A method for artificial intelligence visual recognition and interactive control

Abstract

This invention relates to the field of electronic digital data processing technology, and more particularly to an artificial intelligence visual recognition and interactive control method. It includes the following steps: S1: Obtaining the data acquisition cycle of the camera and the intermittent acquisition cycle of the sensor in the target scene; based on the data acquisition cycle, obtaining the intermittent working cycle of visual recognition; S2: Within the data acquisition cycle, dividing the data features into steady-state data and transient data; based on the steady-state data and the transient data, obtaining variable data; S3: Within the intermittent acquisition cycle, adjusting the intermittent acquisition cycle of the variable data. This invention proposes a visual recognition method based on data feature hierarchical classification. By dynamically dividing data types and combining exponential and logarithmic functions to optimize the cycle, it adaptively adjusts the acquisition and processing strategy; it uses a difference quantization formula to dynamically allocate resources, optimize energy consumption and accuracy, and improve the reliability and adaptability of complex terrain monitoring under low power consumption.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electric digital data processing, and in particular to an artificial intelligence visual recognition and interactive control method. BACKGROUND

[0002] Visual recognition and interactive control in the field environment face severe challenges. Existing visual recognition technology faces problems such as low data processing efficiency and lack of dynamic grading mechanism in complex scenes. Traditional methods rely on fixed acquisition cycles and cannot achieve energy optimization through data feature analysis, and lack adaptive algorithms based on mathematical models, resulting in redundant data occupying computing resources and insufficient target recognition accuracy and real-time performance. For example, continuous monitoring mode cannot distinguish between steady-state and transient data characteristics, resulting in a large amount of invalid data processing and exacerbating device energy consumption pressure. An adaptive method is urgently needed to optimize energy consumption while ensuring recognition accuracy and achieve precise control. SUMMARY

[0003] In order to overcome the shortcomings of low efficiency and high energy consumption in data processing, the present application provides an artificial intelligence visual recognition and interactive control method.

[0004] The technical implementation scheme of the present application is: an artificial intelligence visual recognition and interactive control method, comprising the following steps:

[0005] S1: obtaining the data acquisition cycle of a camera and the interval acquisition cycle of a sensor in a target scene; based on the data acquisition cycle, obtaining an intermittent working cycle of visual recognition;

[0006] S2: dividing data features into steady-state data and transient data within the data acquisition cycle; based on the steady-state data and the transient data, obtaining variable data;

[0007] S3: adjusting the interval acquisition cycle of the variable data within the interval acquisition cycle; based on the adjusted interval acquisition cycle, adjusting the data acquisition cycle of the camera and the intermittent working cycle of visual recognition;

[0008] S4: based on the adjusted data acquisition cycle and intermittent working cycle, constructing a first interactive control sequence and a second interactive control sequence, and determining an interactive control method based on the difference between the first interactive control sequence and the second interactive control sequence.

[0009] Preferably, the data acquisition cycle of the camera and the interval acquisition cycle of the sensor in the target scene are obtained, comprising:

[0010] Obtaining the data acquisition cycle of the camera and the interval acquisition cycle of the sensor forms N data acquisition cycles and interval acquisition cycles;

[0011] During the N data acquisition cycles, interactive control nodes are collected and defined as the first interactive control node;

[0012] During the N interval acquisition cycles, interactive control nodes are collected and defined as the second interactive control node;

[0013] The data acquisition cycle refers to the cyclical time unit in which the camera is regularly started and stopped due to energy limitations. It is usually manifested as an alternation between a fixed-duration activation phase and a dormant phase.

[0014] The interval acquisition cycle refers to the discrete data acquisition rhythm designed by environmental sensors to reduce energy consumption, which is usually manifested as a non-continuous, equal-time sampling triggering mechanism.

[0015] Preferably, obtaining the intermittent working cycle of visual recognition based on the data acquisition cycle includes:

[0016] Within the data acquisition cycle, N-1 intermittent working cycles of visual recognition are formed;

[0017] The intermittent working cycle refers to the intermittent operation mode set to adapt to the camera's power supply cycle, which is usually manifested as selecting only N-1 cycles out of N device working cycles to perform image analysis tasks.

[0018] Preferably, the step of dividing the data features into steady-state data and transient data within the data acquisition period includes:

[0019] Within the data acquisition cycle, based on each intermittent working cycle, the data characteristics are divided into steady-state data and transient data;

[0020] If the position and shape characteristics of the data features do not exceed the preset threshold during the intermittent working cycle, they are classified as steady-state data.

[0021] If the position and shape characteristics of the data features exceed a preset threshold during the intermittent working cycle, they are classified as transient data.

[0022] Preferably, obtaining the variable data based on the steady-state data and the transient data includes:

[0023] If, within the intermittent working cycle, the steady-state data or the transient data in the next intermittent working cycle transforms into transient data or the steady-state data, then the steady-state data or the transient data is classified as variable data.

[0024] Preferably, adjusting the interval collection period of the changing data within the interval collection period includes:

[0025] If the time interval between the first interactive control node and the second interactive control node is greater than a preset interval threshold, and the number of the changing data exceeds a preset number threshold within the time interval between the first interactive control node and the second interactive control node, then the interval collection cycle of the changing data is shortened.

[0026] If the time interval between the first interactive control node and the second interactive control node is less than or equal to a preset interval threshold, and the number of the changing data within the time interval between the first interactive control node and the second interactive control node does not exceed a preset number threshold, then the interval collection period of the changing data is extended.

[0027] Preferably, adjusting the camera's data acquisition cycle and the visual recognition's intermittent working cycle based on the adjusted interval acquisition cycle includes:

[0028] If the interval collection period is shortened, the weather conditions of the monitored area can be obtained;

[0029] If the interval collection period is extended, sudden events in the monitoring area can be acquired;

[0030] Based on the weather conditions, the camera's data acquisition cycle is adjusted according to the first adjustment formula;

[0031] Based on the aforementioned sudden event, the intermittent working cycle of visual recognition is adjusted according to the second adjustment formula.

[0032] Preferably, adjusting the camera's data acquisition cycle based on the weather conditions according to the first adjustment formula includes: the first adjustment formula is as follows,

[0033]

[0034] in, The adjusted data collection cycle, For the data acquisition cycle before adjustment, W severity The degree of weather severity is represented by α, and the weather impact coefficient is represented by α.

[0035] Preferably, adjusting the intermittent working cycle of visual recognition based on the sudden event according to the second adjustment formula includes: the second adjustment formula is as follows,

[0036]

[0037] in, This is the adjusted intermittent work cycle. For the intermittent work cycle before adjustment, E level β represents the level of the emergency, and β is the event response coefficient.

[0038] Preferably, the step of constructing a first interactive control sequence and a second interactive control sequence based on the adjusted data acquisition cycle and intermittent working cycle, and determining the interactive control method based on the degree of difference between the first interactive control sequence and the second interactive control sequence, includes:

[0039] The data acquisition period, the adjusted data acquisition period, and the time interval are combined to form a first interactive control sequence;

[0040] The intermittent work cycle, the adjusted intermittent work cycle, and the time interval are combined to form a second interactive control sequence;

[0041] An interactive control method for determining data features is based on the degree of difference between the first interactive control sequence and the second interactive control sequence; the quantification formula for the degree of difference is as follows.

[0042]

[0043] Where D represents the degree of difference, w1 represents the weighting coefficient of the data collection cycle, w2 represents the weighting coefficient of the intermittent work cycle, and w1+w2=1.

[0044] Beneficial Effects: This invention proposes a visual recognition method based on data feature grading and mathematical modeling. By dynamically dividing steady-state, transient, and changing data, and combining a periodic optimization method driven by exponential and logarithmic functions, it achieves adaptive adjustment of data acquisition and processing strategies, significantly reducing redundant data processing and improving the real-time performance and accuracy of target recognition. Based on a difference quantification formula, it dynamically allocates computing resources, optimizes system energy consumption and recognition accuracy, and balances system stability and real-time response, thereby improving the reliability and adaptability of complex terrain monitoring under low power consumption conditions. Attached Figure Description

[0045] Figure 1 This is a flowchart of the artificial intelligence visual recognition and interactive control method of the present invention;

[0046] Figure 2 This is a schematic diagram illustrating the adjustment process of the interval collection period for variable data in this invention. Detailed Implementation

[0047] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0048] An artificial intelligence visual recognition and interactive control method, such as Figure 1 As shown, it includes the following steps:

[0049] S1: Obtain the data acquisition cycle of the camera and the intermittent acquisition cycle of the sensor in the target scene; based on the data acquisition cycle, obtain the intermittent working cycle of visual recognition;

[0050] S2: Within the data acquisition period, the data characteristics are divided into steady-state data and transient data; based on the steady-state data and the transient data, variable data is obtained;

[0051] S3: Within the interval acquisition period, adjust the interval acquisition period of the changing data; based on the adjusted interval acquisition period, adjust the data acquisition period of the camera and the intermittent working period of visual recognition;

[0052] It should be noted that, based on the adjusted interval acquisition cycle, the data acquisition cycle of the camera and the intermittent working cycle of visual recognition are adjusted according to weather parameters and event levels using the first adjustment formula and the second adjustment formula, respectively.

[0053] S4: Based on the adjusted data acquisition cycle and intermittent working cycle, construct a first interactive control sequence and a second interactive control sequence, and determine the interactive control method based on the degree of difference between the first interactive control sequence and the second interactive control sequence.

[0054] Obtain the data acquisition period of the camera and the interval acquisition period of the sensor in the target scene, including:

[0055] Obtain the camera's data acquisition cycle and the sensor's interval acquisition cycle to form N data acquisition cycles and interval acquisition cycles;

[0056] During the N data acquisition cycles, interactive control nodes are collected and defined as the first interactive control node;

[0057] During the N interval acquisition cycles, interactive control nodes are collected and defined as the second interactive control node;

[0058] The data acquisition cycle refers to the cyclical time unit in which the camera is regularly started and stopped due to energy limitations. It is usually manifested as an alternation between a fixed-duration activation phase and a dormant phase.

[0059] The interval acquisition cycle refers to the discrete data acquisition rhythm designed by environmental sensors to reduce energy consumption, which is usually manifested as a non-continuous, equal-time sampling triggering mechanism.

[0060] It should be noted that in field environments, visual recognition and interactive control require efficient data acquisition under energy-constrained conditions. Traditional methods suffer from excessive energy consumption due to continuous monitoring and are difficult to adapt to complex terrain changes.

[0061] Data acquisition method, camera data acquisition cycle: Due to energy limitations, the camera alternates between activation (working) and sleep (energy saving) according to a fixed cycle to ensure data capture during critical periods.

[0062] Intermittent sampling cycle of sensors: Environmental sensors trigger sampling at a discrete rhythm, reducing energy consumption through discontinuous, equally spaced data acquisition.

[0063] By dividing the camera into periodic working and dormant phases, and combining this with the sensor's discrete triggering mechanism, a balance is struck between data integrity and energy efficiency. The camera's active phase covers key monitoring windows, while the sensor periodically samples to supplement environmental parameters, forming a collaborative acquisition strategy.

[0064] The interactive control node is defined as follows: The data acquisition cycle refers to a cyclical time unit in which the camera is regularly started and stopped due to energy limitations, typically characterized by alternating fixed-duration active and dormant phases. The first interactive control node collects data during the camera's active phase for visual data synchronization and target tracking.

[0065] The second interactive control node: The interval acquisition cycle refers to the discrete data acquisition rhythm designed by the environmental sensor to reduce energy consumption, typically manifested as a non-continuous, equal-interval sampling triggering mechanism. The second interactive control node collects data when the sensor triggers sampling, and is used for environmental parameter fusion and state calibration.

[0066] For example, in a field monitoring scenario: the camera's data acquisition cycle is 10 minutes, with each cycle consisting of a 2-minute activation phase and an 8-minute sleep phase. During the activation phase, images are acquired, forming the first interactive control node. The sensor triggers temperature and humidity sampling every 30 seconds (intermittent acquisition cycle), generating a second interactive control node upon triggering. These two types of nodes collaboratively drive system decision-making. If an anomaly (such as a falling rock) is detected during the activation phase, the sensor is immediately activated for high-frequency sampling, improving real-time response.

[0067] Based on the data acquisition cycle, the intermittent working cycle of visual recognition is obtained, including:

[0068] Within the data acquisition cycle, N-1 intermittent working cycles of visual recognition are formed;

[0069] The intermittent working cycle refers to the intermittent operation mode set to adapt to the camera's power supply cycle, which is usually manifested as selecting only N-1 cycles out of N device working cycles to perform image analysis tasks.

[0070] It should be noted that the intermittent work cycle optimizes energy consumption by proactively skipping some cycles while ensuring critical data analysis, adapting to the camera's power supply rhythm. The core principle is to select only N-1 cycles out of N camera work cycles to perform image analysis tasks, using the skipped cycles for device cooling or low-power standby to avoid energy overload caused by continuous computation.

[0071] For example, suppose the camera's data acquisition cycle is 10 minutes (containing 5 cycles, each with 2 minutes of activation and 8 minutes of sleep). The intermittent working cycle of visual recognition selects 4 of the 5 cycles to perform analysis (N-1=4). For example, analysis is activated in cycles 1-4, and only images are acquired but not processed in cycle 5. Linkage control: During the skipped cycle 5, environmental data (such as sudden changes in temperature and humidity) acquired intermittently by the sensors determines whether to activate emergency analysis.

[0072] Within the data acquisition period, the data characteristics are divided into steady-state data and transient data, including:

[0073] Within the data acquisition cycle, based on each intermittent working cycle, the data characteristics are divided into steady-state data and transient data;

[0074] If the position and shape characteristics of the data features do not exceed the preset threshold during the intermittent working cycle, they are classified as steady-state data.

[0075] If the position and shape characteristics of the data features exceed a preset threshold during the intermittent working cycle, they are classified as transient data.

[0076] It should be noted that, within the data acquisition period, the intermittent work cycle is used as the time base. Preset thresholds are used to quantify the changes in the location and shape of data features, thus classifying steady-state and transient data. Steady-state data represents stable and unchanging targets in the scene (such as fixed field ridges), while transient data reflects dynamically changing targets (such as moving animals).

[0077] The classification criteria are as follows: Steady-state data: within the intermittent working cycle, the target position offset is ≤ the preset threshold and the morphological change rate is ≤ the preset threshold; Transient data: the target position offset is > the threshold or the morphological change rate is > the preset threshold.

[0078] Steady-state data refers to data units whose position and morphological characteristics remain stable within intermittent working cycles. These are typically background targets or long-term stationary objects and require low-frequency maintenance to reduce computational load.

[0079] Transient data refers to data units whose position or morphological characteristics change significantly within an intermittent working cycle. These typically correspond to dynamic targets or sudden disturbances and require high-frequency tracking to ensure real-time control.

[0080] Preset threshold setting method: Historical data analysis: Based on long-term monitoring data, the distribution patterns of target location offset and morphological changes are statistically analyzed, and the upper limit of the 95% confidence interval is taken as the threshold. Environmental adaptation adjustment: The threshold is dynamically adjusted according to the terrain complexity (such as slope and vegetation density), and the threshold is relaxed by 20%-30% in steep slope areas. Online learning mechanism: The threshold is updated in real time through a sliding window algorithm to adapt to seasonal or sudden environmental changes.

[0081] For example, assuming the camera's data acquisition cycle is 10 minutes (containing five 2-minute activation phases), the analysis is performed on four of these intermittent work cycles: Steady-state data: In the first cycle, a field ridge position shift of 0.1 meters (preset threshold 0.5 meters) and a morphological change rate of 2% (preset threshold 5%) are detected, marked as steady-state data, triggering the generation of a monthly maintenance report. Transient data: In the third cycle, a target shift of 1.2 meters (exceeding the preset threshold) and a morphological change rate of 8% (exceeding the preset threshold) are detected, marked as transient data, immediately activating drone tracking and initiating an audio-visual deterrent. Threshold dynamic adjustment: During the rainy season, frequent landslides are detected, automatically increasing the position shift threshold from 0.5 meters to 0.7 meters to reduce the false positive rate.

[0082] Based on the steady-state data and the transient data, variable data is obtained, including:

[0083] If, within the intermittent working cycle, the steady-state data or the transient data in the next intermittent working cycle transforms into transient data or the steady-state data, then the steady-state data or the transient data is classified as variable data.

[0084] It should be noted that variable data is defined as data characteristics that undergo state transitions within adjacent intermittent working cycles. If data marked as steady-state or transient in the current cycle switches to the opposite state (steady-state → transient or transient → steady-state) in the next cycle, it is considered variable data. Such data reflects abnormal changes in the environment or target and should be processed first to trigger a control response.

[0085] The variable data refers to data units that cross the boundary between steady-state and transient classification within intermittent working cycles. They typically exhibit target characteristics of sudden state migration and require dynamic adjustment of control strategies to cope with abnormal changes.

[0086] Example, steady state → transient state: In the first cycle, the position of the field ridge shifted by 0.1 meters (steady state data). In the second cycle, due to heavy rain, the displacement reached 0.8 meters (exceeding the threshold of 0.5 meters), which was marked as variable data, triggering a landslide warning and activating the drainage system.

[0087] Transient to steady state: A sudden increase in cucumber pests was detected in the 3rd cycle (transient data). In the 4th cycle, the target disappeared and the characteristics stabilized (classified as steady state data). It was marked as variable data, triggering the automatic spraying system and generating a pest control report.

[0088] Control Response: Changes in data trigger the camera's data acquisition cycle to be shortened to 5 minutes (from 10 minutes), and additional computing power is allocated to continuously track high-risk targets.

[0089] This mechanism transforms state transition events into control instructions, improving adaptability to dynamic environments.

[0090] Adjusting the interval acquisition period of the changing data within the interval acquisition period includes:

[0091] If the time interval between the first interactive control node and the second interactive control node is greater than a preset interval threshold, and the number of the changing data exceeds a preset number threshold within the time interval between the first interactive control node and the second interactive control node, then the interval collection cycle of the changing data is shortened.

[0092] If the time interval between the first interactive control node and the second interactive control node is less than or equal to a preset interval threshold, and the number of the changing data within the time interval between the first interactive control node and the second interactive control node does not exceed a preset number threshold, then the interval collection period of the changing data is extended.

[0093] It should be noted that, as Figure 2 As shown, the logic for shortening the interval acquisition cycle is as follows: when the time interval between the first interactive control node (the control point within the camera's data acquisition cycle) and the second interactive control node (the control point within the sensor's interval acquisition cycle) exceeds a preset interval threshold, and the number of changing data during this period exceeds a preset quantity threshold, it indicates: Increased environmental disturbance: A large amount of changing data occurs within a long time interval, such as sudden disasters (e.g., landslides) or high-frequency dynamic events (e.g., animal group activity). Increased urgency of response: It is necessary to shorten the changing data acquisition cycle and increase the data update frequency to ensure that the control system can promptly capture anomalies and trigger actions (e.g., initiate drainage or alarms). Conclusion: Shortening the interval acquisition cycle enhances real-time monitoring capabilities and avoids missing critical events.

[0094] The logic for extending the interval acquisition cycle is as follows: when the time interval is less than or equal to a preset interval threshold, and the number of fluctuating data points does not exceed a preset quantity threshold, it indicates that the environment is stable: there is little fluctuation in data within a short time interval, indicating a low-risk state (e.g., sunny weather, no animal activity). Energy consumption optimization requirements: extending the interval acquisition cycle reduces the number of wake-up calls for sensors and cameras, lowering energy consumption and extending device battery life. Conclusion: Extending the interval acquisition cycle prioritizes energy efficiency and reduces redundant data acquisition.

[0095] Threshold setting method, preset interval threshold: Basis: Combining equipment endurance target (e.g., continuous operation for 30 days) and event response delay tolerance (e.g., disaster response must be within 5 minutes). Calculation: Determine the average safe interval through historical data analysis (e.g., the longest interval under no risk is 20 minutes), and take 80% as the threshold (16 minutes).

[0096] Preset threshold: Based on the fluctuation range of statistically normal data (e.g., 10 times the average daily fluctuation on sunny days), take 3 times the standard deviation (e.g., 15 times) as the threshold. Dynamic adjustment: Automatically increase the threshold by 20% (18 times) during the rainy season to avoid frequent adjustments triggered by weather interference.

[0097] Example: The camera's data acquisition cycle is 10 minutes (including five 2-minute activation phases), and the sensor samples every 5 minutes. Conditions to shorten the interval acquisition cycle are triggered: Time interval: 18 minutes between the first interactive control node (end of the camera's data acquisition cycle) and the second interactive control node (end of the sensor's interval acquisition cycle) (preset threshold 15 minutes). Changes in data: 20 changes in data were detected during this period (preset threshold 15), including 12 instances of hillside displacement and 8 instances of sudden increases in cucumber pests. Action: Shorten the interval acquisition cycle of the change data from 5 minutes to 3 minutes, increase monitoring density, and immediately initiate drone spraying operations.

[0098] Extended interval acquisition period trigger condition: Time interval: 12 minutes between nodes (≤15 minutes threshold). Variable data: Only 8 variable data points were detected (<15 times threshold), all of which were minor leaf disturbances. Action: Extend the interval acquisition period from 5 minutes to 8 minutes and shut down secondary sensors to save energy.

[0099] Based on the adjusted interval acquisition period, the data acquisition period of the camera and the intermittent working period of visual recognition are adjusted, including:

[0100] If the interval collection period is shortened, the weather conditions of the monitored area can be obtained;

[0101] If the interval collection period is extended, sudden events in the monitoring area can be acquired;

[0102] Based on the weather conditions, the camera's data acquisition cycle is adjusted according to the first adjustment formula;

[0103] Based on the aforementioned sudden event, the intermittent working cycle of visual recognition is adjusted according to the second adjustment formula.

[0104] It should be noted that by shortening the sensor's interval acquisition cycle, the system acquires weather data (such as heavy rain and sandstorms) in real time, and dynamically optimizes the camera's data acquisition cycle based on the first mathematical model (exponential function). Specifically, the weather severity W... severity Through formula The calculation drives the compression of the camera's data acquisition cycle to increase image data density and ensure the integrity of critical scenes. When acquiring sudden events by extending the interval acquisition cycle, processing resources are adjusted through event priority: Extending the interval acquisition cycle is usually accompanied by low-risk environments, in which case computing power is released to track sudden events (such as the sudden appearance of wild animals) and avoid resource idleness.

[0105] This design overcomes the shortcomings of traditional methods that separate environment and event response. Through a two-way coupling mechanism (weather-driven data collection frequency and event-driven processing intensity), it achieves a dynamic balance between energy consumption and accuracy, rather than a one-way adjustment under fixed rules.

[0106] For example, the basic camera's data acquisition cycle is 10 minutes (including 5 activation phases), and the sensor's intermittent acquisition cycle is 5 minutes. Shortening the intermittent acquisition cycle (weather acquisition): If the sensor detects heavy rain (interval cycle from 5 minutes to 3 minutes), the first adjustment formula is triggered, and the data acquisition cycle changes from 10 minutes to 6 minutes (3 minutes per activation phase). Extending the intermittent acquisition cycle (event acquisition): On a sunny day, the sensor interval is extended to 8 minutes. During this period, the infrared sensor captures a sudden change in banana fruit ripening (a sudden event), triggering the second adjustment formula, and the intermittent working cycle changes from 1 to 0 (full cycle analysis).

[0107] Based on the weather conditions, the camera's data acquisition cycle is adjusted according to a first adjustment formula, including: the first adjustment formula is as follows.

[0108]

[0109] in, The adjusted data collection cycle, For the data acquisition cycle before adjustment, W severity The degree of weather severity is represented by α, and the weather impact coefficient is represented by α.

[0110] It should be noted that the parameter descriptions are as follows: Basic data collection cycle (e.g., 10 minutes by default on sunny days). W severity: Weather severity (0 = sunny, 1 = extremely severe), quantifying the intensity of environmental disturbance (e.g., 0.8 for heavy rain, 0.6 for dust storms). α: Weather impact coefficient (0.1–0.3), controlling the adjustment range of the control cycle; the larger the value, the more sensitive the response.

[0111] The exponential function dynamically compresses the period, and the worse the weather (W) severity ↑), The smaller the value, The shorter the interval, the better. This design proactively increases the data collection frequency under extreme weather conditions, ensuring that critical environmental changes (such as mountain displacement and sudden rises in water levels) are captured in a timely manner, avoiding the risk of missed detections due to sudden environmental changes caused by traditional fixed-cycle methods.

[0112] Example, basic data collection cycle Minutes, α = 0.2. Heavy rain (W) severity =0.8): Minutes, Actions: The data acquisition cycle was shortened from 10 minutes to 8.5 minutes, the proportion of the activation phase increased, and two additional mountain images were acquired per cycle. Dust storm (W severity =0.5): minute.

[0113] Based on the aforementioned sudden event, the intermittent working cycle of visual recognition is adjusted according to a second adjustment formula, including: the second adjustment formula is as follows,

[0114]

[0115] in, This is the adjusted intermittent work cycle. For the intermittent work cycle before adjustment, E level β represents the level of the emergency, and β is the event response coefficient.

[0116] It should be noted that the parameter descriptions are as follows: The basic intermittent duty cycle indicates the default frequency of visual recognition execution (e.g., 4 times every 5 device cycles). level : Emergency level (0 = no event, 1 = low risk, 2 = medium risk, 3 = high risk), quantifying the urgency of the event. β: Event response coefficient (0.1~0.5), controlling the adjustment range of the identification frequency; the larger the value, the more aggressive the response.

[0117] The effect of the logarithmic function: ln(E) level +1) Convert linear event levels into nonlinear gains, with faster growth at low levels and slower growth at high levels, to avoid over-response.

[0118] Formula meaning: The higher the level of the emergency (E) level ↑), visual recognition frequency increased ( (↑), reduce the number of cycle skips to enhance monitoring.

[0119] For example, the camera performs visual recognition 4 times by default every 5 cycles (N=5). β = 0.3.

[0120] Banana plant disease outbreak (E level =2):

[0121] Action: The camera's data acquisition cycle is 10 minutes (including 5 two-minute activation phases). Visual recognition performs 4 analyses every 5 cycles by default (skipping 1 time). When it detects that banana plant disease has spread to adjacent plants (E... level When =3), adjust to perform 5 analyses in the full cycle (cancel cycle skipping).

[0122] Rainstorm landslide (E level =3):

[0123] Action: Continuously perform 5 identifications, simultaneously shorten the camera's data acquisition cycle to 6 minutes (originally 10 minutes), and double the density of mountain displacement monitoring.

[0124] Based on the adjusted data acquisition cycle and intermittent working cycle, a first interactive control sequence and a second interactive control sequence are constructed. The interactive control method is determined by the degree of difference between the first interactive control sequence and the second interactive control sequence, including:

[0125] The data acquisition period, the adjusted data acquisition period, and the time interval are combined to form a first interactive control sequence;

[0126] The intermittent work cycle, the adjusted intermittent work cycle, and the time interval are combined to form a second interactive control sequence;

[0127] An interactive control method for determining data features is based on the degree of difference between the first interactive control sequence and the second interactive control sequence; the quantification formula for the degree of difference is as follows.

[0128]

[0129] Where D represents the degree of difference, w1 represents the weighting coefficient of the data collection cycle, w2 represents the weighting coefficient of the intermittent work cycle, and w1+w2=1.

[0130] It should be noted that the construction and function of the interactive control sequence, the first interactive control sequence: Components: Data acquisition cycle The camera's basic duty cycle (e.g., 10 minutes). Adjusted data acquisition cycle. The cycle is adjusted according to the weather formula (e.g., shortened to 8.5 minutes during heavy rain). Time interval (Δt): The time span between adjacent control nodes within the sequence (e.g., the 2-minute activation phase within each camera cycle). Function: To comprehensively reflect the dynamic adjustment process of data acquisition and record the impact of environmental changes on the camera's working rhythm.

[0131] Second interactive control sequence: Components: intermittent work cycle The default number of skip cycles for visual recognition (e.g., skipping once every 5 cycles). Adjusted intermittent duty cycle.

[0132] The period is adjusted according to the event formula (e.g., skipping periods is canceled during sudden events). Time interval (Δt): The time span synchronized with the first sequence. Function: Characterizes the change in visual recognition processing intensity and records the driving effect of sudden events on computing power allocation.

[0133] Analysis of parameters in the formula for quantifying the degree of difference: w1, w2: weighting coefficients (w1+w2=1), reflecting the optimization priority of data acquisition and visual processing. If the focus is on energy consumption optimization: w1=0.7, w2=0.3 (more attention to camera cycle adjustment). If the focus is on response speed: w1=0.3, w2=0.7 (more attention to visual processing adjustment). D value: degree of difference, measuring the overall magnitude of parameter adjustment. Function: High D value: indicates that the environment or event has caused significant parameter adjustments, requiring the activation of high-intensity control strategies (such as disaster protocols). Low D value: indicates stable operation, maintaining baseline strategies or energy-saving modes.

[0134] The logic for determining the interactive control method is based on the degree of difference D, dynamically selecting the control mode: D>D threshold (High Difference): Action: Simultaneously shorten the camera cycle and visual recognition jump cycle, and initiate multi-device collaboration (e.g., drone + sensor). Example: Heavy rain causes D = 5.2 (threshold D). threshold =4.0), trigger all-weather monitoring and shut down non-core sensors. D≤D threshold (Low Difference): Action: Maintain operation according to the adjusted parameters, responding only to changing data. Example: Sunny D=2.3, maintain camera cycle for 10 minutes, visual recognition jumps once every 5 cycles.

[0135] Example, basic parameters: minute, (1 jump), w1 = 0.6, w2 = 0.4, D threshold =4.0.

[0136] Heavy Rain Weather Adjustment: First Sequence Minutes (shortened by 2 minutes), Δt = 2 minutes. Second sequence: (Cancel cycle skipping), Δt = 2 minutes. Difference calculation:

[0137] Control decision: D < 4.0, maintain the current adjustment parameters, but trigger local responses (such as high-frequency scanning of mountain displacement areas).

[0138] Risk of tree branch breakage + torrential rain (extreme event): First priority: Minutes, Δt = 1.5 minutes. Second sequence: Δt = 1.5 minutes. Difference calculation:

[0139] Control Decisions:

[0140] If D>4.0 is not true, but is close to the threshold, then use backup resources (such as waking up the second camera).

[0141] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An artificial intelligence visual recognition and interaction control method, characterized in that, The method comprises the following steps: S1: obtaining a data acquisition cycle of a camera and an interval acquisition cycle of an environmental sensor in a target scene; based on the data acquisition cycle, obtaining an intermittent working cycle of visual recognition; S2: in the data acquisition cycle, dividing data features into steady-state data and transient data; based on the steady-state data and the transient data, obtaining variable data; S3: in the interval acquisition cycle, adjusting the interval acquisition cycle of the variable data; based on the adjusted interval acquisition cycle, adjusting the data acquisition cycle of the camera and the intermittent working cycle of visual recognition, comprising: if the interval acquisition cycle is shortened, obtaining weather conditions of a monitoring area; if the interval acquisition cycle is lengthened, obtaining a sudden event of a monitoring area; based on the weather conditions, adjusting the data acquisition cycle of the camera according to a first adjustment formula; based on the sudden event, adjusting the intermittent working cycle of visual recognition according to a second adjustment formula; S4: based on the adjusted data acquisition cycle and intermittent working cycle, constructing a first interactive control sequence and a second interactive control sequence, and determining an interactive control method according to the difference degree of the first interactive control sequence and the second interactive control sequence, comprising: composing the data acquisition cycle, the adjusted data acquisition cycle and a time interval into the first interactive control sequence; composing the intermittent working cycle, the adjusted intermittent working cycle and a time interval into the second interactive control sequence; the time interval is the duration of the active stage in the camera data acquisition cycle; based on the difference degree of the first interactive control sequence and the second interactive control sequence, determining the interactive control method of the data features; the quantification formula of the difference degree is as follows, Wherein, D is the difference degree, w1 is the weight coefficient of the data collection period, w2 is the weight coefficient of the intermittent working period, w1+w2=1, is the adjusted data collection period, is the unadjusted data collection period, is the adjusted intermittent working period, is the unadjusted intermittent working period.

2. The artificial intelligence visual recognition and interaction control method according to claim 1, characterized in that, the method comprises the following steps: obtaining a data acquisition cycle of a camera and an interval acquisition cycle of an environmental sensor in a target scene; based on the data acquisition cycle, obtaining an intermittent working cycle of visual recognition; S2: in the data acquisition cycle, dividing data features into steady-state data and transient data; based on the steady-state data and the transient data, obtaining variable data; S3: in the interval acquisition cycle, adjusting the interval acquisition cycle of the variable data; based on the adjusted interval acquisition cycle, adjusting the data acquisition cycle of the camera and the intermittent working cycle of visual recognition, comprising: if the interval acquisition cycle is shortened, obtaining weather conditions of a monitoring area; 3. The artificial intelligence visual recognition and interaction control method according to claim 1, characterized in that, if the interval acquisition cycle is lengthened, obtaining a sudden event of a monitoring area; based on the weather conditions, adjusting the data acquisition cycle of the camera according to a first adjustment formula; 4. The artificial intelligence visual recognition and interaction control method according to claim 1, characterized in that, based on the sudden event, adjusting the intermittent working cycle of visual recognition according to a second adjustment formula; S4: based on the adjusted data acquisition cycle and intermittent working cycle, constructing a first interactive control sequence and a second interactive control sequence, and determining an interactive control method according to the difference degree of the first interactive control sequence and the second interactive control sequence, comprising: composing the data acquisition cycle, the adjusted data acquisition cycle and a time interval into the first interactive control sequence; composing the intermittent working cycle, the adjusted intermittent working cycle and a time interval into the second interactive control sequence; the time interval is the duration of the active stage in the camera data acquisition cycle; based on the difference degree of the first interactive control sequence and the second interactive control sequence, determining the interactive control method of the data features; the quantification formula of the difference degree is as follows, the method comprises the following steps: obtaining a data acquisition cycle of a camera and an interval acquisition cycle of an environmental sensor in a target scene; based on the data acquisition cycle, obtaining an intermittent working cycle of visual recognition; S2: in the data acquisition cycle, dividing data features into steady-state data and transient data; based on the steady-state data and the transient data, obtaining variable data; S3: in the interval acquisition cycle, adjusting the interval acquisition cycle of the variable data; based on the adjusted interval acquisition cycle, adjusting the data acquisition cycle of the camera and the intermittent working cycle of visual recognition, comprising: if the interval acquisition cycle is shortened, obtaining weather conditions of a monitoring area; if the interval acquisition cycle is lengthened, obtaining a sudden event of a monitoring area; based on the weather conditions, adjusting the data acquisition cycle of the camera according to a first adjustment formula; based on the sudden event, adjusting the intermittent working cycle of visual recognition according to a second adjustment formula; S4: based on the adjusted data acquisition cycle and intermittent working cycle, constructing a first interactive control sequence and a second interactive control sequence, and determining an interactive control method according to the difference degree of the first interactive control sequence and the second interactive control sequence, comprising: composing the data acquisition cycle, the adjusted data acquisition cycle and a time interval into the first interactive control sequence; composing the intermittent working cycle, the adjusted intermittent working cycle and a time interval into the second interactive control sequence; the time interval is the duration of the active stage in the camera data acquisition cycle; based on the difference degree of the first interactive control sequence and the second interactive control sequence, determining the interactive control method of the data features; the quantification formula of the difference degree is as follows, In the data acquisition period, the data features are divided into steady-state data and transient data based on each intermittent operation period; If the position and shape features of the data features do not exceed the preset threshold in the intermittent operation period, the data features are divided into steady-state data; If the position and shape features of the data features exceed the preset threshold in the intermittent operation period, the data features are divided into transient data.

5. The artificial intelligence visual recognition and interaction control method according to claim 1, characterized in that, The variation data is obtained based on the steady-state data and the transient data, including: In the intermittent operation period, if the steady-state data or the transient data in the next intermittent operation period is converted into the transient data or the steady-state data, the steady-state data or the transient data is divided into variation data.

6. The artificial intelligence visual recognition and interaction control method according to claim 2, characterized in that, The interval acquisition period of the variation data is adjusted in the interval acquisition period, including: If the time interval between the first interactive control node and the second interactive control node is greater than the preset interval threshold, and the number of variation data existing in the time interval between the first interactive control node and the second interactive control node exceeds the preset number threshold, the interval acquisition period of the variation data is shortened; If the time interval between the first interactive control node and the second interactive control node is less than or equal to the preset interval threshold, and the number of variation data existing in the time interval between the first interactive control node and the second interactive control node does not exceed the preset number threshold, the interval acquisition period of the variation data is lengthened.

7. The artificial intelligence visual recognition and interaction control method according to claim 1, characterized in that, The data acquisition period of the camera is adjusted according to a first adjustment formula based on the weather condition, including: wherein, is the adjusted data collection period, is the unadjusted data collection period, W severity is the weather severity, and a is a weather impact factor.

8. The artificial intelligence visual recognition and interaction control method of claim 1, wherein, The intermittent operation period of visual identification is adjusted according to a second adjustment formula based on the sudden event, including: The first adjustment formula is as follows, The second adjustment formula is as follows, wherein, is the adjusted intermittent duty cycle, is the unadjusted intermittent duty cycle, E level is the event severity, and β is the event response coefficient.

Images