Garden plant irrigation control method and system

Abstract

The invention provides a garden plant irrigation control method and system, and relates to the technical field of garden plant irrigation control, and the method comprises the steps: periodically collecting a canopy image of a target plant in a garden scene through a visual perception device, carrying out the preprocessing and feature extraction of the image, and obtaining the plant visual features of the target plant; then, according to the visual features and water shortage judgment standards preset for different plant types, whether the target plant is in an acute water stress state or not is judged. Once it is confirmed that the plant is in an acute water stress state, the system immediately triggers emergency irrigation of the area where the target plant is located, and the problem that in the prior art, an automatic irrigation system excessively depends on environmental parameter measurement and weather prediction when facing sudden extreme weather, and the irrigation efficiency is high is effectively solved. The acute water stress of the plant cannot be found and responded in time.

Description

Technical Field

[0001] This application relates to the field of irrigation control technology for garden plants, and more specifically, to a method and system for controlling irrigation for garden plants. Background Technology

[0002] In modern landscape management, automated irrigation control systems are widely used to improve water resource utilization efficiency and reduce manual maintenance costs. These systems typically rely on environmental parameter measurements (such as soil moisture) and weather forecasts to make irrigation decisions. For example, a basic automated irrigation system uses soil moisture sensors to detect soil moisture; when the moisture level falls below a preset threshold, the system initiates irrigation until the moisture level returns to a suitable range. This soil moisture feedback-based control method, compared to timed watering, more accurately meets the basic water needs of plants and avoids unnecessary water waste.

[0003] However, in practical landscaping applications, the situation is often more complex. For landscape design and biodiversity considerations, various plant types are often mixed within the same irrigation area, such as deep-rooted trees, shallow-rooted shrubs, and water-demanding flowers. In this case, relying solely on a few soil moisture sensors buried at standard depths cannot comprehensively represent the needs of all plants. For example, when the sensors show sufficient moisture in the deep soil, surface flowers may already be wilting due to lack of water; conversely, frequent watering to care for surface flowers may lead to chronically overwatering of the roots of deep-rooted trees, increasing the risk of rot.

[0004] To address these issues, some improved technical solutions attempt to establish more comprehensive decision-making bases. These solutions combine weather forecasts obtained from external meteorological services, including rainfall, temperature, wind speed, and solar radiation intensity, with internally stored plant biological characteristics such as root depth, water evaporation rate, and drought tolerance, to predict future soil moisture trends and generate more refined irrigation plans. For example, if heavy rain is predicted, the system will delay or cancel irrigation even if current soil moisture is low. This predictive control method effectively optimizes irrigation efficiency in most cases.

[0005] However, the reliability of such a forecast-dependent system rests entirely on the accuracy of weather forecasts. In certain special circumstances, this can introduce new problems. For example, when a weather forecast indicates mild weather, the system may plan infrequent irrigation. However, a sudden, unpredictable heat wave may occur, causing a sharp rise in ambient temperature and a sudden drop in humidity. Under such extreme conditions, plant water evaporation is abnormally amplified, especially for plants with large leaf areas and shallow root systems. These plants will quickly enter a state of water deficiency, exhibiting obvious symptoms of acute dehydration such as wilting and leaf curling. At this point, the irrigation control system continues to operate according to the original plan because soil moisture sensor readings may not have changed significantly, and the system has not received an extreme weather warning and cannot adjust in time. The end result is that the system "believes" everything is normal, but in reality, some sensitive plants have already suffered acute water stress. If this continues, irreversible damage may occur before the next planned irrigation. Summary of the Invention

[0006] This application provides a method and system for controlling irrigation of garden plants, aiming to solve the technical problem that existing automated irrigation systems, when faced with sudden extreme weather, cannot respond in a timely manner to acute water stress in plants due to over-reliance on environmental parameter predictions and soil moisture data, thus causing plant damage.

[0007] On the one hand, this application provides a method for controlling irrigation of garden plants, including:

[0008] Images of the canopy of target plants in a garden scene are periodically collected using visual sensing devices.

[0009] The canopy image is preprocessed and its features are extracted to obtain the visual features of the target plant.

[0010] Based on the plant's visual characteristics and the preset water shortage judgment criteria for different plant species, determine whether the target plant is in an acute water stress state.

[0011] If the target plant is under acute water stress, emergency irrigation is triggered in the area where the target plant is located.

[0012] Optionally, the step of determining whether the target plant is in an acute water stress state based on the plant's visual characteristics and preset water shortage judgment criteria for different plant species includes:

[0013] The light intensity data, ambient temperature data, and soil moisture data of the area where the target plant is located are obtained by using a light intensity sensor, an ambient temperature sensor, and a soil moisture sensor.

[0014] Based on the plant's visual characteristics, the soil moisture data, and the preset water shortage judgment criteria for different plant species, a first-level verification is performed to determine whether the target plant is in an acute water stress state, and the first-level verification result is obtained.

[0015] When the first-layer verification result indicates that the plant's visual characteristics meet the water shortage judgment criteria and the soil moisture data is sufficient, a second-layer verification is performed based on the visual characteristics, the soil moisture data, the light intensity data, and the ambient temperature data to determine whether the target plant is in an acute water stress state, and a second-layer verification result is obtained.

[0016] By combining the results of the first layer of verification and the results of the second layer of verification, it is confirmed whether the target plant is in a state of acute water stress.

[0017] Optionally, the step of determining whether the target plant is in an acute water stress state based on the plant's visual characteristics and preset water shortage judgment criteria for different plant species includes:

[0018] The growth stage of the target plant is identified based on the plant's visual characteristics;

[0019] Based on the growth stage, a water shortage judgment standard matching the growth stage is selected; wherein, when the growth stage of the target plant is the seedling stage, the water shortage judgment standard is leaf morphology, and when the growth stage of the target plant is the flowering stage, the water shortage judgment standard is greenness index.

[0020] Based on the selected water shortage criteria, determine whether the target plant is in a state of acute water stress.

[0021] Optionally, after the step of triggering emergency irrigation of the area where the target plant is located if the target plant is under acute water stress, the following steps are included:

[0022] Continuously monitor the visual characteristics of the target plant;

[0023] Based on the type of the target plant, a recovery judgment criterion corresponding to the type of the target plant is matched;

[0024] The monitored plant visual characteristics were compared with the selected recovery criteria;

[0025] Emergency irrigation should be stopped when the plant's visual characteristics meet the recovery criteria and remain stable.

[0026] Optionally, the step of preprocessing and feature extraction of the canopy image to obtain the visual features of the target plant includes:

[0027] The canopy image is processed to locate the target plant region, and the spatial range of the target plant in the image is identified;

[0028] The located area is segmented into leaf-stem regions to separate the leaf and stem regions.

[0029] Multi-dimensional feature extraction was performed on the segmented leaf and stem regions to obtain the leaf color saturation, hue, greenness index, photochemical reflectance index, as well as the sub-pixel level changes in leaf morphology and the stem micro-drooping angle.

[0030] Optionally, the step of locating the target plant region in the canopy image includes:

[0031] The canopy image is subjected to foreground-background semantic segmentation to distinguish between plant foreground and non-plant background in the image;

[0032] Plant individual target detection processing is performed on the foreground area of ​​plants to identify specific plant individuals in the foreground;

[0033] Based on the morphological, color, and texture characteristics of the plant individual, and in conjunction with a pre-defined target plant feature library, a target plant matching verification process is performed to determine whether the individual is the target plant.

[0034] If the target plant is verified, morphological completion processing of the occluded area is performed—by associating the visible part of the target plant with the occluded part through the motion trajectory of adjacent frame images, the complete canopy area is obtained.

[0035] Optionally, the step of performing multi-dimensional feature extraction processing on the segmented leaf and stem regions includes:

[0036] Obtain the local light intensity distribution information of the area where the target plant is located;

[0037] Based on the distribution information, light compensation processing is performed on the leaf area to correct the color and brightness deviations caused by uneven light.

[0038] For the leaf areas that have undergone the light compensation treatment, color space conversion and standardization are performed to extract the color saturation, hue, and greenness index of the leaves.

[0039] Based on the local light intensity distribution information, the calculation of the photochemical reflectance index is adaptively adjusted to eliminate the influence of light intensity changes on the photochemical reflectance index.

[0040] A motion compensation algorithm based on image sequences is used to eliminate the influence of slight swaying caused by visual sensing device vibration or wind on the sub-pixel level changes in leaf morphology and the extraction of stem micro-drooping angle, thereby obtaining the sub-pixel level changes in leaf morphology and stem micro-drooping angle.

[0041] Optionally, the step of employing an image sequence-based motion compensation algorithm to eliminate the influence of minute swaying caused by visual sensing device vibration or wind on sub-pixel-level changes in leaf morphology and the extraction of stem micro-drooping angle includes:

[0042] Motion features are extracted from the target plant region in the image sequence;

[0043] Based on the extracted motion features, the motion of the target plant is decomposed to separate the overall swaying component and the local morphological change component of the target plant.

[0044] By combining real-time wind speed and direction information, the wind force impact assessment of the overall sway component is performed to determine whether it is consistent with the wind force change trend, thereby eliminating the impact of micro-sway caused by visual sensing device vibration or wind force on the sub-pixel level changes in leaf morphology and the extraction of stem micro-drooping angle.

[0045] Optionally, the step of combining real-time wind speed and direction information to assess the wind impact of the overall oscillation component to determine whether it is consistent with the wind change trend includes:

[0046] When the evaluation results show that the overall sway component is consistent with the wind force change trend, the overall sway component is identified as a non-physiological sway and is removed.

[0047] Extract the subpixel-level changes in leaf morphology and the slight drooping angle of the stem from the remaining local morphological change components.

[0048] On the other hand, this application provides a garden plant irrigation control system, which includes:

[0049] The visual perception module is used to periodically acquire canopy images of target plants in a garden scene through visual perception devices;

[0050] The data processing module is used to preprocess and extract features from the canopy image to obtain the visual features of the target plant.

[0051] The status judgment module is used to determine whether the target plant is in an acute water stress state based on the plant's visual characteristics and preset water shortage judgment criteria for different plant species.

[0052] An irrigation triggering module is used to trigger emergency irrigation of the area where the target plant is located if the target plant is in a state of acute water stress.

[0053] This application discloses a method and system for controlling irrigation of garden plants. It periodically acquires canopy images of target plants in a garden scene using a visual sensing device, and preprocesses and extracts features from the images to obtain the visual characteristics of the target plants. Subsequently, based on these visual characteristics and preset water shortage judgment criteria for different plant species, it determines whether the target plant is under acute water stress. Once confirmed, emergency irrigation of the area containing the target plant is immediately triggered. This effectively solves the problem in existing automated irrigation systems that, when faced with sudden extreme weather events (such as unpredictable heat waves), rely excessively on environmental parameter measurements (such as soil moisture) and weather forecasts, failing to detect and respond to acute water stress in a timely manner. Traditional systems often continue to operate according to the original plan even before soil moisture sensor readings change significantly or extreme weather warnings are received, causing some sensitive plants to suffer irreparable damage before the next scheduled irrigation.

[0054] This application, by directly monitoring the plant's own visual characteristics (such as leaf morphology and color changes), can detect physiological water shortage signals in plants earlier and more directly. Even when soil moisture data has not yet reflected the problem or weather forecasts are inaccurate, it can promptly detect acute water stress in plants. This control method based on plant feedback makes irrigation decisions more precise and real-time, avoiding irrigation delays caused by inaccurate environmental forecasts or sensor lag, thereby effectively protecting garden plants from acute water shortage damage and improving the intelligence and precision of garden management. Attached Figure Description

[0055] To illustrate this application more clearly, the accompanying drawings used in the embodiments will be briefly described below. Obviously, those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0056] Figure 1 The diagram above illustrates a flowchart of an irrigation control method for garden plants in an embodiment.

[0057] Figure 2 The diagram above illustrates a schematic representation of a garden plant irrigation control system in an embodiment.

[0058] Figure reference numerals: 100, Garden plant irrigation control system; 10, Visual perception module; 20, Data processing module; 30, Status judgment module; 40, Irrigation triggering module. Detailed Implementation

[0059] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0060] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0061] Traditional garden irrigation control systems rely primarily on environmental parameter measurements (such as soil moisture) and weather forecasts to make irrigation decisions. However, this approach has limitations in practical applications. For example, it is difficult to fully represent the needs of all plants in mixed planting areas, and in the event of sudden extreme weather, the system may fail to respond promptly to acute water stress in plants, leading to plant damage.

[0062] like Figure 1 The illustration shows a flowchart of an irrigation control method for garden plants in an exemplary embodiment. This application proposes an irrigation control method for garden plants, comprising:

[0063] S10 periodically collects canopy images of target plants in a garden scene using a visual perception device;

[0064] Visual perception devices refer to equipment capable of acquiring image or video data, such as high-definition cameras, multispectral cameras, or visual sensors mounted on drones. These devices can periodically acquire images of the canopy of plants in a garden setting, providing raw data for subsequent plant status analysis.

[0065] Canopy images refer to image data containing the plant canopy area acquired through visual perception devices. The canopy is the collection of the above-ground parts of a plant, and its morphology, color, and other visual characteristics can intuitively reflect the plant's health status and water condition.

[0066] S20, preprocess and extract features from the canopy image to obtain the visual features of the target plant;

[0067] Among them, plant visual characteristics refer to visual information extracted from canopy images that can characterize the physiological state of plants, such as leaf color, shape, and greenness index. These characteristics are important criteria for determining whether a plant is under acute water stress.

[0068] S30, based on the plant's visual characteristics and the preset water shortage judgment criteria for different plant species, determine whether the target plant is in an acute water stress state;

[0069] Among them, the water shortage judgment criteria refer to the thresholds or models preset for different plant species to determine whether a plant is in a state of acute water stress. Different plant species respond differently to water stress, so it is necessary to establish personalized judgment criteria.

[0070] Acute water stress refers to the rapid physiological stress response of plants due to insufficient water supply, such as wilting, curling, and darkening of leaves. This condition requires timely intervention to avoid irreversible damage to the plant.

[0071] S40, if the target plant is under acute water stress, trigger emergency irrigation of the area where the target plant is located.

[0072] When a plant is identified as being under acute water stress, irrigation should be initiated immediately in the area where the plant is located to quickly replenish water and alleviate the stress.

[0073] This application introduces visual perception technology to directly monitor the physiological state of plants, thereby enabling more timely and accurate identification of acute water stress in plants and triggering emergency irrigation, effectively making up for the shortcomings of traditional systems in responding to emergencies and in refined management.

[0074] Firstly, visual sensing devices can be fixed network cameras installed within the garden area, automatically capturing images of the target plant canopy at preset time intervals (e.g., several times per hour or per day). Another approach is to use drones equipped with high-definition cameras, following preset flight paths and schedules to cruise and photograph the garden area, acquiring large-scale images of the plant canopy. Additionally, handheld or vehicle-mounted visual sensing devices can be used, with inspection personnel periodically collecting images of the garden plants.

[0075] Secondly, after acquiring the canopy image, a series of processing steps are required to extract useful visual features. One approach is to first perform preprocessing operations such as denoising and color correction on the image, and then use image processing algorithms (such as edge detection and region growing) to identify the leaf regions of the target plant and calculate features such as the average color and texture of the leaves. Another approach is to use deep learning models, such as convolutional neural networks (CNNs), to directly analyze the canopy image and automatically extract high-dimensional plant visual features. These features can more comprehensively characterize the physiological state of the plant. For example, a CNN model can be trained to identify the greenness index of leaves, the degree of leaf curling, etc.

[0076] Next, after obtaining the plant's visual characteristics, they need to be compared with preset water shortage criteria. One approach is to preset one or more visual characteristic thresholds for each plant. For example, when the leaf greenness index is below a certain threshold, or the degree of leaf curling exceeds a certain threshold, it is judged to be in an acute water stress state. These thresholds can be set through prior experiments and expert experience. Another approach is to build a machine learning-based classification model that takes plant visual characteristics as input and outputs a judgment result on whether the plant is in an acute water stress state. This model can be trained using a large amount of labeled plant image data to accurately identify the water shortage symptoms of different plants. For example, canopy images of different plants in normal and water-shortened states can be collected and labeled with their water status, and then these data can be used to train classifiers such as support vector machines (SVM) or random forests.

[0077] Finally, when the target plant is determined to be under acute water stress, irrigation needs to be initiated immediately. One approach is to send an activation command to a solenoid valve associated with the target plant's location, causing the irrigation sprinklers or drip irrigation system to begin supplying water. The irrigation volume and duration can be preset based on factors such as plant species, stress level, and soil type. Another approach is to dynamically adjust the irrigation intensity and duration based on real-time monitoring of changes in plant visual characteristics until the plant's visual characteristics return to normal. For example, the leaf chlorophyll index can be continuously monitored, and emergency irrigation can automatically stop once it returns to normal and remains at that level for a period of time.

[0078] The irrigation control method for garden plants proposed in this application provides a solution for timely and accurate response to acute water stress in plants by integrating visual perception technology with intelligent judgment mechanism.

[0079] Specifically, this method first uses visual sensing devices to periodically acquire canopy images of target plants in a garden setting. This image data forms the basis for subsequent analysis, providing the most direct information about the plant's current physiological state. Subsequently, the acquired canopy images are preprocessed and feature extracted to obtain plant visual characteristics that quantify the plant's physiological state. These characteristics may include leaf color saturation, greenness index, and morphological changes, which are key indicators for determining whether a plant is water-deficient.

[0080] After acquiring the plant's visual characteristics, the system uses these characteristics and pre-defined water shortage criteria for different plant species to determine whether the target plant is under acute water stress. This determination process is the core of this application. It breaks through the limitations of traditional methods that rely solely on soil moisture or weather forecasts, directly identifying the plant's physiological stress response from its own visual manifestations. For example, when obvious wilting or abnormal coloring of a plant's leaves is detected, even if the soil moisture sensor indicates sufficient soil moisture, it can promptly identify that the plant may be suffering from acute water stress.

[0081] Once a target plant is determined to be under acute water stress, emergency irrigation of the area containing that plant will be triggered immediately. This emergency response mechanism ensures that plants receive timely replenishment when subjected to acute water stress, avoiding irreversible damage caused by waiting for regular irrigation cycles or relying on inaccurate predictions. The entire process forms a closed-loop control system, with each link working closely together from visual perception to data processing, intelligent judgment, and emergency response, jointly addressing the shortcomings of traditional garden irrigation systems in coping with sudden extreme environments and acute water stress in plants.

[0082] Compared to traditional garden irrigation control methods, this application offers significant advantages and innovations. Traditional methods primarily rely on soil moisture sensors and weather forecast data. For example, when a weather forecast indicates mild weather, an infrequent irrigation plan is formulated. However, a sudden, unpredictable heat wave occurs, causing a sharp rise in ambient temperature and a sudden drop in humidity. Under such extreme conditions, plant water evaporation is abnormally aggravated, especially for plants with large leaf areas and shallow root systems, which quickly enter a state of water deficiency, exhibiting obvious symptoms of acute water shortage such as wilting and leaf curling. At this time, the irrigation control system continues to operate according to the original plan because the soil moisture sensor readings may not have changed significantly, and no extreme weather warning has been received, preventing timely adjustments. The end result is that everything is "perceived" to be normal, but in reality, some sensitive plants have already suffered acute water stress, which, if continued, may cause irreparable damage before the next planned irrigation.

[0083] The core innovation of this application lies in its direct assessment of plant water status by analyzing the plant's visual behavior, rather than relying solely on indirect measurements of environmental parameters. This direct perception of plant physiological state enables more timely and accurate identification of acute water stress, especially in cases of rapid water shortage caused by sudden extreme weather or localized environmental anomalies. By triggering emergency irrigation of the area where the target plant is located if it is under acute water stress, this application achieves a rapid response to acute water stress, effectively avoiding plant damage caused by information lag or inaccurate predictions in traditional systems. This irrigation decision-making mechanism, centered on the plant's own state, significantly improves the intelligence and refined management capabilities of garden plant irrigation.

[0084] In some embodiments, the step of determining whether the target plant is in an acute water stress state based on the plant's visual characteristics and preset water shortage judgment criteria for different plant species includes:

[0085] The light intensity data, ambient temperature data, and soil moisture data of the area where the target plant is located are obtained by using a light intensity sensor, an ambient temperature sensor, and a soil moisture sensor.

[0086] Based on the plant's visual characteristics, the soil moisture data, and the preset water shortage judgment criteria for different plant species, a first-level verification is performed to determine whether the target plant is in an acute water stress state, and the first-level verification result is obtained.

[0087] When the first-layer verification result indicates that the plant's visual characteristics meet the water shortage judgment criteria and the soil moisture data is sufficient, a second-layer verification is performed based on the visual characteristics, the soil moisture data, the light intensity data, and the ambient temperature data to determine whether the target plant is in an acute water stress state, and a second-layer verification result is obtained.

[0088] By combining the results of the first layer of verification and the results of the second layer of verification, it is confirmed whether the target plant is in a state of acute water stress.

[0089] Specifically, light intensity sensors, ambient temperature sensors, and soil moisture sensors are configured to monitor the microenvironmental conditions of the target plants in real time. Light intensity data reflects the light energy received by the plants, ambient temperature data indicates the air temperature surrounding the plants, and soil moisture data directly quantifies the amount of water available to the plant roots. Acquiring this multi-dimensional environmental data aims to provide more comprehensive contextual information for assessing the plant's water stress state.

[0090] Furthermore, the first-layer verification is designed as a preliminary and rapid judgment mechanism. In this verification process, plant visual characteristics (such as leaf morphology, color saturation, and greenness index) are combined with soil moisture data and compared with preset water shortage judgment criteria. The aim is to quickly identify acute water stress when plant visual characteristics clearly show signs of water shortage and soil moisture is indeed low.

[0091] As an optional implementation, a second-level verification is initiated when the first-level verification results show that the plant's visual characteristics meet the criteria for water shortage, but the soil moisture data indicates sufficient moisture. This situation usually means that the plant may not be exhibiting visual stress symptoms due to soil water shortage. For example, under high temperature or strong light conditions, the plant may temporarily wilt due to excessive transpiration. The second-level verification further incorporates light intensity and ambient temperature data to conduct a deeper, more comprehensive analysis of the plant's visual characteristics. By assessing the impact of these environmental factors on the plant's physiological state, it is possible to more accurately distinguish between symptoms caused by acute water stress and similar symptoms caused by other environmental factors.

[0092] Therefore, the results of the first and second layers of verification are combined to ultimately confirm whether the target plant is under acute water stress. This hierarchical verification mechanism ensures the rigor and accuracy of the judgment, avoiding misjudgments that may result from a single-dimensional assessment.

[0093] The technical solution of this application effectively addresses the limitations of relying solely on plant visual characteristics to determine water stress by introducing multi-dimensional environmental sensor data and employing a hierarchical verification mechanism. First, light intensity sensors, ambient temperature sensors, and soil moisture sensors provide real-time, objective data on the plant's environment, compensating for the shortcomings of visual characteristics in reflecting the plant's internal physiological state and the influence of the external environment. Second, the first-layer verification combines plant visual characteristics with soil moisture data, enabling rapid identification of typical acute water stress situations where visual symptoms are consistent with soil water shortage. More importantly, when the first-layer verification encounters a "contradictory" situation where visual symptoms and soil moisture data are inconsistent (i.e., visually indicating water shortage but with sufficient soil moisture), the second-layer verification is triggered. At this stage, visual characteristics, soil moisture data, and light intensity and ambient temperature data are comprehensively considered to provide a more refined assessment of the plant's true state. For example, if visual wilting occurs under high temperature and strong light but soil moisture is normal, it may be judged as non-acute water stress, thus avoiding misjudgment. This layered, multi-dimensional judgment logic enables a more accurate distinction between physiological responses caused by genuine dehydration and temporary physiological changes caused by other environmental stresses (such as high temperature and strong light), thereby ensuring the accuracy of the judgment.

[0094] Through the above technical solution, this application can significantly improve the accuracy and reliability of judging the acute water stress state of garden plants. Specifically, by introducing light intensity data, ambient temperature data, and soil moisture data, and adopting a stratified verification mechanism, it can effectively avoid misjudgments that may be caused by judging based on a single visual feature. For example, it reduces the possibility of temporary wilting caused by high temperature or strong light being misjudged as acute water shortage, thereby avoiding unnecessary emergency irrigation and saving water resources. At the same time, this technical solution also improves the ability to identify early or subtle water shortage symptoms, ensuring that irrigation can be triggered in a timely manner when plants truly need it, effectively protecting the healthy growth of garden plants. This comprehensive judgment method makes irrigation decisions more scientific and precise, improving the intelligence level of the entire garden plant irrigation control system.

[0095] In some optional embodiments, suppose that on a hot summer afternoon, the canopy images periodically collected by the visual sensing device show that the leaves of a target plant are slightly drooping and curling, which meets the criteria for water shortage in plant visual characteristics. At this time, the step of determining whether the target plant is in a state of acute water stress will be initiated. First, real-time data of the area is acquired through a light intensity sensor, an ambient temperature sensor, and a soil moisture sensor. For example, the light intensity data is high, the ambient temperature data is 35°C, and the soil moisture data shows a sufficient state.

[0096] Next, the first layer of verification was performed. The plant's visual characteristics (drooping and curled leaves) met the criteria for water shortage, but the soil moisture data showed sufficient moisture, creating a contradiction in the first layer's results. This triggered a second layer of verification. In this second layer, the plant's visual characteristics, sufficient soil moisture data, high light intensity data, and an ambient temperature of 35°C were comprehensively analyzed. Based on this information, it was determined that the drooping and curling of the target plant's leaves were more likely due to excessive transpiration caused by high temperature and strong light, rather than acute water stress. Therefore, the second layer verification confirmed that the target plant was not under acute water stress. Finally, combining the results of the first and second layers of verification, it was confirmed that the target plant was not currently under acute water stress, thus avoiding unnecessary emergency irrigation.

[0097] In contrast, in another scenario, if the visual sensing device also detects drooping and curling leaves on the target plant, and the soil moisture sensor data also shows a low humidity state, then in the first layer of verification, both the plant's visual characteristics and the soil moisture data meet the criteria for water shortage. The first layer of verification directly confirms that the target plant is in a state of acute water stress. In this case, there is no need for a second layer of verification; emergency irrigation of the area where the target plant is located will be triggered directly.

[0098] In some embodiments, the step of determining whether the target plant is in an acute water stress state based on the plant's visual characteristics and preset water shortage judgment criteria for different plant species includes:

[0099] The growth stage of the target plant is identified based on the plant's visual characteristics;

[0100] Based on the growth stage, a water shortage judgment standard matching the growth stage is selected; wherein, when the growth stage of the target plant is the seedling stage, the water shortage judgment standard is leaf morphology, and when the growth stage of the target plant is the flowering stage, the water shortage judgment standard is greenness index.

[0101] Based on the selected water shortage criteria, determine whether the target plant is in a state of acute water stress.

[0102] Specifically, by analyzing plant visual features extracted from canopy images, such as leaf quantity, leaf size, stem height, the appearance and development of flower buds or fruits, and overall canopy structure, and combining this with a pre-defined plant growth model or a machine learning-based growth stage classification model, the current life cycle stage of the target plant can be determined. For example, detecting the formation of flower buds or the opening of flowers can indicate that the plant has entered the flowering period, or the rapid increase in the number and size of leaves can indicate that the plant is in the seedling or vegetative growth stage.

[0103] The selection of water shortage criteria based on the growth stage aims to ensure that the adopted criteria most accurately and sensitively reflect the plant's water status at that specific growth stage. Plants at different growth stages exhibit varying sensitivities and manifestations of water stress. For example, when the target plant is in the seedling stage, the water shortage criterion can be set to leaf morphology. This is because seedling leaves are typically small and fragile, reacting rapidly to water changes; even slight water shortage can lead to noticeable morphological changes such as leaf curling, drooping, or wilting. When the target plant is in the flowering stage, the water shortage criterion can be set to the greenness index. During flowering, plants require a large amount of water to support flower opening and pollination. Changes in the overall leaf greenness index (e.g., calculated by the proportion of green pixels in an image or the intensity of a specific color channel) at this time can more sensitively reflect the plant's overall photosynthetic efficiency and water use status, thus indirectly indicating the occurrence of water stress.

[0104] The technical solution of this application solves the problem of misjudgment that may be caused by single or static judgment criteria in traditional methods by introducing the identification of the target plant's growth stage and dynamically adjusting the water shortage judgment criteria based on the identification results. Specifically, firstly, through in-depth analysis of the plant's visual characteristics, the growth stage of the target plant can be accurately identified, such as the seedling stage, vegetative growth stage, flowering stage, or fruiting stage. Secondly, for different growth stages, specific visual indicators that best reflect the plant's water status at that stage are selected as the water shortage judgment criteria. For example, in the seedling stage, because the plant is sensitive to water changes and leaf morphology changes are obvious, leaf morphology is selected as the main judgment criterion, enabling timely capture of early water stress signals. During the flowering stage, because the plant's physiological activities are vigorous and its water demand is high, changes in the greenness index can more comprehensively and sensitively reflect the overall health status and water supply of the plant, thus avoiding missing the best irrigation time due to focusing only on morphological changes. Therefore, through this adaptive judgment mechanism, it is ensured that the most appropriate indicators can be used to assess the plant's acute water stress state at any growth stage.

[0105] Through the above technical solution, this application can significantly improve the accuracy and timeliness of judging acute water stress in garden plants. By considering the physiological characteristics and water requirements of plants at different growth stages, the selected water shortage judgment criteria are more targeted, effectively avoiding misjudgments or omissions caused by using inappropriate criteria. For example, in the seedling stage, by paying attention to subtle changes in leaf morphology, water stress can be detected earlier, preventing seedling damage due to water shortage; during the flowering period, by monitoring the greenness index, the overall health of the plant can be more comprehensively assessed, ensuring normal flower development and improving ornamental value or fruit yield. This dynamic adaptive judgment mechanism makes irrigation decisions more scientific and rational, not only helping to maintain healthy plant growth but also optimizing water resource utilization efficiency, reducing unnecessary irrigation, thereby lowering operating costs and promoting sustainable development.

[0106] In some optional embodiments, assuming a garden setting requires irrigation control for a batch of roses, visual perception devices periodically acquire images of the rose canopy. These images are then preprocessed and feature extracted to obtain the plant's visual characteristics. Initially, when the roses are identified as seedlings, leaf morphology is used as the criterion for water shortage. When visual feature analysis shows slight curling or drooping of the rose leaves, an acute water stress state is immediately identified, triggering emergency irrigation of the rose area. As the roses grow, their visual characteristics are continuously monitored, identifying when they enter the flowering period, such as bud formation and flower opening. At this point, the water shortage criterion is automatically switched to the greenness index. When the greenness index falls below a preset threshold, indicating a decrease in overall photosynthetic efficiency or leaf chlorosis, an acute water stress state is identified, and emergency irrigation is initiated. This strategy of dynamically adjusting the judgment criteria according to the growth stage ensures the most precise water management at different stages of the rose's life cycle, thereby guaranteeing its healthy growth and good flowering performance.

[0107] In some embodiments, after the step of triggering emergency irrigation of the area where the target plant is located if the target plant is in a state of acute water stress, the following is included:

[0108] Continuously monitor the visual characteristics of the target plant;

[0109] Based on the type of the target plant, a recovery judgment criterion corresponding to the type of the target plant is matched;

[0110] The monitored plant visual characteristics were compared with the selected recovery criteria;

[0111] Emergency irrigation should be stopped when the plant's visual characteristics meet the recovery criteria and remain stable.

[0112] After emergency irrigation is initiated, the visual sensing device continues to acquire canopy images of the target plant at a preset frequency (e.g., every few minutes or in real time). The data processing module then preprocesses and extracts features from these images to obtain real-time visual characteristics of the target plant. These visual characteristics may include, but are not limited to, leaf color saturation, hue, greenness index, photochemical reflectance index, as well as sub-pixel-level changes in leaf morphology and stem drooping angles, used to dynamically track changes in the plant's physiological state.

[0113] For different types of garden plants, one or more sets of visual characteristic thresholds or patterns are pre-set to determine whether a plant has recovered from water stress. This allows for the matching of recovery judgment criteria to the target plant species. For example, for some plants, recovery judgment criteria might involve a return of the greenness index to a specific range and a reduction in leaf wilting to a certain level; for others, the focus might be on the recovery of the stem's slight drooping angle. These criteria are stored in the system and invoked during monitoring based on the target plant's identification results to ensure the accuracy and specificity of the judgment.

[0114] The visual characteristics of the target plant acquired in real time are compared with the matched recovery judgment criteria. This comparison process aims to assess whether the current visual performance of the plant has reached or exceeded the preset recovery status index.

[0115] Once the comparison results show that the plant's visual characteristics have met the recovery criteria, irrigation will not be stopped immediately. Instead, the system will further monitor whether these visual characteristics that meet the recovery criteria remain stable for a period of time (e.g., 30 minutes, 1 hour, or longer). This "continuous stability" mechanism aims to avoid prematurely stopping irrigation due to short-term, non-continuous improvement, ensuring that the plant truly recovers fully from acute water stress and thus preventing secondary stress caused by insufficient irrigation.

[0116] The technical solution of this application continuously monitors the visual characteristics of the target plant after emergency irrigation is triggered, enabling real-time acquisition of changes in the plant's physiological state. By matching appropriate recovery judgment criteria to the target plant species, the accuracy and specificity of the judgment are ensured, avoiding a "one-size-fits-all" irrigation strategy. Comparing the monitored plant visual characteristics with the recovery judgment criteria allows for an objective assessment of whether the plant has recovered from acute water stress. Crucially, emergency irrigation is only stopped when the plant's visual characteristics meet the recovery judgment criteria and remain stable, effectively preventing premature cessation of irrigation due to temporary improvement. This ensures that the plant receives sufficient and appropriate water replenishment, preventing over-irrigation and water waste, while simultaneously guaranteeing the plant's healthy recovery.

[0117] Through the above technical solution, this application enables intelligent closed-loop control of the emergency irrigation process. Compared to basic technical solutions that only trigger irrigation, this application effectively solves the problem of unclear irrigation cessation timing by introducing continuous monitoring and recovery judgment mechanisms, thus avoiding excessive water consumption. Furthermore, this technical solution ensures that plants receive appropriate and sufficient recovery after acute water stress, significantly improving the accuracy and efficiency of irrigation, protecting the healthy growth of garden plants, and reducing the frequency and cost of manual intervention.

[0118] In some optional embodiments, assuming that azaleas in a certain area exhibit acute water stress symptoms such as significant leaf wilting and a marked decrease in greenness index due to prolonged drought, emergency irrigation is triggered. After emergency irrigation is initiated, a visual sensing device continuously acquires images of the azalea canopy. The data processing module continuously extracts the plant's visual characteristics, such as subpixel-level changes in leaf morphology and greenness index. Simultaneously, based on the azalea species, preset recovery judgment criteria are matched. These criteria may include leaf wilting recovering to more than 80% of the normal range, and the greenness index rebounding and stabilizing above a specific threshold. When monitoring data shows that the azalea's leaf morphology and greenness index meet these recovery criteria for 30 consecutive minutes, it is determined that the azalea has recovered from acute water stress, and emergency irrigation for that area is automatically stopped. This ensures that the azalea receives sufficient water recovery while avoiding unnecessary continuous irrigation.

[0119] In some embodiments, the step of preprocessing and feature extraction of the canopy image to obtain the visual features of the target plant includes:

[0120] The canopy image is processed to locate the target plant region, and the spatial range of the target plant in the image is identified;

[0121] The located area is segmented into leaf-stem regions to separate the leaf and stem regions.

[0122] Multi-dimensional feature extraction was performed on the segmented leaf and stem regions to obtain the leaf color saturation, hue, greenness index, photochemical reflectance index, as well as the sub-pixel level changes in leaf morphology and the stem micro-drooping angle.

[0123] The target plant region localization process aims to accurately determine the specific location and boundaries of the target plant in the acquired canopy images, thereby focusing the analysis on the target plant itself and eliminating background interference. Specifically, this process can identify the spatial extent of the target plant in the image, providing a foundation for subsequent refined analysis.

[0124] Furthermore, leaf-stem segmentation refers to further subdividing the plant image within the located region into independent leaf and stem regions after the target plant region has been localized. The purpose of this step is to enable independent feature extraction for different physiological structures of the plant, as the response characteristics of leaves and stems under water stress may differ.

[0125] Therefore, multi-dimensional feature extraction processing was performed on the segmented leaf and stem regions to obtain comprehensive plant visual characteristics. Specifically, these features include leaf color saturation, hue, greenness index, photochemical reflectance index, as well as sub-pixel-level changes in leaf morphology and stem drooping angle. Among these, leaf color saturation, hue, and greenness index are important indicators reflecting the chlorophyll content and photosynthetic status of plants; these color characteristics change significantly when plants are water-deficient. The photochemical reflectance index (PRI) is a sensitive physiological indicator that reflects rapid changes in plant photosynthetic efficiency and is often used for early detection of plant water stress. Sub-pixel-level changes in leaf morphology refer to the capture of subtle deformations at the microscopic level of leaves through high-precision image analysis technology, such as leaf curling and wilting; these changes are direct manifestations of decreased turgor pressure due to water loss. The stem drooping angle refers to the slight drooping of the stem due to weakened support caused by water loss; changes in this angle can also serve as an indicator of plant water stress.

[0126] The technical solution of this application refines the preprocessing and feature extraction of canopy images into target plant region localization, leaf-stem segmentation, and multi-dimensional feature extraction, enabling more comprehensive and accurate acquisition of visual features reflecting the physiological state of the target plant. Traditional preprocessing and feature extraction may only focus on macroscopic color or morphological changes, making it difficult to capture the subtle physiological responses of plants in the early stages of acute water stress. This technical solution, however, avoids interference from background noise by accurately locating the target plant region; it makes feature extraction for different parts more targeted through independent segmentation of leaves and stems; and it constructs a multi-dimensional, high-precision set of plant visual features by extracting physiological indicators such as color saturation, hue, greenness index, and photochemical reflectance index, as well as physical morphological indicators such as sub-pixel changes in leaf morphology and slight drooping angles of stems. These features can reflect the water shortage state of plants from different angles, especially in the early stages of acute water stress, capturing subtle changes that are difficult to detect with the naked eye, thus providing more reliable and richer data support for subsequent water shortage assessment.

[0127] The above technical solutions significantly improve the accuracy and timeliness of assessing acute water stress in garden plants. Specifically, precise target plant region localization effectively eliminates interference from non-target plants or background environment, ensuring data purity. Fine leaf-stem segmentation allows for the independent and accurate capture of physiological responses in different plant organs. More importantly, the extraction of multi-dimensional features, particularly the introduction of sensitive indicators such as photochemical reflectance index, sub-pixel changes in leaf morphology, and stem drooping angle, enables the identification of water shortage signals in plants before visible symptoms appear—in the early stages of acute water stress. This provides a valuable time window for timely emergency irrigation, effectively preventing irreversible plant damage caused by delayed irrigation, thereby improving the health management level and irrigation efficiency of garden plants.

[0128] In some embodiments, the step of performing target plant region localization processing on the canopy image includes:

[0129] The canopy image is subjected to foreground-background semantic segmentation to distinguish between plant foreground and non-plant background in the image;

[0130] Plant individual target detection processing is performed on the foreground area of ​​plants to identify specific plant individuals in the foreground;

[0131] Based on the morphological, color, and texture characteristics of the plant individual, and in conjunction with a pre-defined target plant feature library, a target plant matching verification process is performed to determine whether the individual is the target plant.

[0132] If the target plant is verified, morphological completion processing of the occluded area is performed—by associating the visible part of the target plant with the occluded part through the motion trajectory of adjacent frame images, the complete canopy area is obtained.

[0133] Specifically, the steps for locating target plant regions in canopy images include:

[0134] Specifically, the pixels in the acquired canopy images are divided into two categories: plant foreground and non-plant background. This can be achieved by applying deep learning models, such as semantic segmentation networks based on convolutional neural networks (CNNs). These models are trained on a large number of images with pixel-level annotations and can accurately identify and separate plant regions in the images, thus providing a clean plant foreground for subsequent processing.

[0135] Plant individual target detection is performed on the foreground area of ​​plants to identify specific plant individuals within the foreground. Specifically, after foreground-background semantic segmentation, a target detection algorithm is further employed to identify and locate each individual plant within the identified foreground area. This process generates a bounding box or mask for each plant individual, enabling refined differentiation and management of multiple target plants in a garden scene.

[0136] Based on the morphological, color, and texture characteristics of the individual plant, and in conjunction with a pre-defined target plant feature library, a target plant matching verification process is performed to determine whether the individual plant is the target plant. In practical applications, for each detected plant individual, multi-dimensional visual features are extracted, including but not limited to morphological features such as leaf shape and crown size, color features such as leaf color and saturation, and texture features such as leaf surface texture. These extracted features are then compared with a pre-established "target plant feature library." This feature library stores typical visual features of different types of target plants. Through feature matching algorithms, it is possible to accurately determine whether the currently detected plant individual is the target plant requiring irrigation control.

[0137] If the plant is verified as the target plant, occlusion region morphology completion processing is performed—by using the motion trajectories of adjacent frames, the visible portion of the target plant is associated with the occluded portion to obtain the complete canopy region. Furthermore, once a plant is identified as the target plant, if part of its canopy region is occluded by other objects (such as other plants, building structures, or its own leaves), occlusion region morphology completion processing is initiated. This processing utilizes adjacent frame images acquired by a visual sensing device at different time points. By analyzing the motion trajectory of the target plant in these consecutive frames, the visible portion of the target plant from different viewpoints can be associated and fused with the occluded portion, thereby reconstructing the complete canopy region of the target plant. The purpose is to ensure that subsequent plant visual feature extraction is based on a comprehensive and complete plant morphology.

[0138] The technical solution of this application refines the target plant region localization into multiple steps, including foreground-background semantic segmentation, individual plant target detection, target plant matching verification, and occlusion region morphological completion. This allows for the accurate identification and localization of target plants from complex garden scenes in a progressive manner. Foreground-background semantic segmentation first separates the plant from the environment, individual plant target detection further distinguishes independent plant individuals, target plant matching verification ensures that the identified individuals are the expected target plants, and occlusion region morphological completion solves the occlusion problem commonly encountered in practical applications, ensuring the completeness of subsequent feature extraction.

[0139] Through the above technical solutions, this application can significantly improve the accuracy and robustness of target plant area localization. Especially in complex garden environments, through refined segmentation, detection, matching, and completion processing, challenges such as background interference, mixed growth of multiple plants, and partial occlusion can be effectively overcome, ensuring that subsequent extraction of plant visual features is performed on the correct and complete individual plants. This lays a solid foundation for the accurate judgment of acute water stress, thereby improving the intelligence level and irrigation efficiency of the entire irrigation control system.

[0140] In some embodiments, the step of performing multi-dimensional feature extraction processing on the segmented leaf and stem regions includes:

[0141] Obtain the local light intensity distribution information of the area where the target plant is located;

[0142] Based on the distribution information, light compensation processing is performed on the leaf area to correct the color and brightness deviations caused by uneven light.

[0143] For the leaf areas that have undergone the light compensation treatment, color space conversion and standardization are performed to extract the color saturation, hue, and greenness index of the leaves.

[0144] Based on the local light intensity distribution information, the calculation of the photochemical reflectance index is adaptively adjusted to eliminate the influence of light intensity changes on the photochemical reflectance index.

[0145] A motion compensation algorithm based on image sequences is used to eliminate the influence of slight swaying caused by visual sensing device vibration or wind on the sub-pixel level changes in leaf morphology and the extraction of stem micro-drooping angle, thereby obtaining the sub-pixel level changes in leaf morphology and stem micro-drooping angle.

[0146] Specifically, image processing techniques or auxiliary sensors (e.g., a miniature light sensor array integrated into a visual sensing device) are used to perceive differences in light intensity at different locations within the target plant's canopy. This information can be a light intensity map reflecting the distribution of light intensity within the area. The aim is to provide a precise local light reference for subsequent light compensation and adjustments to the photochemical reflectance index.

[0147] Illumination compensation processing can be understood as an image correction technique. Its purpose is to adjust the pixel values ​​of leaf areas through algorithms to ensure consistent color and brightness under different lighting conditions. For example, grayscale world algorithms, white balance algorithms, or physically based illumination correction methods can be used. This processing can effectively eliminate the interference of uneven lighting caused by shadows, direct sunlight, or cloudy weather on the extraction of leaf color and brightness features.

[0148] In practical applications, color space conversion and standardization refer to converting the light-compensated leaf area image from the original RGB color space to a more suitable color space for color feature analysis, such as HSV or Lab, and standardizing the color components to eliminate differences in brightness or contrast between different images. This allows for the accurate extraction of leaf color saturation, hue, and greenness index, which are important indicators reflecting plant health and water status.

[0149] Furthermore, the influence of local light intensity is considered when calculating the photochemical reflectance index (PRI). PRI is usually calculated as the reflectance ratio of a specific wavelength band, but its value is easily affected by changes in light intensity. By incorporating information on the distribution of local light intensity, the calculation formula or parameters of PRI can be dynamically adjusted. For example, light intensity can be introduced as a weighting factor, or reflectance can be normalized according to light intensity, thereby obtaining a more accurate and stable PRI value. The aim is to eliminate the interference of light intensity changes on PRI, making it more accurately reflect the photosynthetic efficiency and water stress status of plants.

[0150] Furthermore, by utilizing information between consecutive frames, the algorithm identifies and compensates for minute displacements and swaying caused by visual sensing device jitter (e.g., cameras mounted on drones or robots) or wind acting on the plant itself. The algorithm may include techniques such as feature point matching, optical flow, or Kalman filtering to estimate and compensate for motion in the images. The aim is to ensure that subpixel-level changes in leaf morphology (e.g., subtle deformations such as leaf curling and wilting) and slight drooping angles of stems are due to physiological responses of the plant rather than external disturbances, thereby improving the accuracy and reliability of morphological feature extraction.

[0151] The technical solution of this application effectively solves the problem that traditional methods are easily affected by uneven light and changes in light intensity when extracting plant visual features under complex lighting conditions. This is achieved by introducing local light intensity distribution information and performing light compensation processing and adaptive adjustment of the photochemical reflectance index based on this information. Precise correction of light factors ensures that the leaf color saturation, hue, greenness index, and photochemical reflectance index more realistically reflect the physiological state of the target plant, rather than being artifacts of ambient light. Simultaneously, by employing a motion compensation algorithm based on image sequences, this application can identify and eliminate minor plant swaying caused by external mechanical vibrations or wind, thus ensuring that the sub-pixel-level changes in leaf morphology and the slight drooping angle of the stem are manifestations of the plant's own physiological response. The effective elimination of these non-physiological movements allows for the accurate capture of these delicate morphological features, providing a more reliable basis for subsequent assessment of acute water stress.

[0152] Through the above technical solutions, this application significantly improves the accuracy and robustness of plant visual feature extraction in garden plant irrigation control methods. Specifically, by using light compensation and adaptive adjustment of photochemical reflectance index, the interference of complex and variable light conditions on the extraction of color, brightness, and physiological indicators is effectively eliminated, enabling the obtained plant visual features to more accurately reflect the true physiological state of the plant. Furthermore, the introduction of a motion compensation algorithm successfully distinguishes between physiological morphological changes in plants and non-physiological movements caused by external environments (such as wind and equipment vibration), ensuring the extraction accuracy of fine features such as sub-pixel-level changes in leaf morphology and the slight drooping angle of stems. These improvements work together to obtain high-quality, highly reliable plant visual features under various outdoor environments, thereby greatly improving the accuracy of acute water stress assessment, avoiding misjudgments or omissions caused by environmental factors, and thus optimizing the decision-making efficiency and resource utilization rate of emergency irrigation.

[0153] In some alternative embodiments, it is assumed that on a cloudy afternoon with a light breeze, a visual sensing device is periodically acquiring images of the canopy of target plants in a garden area. Due to cloud movement, light intensity changes in a short period of time, while the light breeze causes the plant leaves and stems to sway slightly.

[0154] First, the local light intensity distribution information of the target plant's location is obtained. For example, image analysis identifies that some leaves are in shadow while others are under direct sunlight. Based on this distribution information, the brightness of leaves in the shadowed areas is enhanced and their colors are corrected, while overexposure is suppressed and colors are balanced for leaves in the directly illuminated areas. This process compensates for the light intensity in the leaf area, making the color and brightness of the entire canopy image more consistent.

[0155] Next, the leaf areas treated with light compensation undergo color space conversion (e.g., from RGB to HSV) and standardization to accurately extract the leaf's color saturation, hue, and greenness index. Simultaneously, when calculating the photochemical reflectance index (PRI), the calculation results are adaptively adjusted based on real-time local light intensity distribution information to ensure that the PRI value accurately reflects the plant's photosynthetic efficiency even under fluctuating light conditions.

[0156] Furthermore, a motion compensation algorithm based on image sequences is employed to address plant swaying caused by light winds. By analyzing the motion trajectories of leaves and stems in consecutive frames, the algorithm can distinguish between overall swaying caused by wind and subtle morphological changes resulting from the plant's own physiological wilting. For example, if leaves are detected swaying overall in the wind, but the degree of internal curling or the drooping angle of the stem does not change significantly, the overall swaying is identified as non-physiological swaying and compensated for. Ultimately, sub-pixel-level changes in leaf morphology (e.g., slight curling of leaf edges) and slight drooping angles of stems can be accurately extracted, which are key physiological indicators for assessing acute water stress in plants.

[0157] Through the above processing, even under complex and variable environmental conditions, highly accurate and reliable plant visual characteristics can be obtained, providing a solid data foundation for subsequent assessment of acute water stress.

[0158] In some embodiments, the step of employing an image sequence-based motion compensation algorithm to eliminate the influence of minute swaying caused by visual sensing device vibration or wind on sub-pixel-level changes in leaf morphology and the extraction of stem micro-drooping angle includes:

[0159] Motion features are extracted from the target plant region in the image sequence;

[0160] Based on the extracted motion features, the motion of the target plant is decomposed to separate the overall swaying component and the local morphological change component of the target plant.

[0161] By combining real-time wind speed and direction information, the wind force impact assessment of the overall sway component is performed to determine whether it is consistent with the wind force change trend, thereby eliminating the impact of micro-sway caused by visual sensing device vibration or wind force on the sub-pixel level changes in leaf morphology and the extraction of stem micro-drooping angle.

[0162] Specifically, image processing techniques are used to identify and quantify the displacement and deformation of the target plant between consecutive image frames, enabling motion feature extraction of the target plant region in the image sequence. In particular, optical flow, feature point matching (e.g., SIFT, SURF, or ORB features), or deep learning-based motion estimation models can be employed to capture the motion trajectory and velocity of pixels or feature points within the target plant canopy region, thereby obtaining dynamic change information of the target plant over time.

[0163] Furthermore, based on the extracted motion characteristics, the motion of the target plant is decomposed to separate the overall swaying component and the local morphological change component. The overall swaying component typically refers to the rigid or near-rigid motion of the plant as a whole caused by external factors (such as wind or vibration of visual sensing equipment), manifested as translation or rotation of the plant as a whole. The local morphological change component refers to the non-rigid deformation caused by the plant's own physiological activities (such as leaf drooping or curling due to water stress), manifested as slight bending, twisting, or contraction of leaves or stems. This decomposition can be achieved by establishing a motion model, for example, first fitting a global affine transformation or rigid body transformation to estimate the overall swaying, and then using the residual between the actual motion and the global motion as the local morphological change component.

[0164] Furthermore, by combining real-time wind speed and direction information, a wind force impact assessment is performed on the overall sway component to determine whether it aligns with wind force trends. This assessment aims to distinguish between wind-induced plant swaying and non-wind-induced swaying. Specifically, real-time wind speed and direction data of the target plant's environment can be acquired using sensors and compared with the direction and amplitude of the decomposed overall sway component. For example, when the main direction of the overall sway component is highly consistent with the real-time wind direction, and its amplitude is positively correlated with wind speed, it can be determined that the overall swaying is primarily influenced by wind.

[0165] The technical solution of this application, through refined decomposition of the target plant's movement and verification of external environmental factors, can effectively separate non-physiological swaying caused by visual sensing device vibration or wind from the actual morphological changes of the plant. First, the extraction of movement features provides the basic data for subsequent movement decomposition. Second, decomposing the movement into an overall swaying component and a local morphological change component makes it possible to distinguish between external disturbances and internal physiological changes. Finally, by combining real-time wind speed and direction information to assess the wind impact on the overall swaying component, it further verifies whether the overall swaying is indeed caused by wind, thus avoiding misjudging wind-induced plant swaying as physiological changes.

[0166] The above technical solution significantly improves the accuracy and reliability of extracting sub-pixel-level changes in leaf morphology and the micro-drooping angle of stems. This method effectively filters out interference from environmental noise (such as wind) and equipment noise (such as vibration of visual sensing equipment) on the extraction of plant visual features, ensuring that the acquired plant visual features more realistically reflect the physiological state of the target plant. This provides a more accurate and stable data foundation for subsequent judgments on whether the target plant is under acute water stress, thereby improving the decision-making accuracy and response efficiency of the entire irrigation control system.

[0167] In some embodiments, the step of combining real-time wind speed and wind direction information to assess the wind impact of the overall oscillation component and determine whether it is consistent with the wind change trend includes:

[0168] When the evaluation results show that the overall sway component is consistent with the wind force change trend, the overall sway component is identified as a non-physiological sway and is removed.

[0169] Extract the subpixel-level changes in leaf morphology and the slight drooping angle of the stem from the remaining local morphological change components.

[0170] Specifically, when the evaluation results show that the overall sway component is consistent with the wind force change trend, it means that the macroscopic movement of the target plant is mainly caused by external wind force, rather than physiological changes in the plant itself. In this case, the overall sway component is clearly identified as non-physiological sway. After identification, the non-physiological sway component will be removed from the total motion data, for example, through subtraction, filtering, or model compensation, to eliminate its interference with subsequent feature extraction. After removing the non-physiological sway component, the remaining motion data mainly contains local morphological change components of the target plant, which are more likely to reflect the physiological state of the plant. From these remaining local morphological change components, subpixel-level changes in leaf morphology and stem drooping angles can be accurately extracted. Subpixel-level changes in leaf morphology can refer to microscopic deformations such as leaf curling, drooping, and wilting, while stem drooping angles reflect the decrease in stem support caused by water stress.

[0171] The technical solution of this application solves the accuracy problem that may exist in the above-mentioned technical basis by introducing a clear processing mechanism for the wind force impact assessment results of the overall sway component. Specifically, when the assessment confirms that the overall sway component is consistent with the wind force change trend, this component is identified as a non-physiological movement caused by wind force. By removing this non-physiological sway component from the total movement data, the interference of environmental wind force on the extraction of plant morphological features can be effectively eliminated. As a result, the remaining local morphological change components can more purely reflect the physiological changes of the target plant, thereby ensuring that the sub-pixel-level changes in leaf morphology and the slight drooping angle of the stem extracted subsequently are physiological indicators that truly reflect the plant's water status, rather than illusions influenced by the external environment.

[0172] Through the above technical solution, this application can significantly improve the accuracy and reliability of plant visual feature extraction. By accurately identifying and eliminating non-physiological swaying caused by wind, it avoids misjudging environmental disturbances as physiological water shortage symptoms in plants. This allows the extracted sub-pixel-level changes in leaf morphology and the slight drooping angle of stems to more realistically and accurately reflect the actual water stress state of the target plant. This provides more accurate data support for subsequent water shortage assessment and emergency irrigation decisions, effectively avoiding over-irrigation or under-irrigation caused by misjudgments due to environmental factors, and improving the intelligence level of irrigation control and resource utilization efficiency.

[0173] In some optional embodiments, a specific example is illustrated below. Suppose that in a garden setting, a visual perception device periodically acquires images of the canopy of a target plant. At a certain moment, a noticeable overall swaying of the target plant is detected. At this point, data from real-time wind speed and direction sensors are used to assess the wind impact on this overall swaying component. If the assessment shows that the direction and frequency of this overall swaying component are highly consistent with the changing trends of real-time wind speed and direction—for example, the swaying amplitude increases with increasing wind speed, and the swaying direction is consistent with the wind direction—then this overall swaying component is identified as a non-physiological swaying caused by wind. Subsequently, this non-physiological swaying component is effectively removed from the image sequence using digital filtering or motion compensation algorithms. After removal, sub-pixel-level changes in leaf morphology, such as slight curling of leaf edges and slight drooping angles of stems, are precisely extracted from the remaining, wind-independent local morphological change components. These precisely processed visual features will be used for subsequent water shortage assessments, thereby avoiding misjudgments of plants being in an acute water stress state due to wind interference and ensuring the accuracy of irrigation decisions.

[0174] This application also proposes a garden plant irrigation control system, such as... Figure 2 As shown, a garden plant irrigation control system 100 includes:

[0175] The visual perception module 10 is used to periodically acquire canopy images of target plants in a garden scene through a visual perception device.

[0176] Data processing module 20 is used to preprocess and extract features from the canopy image to obtain the visual features of the target plant.

[0177] The status judgment module 30 is used to determine whether the target plant is in an acute water stress state based on the plant's visual characteristics and the water shortage judgment criteria preset for different plant species.

[0178] The irrigation triggering module 40 is used to trigger emergency irrigation of the area where the target plant is located if the target plant is in an acute water stress state.

[0179] The garden plant irrigation control system disclosed in this application integrates a visual perception module, a data processing module, a status judgment module, and an irrigation triggering module to construct an intelligent irrigation system capable of directly sensing the physiological state of plants and responding promptly to acute water stress. This system overcomes the limitations of traditional irrigation systems that rely solely on environmental parameters or weather forecasts. It can accurately identify water shortage symptoms in plants, especially in cases of sudden extreme weather or local environmental anomalies leading to rapid water shortages. This ensures timely and effective emergency irrigation, thereby avoiding plant damage caused by information delays or inaccurate predictions and significantly improving the precision and intelligence of garden plant management.

[0180] First, the visual perception module is used to periodically acquire canopy images of target plants in a garden scene using visual perception devices. The specific methods for image acquisition have already been described in the above embodiments and will not be repeated here. It is important to emphasize that the visual perception module is a combination of hardware and software to achieve this function. Specifically, the visual perception module may include one or more visual perception devices, such as high-definition cameras, multispectral cameras, or visual sensors mounted on drones, as well as control units and data interfaces for controlling these devices to acquire images. As an optional implementation, the visual perception module can be configured to periodically activate the visual perception devices, automatically acquire canopy images of target plants according to a preset acquisition frequency and range, and transmit the raw image data to a subsequent processing module. For example, the visual perception module can be integrated into a garden inspection robot for mobile acquisition along a preset path, or used as a fixed monitoring point to continuously monitor a specific area.

[0181] Secondly, the data processing module is used to preprocess and extract features from the canopy image to obtain the plant visual features of the target plant. The specific methods for image preprocessing and feature extraction have been described in the above embodiments and will not be repeated here. It is important to emphasize that the data processing module is responsible for processing and analyzing the raw image data collected by the visual perception module. Specifically, the data processing module may include an image preprocessing unit and a feature extraction unit. The image preprocessing unit performs denoising, color correction, and geometric correction on the canopy image to improve image quality and the accuracy of subsequent analysis. The feature extraction unit uses image processing algorithms or deep learning models to extract plant visual features that characterize the plant's physiological state from the preprocessed image, such as leaf color saturation, hue, greenness index, and leaf morphological changes. For example, the data processing module can use algorithms based on traditional computer vision, such as threshold segmentation and edge detection, to identify leaf regions and calculate their color features; or, it can deploy a pre-trained convolutional neural network model to automatically identify and quantify various visual features of the plant.

[0182] Next, the state judgment module determines whether the target plant is under acute water stress based on the plant's visual characteristics and preset water shortage judgment criteria for different plant species. The specific method for determining whether a plant is under acute water stress has already been described in the above embodiments and will not be repeated here. It is important to emphasize that the state judgment module is the core of the system's decision-making process. It assesses the plant's water status based on the plant's visual characteristics provided by the data processing module and the preset judgment criteria. Specifically, the state judgment module can have a built-in knowledge base storing water shortage judgment criteria corresponding to different plant species. These criteria can be a set of thresholds for visual characteristics or a classification model trained based on machine learning. Upon receiving plant visual characteristics, the state judgment module matches the corresponding judgment criteria according to the target plant species and performs comparative analysis. For example, the state judgment module can use a rule engine; when the leaf greenness index is below a specific threshold and the leaves show obvious curling, it is judged as acute water stress. Alternatively, it can use a support vector machine or neural network model to comprehensively judge multi-dimensional plant visual characteristics.

[0183] Finally, the irrigation trigger module is used to trigger emergency irrigation of the area where the target plant is located if the target plant is under acute water stress. The specific method for triggering emergency irrigation has been described in the above embodiments and will not be repeated here. It is important to emphasize that the irrigation trigger module is the execution unit for the system to perform irrigation operations. Specifically, the irrigation trigger module can be connected to actuators such as solenoid valves and water pumps in the garden irrigation system. When the status judgment module issues an emergency irrigation command, the irrigation trigger module receives the command and, according to a preset irrigation strategy (e.g., irrigation area, irrigation volume, irrigation duration), controls the corresponding irrigation equipment to start water supply. For example, the irrigation trigger module can send an opening signal to the solenoid valve in a specific area and continue for a period of time to ensure that the target plant receives sufficient water replenishment. In some embodiments, the irrigation trigger module can also be linked with flow sensors, pressure sensors, etc., to monitor the irrigation effect in real time and adjust irrigation parameters based on feedback information.

[0184] Compared to traditional irrigation control systems for garden plants, the system proposed in this application has significant advantages and innovations. Traditional systems mainly rely on soil moisture sensors and meteorological forecast data for irrigation decisions. However, this indirect monitoring method has limitations in practical applications. For example, in mixed planting areas, it is difficult to fully reflect the actual water requirements of all plants, and when sudden extreme weather (such as unforeseen heat waves) causes plants to rapidly enter a state of acute water stress, traditional systems often fail to respond in time due to information lag or inaccurate predictions, which may lead to irreversible damage to plants.

[0185] The core innovation of this application lies in constructing a closed-loop control system with a visual perception module as the front-end sensing unit, a data processing module for information transformation, a "state judgment module" for intelligent decision-making, and an "irrigation trigger module" for execution response. This system breaks through the traditional method of relying solely on environmental parameters for indirect judgment. Instead, it directly acquires canopy images of the target plant through the visual perception module, extracts the plant's visual features through the data processing module, and then the state judgment module determines whether the plant is under acute water stress based on its physiological characteristics. This direct perception of plant physiological state enables the system to identify acute water stress more promptly and accurately, especially in cases of sudden extreme environmental changes or local anomalies leading to rapid water shortage. Through the emergency response of the irrigation trigger module, this application ensures that plants receive rapid and precise water replenishment when suffering from acute water stress, effectively avoiding plant damage caused by decision-making delays in traditional systems. Therefore, this system significantly improves the intelligence level and refined management capabilities of garden plant irrigation, providing a more reliable guarantee for healthy plant growth.

[0186] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for controlling irrigation of garden plants, characterized in that, include: Images of the canopy of target plants in a garden scene are periodically collected using visual sensing devices. The canopy image is preprocessed and its features are extracted to obtain the visual features of the target plant. Based on the plant's visual characteristics and the preset water shortage judgment criteria for different plant species, determine whether the target plant is in an acute water stress state. If the target plant is under acute water stress, emergency irrigation is triggered in the area where the target plant is located.

2. The irrigation control method for garden plants according to claim 1, characterized in that, The step of determining whether the target plant is in an acute water stress state based on the plant's visual characteristics and preset water shortage judgment criteria for different plant species includes: The light intensity data, ambient temperature data, and soil moisture data of the area where the target plant is located are obtained by using a light intensity sensor, an ambient temperature sensor, and a soil moisture sensor. Based on the plant's visual characteristics, the soil moisture data, and the preset water shortage judgment criteria for different plant species, a first-level verification is performed to determine whether the target plant is in an acute water stress state, and the first-level verification result is obtained. When the first-layer verification result indicates that the plant's visual characteristics meet the water shortage judgment criteria and the soil moisture data is sufficient, a second-layer verification is performed based on the visual characteristics, the soil moisture data, the light intensity data, and the ambient temperature data to determine whether the target plant is in an acute water stress state, and a second-layer verification result is obtained. By combining the results of the first layer of verification and the results of the second layer of verification, it is confirmed whether the target plant is in a state of acute water stress.

3. The irrigation control method for garden plants according to claim 1, characterized in that, The step of determining whether the target plant is in an acute water stress state based on the plant's visual characteristics and preset water shortage judgment criteria for different plant species includes: The growth stage of the target plant is identified based on the plant's visual characteristics; Based on the growth stage, a water shortage judgment standard matching the growth stage is selected; wherein, when the growth stage of the target plant is the seedling stage, the water shortage judgment standard is leaf morphology, and when the growth stage of the target plant is the flowering stage, the water shortage judgment standard is greenness index. Based on the selected water shortage criteria, determine whether the target plant is in a state of acute water stress.

4. The irrigation control method for garden plants according to claim 1, characterized in that, Following the step of triggering emergency irrigation of the area where the target plant is located if the target plant is under acute water stress, the following steps are included: Continuously monitor the visual characteristics of the target plant; Based on the type of the target plant, a recovery judgment criterion corresponding to the type of the target plant is matched; The monitored plant visual characteristics were compared with the selected recovery criteria; Emergency irrigation should be stopped when the plant's visual characteristics meet the recovery criteria and remain stable.

5. The irrigation control method for garden plants according to claim 1, characterized in that, The steps of preprocessing and feature extraction of the canopy image to obtain the visual features of the target plant include: The canopy image is processed to locate the target plant region, and the spatial range of the target plant in the image is identified; The located area is segmented into leaf-stem regions to separate the leaf and stem regions. Multi-dimensional feature extraction was performed on the segmented leaf and stem regions to obtain the leaf color saturation, hue, greenness index, photochemical reflectance index, as well as the sub-pixel level changes in leaf morphology and the stem micro-drooping angle.

6. The irrigation control method for garden plants according to claim 5, characterized in that, The step of locating the target plant region in the canopy image includes: The canopy image is subjected to foreground-background semantic segmentation to distinguish between plant foreground and non-plant background in the image; Plant individual target detection processing is performed on the foreground area of ​​plants to identify specific plant individuals in the foreground; Based on the morphological, color, and texture characteristics of the plant individual, and in conjunction with a pre-defined target plant feature library, a target plant matching verification process is performed to determine whether the individual is the target plant. If the target plant is verified, morphological completion processing of the occluded area is performed—by associating the visible part of the target plant with the occluded part through the motion trajectory of adjacent frame images, the complete canopy area is obtained.

7. The irrigation control method for garden plants according to claim 6, characterized in that, The steps for performing multi-dimensional feature extraction on the segmented leaf and stem regions include: Obtain the local light intensity distribution information of the area where the target plant is located; Based on the distribution information, light compensation processing is performed on the leaf area to correct the color and brightness deviations caused by uneven light. For the leaf areas that have undergone the light compensation treatment, color space conversion and standardization are performed to extract the color saturation, hue, and greenness index of the leaves. Based on the local light intensity distribution information, the calculation of the photochemical reflectance index is adaptively adjusted to eliminate the influence of light intensity changes on the photochemical reflectance index. A motion compensation algorithm based on image sequences is used to eliminate the influence of slight swaying caused by visual sensing device vibration or wind on the sub-pixel level changes in leaf morphology and the extraction of stem micro-drooping angle, thereby obtaining the sub-pixel level changes in leaf morphology and stem micro-drooping angle.

8. The irrigation control method for garden plants according to claim 7, characterized in that, The steps of employing an image sequence-based motion compensation algorithm to eliminate the influence of minute swaying caused by visual sensing device vibration or wind on sub-pixel-level changes in leaf morphology and the extraction of stem micro-drooping angle include: Motion features are extracted from the target plant region in the image sequence; Based on the extracted motion features, the motion of the target plant is decomposed to separate the overall swaying component and the local morphological change component of the target plant. By combining real-time wind speed and direction information, the wind force impact assessment of the overall sway component is performed to determine whether it is consistent with the wind force change trend, thereby eliminating the impact of micro-sway caused by visual sensing device vibration or wind force on the sub-pixel level changes in leaf morphology and the extraction of stem micro-drooping angle.

9. The irrigation control method for garden plants according to claim 8, characterized in that, The step of combining real-time wind speed and direction information to assess the wind impact of the overall sway component and determine whether it is consistent with the wind change trend includes: When the evaluation results show that the overall sway component is consistent with the wind force change trend, the overall sway component is identified as a non-physiological sway and is removed. Extract the subpixel-level changes in leaf morphology and the slight drooping angle of the stem from the remaining local morphological change components.

10. A garden plant irrigation control system, characterized in that, The system includes: The visual perception module is used to periodically acquire canopy images of target plants in a garden scene through visual perception devices; The data processing module is used to preprocess and extract features from the canopy image to obtain the visual features of the target plant. The status judgment module is used to determine whether the target plant is in an acute water stress state based on the plant's visual characteristics and preset water shortage judgment criteria for different plant species. An irrigation triggering module is used to trigger emergency irrigation of the area where the target plant is located if the target plant is in a state of acute water stress.

Images