A regional akebia trifoliata planting intelligent monitoring and disease diagnosis method and system

Abstract

This invention relates to the field of image recognition technology and discloses an intelligent monitoring and disease diagnosis method and system for Akebia trifoliata planting in a region. The method includes the following steps: acquiring environmental monitoring data for each Akebia trifoliata planting area, and determining the disease-sensitive time window for each planting area based on the environmental monitoring data; collecting plant image data for the corresponding Akebia trifoliata planting area based on the disease-sensitive time window, and extracting apparent anomaly indicators based on the plant image data; generating disease risk results for the corresponding Akebia trifoliata planting area and establishing a record of suspected disease objects; performing duplicate filtering on the current suspected disease objects based on the suspected disease object record, and outputting disease diagnosis results and on-site verification prompts. This solution reduces invalid image acquisition and duplicate alarms by acquiring data through disease-sensitive time windows, identifying apparent anomalies, and filtering duplicate suspected disease objects, thereby improving the efficiency of regional Akebia trifoliata disease monitoring and on-site verification.

Description

Technical Field

[0001] This invention relates to the field of image recognition technology, and in particular to an intelligent monitoring and disease diagnosis method and system for regional Akebia trifoliata planting. Background Technology

[0002] Akebia trifoliata is mostly planted in mountainous, forest, or sloping environments, and planting areas are usually quite scattered, with significant differences in light, humidity, drainage, and ventilation between different areas. In actual management, growers mainly rely on regular garden visits to observe changes in leaf color, lesions, leaf curling, and growth to determine the condition of diseases. While this method is intuitive, it is greatly affected by the frequency of inspections, the experience of the growers, and weather conditions. In cases of continuous rain, high humidity, or severe shade under the trees, leaf lesions and abnormal yellowing often go undetected, easily missing the opportunity for early treatment.

[0003] Current agricultural monitoring solutions typically employ environmental sensors and cameras for timed data collection, followed by image recognition models to diagnose crop diseases. While this approach has some value in facility agriculture or intensive planting settings, it has limitations when directly applied to Akebia trifoliata cultivation areas. Firstly, fixed-period data collection generates a large number of normal images, placing significant strain on network transmission and power supply in mountainous regions. Secondly, disease characteristics are not always clearly visible; images taken after rain, during morning dew, or after prolonged periods of high humidity are often more valuable for diagnosis. Furthermore, consistently shooting at fixed times can easily lead to misalignment between the data collection period and the disease's apparent manifestation.

[0004] Furthermore, the intertwined vines and overlapping leaves of Akebia trifoliata can cause the same diseased leaf to appear repeatedly in images from multiple days. Current monitoring methods, which generate alarms based solely on single image anomalies, are prone to repeatedly alerting the same diseased organism, leading to a backlog of inspection tasks and hindering growers' ability to assess the true extent of new risks. For small-scale planting areas with limited technical maintenance capabilities, over-reliance on complex models is impractical. A more suitable monitoring and diagnostic method is needed, one that selects the timing of data collection based on environmental changes, performs tiered assessments based on apparent anomalies, and can filter out recurring diseased organisms.

[0005] Therefore, there is an urgent need to propose an intelligent monitoring and disease diagnosis method and system for Akebia trifoliata planting in a region, so as to at least solve the problems of excessive invalid image collection, inaccurate capture of disease manifestation time, and repeated alarms for the same diseased object under the existing fixed-period monitoring method. Summary of the Invention

[0006] This invention provides an intelligent monitoring and disease diagnosis method and system for regional Akebia trifoliata planting, which at least solves the problems of existing technologies such as the generation of many invalid images from fixed-period monitoring, inaccurate capture of disease manifestation periods, and repeated alarms for the same diseased object.

[0007] To achieve the above objectives, a first aspect of the present invention provides an intelligent monitoring and disease diagnosis method for regional Akebia trifoliata planting, the method comprising the following steps: Environmental monitoring data of each Akebia trifoliata planting area were obtained, and disease-sensitive time windows were determined for each Akebia trifoliata planting area based on the environmental monitoring data. Based on the disease-sensitive time window, plant image data of the corresponding Akebia trifoliata planting area were collected, and appearance abnormality indicators were extracted based on the plant image data. Based on the environmental monitoring data and the apparent anomaly indicators, the disease risk results for the corresponding Akebia trifoliata planting area are generated, and a record of suspected disease objects is established. Based on the suspected disease object record, duplicate filtering is performed on the current suspected disease object, and the disease diagnosis result and on-site verification prompt information are output based on the duplicate filtering result.

[0008] To achieve the above objectives, a second aspect of the present invention also provides an intelligent monitoring and disease diagnosis system for regional Akebia trifoliata planting, the system comprising: The first unit is used to acquire environmental monitoring data of each Akebia trifoliata planting area and determine the disease-sensitive time window for each Akebia trifoliata planting area based on the environmental monitoring data. The second unit is used to collect plant image data of the corresponding Akebia trifoliata planting area based on the disease-sensitive time window, and to extract appearance abnormality indicators based on the plant image data. The third unit is used to generate disease risk results for the corresponding Akebia trifoliata planting area based on the environmental monitoring data and the apparent anomaly indicators, and to establish a record of suspected disease objects. The fourth unit is used to perform duplicate filtering on the current suspected disease object based on the suspected disease object record, and output disease diagnosis results and on-site verification prompts based on the duplicate filtering results.

[0009] The beneficial effects of this invention are as follows: It determines the sensitive time window for diseases based on environmental monitoring data, concentrating plant image acquisition during periods when disease characteristics are more easily apparent, thus reducing invalid images generated by fixed-period acquisition; it generates disease risk results by combining environmental monitoring data with apparent anomaly indicators, improving the matching degree between disease diagnosis and the actual planting environment; and it uses suspected disease object records to repeatedly filter currently suspected disease objects, avoiding repeated alarms for the same disease object, thereby improving the effectiveness of regional Akebia trifoliata planting monitoring and the efficiency of on-site verification. Attached Figure Description

[0010] Figure 1 This is a schematic diagram illustrating the process of intelligent monitoring and disease diagnosis of Akebia trifoliata planting in a region, as described in this embodiment of the invention. Figure 2 This is a schematic diagram of the detection results of apparent anomalies in the leaves of Akebia trifoliata in an embodiment of the present invention; Figure 3 This is a schematic diagram of the operational architecture of the regional Akebia trifoliata planting intelligent monitoring and disease diagnosis system in an embodiment of the present invention. Detailed Implementation

[0011] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0012] This invention provides an intelligent monitoring and disease diagnosis method for Akebia trifoliata planting in a specific region, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the intelligent monitoring and disease diagnosis method for Akebia trifoliata planting in a specific region, as described in an embodiment of the present invention.

[0013] In this embodiment, a method for intelligent monitoring and disease diagnosis of Akebia trifoliata planting in a region includes the following steps: S10: Obtain environmental monitoring data for each Akebia trifoliata planting area, and determine the disease-sensitive time window for each Akebia trifoliata planting area based on the environmental monitoring data.

[0014] Specifically, the disease-sensitive time window for each Akebia trifoliata planting area is determined based on the environmental monitoring data, including: extracting air humidity, leaf surface humidity, soil moisture, light intensity, and ambient temperature from the environmental monitoring data; determining the duration of leaf surface wetting based on the air humidity and leaf surface humidity, determining the change in soil moisture based on the soil moisture, determining the duration of weak light based on the light intensity, and determining the change in diurnal temperature range based on the ambient temperature; determining the environmental disturbance value for the corresponding Akebia trifoliata planting area based on the duration of leaf surface wetting, the change in soil moisture, the duration of weak light, and the change in diurnal temperature range; comparing the environmental disturbance value with a preset disturbance threshold, and selecting the disease-sensitive time window for the corresponding Akebia trifoliata planting area from the preset time window set according to the comparison result.

[0015] In this embodiment of the invention, each Akebia trifoliata planting area can be divided according to the actual planting terrain, management scope, or monitoring equipment deployment range. At least one environmental monitoring node is set up within each Akebia trifoliata planting area. The environmental monitoring node continuously collects air humidity, leaf surface humidity, soil moisture, light intensity, and ambient temperature to form environmental monitoring data for the corresponding Akebia trifoliata planting area. The aforementioned environmental monitoring data is not only used to reflect the current environmental status but also to determine whether the Akebia trifoliata planting area has entered a stage where diseases are easily manifested.

[0016] In practice, air humidity and leaf surface humidity are continuously recorded at preset sampling intervals. When air humidity remains high and leaf surface humidity remains moist, the corresponding duration of leaf surface moisture is calculated. This duration of leaf surface moisture can be used to characterize the continuity of the leaf surface being in a moist environment. For soil moisture, the difference in soil moisture between adjacent monitoring periods or within a preset statistical period can be selected to obtain the change in soil moisture, which reflects changes in the root environment after rainfall, waterlogging, or irrigation. For light intensity, the duration of light intensity below a preset light threshold can be counted to obtain the duration of weak light. For ambient temperature, the diurnal temperature variation can be determined based on the difference between the average daytime temperature and the average nighttime temperature.

[0017] After obtaining data on the duration of leaf moisture, changes in soil moisture, duration of low light, and diurnal temperature variation, these data can be converted into perturbation sub-values ​​according to preset grading rules. These sub-values ​​are then combined to form the environmental perturbation value for the corresponding Akebia trifoliata planting area. The preset grading rules can be set based on the seasons when common diseases of Akebia trifoliata occur in the local area, historical orchard inspection records, and planting management experience. A higher environmental perturbation value indicates that the Akebia trifoliata planting area is closer to an environmental state where diseases are easily observed or likely to occur within the current period.

[0018] The environmental disturbance value is compared with a preset disturbance threshold, and a disease-sensitive time window is selected from a preset time window set based on the comparison result. For example, when the environmental disturbance value is low, the routine inspection time window can be selected; when the environmental disturbance value is high and the leaf surface remains moist for a long time, the time window before and after the morning dew disappears can be selected; when the environmental disturbance value is high and the soil moisture changes significantly, a preset period after rain or irrigation can be selected as the disease-sensitive time window. In this way, the acquisition time of subsequent plant image data is matched with the time when disease phenotypic characteristics are more likely to appear.

[0019] Step S20: Collect plant image data of the corresponding Akebia trifoliata planting area based on the disease-sensitive time window, and extract appearance abnormality indicators based on the plant image data.

[0020] Specifically, the process of collecting plant image data for the corresponding Akebia trifoliata planting area based on the disease-sensitive time window includes: acquiring the terminal location information and shooting orientation information of the image acquisition terminal within the corresponding Akebia trifoliata planting area; generating an image acquisition instruction for the image acquisition terminal based on the disease-sensitive time window, the image acquisition instruction including the acquisition start time, the number of acquisitions, and the acquisition interval; controlling the image acquisition terminal to perform fixed-point shooting on representative plant areas within the corresponding Akebia trifoliata planting area according to the image acquisition instruction to obtain original plant images; and binding the terminal location information, the shooting orientation information, the acquisition start time, and the original plant images to generate plant image data for the corresponding Akebia trifoliata planting area.

[0021] Specifically, extracting appearance anomaly indicators based on the plant image data includes: classifying the corresponding original plant images based on the terminal location information, shooting orientation information, and acquisition start time in the plant image data to obtain image groups corresponding to the same Akebia trifoliata planting area and the same representative plant area; segmenting the original plant images in the image group into leaf regions to obtain leaf target regions; extracting leaf yellowing areas, brown spot areas, yellowed leaf margin areas, and curled leaf areas based on the leaf target regions; calculating the proportions of the leaf yellowing areas, brown spot areas, yellowed leaf margin areas, and curled leaf areas in the leaf target regions to generate appearance anomaly indicators corresponding to the Akebia trifoliata planting area.

[0022] Specifically, the extraction of leaf yellowing areas, brown spot areas, yellowing leaf margin areas, and leaf curling areas based on the target leaf region includes: inputting the target leaf region into a pre-trained YOLO disease appearance detection model, which uses images of Akebia trifoliata leaves labeled with yellowing, brown spot, yellowing leaf margin, and leaf curling areas as training samples; performing multi-scale feature extraction on the target leaf region using the YOLO disease appearance detection model to generate candidate anomaly boxes corresponding to the target leaf region; filtering the candidate anomaly boxes based on their category confidence and overlap relationship to obtain yellowing detection boxes, brown spot detection boxes, yellowing leaf margin detection boxes, and leaf curling detection boxes; and determining the yellowing, brown spot, yellowing leaf margin, and leaf curling areas based on their positions and sizes within the target leaf region.

[0023] In this embodiment of the invention, after determining the disease-sensitive time window for a corresponding Akebia trifoliata planting area, the image acquisition process within that area revolves around a preset representative plant region. The representative plant region can be selected based on Akebia trifoliata planting density, slope aspect, shading conditions, or historical disease distribution. During installation, the image acquisition terminal records its location and shooting orientation information. The terminal location information can be marked using a combination of area number and relative location number, while the shooting orientation information indicates the shooting direction and coverage area of ​​the corresponding plant region. Since images of the same area within different monitoring periods need to be compared subsequently, the terminal location and shooting orientation information remain fixed throughout the entire monitoring period.

[0024] After the disease-sensitive time window is generated, an image acquisition command is prepared based on the corresponding time window. The image acquisition command includes the acquisition start time, the number of acquisitions, and the acquisition interval. For Akebia trifoliata planting areas with low environmental disturbance values, a lower acquisition frequency can be used; for Akebia trifoliata planting areas with high environmental disturbance values ​​and continuous high humidity, the number of acquisitions within the corresponding disease-sensitive time window can be increased. The acquisition interval can be adjusted according to the rate of leaf condition change. For example, during the high humidity period after rain, the acquisition interval can be shortened to record changes in the edges of leaf lesions, water-soaked areas, or leaf mold.

[0025] Following the image acquisition instructions, fixed-point shooting is performed to obtain raw plant images of the corresponding representative plant area. The raw plant images may include leaf, vine, and fruit areas, with the leaf area being the primary focus of analysis. After shooting, the terminal location information, shooting orientation information, and acquisition start time are bound to the corresponding raw plant image to form plant image data. Subsequent images acquired in each monitoring cycle are bound in the same way to ensure that images of the same representative plant area at different time periods maintain a correspondence.

[0026] After obtaining plant image data, the data is categorized. During categorization, the source range of the original plant images is first determined based on the area number, terminal location information, and shooting orientation. Then, the monitoring cycle is marked and time-sorted for original plant images within the same source range using the acquisition start time. This groups original plant images corresponding to the same Akebia trifoliata planting area and the same representative plant area into the same image group. In other words, the acquisition start time is primarily used to distinguish different monitoring cycles and form time series, rather than as the sole criterion for excluding historical images of the same representative plant area. This method facilitates subsequent analysis of abnormal changes in the same leaf area over different time periods.

[0027] For the original plant images in the image set, leaf region segmentation is performed to obtain the target leaf regions. Leaf region segmentation can be achieved using color thresholding, edge segmentation, or semantic segmentation. Since Akebia trifoliata leaves are easily obscured by vines, dappled light, and interference from background branches and leaves in natural environments, some redundant leaf margin regions can be retained during leaf region segmentation to avoid accidental cropping of leaf margin lesions.

[0028] After obtaining the target area of ​​the leaf, the abnormal areas within the target area are identified. Specifically, the target area of ​​the leaf is input into a pre-trained YOLO (You Only Look Once, target detection model) disease appearance detection model. The YOLO disease appearance detection model uses images of Akebia trifoliata leaves labeled with yellowing areas, brown spot areas, yellowing leaf margins, and leaf curling areas as training samples. The labeled areas in the training samples can be compiled from manual inspection records and historical diseased leaf images. The sample images are allowed to include leaf images under different light conditions, humidity conditions, and shooting distances to improve the model's adaptability to natural planting environments.

[0029] The YOLO disease appearance detection model performs multi-scale feature extraction on the target area of ​​the leaf, generating candidate anomaly boxes for the corresponding target area. Since multiple anomaly areas may exist simultaneously on the same leaf, the candidate anomaly boxes can correspond to different anomaly categories. Then, based on the category confidence and overlap of the candidate anomaly boxes, they are filtered to obtain yellowing detection boxes, brown spot detection boxes, leaf margin yellowing detection boxes, and leaf curling detection boxes. For candidate anomaly boxes with high overlap and the same category, the candidate anomaly box with higher confidence can be retained as the final detection result.

[0030] like Figure 2 After determining the disease-sensitive time window, fixed-point image acquisition was performed on representative plant areas in different Akebia trifoliata planting areas, and corresponding appearance abnormality indicators were extracted based on the acquired plant image data. Figure 2 The image shows the detection results corresponding to the yellowing detection frame, brown spot detection frame, leaf margin yellowing detection frame, and leaf curling detection frame, respectively. After the collected leaf target area is input into the YOLO disease appearance detection model, abnormal areas in the leaf are located, and the corresponding abnormality category and detection frame position are output. Then, using the detection frame as the location range of candidate abnormal areas, the actual abnormal areas that meet the abnormal characteristics in terms of color, texture, or edge morphology are further extracted within the corresponding detection frame. Based on the area ratio and position distribution of the actual abnormal areas in the leaf target area, the appearance abnormality index of the corresponding plant area is generated for subsequent disease risk result analysis and the establishment of records of suspected disease objects.

[0031] After obtaining the detection frames, based on their position and size within the target area of ​​the leaf, candidate leaf yellowing areas, candidate brown spot areas, candidate leaf margin yellowing areas, and candidate leaf curling areas are first determined. Subsequently, actual abnormal areas satisfying the corresponding abnormality category characteristics are further extracted from each candidate area. For example, the actual abnormal range is determined based on yellowing color difference, brown spot boundaries, leaf margin yellowing distribution, or curling contour changes. Furthermore, the area ratio of the aforementioned actual abnormal areas within the target area of ​​the leaf is calculated to form the corresponding apparent abnormality index for the Akebia trifoliata planting area. These apparent abnormality indicators are used not only for disease risk analysis within the current monitoring period but also for comparing abnormal changes between subsequent monitoring periods and filtering duplicate disease targets.

[0032] In another possible implementation, when Akebia trifoliata is planted under forest canopies, the leaves are often obscured by vines and upper leaves, making it difficult to reliably identify slight yellowing and early brown spots in frontal images. A supplementary backlit image can be acquired for the same representative plant area within the disease-sensitive time window, during periods when light enters from the underside of the leaves, such as early morning side light or afternoon oblique light. After acquisition, the backlit image is bound to the regular frontal image according to the terminal location information, shooting orientation information, and acquisition start time, and areas with abnormal light transmission are extracted within the target area of ​​the leaf. If an area shows only slight color difference in the frontal image but exhibits uneven light transmission, dark spot boundaries, or abnormalities around leaf veins in the backlit image, this area is considered a candidate area for subsequent detection box generation. This implementation does not add complex equipment; it only utilizes natural oblique light to improve the visibility of leaves with weak symptoms, making it suitable for low-cost monitoring scenarios of Akebia trifoliata under forest canopies.

[0033] Step S30: Based on the environmental monitoring data and the apparent anomaly indicators, generate disease risk results for the corresponding Akebia trifoliata planting area and establish a record of suspected disease objects.

[0034] Specifically, the environmental risk status of the corresponding Akebia trifoliata planting area is determined based on the environmental monitoring data; the image anomaly status of the corresponding Akebia trifoliata planting area is determined based on the apparent anomaly indicators; the environmental risk status and the image anomaly status are compared with a preset disease discrimination rule table to determine the initial disease risk level of the corresponding Akebia trifoliata planting area; based on the duration of the environmental risk status and the changing trend of the image anomaly status, the initial disease risk level is adjusted by upgrading or downgrading to obtain the disease risk result of the corresponding Akebia trifoliata planting area; when the disease risk result meets the preset recording conditions, the area number, terminal location information, shooting orientation information, collection start time, anomaly area category, anomaly area location, and anomaly area area are extracted from the corresponding plant image data to establish a record of suspected disease objects.

[0035] In this embodiment of the invention, in step S30, after obtaining the environmental monitoring data and apparent anomaly indicators of the corresponding Akebia trifoliata planting area, a comprehensive analysis of the disease risk within the current monitoring period is performed. The environmental monitoring data mainly reflects whether the current planting environment is in a disease-prone state, while the apparent anomaly indicators reflect whether abnormal changes have already occurred on the leaf surface. Both types of data participate in the subsequent generation of disease risk results.

[0036] In practice, the environmental risk status of the corresponding Akebia trifoliata planting area is determined based on air humidity, leaf surface humidity, soil moisture, duration of low light, and diurnal temperature variation. For example, continuous high humidity, prolonged leaf surface moisture, and a significant increase in soil moisture within a short period can classify the corresponding Akebia trifoliata planting area as a high-risk environment; when environmental parameters fluctuate little and remain within the normal range, it can be classified as a low-risk environment. The environmental risk status can be represented by a classification system, such as low-risk, medium-risk, and high-risk.

[0037] For visual anomalies, the image anomaly status can be determined based on the proportions of yellowed areas, brown spots, yellowed leaf margins, and curled areas. When the yellowed and brown spot areas are small and show little change, it can be classified as a low-anomaly state; when the anomaly area continues to expand or multiple anomaly areas appear simultaneously, it can be classified as a high-anomaly state. For leaf areas exhibiting abnormal changes across multiple consecutive monitoring periods, the corresponding image anomaly status level can be increased.

[0038] After obtaining the environmental risk status and image anomaly status, they are compared with a preset disease discrimination rule table. The preset disease discrimination rule table establishes a pre-defined correspondence between different environmental conditions and different apparent anomalies. For example, a high humidity environment corresponding to an expansion of brown spot areas can correspond to a risk of leaf spot diseases; a significant increase in soil moisture and an increase in yellowing areas of leaves can correspond to a risk of root stress; continuous low light and an expansion of leaf curling areas can correspond to a risk of abnormal growth. Based on the rule table matching results, the initial disease risk level for the corresponding Akebia trifoliata planting area is determined. In one embodiment, the preset disease discrimination rule table is shown in Table 1.

[0039] Table 1 Pre-set Disease Identification Rules ; Because some diseases continue to develop across different monitoring periods, the initial disease risk level is dynamically adjusted after it is determined. Specifically, the initial disease risk level is adjusted based on the duration of the environmental risk state and the changing trend of image anomalies. For example, if the high environmental risk state lasts for a long time and the area of ​​brown spots increases for several consecutive monitoring periods, the corresponding disease risk level can be increased; if the environmental risk state has decreased and the area of ​​the abnormal region has not significantly expanded, the corresponding disease risk level can be decreased. The changing trend of image anomalies can be determined based on the change in the area of ​​the abnormal region, the change in the number of abnormal regions, or the direction of expansion of the abnormal region between adjacent monitoring periods.

[0040] After obtaining the disease risk results for the corresponding Akebia trifoliata planting area, it is determined whether the disease risk results meet the preset recording conditions. The preset recording conditions can be set based on the disease risk level, abnormal area area, or abnormal change rate. When the preset recording conditions are met, the area number, terminal location information, shooting orientation information, data collection start time, abnormal area category, abnormal area location, and abnormal area area are extracted from the corresponding plant image data to generate a current suspected disease object record, which is then marked as pending filtering. The current suspected disease object record in the pending filtering state is used for matching with historical suspected disease object records in subsequent steps. Before repeated filtering is completed, it will not participate in the current matching as a historical suspected disease object.

[0041] When the disease risk result does not meet the preset recording conditions, a low-risk monitoring result is output or the normal monitoring status is maintained, and no record of the current suspected disease object is generated. Environmental monitoring data and plant image data within this monitoring period can be archived according to area number and collection time for subsequent trend analysis. Therefore, monitoring results that do not meet the recording conditions will not enter the duplicate filtering process, avoiding the problem of no currently suspected disease object to match in subsequent steps.

[0042] The location of abnormal areas can be represented using relative coordinates within the leaf region, and the area of ​​abnormal areas can be represented by the proportion of the corresponding abnormal area within the target leaf region. By classifying current suspected disease records into pending filtering, continuing, newly added, or false alarm states, continuous tracking of abnormal changes in the same leaf region can be achieved in subsequent monitoring cycles, providing a corresponding data foundation for filtering recurring disease objects. Records that have already been filtered can be used as historical suspected disease records in subsequent monitoring cycles; newly generated pending filtering records in the current monitoring cycle are only used as current matching objects.

[0043] In another possible implementation, when Akebia trifoliata leaves are affected by rainwater erosion and vein conduction under natural conditions, the expansion direction of different diseases varies. For example, some leaf spot diseases expand along the veins, while some abnormal yellowing of leaf margins spreads from the leaf edge towards the center. Based on this, after obtaining the abnormal areas in the current monitoring period, the directional parameters of the abnormal area's edge contour are statistically analyzed to obtain the edge expansion direction parameters of the corresponding abnormal areas. These edge expansion direction parameters for the current monitoring period are then compared with the corresponding parameters from historical monitoring periods.

[0044] When the direction of edge expansion remains consistent across multiple consecutive monitoring periods, and the area of ​​the abnormal region continues to increase, the corresponding disease risk level is increased; when the direction of edge expansion of the abnormal region changes irregularly, and the area of ​​the abnormal region changes only slightly, the corresponding disease risk level is decreased. This implementation method can distinguish between natural insect bites, mechanical abrasions, or occasional leaf damage, reducing misjudgments of disease risk caused by localized random defects.

[0045] Step S40: Perform duplicate filtering on the current suspected disease object based on the suspected disease object record, and output the disease diagnosis result and on-site verification prompt information based on the duplicate filtering result.

[0046] Specifically, based on the suspected disease object records, duplicate filtering of the current suspected disease object includes: filtering historical suspected disease object records corresponding to the same representative plant area from the suspected disease object records based on the current area number, the current terminal location information, and the current shooting orientation information; matching the current abnormal area category, the current abnormal area location, and the current abnormal area area with the historical abnormal area category, historical abnormal area location, and historical abnormal area area in the historical suspected disease object records, respectively, to obtain object duplicate matching results; when the object duplicate matching results meet the preset duplicate conditions, the current suspected disease object is marked as a continuation object of the historical suspected disease object; when the object duplicate matching results do not meet the preset duplicate conditions, the current suspected disease object is marked as a newly added suspected disease object, and the current object information is written into the suspected disease object record.

[0047] Furthermore, based on the duplicate filtering results, the system outputs disease diagnosis results and on-site verification prompts, including: acquiring the duplicate filtering results and the corresponding disease risk results for the Akebia trifoliata planting area; when the duplicate filtering result is a continuing disease and the disease risk result does not meet the preset upgrade conditions, outputting a disease diagnosis result for maintaining observation; when the duplicate filtering result is a continuing disease and the disease risk result meets the preset upgrade conditions, outputting a disease diagnosis result for continuously aggravated disease and generating on-site verification prompts; when the duplicate filtering result is a newly added disease and the disease risk result meets the preset verification conditions, outputting a newly added suspected disease diagnosis result and generating on-site verification prompts.

[0048] In this embodiment of the invention, after identifying a suspected disease object in the current monitoring cycle, duplicate filtering is performed on the current suspected disease object based on the established suspected disease object records. This step is mainly used to distinguish whether the current suspected disease object is a continuous manifestation of a historical suspected disease object or a newly emerging abnormal object in the current monitoring cycle, thereby avoiding the same disease object being repeatedly used as a new risk warning in continuous monitoring cycles.

[0049] In practice, the process begins by identifying records from the suspected disease object records that have been historically entered into the database or continuously updated, while excluding current suspected disease object records in the current batch that are awaiting filtering. Then, based on the current area number, current terminal location information, and current shooting orientation information, historical suspected disease object records corresponding to the same representative plant area are selected from the suspected disease object records after excluding the current batch. Since the image acquisition terminal uses a fixed-point shooting method, the area number, terminal location information, and shooting orientation information can limit the image source range of the current suspected disease object, ensuring that subsequent matching is only performed within the same representative plant area, reducing mismatches between different plant areas. Considering that the vines and leaves of *Akebia trifoliata* may experience slight displacement due to wind, growth, or traction in the natural environment, before performing abnormal area location matching, local registration can be performed on the current and historical images of the same representative plant area based on the support edge, main vine intersection point, fixed marker point, or stable leaf vein characteristics. After registration, abnormal area matching is performed based on the relative position within the leaf target area, allowing for a preset position tolerance range.

[0050] After obtaining historical records of suspected disease targets, the current abnormal area category, current abnormal area location, and current abnormal area area are matched with the historical abnormal area categories, locations, and areas in the historical records of suspected disease targets, respectively. The abnormal area category is used to determine if the abnormality type is consistent; the abnormal area location is used to determine if the current abnormal area is near a historical abnormal area; and the abnormal area area is used to determine if the current abnormal area falls within the normal range of variation of the historical abnormal area. The abnormal area location can be represented by relative coordinates within the leaf target area, and the abnormal area area can be represented by the proportion of the abnormal area within the leaf target area. For cases where there is a slight offset during fixed-point shooting, a position tolerance range can be set to allow the position comparison to adapt to slight swaying of branches and leaves in natural environments.

[0051] When the current abnormal region category matches the historical abnormal region category, the deviation between the current abnormal region location and the historical abnormal region location meets the preset location condition, and the change between the current abnormal region area and the historical abnormal region area meets the preset area condition, the object duplication matching result can be considered to meet the preset duplication condition. In this case, the current suspected disease object is marked as a continuation of the historical suspected disease object, and the current abnormal region area, current collection time, and current disease risk result are added to the corresponding historical suspected disease object record. This preserves the disease change process while preventing the same object from being flagged as a newly added abnormal duplication.

[0052] When the duplicate matching result does not meet the preset duplicate criteria, the current suspected disease object is marked as a new suspected disease object, and the current object information is written to the suspected disease object record. The current object information may include the current area number, current terminal location information, current shooting orientation information, current abnormal area category, current abnormal area location, current abnormal area area, and current acquisition time. After the new suspected disease object is written, it serves as the historical basis for duplicate filtering in subsequent monitoring cycles.

[0053] After obtaining the duplicate filtering results, the disease diagnosis results are output based on the disease risk results of the corresponding Akebia trifoliata planting area. These disease diagnosis results serve primarily as preliminary or suspected diagnoses before on-site verification, prompting growers to determine whether on-site inspection is necessary. Whether the disease is ultimately confirmed can be updated based on subsequent manual verification results. If the duplicate filtering results are for a continuing disease and the disease risk results do not meet the preset escalation conditions, it indicates that the current suspected disease is essentially consistent with historical suspected diseases, and the abnormal changes have not significantly expanded; a disease diagnosis result for maintaining observation can be output. If the duplicate filtering results are for a continuing disease and the disease risk results meet the preset escalation conditions, it indicates that historical suspected diseases show a continuous worsening trend; a disease diagnosis result for continuously worsening diseases can be output, and on-site verification prompts can be generated.

[0054] When duplicate filtering results identify newly added objects, and the disease risk results meet preset verification conditions, a new suspected disease diagnosis result is output, along with on-site verification prompts. These prompts may include the area number, terminal location information, shooting orientation, abnormal area category, abnormal area location, and suggested verification content. Growers can use this information to locate representative plant areas and conduct on-site verification of the upper and lower surfaces of leaves, the base of branches, or surrounding drainage. This approach ensures that disease diagnosis results retain continuous tracking of historical objects while highlighting truly newly added or aggravated risk objects.

[0055] When the repeated filtering results show a newly added object, but the disease risk result does not meet the preset verification conditions, the newly added observation-type disease diagnosis result is output, and the newly added suspected disease object is retained as an observation status record. For the observation status record, the changes in the area, location, and type of abnormal area are continued to be compared in subsequent monitoring cycles; if the subsequent disease risk result meets the preset verification conditions, on-site verification prompt information is generated. This avoids directly triggering on-site verification for minor, occasional new anomalies, while preserving their subsequent development process.

[0056] Preferably, the method further includes: obtaining the manual verification result corresponding to the on-site verification prompt information, wherein the manual verification result includes a confirmed disease result, a non-disease abnormal result, or a continued observation result; when the manual verification result is a confirmed disease result, increasing the image acquisition frequency of the corresponding Akebia trifoliata planting area in subsequent monitoring cycles; when the manual verification result is a non-disease abnormal result, marking the corresponding suspected disease object as a false alarm record; when the manual verification result is a continued observation result, maintaining the disease-sensitive time window of the corresponding Akebia trifoliata planting area, and continuing to acquire plant image data in subsequent monitoring cycles.

[0057] In this embodiment of the invention, the manual verification results can be entered by the grower after completing the on-site inspection. Specifically, this includes confirming disease results, non-disease abnormal results, or results requiring further observation. The on-site inspection may include lesions on the upper surface of leaves, mold on the lower surface of leaves, the condition of the base of branches and vines, water accumulation around the plant, and whether there are similar abnormalities in adjacent plants.

[0058] When the manual verification result confirms the disease, it means that the suspected disease object is consistent with the actual situation on site. At this point, increase the image acquisition frequency for the corresponding Akebia trifoliata planting area in subsequent monitoring periods. For example, change the original once-daily fixed-point shooting to multiple shootings per day, or perform image acquisition within several consecutive disease-sensitive time windows. Manual confirmation marks and confirmation times can also be added to the records of suspected disease objects to facilitate continuous tracking of the changes in this abnormal object.

[0059] When manual verification indicates a non-disease-related abnormality, it suggests that the suspected disease may originate from insect bites, mechanical abrasions, light reflection, leaf shading, or natural discoloration of older leaves. In this case, the corresponding suspected disease object is marked as a false alarm. In subsequent monitoring cycles, if objects similar to this false alarm in terms of abnormal area category, location, and area reappear, their priority for triggering on-site verification alerts can be reduced to avoid redundant and ineffective inspections.

[0060] When the manual review result indicates continued observation, it means that the on-site symptoms are not yet sufficient to directly confirm the disease, but still have tracking value. At this point, maintain the disease-sensitive time window for the corresponding Akebia trifoliata planting area and continue collecting plant image data in subsequent monitoring cycles. During the continued observation period, the observation status of the corresponding suspected disease-affected objects can be retained, and subsequently collected data on the area of ​​abnormal regions, changes in the location of abnormal regions, and disease risk results can be added to the same record.

[0061] Through the above processing, the results of manual verification are not only used as one-time confirmation information, but also participate in the adjustment of subsequent image acquisition frequency, recording status of suspected diseased objects, and on-site verification prompt strategies. For confirmed disease results, the attention level of the same area and the same abnormal category can be increased in subsequent monitoring cycles; for non-disease abnormal results, they can be used as false alarm samples for subsequent duplicate filtering and priority adjustment of verification prompts; for results requiring continued observation, the observation status of the corresponding object can be maintained, and information on abnormal changes in subsequent monitoring cycles can be continuously added. Thus, the regional Akebia trifoliata planting monitoring process can be corrected based on on-site feedback.

[0062] Reference Figure 3 , Figure 3 This is a schematic diagram of the structure of the intelligent monitoring and disease diagnosis system for Akebia trifoliata planting in a specific region, according to an embodiment of the present invention.

[0063] like Figure 3 As shown in the embodiment of the present invention, the intelligent monitoring and disease diagnosis system for regional Akebia trifoliata planting includes: The first unit is used to acquire environmental monitoring data of each Akebia trifoliata planting area and determine the disease-sensitive time window for each Akebia trifoliata planting area based on the environmental monitoring data. The second unit is used to collect plant image data of the corresponding Akebia trifoliata planting area based on the disease-sensitive time window, and to extract appearance abnormality indicators based on the plant image data. The third unit is used to generate disease risk results for the corresponding Akebia trifoliata planting area based on the environmental monitoring data and the apparent anomaly indicators, and to establish a record of suspected disease objects. The fourth unit is used to perform duplicate filtering on the current suspected disease object based on the suspected disease object record, and output disease diagnosis results and on-site verification prompts based on the duplicate filtering results.

[0064] It is understood that in the description of this specification, references to terms such as "one embodiment," "another embodiment," "other embodiments," or "first embodiment to Nth embodiment," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0065] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0066] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for intelligent monitoring and disease diagnosis of Akebia trifoliata planting in a region, characterized in that, The method includes the following steps: Environmental monitoring data of each Akebia trifoliata planting area were obtained, and disease-sensitive time windows were determined for each Akebia trifoliata planting area based on the environmental monitoring data. Based on the disease-sensitive time window, plant image data of the corresponding Akebia trifoliata planting area were collected, and appearance abnormality indicators were extracted based on the plant image data. Based on the environmental monitoring data and the apparent anomaly indicators, the disease risk results for the corresponding Akebia trifoliata planting area are generated, and a record of suspected disease objects is established. Based on the suspected disease object record, duplicate filtering is performed on the current suspected disease object, and the disease diagnosis result and on-site verification prompt information are output based on the duplicate filtering result.

2. The intelligent monitoring and disease diagnosis method for regional Akebia trifoliata planting as described in claim 1, characterized in that, Based on the environmental monitoring data, the disease-sensitive time windows for each Akebia trifoliata planting area were determined, including: Air humidity, leaf surface humidity, soil moisture, light intensity, and ambient temperature are extracted from the environmental monitoring data. The duration of leaf surface humidification is determined based on the air humidity and the leaf surface humidity; the change in soil humidity is determined based on the soil humidity; the duration of weak light is determined based on the light intensity; and the change in diurnal temperature range is determined based on the ambient temperature. Based on the duration of leaf surface wetting, the change in soil moisture, the duration of weak light, and the change in diurnal temperature range, the environmental disturbance value of the corresponding Akebia trifoliata planting area is determined. The environmental disturbance value is compared with a preset disturbance threshold, and the disease-sensitive time window for the corresponding Akebia trifoliata planting area is selected from the preset time window set based on the comparison result.

3. The method for intelligent monitoring and disease diagnosis of Akebia trifoliata planting in a specific region as described in claim 1, characterized in that, Based on the disease-sensitive time window, plant image data of the corresponding Akebia trifoliata planting area was collected, including: Obtain the terminal location information and shooting orientation information of the image acquisition terminal in the corresponding Akebia trifoliata planting area; Based on the disease-sensitive time window, an image acquisition instruction corresponding to the image acquisition terminal is generated. The image acquisition instruction includes the acquisition start time, the number of acquisitions, and the acquisition interval. According to the image acquisition command, the image acquisition terminal is controlled to perform fixed-point shooting on the representative plant area in the corresponding Akebia trifoliata planting area to obtain the original plant image; The terminal location information, the shooting orientation information, the acquisition start time, and the original plant image are bound together to generate plant image data for the corresponding Akebia trifoliata planting area.

4. The intelligent monitoring and disease diagnosis method for planting Akebia trifoliata in a specific region as described in claim 3, characterized in that, Based on the plant image data, extract appearance anomaly indicators, including: Based on the terminal location information, shooting orientation information and acquisition start time in the plant image data, the corresponding original plant images are classified to obtain image groups corresponding to the same Akebia trifoliata planting area and the same representative plant area. Leaf region segmentation is performed on the original plant images in the image group to obtain the leaf target region; Based on the target area of ​​the leaf, extract the yellowing area, brown spot area, yellowing area of ​​leaf margin and curling area of ​​leaf; The proportions of the yellowing area, brown spot area, yellowing leaf margin area, and curled leaf area in the target area of ​​the leaf are calculated respectively to generate the corresponding appearance abnormality index of the Akebia trifoliata planting area.

5. The intelligent monitoring and disease diagnosis method for planting Akebia trifoliata in a specific region as described in claim 4, characterized in that, Based on the target area of ​​the leaf, the yellowing area, brown spot area, yellowing area of ​​the leaf margin, and curling area of ​​the leaf are extracted, including: The target area of ​​the leaf is input into a pre-trained YOLO disease appearance detection model, which uses images of Akebia trifoliata leaves labeled with yellowing areas, brown spot areas, yellowing leaf margin areas, and leaf curling areas as training samples. The YOLO disease appearance detection model is used to extract multi-scale features from the target area of ​​the leaf to generate candidate anomaly boxes corresponding to the target area of ​​the leaf. Based on the category confidence and frame overlap of the candidate anomaly boxes, the candidate anomaly boxes are filtered to obtain yellowing detection boxes, brown spot detection boxes, leaf margin yellowing detection boxes, and leaf curling detection boxes. Based on the position and size of the yellowing detection frame, the brown spot detection frame, the leaf margin yellowing detection frame, and the leaf curling detection frame in the target area of ​​the leaf, the yellowing area, the brown spot area, the leaf margin yellowing area, and the leaf curling area are determined respectively.

6. The intelligent monitoring and disease diagnosis method for planting Akebia trifoliata in a specific region as described in claim 1, characterized in that, Based on the environmental monitoring data and the apparent anomaly indicators, disease risk results are generated for the corresponding Akebia trifoliata planting areas, and records of suspected disease-affected objects are established, including: The environmental risk status of the corresponding Akebia trifoliata planting area is determined based on the environmental monitoring data. Based on the aforementioned apparent anomaly indicators, determine the image anomaly status of the corresponding Akebia trifoliata planting area; The environmental risk status and the image anomaly status are compared with the preset disease discrimination rule table to determine the initial disease risk level of the corresponding Akebia trifoliata planting area. Based on the duration of the environmental risk state and the changing trend of the image anomaly state, the initial disease risk level is adjusted by upgrading or downgrading to obtain the disease risk results for the corresponding Akebia trifoliata planting area. When the disease risk result meets the preset recording conditions, the area number, terminal location information, shooting orientation information, collection start time, abnormal area category, abnormal area location and abnormal area area are extracted from the corresponding plant image data to establish a record of suspected disease objects.

7. The method for intelligent monitoring and disease diagnosis of Akebia trifoliata planting in a specific region as described in claim 6, characterized in that, Based on the suspected disease object records, perform duplicate filtering on the current suspected disease objects, including: Based on the current area number, the current terminal location information, and the current shooting orientation information, historical suspected disease object records corresponding to the same representative plant area are selected from the suspected disease object records; The current abnormal region category, the current abnormal region location, and the current abnormal region area are matched with the historical abnormal region category, historical abnormal region location, and historical abnormal region area in the historical suspected disease object record, respectively, to obtain the object duplicate matching result; When the object repeat matching result meets the preset repeat condition, the current suspected disease object is marked as a continuation object of the historical suspected disease object; When the object duplicate matching result does not meet the preset duplicate condition, the current suspected disease object is marked as a new suspected disease object, and the current object information is written into the suspected disease object record.

8. The intelligent monitoring and disease diagnosis method for planting Akebia trifoliata in a specific region as described in claim 7, characterized in that, Based on the duplicate filtering results, the system outputs disease diagnosis results and on-site verification prompts, including: Obtain the duplicate filtering results and the corresponding disease risk results for the Akebia trifoliata planting area; When the repeated filtering result is a continuation object and the disease risk result does not meet the preset upgrade conditions, output the disease diagnosis result for the maintenance observation category. When the repeated filtering result is a continuing object and the disease risk result meets the preset upgrade conditions, the diagnostic result of the continuously aggravated disease is output, and on-site verification prompt information is generated; When the duplicate filtering result is a newly added object and the disease risk result meets the preset review conditions, the newly added suspected disease diagnosis result is output, and on-site review prompt information is generated.

9. The intelligent monitoring and disease diagnosis method for planting Akebia trifoliata in a specific region as described in claim 1, characterized in that, The method further includes: Obtain the manual review result corresponding to the on-site review prompt information. The manual review result includes confirmed disease results, non-disease abnormal results, or continued observation results. When the manual review result confirms the disease, the image acquisition frequency of the corresponding Akebia trifoliata planting area will be increased in subsequent monitoring cycles. When the manual review result is a non-disease abnormality result, the corresponding suspected disease object will be marked as a false alarm record; When the manual review result is a continued observation result, the disease-sensitive time window for the corresponding Akebia trifoliata planting area is maintained, and plant image data continues to be collected in subsequent monitoring cycles.

10. A regional intelligent monitoring and disease diagnosis system for Akebia trifoliata cultivation, characterized in that, The system includes: The first unit is used to acquire environmental monitoring data of each Akebia trifoliata planting area and determine the disease-sensitive time window for each Akebia trifoliata planting area based on the environmental monitoring data. The second unit is used to collect plant image data of the corresponding Akebia trifoliata planting area based on the disease-sensitive time window, and to extract appearance abnormality indicators based on the plant image data. The third unit is used to generate disease risk results for the corresponding Akebia trifoliata planting area based on the environmental monitoring data and the apparent anomaly indicators, and to establish a record of suspected disease objects. The fourth unit is used to perform duplicate filtering on the current suspected disease object based on the suspected disease object record, and output disease diagnosis results and on-site verification prompts based on the duplicate filtering results.

Images