature is recorded to obtain the multi-channel comparison results.

[0087] Based on the results of multi-channel comparison, the overall consistency is judged comprehensively. If the core feature types of all channels match, it is determined that the core feature types are completely consistent. If individual channels have slight deviations that are within the range of environmental fluctuations, it is determined that the core feature types are basically consistent. If feature misalignment occurs in multiple channels, such as a shift in reflection peaks, it is determined that the core feature types are significantly deviated, and the criteria for different levels of judgment are clarified.

[0088] The system generates feature comparison results, clearly indicating the consistency level and providing detailed information on the comparison of each channel, such as the matching of absorption valleys in the ultraviolet light channel. It analyzes possible causes of deviations, such as temporary environmental fluctuations or abnormal sample growth, and links them with real-time spectral feature extraction records and environmental calibration data to form a complete traceability chain for the comparison results, ensuring that the results are interpretable and verifiable.

[0089] Each channel's dedicated gain control module receives the status flag to be adjusted and the feature comparison results. If the reflectance of a certain channel exceeds the standard range, only the gain of that channel is adjusted. For example, if the light intensity of the ultraviolet channel suddenly increases, causing the reflectance to be too low, the upper limit of the gain of that channel is lowered, while other channels maintain the initial gain to ensure that the hyphal features of key channels are clear.

[0090] Methods for adjusting channel gain based on the channel's state to be adjusted and feature comparison results include:

[0091] Retrieve the channel information marked as being under control, and combine it with the feature comparison results of the channel to clarify the specific manifestation of reflectivity exceeding the preset standard range, determine whether the overall reflectivity is too high or too low, and at the same time confirm whether the deviation is directly related to the triggering cause of the under-control state, so as to lock the core problem for gain adjustment. The triggering cause is, for example, a sudden increase in light intensity, and obtain the type of reflectivity deviation of the channel under control.

[0092] Based on the type of the channel to be adjusted and the type of reflectivity deviation, the direction of gain adjustment is determined: if the ultraviolet channel has a high reflectivity due to a sudden increase in light intensity, the gain is adjusted downward; if the near-infrared channel has a low reflectivity due to a sudden decrease in light intensity, the gain is adjusted upward. This ensures that the adjustment direction is adapted to the channel characteristics and the cause of the deviation. The types of channels to be adjusted include visible light channels, ultraviolet light channels, and near-infrared light channels.

[0093] Select non-critical monitoring areas within the channel to be regulated, such as areas without core mycelia at the edge of the material surface. Based on the type of channel to be regulated and the type of reflectance deviation, make small-amplitude pre-adjustments to the corresponding channel's gain. Collect the pre-adjusted reflectance data in real time and observe how the data approaches the preset standard range to avoid monitoring errors caused by directly adjusting the critical area.

[0094] Based on the pre-adjustment results, optimize the gain adjustment range: if the reflectance after pre-adjustment is close to the standard range, use this range to adjust the gain for the entire channel area; if the reflectance still deviates from the standard range, fine-tune the range based on the degree of deviation in the feature comparison, while ensuring that the spectral response characteristics of the channel are not affected during the adjustment process.

[0095] After completing the gain adjustment of the entire region, the reflectance data of this channel is reacquired and compared with the preset standard range in the feature library to verify whether the reflectance meets the standard. If the reflectance meets the standard, the channel gain adjustment is completed. At the same time, the reflectance synergy between this channel and other non-adjustable channels is checked to avoid the overall image feature imbalance caused by adjusting a single channel and to ensure the consistency of spectral features of each channel.

[0096] Record complete information about this gain adjustment, including the triggering reason of the state to be adjusted, details of feature comparison deviation, adjustment direction and magnitude, pre-adjustment verification results and final verification data, establish a gain adjustment file for this channel, provide a reference for subsequent gain adjustments in similar scenarios to be adjusted, and form a closed-loop optimization mechanism.

[0097] After adjusting the target channel gain, real-time image data from multiple channels are acquired and a complete material surface image is synthesized. The mycelial features in the synthesized image are compared with the feature library standard. If local features are blurred due to the adjustment of a single channel, the gains of other channels are finely adjusted (not exceeding ±10% of the initial parameters) to ensure that the overall image is free from overexposure or underexposure.

[0098] It should be noted that methods for acquiring real-time image data from multiple channels and synthesizing a complete fabric image include:

[0099] Define the physical range of image acquisition for multiple spectral channels, mark the splicing boundary on each channel image, and delineate a texture observation area 1 cm wide on each side of the boundary to ensure subsequent focus on key areas for observing mycelial features. The multiple spectral channels include visible light channels, ultraviolet light channels, and near-infrared light channels. For example, the splicing boundary is the right boundary of the visible light channel and the left boundary of the ultraviolet light channel.

[0100] In the texture observation area of ​​each channel, the mycelium of *Agaricus esculentus* was magnified and observed using the image acquisition unit. The texture direction of the mycelium was recorded, such as horizontal, vertical, and 45° oblique. At the same time, the observation area was divided into 10×10 mm grids using the grid method, and the number of mycelium in each grid was counted to obtain a reference table of texture direction and mycelium number.

[0101] Compare the reference table of texture observation areas of multiple channels; if the visible light channel has horizontal hyphae with 8 hyphae per small square, and the ultraviolet light channel has 30° oblique hyphae with 5 hyphae per small square, analyze the cause of the difference. Ultraviolet light can easily penetrate the surface hyphae, and may capture the deep sparse hyphae rather than the true deviation in direction. It is necessary to distinguish between apparent difference and true difference.

[0102] To address the deviation in orientation, a physical reference was used for verification. The boundary between two channels of the same continuous hyphae in the observation area was selected. If it was transverse in the visible light channel and obliquely cut off in the ultraviolet light channel, it was determined to be an image rotation deviation. The rotation angle of the ultraviolet light channel was finely adjusted with the hyphae as the axis until the hyphae were continuous.

[0103] To address density differences, calibration is performed based on channel characteristics. Since ultraviolet light is sensitive to deep hyphae, the density ratio of surface and deep hyphae in the two channels is calculated. For example, the visible light surface accounts for 80% and the ultraviolet light deep accounts for 60%. The image contrast of the ultraviolet channel is adjusted proportionally to make the deep hyphae clearly visible, rather than simply increasing the brightness, to ensure that the density reflects the true distribution.

[0104] Based on the rotation angle and image contrast, the images of the two channels are stitched together. After stitching the two channel images, three continuous hyphae across the channel are selected at the boundary, and their direction and density changes are manually tracked. If a hyphae extends laterally in the visible light channel and remains continuous and has a gradual change in density after entering the ultraviolet light channel, the stitching is considered valid. Otherwise, the fourth and fifth steps are repeated until all the hyphae across the channel have continuous characteristics. Then, the near-infrared light channel is added for repeated verification to complete the synthesis and obtain a complete fabric image.

[0105] Specifically, methods for obtaining the texture and direction of hyphae include:

[0106] A polarizing filter is installed in each spectral channel, with filter angles divided into 0°, 45° and 90°. Multi-polarization images of the same area are acquired simultaneously. By utilizing the birefringence properties of the hyphal cell wall, hyphae with different orientations exhibit brightness differences under specific polarization angles.

[0107] By comparing the brightness distribution of images with different polarization angles in each channel, the polarization angle direction corresponding to the brightness peak is the dominant direction of the hyphal texture. For example, the vertical hyphae have the highest brightness under 0° polarization, thus obtaining the dominant direction of the hyphal texture in the corresponding channel.

[0108] By comparing the dominant directions of hyphal textures in multiple channels, the polarization angle brightness peak directions of each channel are cross-validated to eliminate outliers caused by ambient light interference and obtain the true direction of the hyphal texture in each channel.

[0109] It should be noted that the methods for verifying the accuracy of complete fabric images include:

[0110] In the multiple channel images before splicing, select 3 to 5 continuous hyphae that cross the boundary as judgment samples. It is necessary to ensure that one end of these hyphae is in the visible light channel and the other end is in the channel to be spliced, such as the ultraviolet light channel, and that the hyphae do not have natural breaks, so as to provide a clear reference for subsequent direction comparison.

[0111] Based on the judgment samples, in the real-time images of the visible light channels of multiple channels, a directional baseline is drawn along the natural direction of each judgment sample hyphae. The angle between the baseline and the horizontal axis of the image is measured with a ruler, and the coordinates of three evenly distributed feature points on the hyphae are marked as comparison standards. Feature points include, for example, the bifurcation of hyphae and the changes in thickness.

[0112] The channel image to be stitched is initially aligned with the visible light channel image to obtain the hyphal direction and coordinates of multiple feature points of the mycelium in the real-time image of the channel to be stitched. The hyphal direction and coordinates of multiple feature points of the mycelium in the real-time image of the channel to be stitched are judged to obtain the judgment result. If the angle between the hyphal direction and the direction baseline in the real-time image of the channel to be stitched is less than a preset angle of 5°, or the coordinate deviation of multiple feature points is less than a preset length of 1 mm, then the real-time image of the channel to be stitched is accurately stitched. Otherwise, it is determined that there is a misalignment in the stitching.

[0113] To address misalignment, fine-tune the rotation angle or position of the image to be spliced: After each adjustment, re-observe and judge whether the sample hyphae extend along the baseline until the coordinates of the feature points of the sample hyphae in the spliced ​​channel deviate from the corresponding points of the visible light channel by ≤1 mm, and the angle of orientation is ≤5°, to ensure the continuity of a single hyphae.

[0114] After completing the alignment of a single sample hyphae, extend the observation to all hyphae within a 5 cm range on both sides of the splicing boundary: if the direction of these hyphae is consistent with the direction baseline and there is no sudden change, such as a sudden change from horizontal to oblique, the overall texture direction is determined to be continuous, and the real-time image of the channel to be spliced ​​is accurate.

[0115] Two operators independently repeat the above steps to verify the continuity of hyphal direction and boundary texture of all samples: if the two operators' judgments are completely consistent and the records are accurately spliced, the complete fabric image meets the accuracy verification; if there are differences, the coordinates of the direction baseline mark and feature points are re-checked until they are consistent to ensure that the judgment results are reliable.

[0116] It should be noted that methods for adjusting the gain of other channels include:

[0117] Locate the local feature blurring area that appears after single-channel adjustment, clarify the blurring behavior, such as mycelial texture breakage and edge blurring, associate the original feature state of this area in other unadjusted channels, determine whether the blurring is caused by the feature balance being broken by the single-channel gain change, and obtain the associated unadjusted channels, where the associated unadjusted channels and other unadjusted channels are the unadjusted channels in the visible light channel, ultraviolet light channel, and near-infrared light channel.

[0118] Based on the identified local feature fuzzy regions and associated non-adjusted channels, the existing feature sharpness is analyzed for the local feature fuzzy regions of these channels. If the spectral features of the corresponding region of the ultraviolet channel are complete but the brightness is insufficient, it is determined that the channel needs to be improved by gain adjustment, i.e., the channel is an auxiliary adjustment channel. If the spectral features of the near-infrared channel are continuous but the contrast is low, the channel is locked as an auxiliary adjustment channel.

[0119] For the auxiliary adjustment channel, a pre-adjustment confirmation direction is formulated based on the feature requirements of the local feature blurring area; if the local feature blurring area needs to enhance texture details, the auxiliary channel gain is increased to improve feature sharpness; if it needs to balance brightness transition, the auxiliary channel gain is decreased to weaken excessive reflection. The pre-adjustment confirmation direction needs to be adapted to the spectral response characteristics of the channel.

[0120] In the auxiliary adjustment channel, select a non-sensitive region adjacent to the local feature blurry region. For example, a region without important hyphae distribution. Perform a small-scale pre-adjustment as above, observe the feature changes of the blurry region in real time, and verify whether the adjustment can reduce the blur without destroying the original features of the auxiliary channel to obtain the pre-adjustment confirmation range.

[0121] Apply the pre-adjustment confirmation direction and pre-adjustment confirmation magnitude to the fuzzy correlation region of the auxiliary channel, re-acquire the multi-channel composite image of the fuzzy correlation region, check whether the local feature fuzzy region in the multi-channel composite image is eliminated and whether the spectral features of each channel are consistent. If the local feature fuzzy region is eliminated and the spectral features of each channel are consistent, such as the hyphal texture without breakage across channels, then the gain adjustment of other channels is finished, while ensuring that the features of other regions of the auxiliary channel are not disturbed.

[0122] Record the basis, direction, and effect of this auxiliary channel adjustment, link it to the previous single-channel adjustment records, update the gain coordination threshold of each channel, and improve the accuracy of subsequent adjustments. The basis includes the blurred region features of local feature blurred areas and the original state of the associated channels.

[0123] Specifically, methods for performing existing feature clarity analysis include:

[0124] Identify the range of local feature blurring areas that appear after single-channel adjustment, retrieve the associated non-adjusted channels corresponding to the local feature blurring areas, determine the channel objects to be analyzed, and lock in the target range for subsequent feature clarity checks;

[0125] From the spectral reflectance feature library of velvet mushroom mycelium, standard features of the associated non-adjusted channels at the corresponding growth stages were retrieved to ensure that there are clear standard references during analysis and to avoid subjective judgment bias. The standard features include the clarity of mycelial texture direction, reflectance stability and feature edge sharpness.

[0126] Focus on the region corresponding to the blurred local feature area in the associated unadjusted channel, and obtain the spectral features of the region corresponding to the blurred local feature area in the associated unadjusted channel one by one. For example, in the visible light channel, focus on whether the hyphal texture is continuous and unbroken; in the ultraviolet channel, focus on whether the absorption valley feature is clearly distinguishable; and in the near-infrared channel, focus on whether the reflection trend of the characteristic band is stable. Record the actual presentation state of the features of each channel.

[0127] Compare the spectral features with the standard features to identify differences in sharpness: if the edges of the absorption valleys in the actual features of the ultraviolet channel are blurred and do not match the sharp edges of the standard features, it is determined that the feature edges of this channel are not sharp enough; if the reflection trend of the near-infrared channel is stable and consistent with the standard features, it is determined that the sharpness of the reflection features meets the standard.

[0128] Based on the difference results, the auxiliary value of each associated non-adjustment channel is evaluated: if a channel has only slight brightness deficiency but complete core features, such as a visible light channel with continuous texture but darker appearance, it is determined that the blurred area of ​​the channel can be improved by gain adjustment; if a channel has severe feature breakage, it is excluded as an auxiliary adjustment channel, providing a clear basis for subsequent screening of auxiliary adjustment channels.

[0129] Establish a closed-loop data feedback mechanism for channel gain adjustment, record the light intensity data, gain parameters, feature comparison results and synthetic image quality for each adjustment, and periodically input the data into the spectral reflectance feature library of *Agaricus esculentus* mycelium and the initial gain parameter library to update the gain adjustment thresholds for each channel and improve the accuracy of gain control under extreme lighting conditions.

[0130] Working Principle: When a sudden burst of intense light occurs during lightning strikes inside the greenhouse, the channel segmentation module of the coupled IoT control system based on multispectral and mycelial feature libraries first identifies the spectral band corresponding to the intense light. If the intense light is concentrated in the ultraviolet (UV) band, the data acquisition module immediately increases the acquisition frequency of the UV channel, acquiring light intensity and spectral data of the UV channel every microsecond. The data acquisition module compares the real-time light intensity data with the judgment threshold of the UV channel. If the real-time light intensity data is greater than the judgment threshold for three consecutive times, and the sterilization lamp's operating parameters are displayed normally, the data acquisition module marks the UV channel as a state to be controlled. The feature library module retrieves the standard reflectance data of the UV channel from the *Agaricus esculentus* mycelial spectral reflectance feature library, compares the real-time reflectance data of the UV channel with the standard reflectance data, and obtains the feature comparison result. If the real-time reflectance is lower than the standard reflectance range, the gain control module selects a non-critical monitoring area at the edge of the substrate surface within the UV channel, lowers the gain parameter of that area, and acquires the pre-adjusted reflectance data. If the pre-adjusted reflectance data approaches the standard reflectance range, the gain control module applies the downward adjustment to the entire area of ​​the ultraviolet channel and re-acquires reflectance data for the entire area. If the reflectance data meets the standard, the gain control module completes the gain adjustment of the ultraviolet channel and checks the reflectance synergy between the ultraviolet channel, the visible light channel, and the near-infrared light channel to ensure that the spectral characteristics of the three channels are not unbalanced.

[0131] Judgment and regulation of the mycelial stage in the early fruiting stage:

[0132] During the pre-fruiting mycelial culture stage, spectral reflectance data of the mycelium were simultaneously collected through multiple independent channels including visible light, ultraviolet, and near-infrared. Standard reflectance data for this stage were retrieved from the *Pleurotus eryngii* mycelial spectral reflectance feature library. The real-time spectral reflectance characteristics were compared to determine whether the mycelial vigor and growth density met the standards. If insufficient mycelial vigor was found, manifested as a real-time spectral reflectance lower than the standard range in the feature library, the spectral output intensity of the supplementary light in the visible light channel was adjusted, while maintaining a stable gain in the ultraviolet channel to avoid inhibiting mycelial activity, ensuring that the mycelium forms a uniform and robust growth foundation before fruiting. During this period, the mycelial substrate images were continuously stitched together using a multi-channel image synthesis module to verify whether the mycelial growth status gradually improved with spectral adjustment, avoiding misjudgments of mycelial status caused by traditional single-channel monitoring.

[0133] Judgment and regulation application in the primordium stage:

[0134] After entering the primordium formation stage, the near-infrared channel is used to capture the spectral reflectance signal of the primordium. Combined with the standard spectral characteristics of the primordium in the spectral reflectance feature library of *Pleurotus ostreatus* mycelium, the appearance time, distribution density, and morphological integrity of the primordium are determined. If the primordium density is detected to be sparse, that is, the number of near-infrared reflectance peaks in the real-time spectrum is less than the standard value in the feature library, the irradiation intensity of the ultraviolet light channel is appropriately reduced. At the same time, the image clarity is optimized through the dedicated gain control module of the visible light channel to accurately observe the dynamics of primordium development and avoid abnormal primordium differentiation due to unsuitable spectrum. When the primordium morphology is irregular, that is, the edges of the primordium are blurred in the multi-channel composite image, the spectral proportion of the supplementary light in the visible light channel is finely adjusted to promote the development of the primordium into a regular morphology and ensure that the growth of the primordium stage meets the requirements of subsequent fruiting.

[0135] Judgment and regulation application in the small mushroom stage:

[0136] During the growth stage of the small mushrooms, spectral data of the mushroom surface color are collected through the visible light channel and compared with the standard color spectrum of the small mushroom stage in the spectral reflectance feature library of *Pleurotus ostreatus* mycelium to determine whether the mushrooms show abnormal colors such as yellowing or darkening. A multi-channel image synthesis module is used to stitch together complete mushroom surface images to observe the cap expansion and stipe thickness. If the stipe is found to be too thin (i.e., the spectral reflectance intensity corresponding to the stipe diameter in the image is lower than the standard in the feature library), the spectral proportion of the supplementary light in the near-infrared channel is adjusted to improve the mushrooms' nutrient absorption efficiency and ensure stable growth. During this period, the ultraviolet channel is used to assist in monitoring for early contamination of the mushroom surface, which is manifested as abnormal fluctuations in ultraviolet spectral reflectance. Spectral modulation is used in a timely manner to inhibit the growth of contaminants and avoid affecting the development of the small mushrooms.

[0137] Judgment and regulation application in the mid-mushroom stage:

[0138] During the mid-stage of mushroom growth, the focus is on monitoring the mushroom's condition and shape development. Spectral data acquired via the ultraviolet (UV) channel is compared with the UV reflectance characteristics of healthy mid-stage mushrooms in the feature library. The UV data is used to determine if there are any spectral anomalies on the mushroom surface caused by contamination by other microorganisms. Combined with detailed image observations from the visible light channel, the neatness of the cap edge and the uprightness of the stem are examined. If the mushroom shape is skewed (i.e., the angle between the mushroom's central axis and the horizontal axis in the multi-channel composite image deviates from the standard range in the feature library), the spectral distribution of the light source in the greenhouse is adjusted, increasing the output of specific wavelengths in the visible light channel to optimize the light environment for mid-stage mushroom growth. Simultaneously, the gain parameters of each channel are kept coordinated to ensure that image monitoring provides continuous and accurate feedback on the mushroom shape adjustment effect. When the mushroom appears shriveled (i.e., the near-infrared spectral reflectance is higher than the feature library standard, indicating insufficient moisture), the spectrum of the supplemental lighting is fine-tuned to reduce excessive moisture evaporation and maintain the plumpness of the mid-stage mushrooms.

[0139] Judgment and regulation of the large mushroom stage:

[0140] During the maturation stage of large mushrooms, the focus is on judging the uniformity of mushroom color, overall fullness, and integrity of mushroom shape. Spectral data collected through the visible light channel confirms whether the cap color meets the standard color for commercial mushrooms. This involves comparing the spectral data collected through the visible light channel with the standard color spectrum for large mushrooms in the feature library to avoid uneven color distribution. Spectral data collected through the near-infrared channel helps determine if the internal moisture content of the mushroom is suitable. When moisture is sufficient, the near-infrared reflectance is within the preset range of the feature library. If the edges of the large mushroom caps are found to be curled, meaning the edge morphology of the cap deviates from the standard in the multi-channel composite image, the ratio of visible light to near-infrared spectrum in the supplementary lighting is adjusted to enhance the light adaptability of the mushroom edge growth. The adjustment effect is verified based on the multi-channel image synthesis results. When mushroom deformities occur, such as bent stems or asymmetrical caps, the clarity of each channel image is optimized through a dedicated gain control module to more accurately observe deformed areas. The local light source spectrum is adjusted accordingly to ensure that the large mushrooms maintain a good commercial shape before harvest.

[0141] Example 2:

[0142] See Figures 1-2A coupled IoT control system based on multispectral and mycelial feature libraries includes: a channel segmentation module, which, when segmenting the photosensitive area of ​​the image sensor, collects full-spectrum data of all light sources in the greenhouse, reflectance data of *Pleurotus ostreatus* mycelium in each spectral band, absorbance data of *Pleurotus ostreatus* mycelium in each spectral band, and location data of all light sources in the greenhouse. Based on the full-spectrum data of all light sources in the greenhouse and the reflectance data of *Pleurotus ostreatus* mycelium in each spectral band, the visible light channel band range, ultraviolet light channel band range, and near-infrared light channel band range are obtained. Based on the visible light channel band range, ultraviolet light channel band range, and near-infrared light channel band range, combined with the location data of all light sources in the greenhouse and dedicated photosensitive elements and nanoscale filter strips, spectrally isolated visible light channels, ultraviolet light channels, and near-infrared light channels are obtained.

[0143] When a sudden burst of intense light occurs inside the greenhouse during a lightning strike, the channel segmentation module of the coupled IoT control system based on a multispectral and mycelial feature library first identifies the spectral band corresponding to the intense light. If the intense light is concentrated in the ultraviolet (UV) band, the data acquisition module immediately increases the acquisition frequency of the UV channel, acquiring light intensity and spectral data every microsecond. The data acquisition module compares the real-time light intensity data with the judgment threshold of the UV channel. If three consecutive real-time light intensity data are greater than the judgment threshold, and the sterilization lamp's operating parameters are displayed normally, the data acquisition module marks the UV channel as needing adjustment. The feature library module retrieves the standard reflectance data of the UV channel from the *Agaricus esculentus* mycelial spectral reflectance feature library, compares the real-time reflectance data of the UV channel with the standard reflectance data, and obtains the feature comparison result. If the real-time reflectance is lower than the standard reflectance range, the gain control module selects a non-critical monitoring area at the edge of the substrate surface within the UV channel, lowers the gain parameter of that area, and acquires the pre-adjusted reflectance data. If the pre-adjusted reflectance data approaches the standard reflectance range, the gain control module applies the downward adjustment to the entire area of ​​the ultraviolet channel and re-acquires reflectance data for the entire area. If the reflectance data meets the standard, the gain control module completes the gain adjustment of the ultraviolet channel and checks the reflectance synergy between the ultraviolet channel, the visible light channel, and the near-infrared light channel to ensure that the spectral characteristics of the three channels are not unbalanced.

[0144] The data acquisition module, when monitoring the light intensity and spectrum of each channel, collects light intensity distribution data, spectral distribution data, and corresponding light source operating parameters for each channel within a preset time period. Based on the light intensity distribution data, the average light intensity of each channel is obtained. Based on the spectral distribution data, the target spectral proportion for each channel is obtained. Based on the average light intensity and target spectral proportion of each channel, combined with the corresponding light source operating parameters, real-time light intensity data, and real-time spectral data, the desired control state is determined.

[0145] The feature library module, when constructing the spectral reflectance characteristics of *Pleurotus ostreatus* mycelium, collects healthy mycelium samples at each growth stage of *Pleurotus ostreatus*, corresponding greenhouse environmental data for each growth stage, and spectral reflectance data of healthy mycelium samples at each growth stage in each channel. Based on the healthy mycelium samples at each growth stage, growth stage classification labels are obtained. Based on the health status of the healthy mycelium samples at each growth stage, health status classification labels are obtained. Based on the spectral reflectance data of healthy mycelium samples at each growth stage in each channel, and combining the growth stage classification labels and health status classification labels, the *Pleurotus ostreatus* mycelium spectral reflectance feature library is obtained.

[0146] The gain control module, when adjusting channel gain, collects real-time reflectance data of the channel to be controlled, reflectance data of the non-critical monitoring area of ​​the channel to be controlled, and standard reflectance data of the channel to be controlled from the *Pleurotus ostreatus* mycelial spectral reflectance feature library. Based on the real-time reflectance data of the channel to be controlled and the standard reflectance data of the channel to be controlled from the *Pleurotus ostreatus* mycelial spectral reflectance feature library, the reflectance deviation type of the channel to be controlled is obtained. Based on the reflectance deviation type of the channel to be controlled, combined with the channel type, the direction of gain adjustment of the channel to be controlled is obtained. Based on the reflectance data of the non-critical monitoring area of ​​the channel to be controlled, combined with the direction of gain adjustment of the channel to be controlled, the magnitude of gain adjustment of the channel to be controlled is obtained. Based on the magnitude of gain adjustment of the channel to be controlled, combined with the reflectance data of the entire area of ​​the channel to be controlled and the standard reflectance data of the channel to be controlled from the *Pleurotus ostreatus* mycelial spectral reflectance feature library, the channel to be controlled after gain adjustment is obtained.

[0147] The image synthesis module, when synthesizing a complete material surface image, acquires the physical acquisition range of each channel image, the stitching boundary data of each channel image, the mycelial texture direction data of each channel texture observation area, the mycelial quantity data within the grid of each channel texture observation area, multiple cross-channel mycelial samples, the visible light channel direction baseline, and the visible light channel feature point coordinates. Based on the physical acquisition range and stitching boundary data of each channel image, the texture observation area is obtained. Based on the mycelial texture direction data of each channel texture observation area, the rotation angle of each channel image is obtained. Based on the mycelial quantity data within the grid of each channel texture observation area, the contrast of each channel image is obtained. Based on the rotation angle and contrast of each channel image, combined with each channel image, a complete material surface image is obtained. Based on the complete material surface image and multiple cross-channel mycelial samples, combined with the visible light channel direction baseline and the visible light channel feature point coordinates, a complete material surface image that has passed accuracy verification is obtained.

[0148] The collaborative adjustment module, when processing local feature blurring, collects data on the blurred areas of a single-channel adjusted region, spectral feature data of the associated unadjusted channel, standard feature data of the associated unadjusted channel from the *Pleurotus eryngii* mycelial spectral reflectance feature library, and reflectance data of the non-sensitive areas of the auxiliary adjustment channel. Based on the local feature blurring data of the single-channel adjusted region, the associated unadjusted channel is obtained. Based on the spectral feature data of the associated unadjusted channel and the standard feature data of the associated unadjusted channel from the *Pleurotus eryngii* mycelial spectral reflectance feature library, the auxiliary adjustment channel is obtained. Based on the reflectance data of the non-sensitive areas of the auxiliary adjustment channel, combined with the local feature blurring data of the single-channel adjusted region, the pre-adjustment direction and pre-adjustment magnitude of the auxiliary adjustment channel are obtained. Based on the pre-adjustment direction and pre-adjustment magnitude of the auxiliary adjustment channel, combined with the blurred associated region data of the auxiliary adjustment channel, a multi-channel composite image with local feature elimination and coherent spectral features is obtained.

[0149] When the collaborative adjustment module handles local feature blurring, it first locates the blurred area after single-channel adjustment using an image recognition device. If the blurred area exhibits broken hyphal texture, the collaborative adjustment module retrieves the original spectral feature data of that area in the associated non-adjusted channel (such as the visible light channel). The collaborative adjustment module then uses a spectral comparison device to obtain the standard feature data of the associated non-adjusted channel from the *Agaricus esculentus* hyphal spectral reflectance feature library. It compares the real-time spectral feature data of the associated non-adjusted channel with the standard feature data to perform an existing feature clarity analysis. If the analysis shows that the spectral features of the visible light channel are complete but the brightness is insufficient, the collaborative adjustment module determines the visible light channel as the auxiliary adjustment channel. The collaborative adjustment module selects a non-sensitive area adjacent to the blurred area with no significant hyphal distribution, increases the gain parameter of the non-sensitive area in the visible light channel, and observes the changes in hyphal texture in the blurred area in real time. If the hyphal texture in the blurred area gradually becomes clearer, the collaborative adjustment module records the increase magnitude as the pre-adjustment confirmation magnitude and the increase direction as the pre-adjustment confirmation direction. The collaborative adjustment module applies the pre-adjustment confirmation direction and pre-adjustment confirmation magnitude to the fuzzy correlation region of the visible light channel. It uses an image acquisition device to acquire a multi-channel composite image of the fuzzy correlation region. If the local feature fuzzy region in the image is eliminated and the spectral features of each channel are consistent, the collaborative adjustment module ends the gain adjustment of other channels.

[0150] The data closed-loop module collects the light intensity data, gain parameters, feature comparison results, synthetic image quality data, and the *Pleurotus eryngii* mycelial spectral reflectance feature library for each adjustment when updating control parameters. Based on the light intensity data and gain parameters adjusted each time, an adjustment data record is obtained. Based on the adjustment data record and feature comparison results, combined with the synthetic image quality data and the *Pleurotus eryngii* mycelial spectral reflectance feature library, an updated *Pleurotus eryngii* mycelial spectral reflectance feature library is obtained. Based on the updated *Pleurotus eryngii* mycelial spectral reflectance feature library, updated judgment thresholds and initial gain parameters for each channel are obtained.

[0151] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention based on the Internet of Things control system of multispectral and mycelial feature library and its inventive concept should be covered within the scope of protection of the present invention.

Claims

1. A coupled Internet of Things (IoT) control system based on multispectral and mycelial feature libraries, characterized in that, include: The channel division module is used to divide the photosensitive area of ​​the image sensor into multiple channels based on the spectral characteristics of the light source. The data acquisition module is used to collect light intensity distribution data and spectral distribution data from multiple channels. When the real-time light intensity or real-time spectrum of the corresponding channel is detected to be greater than the judgment threshold, the corresponding channel is marked as a state to be adjusted. The feature library module is used to construct a spectral reflectance feature library of mushroom mycelium. When the corresponding channel is marked as a state to be adjusted, the standard reflectance data of the corresponding channel in the feature library is retrieved, and the real-time reflectance is compared with the standard value to obtain the feature comparison result. The gain control module adjusts the gain of the corresponding channel based on the channel's state to be controlled and the feature comparison results. The image synthesis module is used to collect real-time image data from multiple channels, synthesize a complete fabric image, verify the accuracy of the complete fabric image, and if it meets the requirements, compare the mycelial features of the complete fabric image with the feature library standard. The collaborative adjustment module adjusts the gain of other channels if local features become blurred due to adjustment of a single channel. The data closed-loop module is used to record the light intensity data, gain parameters, feature comparison results and synthetic image quality of each adjustment, and input the data into the spectral reflectance feature library of *Agaricus esculentus* mycelium.

2. The coupled IoT control system based on multispectral and mycelial feature library according to claim 1, characterized in that, The method for acquiring real-time image data from multiple channels and synthesizing a complete fabric image includes: Define the physical range of image acquisition for multiple channels; mark the stitching boundary on the images of the corresponding channels, and delineate the texture observation area on both sides of the stitching boundary; Within the texture observation area of ​​the corresponding channel, the texture direction of the hyphae is obtained. The observation area is divided into multiple grids using the grid method, and the number of hyphae in each grid is recorded to obtain a reference table of texture direction and hyphae number. By comparing the reference tables of texture observation areas in multiple channels, the orientation deviation and density difference are obtained; Based on the orientation deviation, the rotation angle of the corresponding channel image is obtained, and based on the density difference, the image contrast of the corresponding channel is obtained; A complete fabric image is obtained by stitching together channel images based on rotation angle and image contrast.

3. The coupled IoT control system based on multispectral and mycelial feature library according to claim 2, characterized in that, The method for obtaining the texture direction of hyphae includes: A polarizing filter is installed in each channel, and the filter angle is divided into multiple angles to acquire multi-polarization images of the same area. By comparing the brightness distribution of the polarization angle image in each channel, the dominant direction of the hyphal texture in the corresponding channel can be obtained; By comparing the dominant directions of hyphal textures in multiple channels, the texture direction of the hyphal texture is obtained.

4. The coupled IoT control system based on multispectral and mycelial feature library according to claim 2, characterized in that, The method for verifying the accuracy of a complete fabric image includes: Multiple mycelia are selected as judgment samples from real-time images across multiple channels. Based on the judgment samples, a directional baseline is drawn in the real-time image of the visible light channel of multiple channels, and the angle between the baseline and the horizontal axis of the real-time image is measured, and the coordinates of multiple feature points are marked. Align the real-time image of the channel to be stitched with the visible light channel image to obtain the hyphal direction and coordinates of multiple feature points of the mycelium of the judgment sample in the real-time image of the channel to be stitched; and judge the hyphal direction and coordinates of multiple feature points of the mycelium of the judgment sample in the real-time image of the channel to be stitched to obtain the judgment result. If the angle between the hyphae direction and the baseline in the real-time image of the channel to be spliced ​​is lower than the preset angle and the coordinate deviation of multiple feature points is lower than the preset length, then the real-time image of the channel to be spliced ​​is accurate, and the complete fabric image meets the accuracy verification.

5. The coupled IoT control system based on multispectral and mycelial feature library according to claim 1, characterized in that, The method for dividing the photosensitive area of ​​an image sensor into multiple channels based on the spectral characteristics of a light source includes: Full-spectrum data of all light sources in the greenhouse were collected, and the spectral range and intensity peak of natural light, germicidal lamps and supplemental lights were recorded. The reflectance and absorptivity of the antler mushroom mycelium in each spectral band were measured to obtain the dataset. Based on the dataset and the location of all light sources in the studio, the wavelengths are divided into visible light channels, ultraviolet light channels, and near-infrared light channels.

6. The coupled IoT control system based on multispectral and mycelial feature library according to claim 4, characterized in that, Methods for acquiring light intensity distribution data and spectral distribution data from multiple channels include: Collect light intensity distribution data and spectral distribution data within a preset time period, and calculate the average light intensity and spectral target proportion of multiple channels; Collect the operating parameters of the light source corresponding to multiple channels; Based on the average light intensity of multiple channels and the proportion of spectral targets, the judgment thresholds for multiple channels are obtained; Collect real-time light intensity and real-time spectral data from multiple channels within a preset time period, and compare the real-time light intensity and real-time spectral data from multiple channels with a judgment threshold. If the real-time light intensity or real-time spectrum of the corresponding channel is greater than the judgment threshold and the working parameters are displayed normally, then the corresponding channel is marked as a candidate state to be controlled. The real-time light intensity and real-time spectral data of the channel marked as a candidate state to be regulated are compared with the judgment threshold multiple times. If the real-time light intensity and real-time spectral data are all greater than the judgment threshold multiple times, the corresponding channel is marked as a state to be regulated.

7. The coupled IoT control system based on multispectral and mycelial feature library according to claim 1, characterized in that, Methods for obtaining feature alignment results include: Based on the mycelial spectral reflectance feature library of *Pleurotus ostreatus*, the baseline spectral features and core feature types of multiple channels of the mycelial samples to be compared are obtained; the real-time spectral features of the mycelial samples to be compared are obtained. Environmental consistency calibration is performed on the real-time spectral features to obtain calibrated real-time spectral features. The calibrated real-time spectral features are then compared with the core feature types to obtain multi-channel comparison results. Based on the multi-channel comparison results, the overall consistency is judged; and the feature comparison results are obtained.

8. The coupled IoT control system based on multispectral and mycelial feature library according to claim 1, characterized in that, Methods for adjusting channel gain based on the channel's unadjusted state and feature comparison results include: Obtain the reflectivity deviation type of the channel to be controlled, and adjust the gain of the corresponding channel according to the type of the channel to be controlled and the reflectivity deviation type; Select non-critical monitoring areas within the channel to be regulated, adjust the gain of the corresponding channel according to the type of the channel to be regulated and the type of reflectivity deviation, collect reflectivity data of the channel to be regulated after gain adjustment, and obtain the pre-adjustment result; Based on the pre-adjustment results, the gain adjustment range is adjusted, and the reflectivity data of the channel to be adjusted is re-acquired and compared with the preset standard range in the feature library to verify whether the reflectivity meets the standard. If the reflectivity meets the standard, the gain adjustment of the channel to be adjusted is completed.

9. The coupled IoT control system based on multispectral and mycelial feature library according to claim 1, characterized in that, Methods for adjusting the gain of other channels include: Obtain the local feature blurring region and associated non-adjusted channel that appears after single-channel adjustment; Based on the blurred regions of local features and the associated non-adjusted channels, the sharpness of existing features is analyzed to obtain auxiliary adjusted channels; Based on the local feature fuzzy region and the adjacent non-sensitive region of the local feature fuzzy region, the pre-adjustment confirmation direction and pre-adjustment confirmation magnitude of the auxiliary adjustment channel are obtained; The pre-adjustment confirmation direction and pre-adjustment confirmation magnitude are applied to the fuzzy correlation region of the auxiliary channel, and a multi-channel composite image of the fuzzy correlation region is acquired; if the local feature fuzzy region in the multi-channel composite image is eliminated and the spectral features of each channel are consistent, then the gain adjustment of other channels ends.

10. The coupled IoT control system based on multispectral and mycelial feature library according to claim 9, characterized in that, Methods for performing existing feature sharpness analysis include: Obtain standard features from the spectral reflectance feature library of *Flammulina velutipes* mycelium, and obtain spectral features of regions corresponding to the blurred regions of local features in the associated unadjusted channels; By comparing spectral features with standard features, the differences in the sharpness analysis of existing features are obtained.

Images