Iot edge computing data dynamic integration method and system based on end-cloud cooperation

Abstract

The application provides an Internet of Things edge computing data dynamic integration method and system based on end-cloud cooperation, which comprises the following steps: an edge node receives agricultural monitoring data uploaded by a sensing device and packs the data for uploading according to a default time window; when a physical state mutation of a source sensing device is monitored, an initial trigger timestamp is extracted; a spatial topological relationship is obtained, and a physical propagation vector of a state mutation event spread is deduced in combination with the initial trigger timestamp; a physical time delay bias of an event reaching a downstream sensing device is calculated based on the physical propagation vector; the physical time delay bias is applied to the default time window to generate a phase alignment integral window with time axis translation characteristics; and the agricultural monitoring data of the source and downstream devices are respectively intercepted according to the window to assemble an event evolution spatial sequence package and upload the package. The application realizes dynamic alignment of multi-node agricultural monitoring data in a spatial spread period of a physical event, and improves the accuracy of data integration.

Description

Technical Field

[0001] This invention relates to the field of IoT edge computing and data processing technology, and in particular to a method and system for dynamic integration of IoT edge computing data based on edge-cloud collaboration. Background Technology

[0002] In agricultural IoT applications, numerous sensing devices are typically deployed in farmland to discretely sample environmental physical parameters. The underlying sensing network, using edge computing nodes, aggregates the collected agricultural monitoring data and uploads it to a cloud server. Existing edge data aggregation solutions primarily perform data slicing and packaging based on a pre-defined absolute timeline. The system assembles multi-node data collected within the same time period into a time-series matrix over a fixed time span. This static alignment mechanism, based on a fixed physical clock, maintains regular data transmission order and ensures the synchronization of data acquired from the cloud even when environmental parameters fluctuate smoothly.

[0003] In real farmland environments, physical state changes objectively exhibit spatially spreading characteristics. When a sudden change in physical state occurs locally, the related physical parameter changes will spread and propagate outwards through the environmental medium. Downstream sensing devices along the spread path are affected by geographical distance and terrain damping, resulting in an objective physical time difference in the moment they capture the same abrupt change feature. Using a fixed absolute time window to segment the data of the entire regional network will objectively disperse the physical evolution process with coherent causal relationships into different periodic communication messages.

[0004] When the source sensing device is in a state surge phase, the data from downstream devices may not yet have fluctuated. By the time the downstream devices exhibit characteristic changes, the time capture window of the source device has already switched. This method of segmenting data by absolute time makes it difficult for cloud servers to directly reconstruct the coherent trajectory of a sudden event's propagation along physical space using the original time stamps after receiving the data. The backend system needs to retrieve related data across multiple historical message cycles to reconstruct the event's spread, which objectively increases the processing load for cloud data feature stitching and analysis. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for dynamic integration of IoT edge computing data based on edge-cloud collaboration, in order to solve the technical problem that the existing technology uses fixed time windows to package data, which leads to the fragmentation and dispersion of associated monitoring data in the spatial spread of physical events, thereby increasing the computational load of cross-cycle retrieval and feature reconstruction in the cloud system.

[0006] To achieve the above objectives, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for dynamic integration of IoT edge computing data based on edge-cloud collaboration, applicable to agricultural IoT, including: Edge nodes receive agricultural monitoring data uploaded by sensing devices deployed in different farmland locations, package the agricultural monitoring data according to the default time sequence window, and upload it to the cloud server. When the edge node detects a sudden change in the physical state of the source sensing device, it extracts the initial trigger timestamp of the state change event; The edge node acquires the spatial topological relationship between the source sensing device and the downstream sensing device, and deduces the physical propagation vector of the state mutation event by combining the initial trigger timestamp; Calculate the physical delay offset of the state change event reaching different downstream sensing devices based on the physical propagation vector; The edge node applies the physical delay offset to the default timing window to generate a phase-aligned integral window with time axis translation characteristics; The edge nodes extract the agricultural monitoring data from the source sensing device and the downstream sensing device according to the phase-aligned integral window, assemble them into an event evolution spatial sequence package, and upload them to the cloud server.

[0007] Optionally, the step of packaging the agricultural monitoring data according to the default time-series window includes: Extract the absolute physical timestamp attached to the received agricultural monitoring data; The absolute physical timestamps are assigned to a fixed time interval according to the absolute physical timeline to form the default time window; Agricultural monitoring data belonging to the same default time window are concatenated into a time-aligned two-dimensional matrix, and the two-dimensional matrix is ​​encapsulated into the payload field of a regular periodic message.

[0008] Optionally, the edge node detects a sudden change in the physical state of the source sensing device, including: Construct a moving average calculation queue and update the dynamic baseline of the output values ​​of the source sensing device in real time; Calculate the absolute deviation difference between the latest output value and the dynamic baseline base value; When the absolute deviation exceeds the preset fluctuation tolerance limit and the duration exceeds the false alarm filtering window, it is determined that the source sensing device has experienced a sudden change in physical state.

[0009] Optionally, the deduction of the physical propagation vector of the state mutation event includes: Send a request to the cloud server to retrieve the 3D digital elevation model of the farmland; Locate the three-dimensional spatial nodes corresponding to the source sensing device and the downstream sensing device in the three-dimensional digital elevation model of the farmland; Calculate the straight-line Euclidean distance and elevation difference / slope between the nodes in the three-dimensional space; Extract the default planar diffusion rate corresponding to the category to which the state mutation event belongs, and use the elevation difference slope to perform vector correction on the default planar diffusion rate to obtain the physical propagation vector.

[0010] Optionally, calculating the physical delay offset of the state change event reaching different downstream sensing devices based on the physical propagation vector includes: Obtain the ground vegetation friction resistance coefficient between the source sensing device and the downstream sensing device; The basic propagation time is obtained by dividing the linear Euclidean distance by the velocity component in the physical propagation vector. The damping additional time is obtained by multiplying the basic propagation time by the friction resistance coefficient of the ground vegetation; The physical delay offset is obtained by superimposing the basic propagation time with the damping additional time.

[0011] Optionally, applying the physical delay offset to the default timing window to generate a phase-aligned integration window with time-axis translation characteristics includes: The start and end time spans of the default timing window corresponding to the source sensing device remain unchanged. For any of the downstream sensing devices, the initial trigger timestamp is added to the corresponding physical delay offset to serve as the starting trigger reference for the downstream dedicated integration window. According to the preset physical evolution observation length, the phase-aligned integration window with time axis translation characteristics is generated by extending backward from the starting trigger reference.

[0012] Optionally, assembling the event evolution space sequence package and uploading it to the cloud server includes: Add tags containing spatial topological hierarchy and phase state type to the extracted agricultural monitoring data; According to the spatial topological hierarchy, the agricultural monitoring data with added tags are serialized and concatenated to form the event evolution spatial sequence package that records the entire life cycle of the state change event along its evolution path; The event evolution spatial sequence packet is compressed using a lossless data compression algorithm and then pushed to the cloud server through an encrypted tunnel.

[0013] Optionally, the event evolution space sequence package uploaded to the cloud server is used for the cloud server to parse in order to compare the change amplitude of the environmental peak under the same phase state type of the source sensing device and the downstream sensing device. When the environmental peak change amplitude corresponding to the downstream sensing device decreases and falls below the warning attenuation threshold, the cloud server determines that the state change event has been blocked in the physical propagation of the farmland, and generates an on-site investigation work order based on the spatial location of the blocked downstream sensing device and sends it to the associated agricultural operation and maintenance terminal.

[0014] Optionally, after generating the phase-aligned integration window with time-axis translation characteristics, the method further includes: The edge node continuously monitors the agricultural monitoring data output by all downstream sensing devices; When the agricultural monitoring data from all downstream sensing devices have fallen back to the stable baseline level before the physical state abrupt change, the phase alignment integral window assigned to each downstream sensing device is cancelled. Resume execution of the step of packaging the agricultural monitoring data according to the default timing window.

[0015] Secondly, this invention provides an IoT edge computing data dynamic integration system based on edge-cloud collaboration, applicable to agricultural IoT, comprising: The cloud server is used to receive and process event evolution space sequence packets; Sensing devices are deployed in different locations in farmland to collect and upload agricultural monitoring data; Edge nodes are communicatively connected to both the cloud server and the sensing devices. These edge nodes receive agricultural monitoring data uploaded by sensing devices deployed in different farmland locations, package the data according to a default time-series window, and upload it to the cloud server. When a physical state change is detected at the source sensing device, the initial trigger timestamp of the state change event is extracted. The spatial topology relationship between the source and downstream sensing devices is obtained, and the physical propagation vector of the state change event is deduced based on the initial trigger timestamp. The physical delay offset of the state change event reaching different downstream sensing devices is calculated based on the physical propagation vector. The physical delay offset is applied to the default time-series window to generate a phase-aligned integral window with time-axis translation characteristics. The agricultural monitoring data from the source and downstream sensing devices are extracted according to the phase-aligned integral window, assembled into an event evolution spatial sequence package, and uploaded to the cloud server. The edge node works in conjunction with the cloud server to perform data processing steps as defined in any of the first aspects.

[0016] The present invention has achieved the following beneficial effects: This invention monitors abrupt changes in the physical state of source sensing devices at edge nodes. By combining spatial topological relationships with deduced physical propagation vectors, it calculates the physical delay offset of the abrupt event reaching downstream sensing devices, thereby generating a phase-aligned integral window with time-axis translation characteristics for downstream nodes. This dynamic interception mechanism conforms to the objectively existing spatial propagation delay law of physical events, enabling edge nodes to synchronously intercept multi-source monitoring data according to the natural spread of events in the real environment. This scheme achieves precise temporal phase alignment of environmental parameter change data generated when the same physical event affects different spatial nodes at the edge side and assembles them into a unified event evolution spatial sequence packet for uploading. This data integration logic based on physical propagation characteristics avoids the fragmentation and dispersion of associated evolution data by fixed time-series windows, ensuring that the multi-node data received by the cloud server maintains logical matching in phase state. This reduces the cross-message period retrieval and feature splicing steps required by the cloud to reconstruct the entire life cycle of spatial event evolution, optimizing the execution efficiency of the backend system in processing spatial evolution data.

[0017] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0018] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0019] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a diagram of the overall network physical architecture of the IoT edge computing data dynamic integration system based on edge-cloud collaboration in this embodiment of the invention. Figure 2 This is the main flowchart of the IoT edge computing data dynamic integration method based on edge-cloud collaboration in this embodiment of the invention; Figure 3 This is a flowchart illustrating how agricultural monitoring data is packaged according to a default time-series window in an embodiment of the present invention; Figure 4 This is a flowchart illustrating the physical state change of the edge node monitoring source sensing device in an embodiment of the present invention. Figure 5 This is a flowchart illustrating the physical propagation vector of a state mutation event propagation in an embodiment of the present invention; Figure 6This is a flowchart illustrating the calculation of physical delay offset based on physical propagation vectors in an embodiment of the present invention. Detailed Implementation

[0020] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0021] like Figure 1 As shown, the overall network physical architecture of the IoT edge computing data dynamic integration system based on edge-cloud collaboration is first described. In this embodiment, the agricultural IoT system topology is divided into three layers: a bottom-layer data acquisition network composed of sensing devices, edge nodes located on the data aggregation side, and cloud servers deployed remotely. Sensing devices include soil volumetric moisture sensors, temperature and humidity detectors, specific gas concentration transmitters, and level gauges. These sensing devices are equipped with microcontrollers, analog-to-digital converters, and wireless transceivers, and perform discrete sampling of farmland physical parameters according to a set sampling frequency, converting analog signals into digital agricultural monitoring data. Edge nodes are deployed in communication base stations or control rooms at the farmland site, serving as computing entities connecting the bottom-layer sensor network and the upper-layer wide area network. They are equipped with processors, memory, and network communication interfaces, performing data buffering, time-series reconstruction, and logical computation tasks locally. The cloud server establishes data connections with each edge node through a communication network, responsible for receiving structured data streams and performing backend data analysis and processing. Based on this architecture, this embodiment introduces a data alignment mechanism based on physical space characteristics at the edge side to achieve dynamic integration of discrete time-series data.

[0022] Reference Figure 2 As shown, this embodiment discloses a dynamic data integration method for IoT edge computing based on edge-cloud collaboration, applied to agricultural IoT, including the following data processing steps: Step S11: The edge node receives agricultural monitoring data uploaded by sensing devices deployed in different farmland locations, packages the agricultural monitoring data according to the default time window, and uploads it to the cloud server.

[0023] like Figure 3 As shown, in this embodiment, the step of packaging the agricultural monitoring data according to the default time sequence window includes: extracting the absolute physical timestamps attached to the received agricultural monitoring data; classifying the absolute physical timestamps into a fixed span time interval according to the absolute physical time axis to form the default time sequence window; splicing agricultural monitoring data belonging to the same default time sequence window into a time-aligned two-dimensional matrix; and encapsulating the two-dimensional matrix into the payload field of a regular periodic message.

[0024] Edge nodes run a time synchronization protocol process to calibrate their time with a standard network time source and maintain a globally unified absolute physical timeline in their local operating system. When a sensing device performs physical parameter sampling, it acquires local time parameters, converts them into absolute physical timestamps, and writes them into the protocol header of the agricultural monitoring data message. Upon receiving the message, the edge node parses the protocol and extracts the valid measurement payload and its corresponding absolute physical timestamp. Simultaneously, the edge node divides the monotonically increasing absolute physical timeline into continuous time intervals in its memory, using a preset time span constant as the step size. Each fixed time interval is logically assigned a default timing window with a buffered address. This preset time span constant is not an arbitrarily specified software variable; its calibration is strictly constrained by the maximum transmission unit (MTU) allowed by the underlying wireless local area communication module (such as LoRa or NB-IoT) media access control layer. The microcontroller extracts the single-packet payload byte limit set by the underlying protocol stack and performs a division operation based on the total number of online sensing devices in the current array and the basic discrete sampling frequency. The system accordingly truncates the time span constant and clamps it within a physically permissible range of 15 to 30 minutes. This underlying network parameter tracing mechanism ensures that the total number of bytes in the segmented and packaged two-dimensional matrix approaches the MTU threshold infinitely but avoids overflowing, thus blocking the risk of edge node static random access memory (SRAM) stack overflow caused by network layer IP fragmentation and reassembly backlog at the hardware communication root. The edge node's processor extracts the absolute physical timestamp value from the data packet and compares it with the start and end boundary parameters of each time interval. When the absolute physical timestamp is greater than or equal to the start boundary of a time interval and less than the end boundary, the edge node stores the agricultural monitoring data in the corresponding default time-series window buffer storage area.

[0025] When a default timing window reaches its set termination time boundary, the edge node triggers the matrix splicing process. The edge node extracts the network identifiers of registered online sensing devices in the current system, converts them into continuous integer index values, and uses them as row vector indices to construct the two-dimensional matrix. Based on the system's configured basic sampling frequency, the total duration of the default timing window is divided into continuous time slots, and the slot numbers are converted into column vector indices of the two-dimensional matrix. The edge node traverses all agricultural monitoring data in the default timing window's buffer storage area, calculates the corresponding row and column address coordinates based on the extracted device identifiers and timestamps, and fills the monitoring values ​​into the corresponding address units of the two-dimensional matrix. For missing data positions caused by network latency or device sleep, the edge node writes preset null placeholders. After the two-dimensional matrix splicing is completed, the serialization module converts it into a unidirectional continuous byte stream structure, encapsulates it into the payload field of a regular periodic message, attaches the network routing address and checksum, and uploads it to the cloud server via the communication interface.

[0026] Step S12: When the edge node detects a physical state change in the source sensing device, it extracts the initial trigger timestamp of the state change event.

[0027] like Figure 4 As shown, in this embodiment, the edge node detects a sudden change in the physical state of the source sensing device, including: constructing a moving average calculation queue and updating the dynamic baseline of the output value of the source sensing device in real time; calculating the absolute deviation difference between the latest output value and the dynamic baseline; and determining that the source sensing device has experienced a sudden change in physical state when the absolute deviation difference exceeds a preset fluctuation tolerance limit and the duration exceeds the false alarm filtering window.

[0028] The system internally operates a baseline noise assessment routine to dynamically calibrate preset fluctuation tolerance limits. The processor extracts 100 consecutive historical discrete sampling sequences from each sensing device in a stable sleep state and calculates the standard deviation of its background fluctuations. The preset fluctuation tolerance limit is strictly constrained to The duration of the false alarm filtering window is configured based on the thermodynamic or kinetic hysteresis characteristics of different agricultural parameters: for slowly changing physical quantities such as soil moisture content, the window is fixed at 5 to 8 basic sampling cycles; for rapidly changing parameters such as micrometeorological wind speed, it is shortened to 2 to 3 cycles. This threshold derivation mechanism, which combines dynamic noise floor and parametric physical inertia, blocks the false positive triggering link caused by sensor power supply ripple or high-frequency gusts at the underlying hardware logic level.

[0029] Specifically, for each online sensing device, the edge node allocates a contiguous address space in its memory to establish an independent moving average calculation queue. This queue follows a first-in, first-out (FIFO) logic for data scheduling and has fixed capacity parameters. Whenever a sensing device uploads the latest frame of agricultural monitoring data, the edge node writes it to the tail address of the queue and removes the oldest historical data from the head of the queue. Subsequently, the edge node processor accumulates the valid values ​​in the current queue and calculates the arithmetic mean. This average value is then overwritten into the corresponding baseline parameter variable, updating it to the current dynamic baseline of the source sensing device. This update calculation smooths out the low-frequency, slowly varying parameters of the environmental physical background values.

[0030] When a sensing device outputs a new monitoring value, the edge node reads the value, calculates the algebraic difference between the latest output value and the corresponding dynamic baseline, and performs an absolute value operation to obtain the absolute deviation difference. The edge node internally stores preset fluctuation tolerance limits for different environmental parameter attributes. The processor compares the calculated absolute deviation difference with the preset fluctuation tolerance limits. If the absolute deviation difference is confirmed to be greater than the preset fluctuation tolerance limit, the edge node starts a state detection timer, and the program execution flow transitions to the false alarm filtering process. Within the time span defined by the false alarm filtering window parameter, the edge node continuously calculates the absolute deviation difference for each subsequent latest monitoring value reported by the sensing device and performs a limit comparison. If, within this time window, any calculated absolute deviation difference is less than or equal to the preset fluctuation tolerance limit, the system resets the state timer and resets the judgment logic flag. If, within the continuous time statistical dimension of the false alarm filtering window, each calculated absolute deviation difference is greater than the preset fluctuation tolerance limit, the edge node confirms that a physical state change driven by a continuous physical quantity has occurred in the micro-environment area monitored by the source sensing device. Subsequently, the edge node retrieves the historical agricultural monitoring data record that caused the absolute deviation difference to exceed the preset fluctuation tolerance limit for the first time in the buffer data stack, extracts the time attribute parameter in its message header, and sets it as the initial trigger timestamp of the state change event.

[0031] Step S13: The edge node obtains the spatial topological relationship between the source sensing device and the downstream sensing device, and deduces the physical propagation vector of the state change event by combining the initial trigger timestamp.

[0032] like Figure 5 As shown, in this embodiment, the deduction of the physical propagation vector of the state mutation event includes: sending a request to the cloud server to retrieve the farmland digital elevation 3D model; locating the 3D spatial nodes corresponding to the source sensing device and the downstream sensing device in the farmland digital elevation 3D model; calculating the straight-line Euclidean distance and elevation difference slope between the 3D spatial nodes; extracting the default plane diffusion rate corresponding to the category to which the state mutation event belongs; and using the elevation difference slope to perform vector correction on the default plane diffusion rate to obtain the physical propagation vector.

[0033] The processor's built-in first-order algebraic correction model is defined as In the formula, This is the corrected rate scalar. This is the default planar diffusion rate. The positive or negative sign of the elevation difference slope scalar directly indicates whether the terrain slopes down or up. The environmental gravity coupling constant is used to characterize the fluid potential energy conversion efficiency. For flood irrigation overflow events, where water flow is strongly dominated by gravity, the constant is extracted... It is 0.15 m / s. The value range is calibrated to 0.8 to 1.2; for gaseous diffusion events such as greenhouse gas leaks, the influence of terrain traction is relatively weak, and extraction... Take values ​​ranging from 0.15 to 0.30. Environmental gravity coupling constant. The values ​​are based on a static mapping table: for flood irrigation events, 0.8 is used for sandy soil, 1.0 for loam, and 1.2 for clay; for gaseous diffusion events, 0.15 is used for calm winds (wind speed < 0.5 m / s), 0.22 for light winds (0.5-1.5 m / s), and 0.30 for gentle winds (wind speed > 1.5 m / s). The obtained velocity scalar is combined with its corresponding three-dimensional spatial direction vector to generate the physical propagation vector. This algebraic model avoids the computational burden of partial differential equations on the edge side, accurately adapting to the low-power computing boundaries of agricultural IoT gateways.

[0034] Specifically, after the edge node confirms a sudden change in physical state through its state judgment logic, it sends a data request message containing the local area's grid code to the cloud server. The cloud server retrieves the corresponding farmland digital elevation 3D model data stream file from the repository and distributes it to the edge node. Upon receiving the file stream, the edge node runs a spatial parsing program on its local processor to reconstruct the structure of the farmland digital elevation 3D model. The edge node reads the pre-configured 2D latitude and longitude coordinate parameters of each sensing device from its memory, maps and projects these coordinate parameters into the grid coordinate system of the reconstructed farmland digital elevation 3D model, and uses a bilinear interpolation algorithm to calculate the altitude parameters of the corresponding coordinate projections onto the model's surface. This generates 3D spatial nodes containing longitude, latitude, and altitude depth for the source sensing device and each downstream sensing device.

[0035] The edge node processor extracts the spatial distance components of the 3D spatial nodes corresponding to the source sensing device and the 3D spatial nodes corresponding to the downstream sensing device on three orthogonal spatial axes, and calculates the straight-line Euclidean distance between the two nodes according to the 3D geometric distance formula. Simultaneously, the processor extracts the altitude parameters of the two nodes and performs a subtraction operation to obtain the vertical elevation difference value. This vertical elevation difference is used as the numerator, divided by the projected straight-line distance between the two nodes on the 2D horizontal reference plane as the denominator, to obtain the elevation difference slope scalar. The positive or negative sign attribute of this elevation difference slope parameter is used to characterize the tilt direction parameter of the terrain between the two nodes. The edge node extracts the default planar diffusion rate parameter configured for the event category from the system attribute table according to the mutation event type code. The processor substitutes the calculated elevation difference slope into the physical correction logic equation set by the system to perform algebraic calculations: if the elevation difference slope is a positive value, the calculated positive acceleration compensation parameter is added to the default planar diffusion rate parameter; if the elevation difference slope is a negative value, the corresponding negative attenuation hysteresis parameter is subtracted. The processor combines the final motion rate value after gravity scalar correction with the geometric direction vector pointing to the corresponding downstream node in three-dimensional space to generate the physical propagation vector.

[0036] Step S14: Calculate the physical delay offset of the state change event reaching different downstream sensing devices based on the physical propagation vector.

[0037] like Figure 6 As shown, in this embodiment, the step of calculating the physical delay offset of the state change event reaching different downstream sensing devices based on the physical propagation vector includes: obtaining the ground vegetation friction resistance coefficient between the source sensing device and the downstream sensing device; obtaining the basic propagation time by dividing the linear Euclidean distance by the velocity component in the physical propagation vector; multiplying the basic propagation time by the ground vegetation friction resistance coefficient to obtain the damping additional time; and superimposing the basic propagation time and the damping additional time to obtain the physical delay offset.

[0038] The extraction of the ground vegetation friction coefficient relies on a quantitative dictionary of surface roughness corresponding to agronomic phenological stages. Edge nodes extract the crop growth stage corresponding to the current natural calendar of the target monitoring area, retrieving the damping mapping table from the read-only storage area. For surfaces during winter fallow or seedling stages, the stem and leaf obstruction cross-sectional area is small, and the baseline value of the drag coefficient is calibrated in the range of 0.05 to 0.12. For field crops (such as corn or sorghum) in the jointing and heading stages with a plant height exceeding 1 meter, the mechanical interception effect of the canopy on medium flow increases sharply, and the drag coefficient jumps and locks in at 0.45 to 0.65. Specifically, the extraction is based on the actual plant height of the crop. Perform linear interpolation calculations. Let the desired ground vegetation friction coefficient be... The formula is: If the limit is exceeded, a hard clamp is applied to 0.45 or 0.65. If the spatial physical connection from the source to the downstream crosses multiple grids with differentiated crop attributes, the processor extracts the weight ratio of each grid intercept to the total connection distance and performs a linear weighted summation operation on the multiple drag coefficients obtained from the dictionary lookup. The table lookup mechanism effectively reduces the highly complex aerodynamic drag problem of plant canopy to a discrete weight extraction operation that can be responded to in real time by the microcontroller.

[0039] Specifically, the edge node executes a geographic information query process, using the coordinates of the spatial connection range between the source sensing device and the downstream sensing device as input conditions. It retrieves the local geographic information mapping table, reads the crop type parameter configuration within the path crossing the grid area, and obtains the matching ground vegetation friction resistance coefficient. This ground vegetation friction resistance coefficient is a real-number proportional weight without physical units. The edge node processor calls the division calculation logic unit, using the straight-line Euclidean distance value obtained in step S13 as the dividend, extracting the velocity component scalar from the physical propagation vector as the divisor, and performing division to obtain the basic propagation time parameter with time dimensions. This basic propagation time characterizes the theoretical spatial transmission delay of the medium under the physical condition of excluding vegetation friction resistance.

[0040] The processor invokes the multiplication logic unit to multiply the obtained basic propagation time parameter with the queried ground vegetation friction resistance coefficient configuration item, calculating the damping additional time parameter with time dimensions. This damping additional time reflects the objective delay increment caused by the mechanical obstruction of vegetation during the outward expansion of the medium. Finally, the processor executes an algebraic addition instruction to sum the basic propagation time parameter and the damping additional time parameter, and the resulting sum parameter is recorded and configured as the physical delay bias for this specific downstream sensing device. The execution thread of the edge node traverses all connected downstream nodes within the current topology calculation ripple path graph, repeating the above coefficient retrieval and arithmetic calculation process to generate the physical delay bias configured specifically for each downstream sensing device.

[0041] Step S15: The edge node applies the physical delay offset to the default timing window to generate a phase-aligned integral window with time axis translation characteristics.

[0042] In this embodiment, applying the physical delay offset to the default timing window to generate a phase-aligned integration window with time-axis translation characteristics includes: maintaining the start and end time span of the default timing window corresponding to the source sensing device unchanged; for any corresponding downstream sensing device, adding the corresponding physical delay offset to the initial trigger timestamp as the starting trigger reference for the downstream dedicated integration window; and extending backward from the starting trigger reference according to a preset physical evolution observation length to generate the phase-aligned integration window with time-axis translation characteristics.

[0043] The determination of the physical evolution observation length abandons the static dead zone constant, which easily leads to the forced truncation of the waveform tail. Instead, the system initiates an algebraic calculation process for wave packet broadening based on spatial damping. When the physical medium propagates in the porous environment of farmland, it objectively follows the dispersion diffusion effect; the farther away from the wave source, the smoother the energy attenuation tail. Edge nodes calculate the time step base required for the source sensing device to transition from a state surge to an extreme value maintenance phase, extract the previously calculated three-dimensional linear Euclidean distance, and multiply it by the hysteresis expansion coefficient of the environmental medium (the soil macroporous water conduction expansion coefficient is taken as 1.5, and the micro-meteorological airflow coefficient as 0.3). Finally, the algebraic sum of this product and the time step base is used as the independent physical evolution observation length sent to the downstream node. This adaptive elastic scaling mechanism ensures that the allocated memory addressing space precisely encompasses the far-end, gently stretched physical attenuation tail wave.

[0044] The edge node's memory controller protects the time attribute parameters of the default timing window currently associated with the source sensing device, maintaining the absolute start time and end time span parameters of the data extraction window unchanged. For downstream sensing devices within the event propagation logic calculation range, the edge node processor reads the generated physical delay bias parameter corresponding to the device from the storage area and performs floating-point addition with the initial trigger timestamp value extracted from the buffer record. A new time point value based on delay estimation is calculated, and the system assigns this time point value to the extraction parameter control block of the downstream sensing device, setting it as the starting trigger reference for the downstream dedicated integration window. The edge node reads the event physical evolution observation length parameter set in the initialization configuration file. Using the previously configured starting trigger references of each downstream dedicated integration window as the starting endpoint of the data access time, the physical evolution observation length is added to the time axis record variable inside the processor, generating the end time boundary parameter of the access action. Based on the generated independent start triggering reference and termination time boundary parameter pair, the system memory manager allocates and generates a range structure with dynamic extraction pointers for each corresponding downstream sensing device, namely the phase-aligned integral window with time axis translation characteristics.

[0045] Step S16: The edge node extracts the agricultural monitoring data from the source sensing device and the downstream sensing device according to the phase-aligned integral window, assembles them into an event evolution spatial sequence package, and uploads it to the cloud server.

[0046] In this embodiment, assembling the event evolution spatial sequence package and uploading it to the cloud server includes: adding tags containing spatial topological hierarchy order and phase state type to the extracted agricultural monitoring data; serializing and concatenating the tagged agricultural monitoring data according to the spatial topological hierarchy order to form the event evolution spatial sequence package that records the entire life cycle of the state mutation event along its evolution path; performing volume compression on the event evolution spatial sequence package using a lossless data compression algorithm, and pushing it to the cloud server through an encrypted tunnel.

[0047] The lossless data compression algorithm reuses the waveform isomorphism created by the preceding phase alignment operation at the execution level. Because the physical time delay bias of agricultural monitoring data at each level has been smoothed out in advance by the memory extraction pointer, the same spatial mutation event achieves high time-axis synchronization of waveform fluctuations within the sequence packet. The microcontroller directly extracts the source measurement array with zero layer values ​​as the reference base. Subsequent downstream node data arrays abandon independent full-width floating-point encoding and directly perform point-by-point algebraic subtraction with the corresponding in-phase time slot of the reference base. Under the physical premise of strong waveform correlation, the output spatial difference matrix exhibits super-... The elements collapse to a tiny floating-point band approaching zero, and the edge compression engine only needs to call a very short mapping instruction of 2 to 4 bits to accurately replace these collapse differences. This mapping instruction is stored in a static dictionary table. Let the baseline measurement value be... Relative collapse difference rate The table lookup rules are as follows: Mapped to instruction 00; Mapped to 011; Mapped to 1000; The mapping is 1001. If... Exceeding If the decision band is selected, the escape instruction 1111 is written directly and the original 16-bit floating-point value sequence is appended. This computing power collaboration mechanism effectively eliminates the overhead of dynamically constructing a high-energy-consuming dictionary tree in the microcontroller using the conventional Huffman algorithm, and breaks through the bottleneck of the transmission volume of the full floating-point matrix with extremely low computing power.

[0048] Specifically, the data scheduling process at the edge nodes sends read commands to the memory addressing space based on the time boundary parameters within the phase alignment integration window specific to each sensing device, generated by the configuration. This extracts agricultural monitoring data from the source sensing devices and downstream sensing devices within each window's time span, forming discrete memory data buffer blocks. The edge nodes read the generated network spatial topology model structure and perform a breadth-first search algorithm traversal calculation with the source sensing device node as the origin. The system assigns a spatial topology hierarchy order parameter of zero to the source sensing device; it calculates and selects the first-level downstream devices that directly establish hierarchical connections with the source device, assigning them a spatial topology hierarchy order parameter of one; and so on, incrementing the value to assign the corresponding spatial topology hierarchy order to all sensing nodes involved in the calculation.

[0049] Simultaneously, the edge node processor calls the timing processing subroutine to perform time difference operations on the continuous numerical sequences contained in each extracted discrete data buffer block, extracting the slope change parameter of the data difference between adjacent sampling points. The system assigns state recognition codes to the data array based on the set and saved slope interval threshold judgment conditions: when the calculated difference slope value is continuously positive and greater than the configured limit, a feature code indicating a sudden surge phase is written to the head of this segment of the array; when the difference slope fluctuates slightly near zero and the data value is in the extreme value configuration band, a feature code indicating an extreme value maintenance phase is written; when the difference slope is negative and less than the negative decay set limit, a feature code indicating a decline and decay phase is written.

[0050] The system merges and defines the above state identifiers as phase state types. The threshold limits of the above judgment logic are derived from the real-time calculus state machine of the underlying data stream. Let the absolute time span of adjacent sampling points be... The difference slope of the latest step size is The processor calls the standard deviation of the background noise calculated in the early stage. The slope configuration limit for determining the mutation surge phase is defined as follows: The detection baseline of the extreme value configuration band is anchored to the local physical maximum extreme value within the observation window. When satisfied When the micro-oscillation state is maintained for three consecutive hardware cycles, the state machine locks the extreme value maintenance phase. The negative decision boundary constraint for the decline and decay phase is... Adaptive boundary detection based on dynamic standard deviation eliminates the problem of fixed constant threshold failure caused by environmental temperature drift at the root of computation, ensuring that memory truncation pointers accurately cut the evolution cycle of real physical events. Edge nodes merge and encapsulate the allocated spatial topology hierarchy order parameter values ​​and phase state type codes, writing them into the data packet header of each corresponding truncation data buffer block as parameter markers.

[0051] After completing the attribute injection operation for the data blocks, the internal program of the edge node executes an ascending sorting algorithm on the extracted data buffer blocks according to the spatial topology hierarchy order parameters. Within the allocated contiguous dynamic heap memory space, the system, following the sorted generation order, prioritizes copying and writing the source sensing device data buffer block with a hierarchy value of zero. Then, following a monotonically increasing hierarchy value, the data buffer blocks of downstream sensing devices are sequentially appended and concatenated to the address boundary ending with the previous memory block. Predefined protocol frame delimiting character codes are inserted into the seam areas between the data buffer blocks of different physical devices. The discrete array fragments are serialized and combined, then reconstructed in memory into a byte stream sequence structure with a linear read structure. This reconstructed dataset instance is the event evolution space sequence packet. Subsequently, the edge node calls a lossless data compression algorithm processing routine containing dictionary search matching functionality to scan the reconstructed event evolution space sequence packet and replace internal redundant similar values ​​with short-width encoding, outputting a compressed byte stream. The edge node processor uses a predefined symmetric communication cryptography key to perform data obfuscation encryption on the compressed byte stream. The encrypted message payload is pushed to the cloud server via the network transmission control layer.

[0052] Step S17: After uploading the event evolution spatial sequence package, the method further includes: the event evolution spatial sequence package uploaded to the cloud server is used for the cloud server to parse, so as to compare the environmental peak change amplitude of the source sensing device and the downstream sensing device under the same phase state type; and when the environmental peak change amplitude corresponding to the downstream sensing device decreases and is lower than the warning attenuation threshold, the cloud server determines that the state change event is blocked in the physical propagation of the farmland, and generates a field investigation work order based on the spatial location of the blocked downstream sensing device and sends it to the associated agricultural operation and maintenance terminal.

[0053] The warning attenuation threshold is dynamically calculated based on the negative exponential decay law of kinetic energy dissipation in open farmland space, and its algebraic expression is: In the formula, The warning attenuation threshold, Represents the linear Euclidean distance between the corresponding three-dimensional spatial nodes of the source sensing device and the downstream sensing device; a constant. This refers to the intrinsic dissipation damping rate of the environmental medium in a specific monitoring area. For surface runoff sweep events, Based on the average soil porosity, it was calibrated to be 0.05 to 0.08; for aerosol diffusion-related events, Based on absolute humidity of the air, it is calibrated to be 0.02 to 0.04. Dissipation damping rate. Calculated via linear mapping. Surface runoff events are based on soil porosity. (Configuration range 0.3 to 0.6), the calculation formula is: Aerosol diffusion events are based on absolute air humidity (Configuration range 0 to 25 g / m³), the calculation formula is: When the actual residual ratio falls below the baseline calculated by the above exponential equation... At that time, the cloud engine determined that the kinetic energy drop of the event's spread exceeded the theoretical upper limit of the natural distance decay law, thus accurately classifying the phenomenon as encountering non-natural physical and mechanical obstruction such as accidental closure of agricultural mulch, interception of windbreaks, or collapse of drainage and irrigation ditches.

[0054] After receiving the encrypted payload data stream uploaded by the edge node through the network interface, the cloud server calls the decoding component to perform decryption operations and decompression algorithms to reconstruct the event evolution spatial sequence packet containing the multi-device structure. The cloud-based data parsing daemon scans and reads the protocol frame delimiter character codes set above, and disassembles and segments the reconstructed one-dimensional sequence packet file into multiple data sets belonging to specific sensing devices. The cloud server process extracts the phase state type label and spatial topology hierarchy order parameter values ​​contained in the header of each segmented data set. Based on the phase state type label information, parameter judgment and filtering are performed, and the cloud server processor only extracts specific measurement value fragments labeled with extreme value maintenance phase characteristic codes for subsequent calculations.

[0055] In the data extraction segment marked as the extreme value maintenance phase of the source sensing device, the cloud server calls its internal numerical sorting operator to extract the environmental physical quantity parameter with the largest measured value, and queries and retrieves the dynamic baseline constant updated before the sudden anomaly of the source node stored in the cloud database table. The processor executes floating-point subtraction instructions to calculate the absolute value of the algebraic difference between the aforementioned maximum value and the dynamic baseline constant. This calculated difference value is recorded as the peak change amplitude of the source environment. Based on the increasing logic of the spatial topology hierarchy parameters, the cloud server sequentially performs maximum extreme value extraction and subtraction operations on each downstream sensing device data extraction segment at each level, and outputs the downstream environment peak change amplitude corresponding to the characteristics of each node.

[0056] The cloud server's computing logic unit sets the environmental peak change amplitude parameter of the specific downstream sensing device iterates through as the divisor, retrieves the environmental peak change amplitude parameter of the upstream path node with which it has a topological connection as the divisor, and obtains the actual residual ratio coefficient indicating the state decay characteristics by triggering the division operation instruction logic. The cloud server extracts the preset associated early warning decay threshold under the constraint of the straight-line Euclidean distance between the corresponding three-dimensional spatial nodes. The system's threshold comparator compares the actual residual ratio coefficient with the retrieved early warning decay threshold. When the input comparison condition determines that the decrease in the actual residual ratio coefficient value is lower than the lower limit of the early warning decay threshold, the cloud server's judgment algorithm outputs a conclusion that the environmental propagation has encountered non-natural obstruction barriers in physical space, and determines in the system state parameters that the state change event has been blocked during the physical propagation of the farmland. When the cloud process reads the blocking conclusion signal and is triggered, it automatically associates the operation dispatch mechanism process, reads the deployment coordinate system data of the blocked downstream sensing device and the grid positioning code of the monitoring area from the equipment mapping information database. The cloud server assembles the work order message structure parameters and nests and combines data elements such as coordinate parameters to generate a field inspection work order with a specific application layer protocol structure. Finally, the business gateway component pushes the generated field inspection work order data packet to the agricultural operation and maintenance terminal equipment that has reserved and registered the corresponding monitoring area information via a wide area transmission path.

[0057] Step S18: After generating the phase-aligned integral window with time axis translation characteristics, the method further includes: the edge node continuously monitors the agricultural monitoring data output by all downstream sensing devices; when the agricultural monitoring data of all downstream sensing devices falls back to the stable baseline level before the physical state change, the phase-aligned integral window assigned to each downstream sensing device is cancelled; and the step of packaging the agricultural monitoring data according to the default timing window is resumed.

[0058] During the active period of the abnormal alignment mode, the resident monitoring process of the edge node follows the set polling base sampling cycle to extract the newly acquired real-time agricultural monitoring data values ​​transmitted to the channel by all downstream sensing devices included in the sweep effect graph. The processor performs an algebraic difference comparison between each latest acquired data value and the dynamic baseline value corresponding to the device recorded before the event burst in the storage, and performs absolute value conversion. When the absolute value of the algebraic deviation of a certain downstream sensing device is detected to fall within the boundary limit of the low-amplitude normal fallback tolerance discrimination interval preset in the edge system configuration file, and maintains this low difference fluctuation within the set continuous regression observation time parameter boundary, the system register configuration manager overwrites and modifies the running status parameter flag of the downstream sensing device to the fallback completion attribute flag.

[0059] The decision logic for exiting the deregistration mechanism and the noise floor calculation during the initial anomaly triggering phase form a rigorous mathematical closed loop. The edge processor reads the globally locked dynamic baseline and its baseline standard deviation variable from the temporary register before the event occurs. The boundary limit values ​​of the discrimination interval used to characterize the tolerance of low-amplitude normal decline are converged and clamped at... Within the small envelope band, an asymmetric hysteresis dead zone significantly below the trigger threshold for mutation is formed. Simultaneously, the underlying hardware watchdog timer forcibly sets the boundary of the continuous regression observation time parameter to 5 to 8 basic communication sampling polling cycles. This complementary Schmidt hysteresis verification lock, at the underlying arithmetic level, confirms that the spatial wave kinetic energy has been fully absorbed by the natural background, effectively avoiding frequent false shutdowns and ping-pong oscillations of the window state machine induced by localized capillary water recirculation or high-frequency electromagnetic pulse spikes.

[0060] The edge node main monitoring calculation logic program performs judgment and identification operations on the status parameter tag table record array matrix, which includes all downstream target devices related to the event-affected link range, based on the cyclic query frequency. When the attribute words of each object node included in the array matrix are confirmed to be reset to the fallback attribute tag value words, the edge system kernel confirms that the agricultural monitoring data information of all downstream areas affected by the physical dynamics of the source event has completely converged and subsided, meeting the judgment exit criteria of falling back to the stable baseline level before the physical state change. At this time, the address allocation scheduling manager of the edge node system calls the address clearing control primitive release instruction to erase and destroy the phase alignment integral window control descriptor object carrying the time shift variable address offset amount generated for all downstream nodes in the cache pool control domain and exit the allocation scheduling record list table; the corresponding storage address offset boundary addressing reference restriction state is released, and the reserved cache area is returned to empty and recycled to the public address callable allocation storage pool. The system executes the counter and flow control stack flag instruction to clear and zero the jump execution reset, terminates the call to the burst offset variable control interception judgment calculation module process action; the network task dispatcher restarts and resumes the background resident listening task processing mechanism, and performs two-dimensional time slot slicing, merging and encapsulation and array packaging operations on the received messages in accordance with the default timing window program with fixed system length configuration and absolute physical clock uniform alignment constraint characteristics.

[0061] This invention provides an IoT edge computing data dynamic integration system based on edge-cloud collaboration, applicable to agricultural IoT, comprising: The cloud server is used to receive and process event evolution space sequence packets; Sensing devices are deployed in different locations in farmland to collect and upload agricultural monitoring data; Edge nodes are communicatively connected to both the cloud server and the sensing devices. These edge nodes receive agricultural monitoring data uploaded by sensing devices deployed in different farmland locations, package the data according to a default time-series window, and upload it to the cloud server. When a physical state change is detected at the source sensing device, the initial trigger timestamp of the state change event is extracted. The spatial topology relationship between the source and downstream sensing devices is obtained, and the physical propagation vector of the state change event is deduced based on the initial trigger timestamp. The physical delay offset of the state change event reaching different downstream sensing devices is calculated based on the physical propagation vector. The physical delay offset is applied to the default time-series window to generate a phase-aligned integral window with time-axis translation characteristics. The agricultural monitoring data from the source and downstream sensing devices are extracted according to the phase-aligned integral window, assembled into an event evolution spatial sequence package, and uploaded to the cloud server. The edge node works in conjunction with the cloud server to perform the above data processing steps.

[0062] The aforementioned three layers of hardware physical components achieve interconnected collaboration through the execution of data control instructions between the various underlying processing logic steps. This constitutes the feature extraction, identification, judgment, and offset integration encapsulation logic for massive discrete time-series data of farmland monitoring, supporting the system hardware architecture. The sensing device structure layer includes a physically isolated protective shell and integrates, within the printed circuit board, sensor components with surface-mount soldered encapsulation of probe structures sensitive to changes in various trace environmental elements. Coupled with a multi-stage low-frequency clutter isolation analog signal conditioning hardware amplification and filtering chip network module, serial communication driver access to a multi-channel parallel sampling analog-to-digital converter chip, and internally configured microcontroller processing core management system for data acquisition. Based on its integrated low-power near-field antenna transmission hardware, it sends packet protocol message frame payload data. Edge nodes are integrated relay devices within the base station hardware environment, equipped with a computing platform and a physical operating system. They rely on a high-speed computing motherboard structure, including a high-frequency network control multi-core architecture application processor, a multi-channel dynamic random access addressing (DRAM) array module, and a static solid-state memory (SSD) driver chip containing mapped configuration files, configuration data, and characteristic information, forming the foundation for the computing platform. Through its instruction architecture, it performs register assignment operations to acquire basic window timing, splitting, packing, and alignment actions. It then calls memory to execute sliding difference baseline calculation, iteration comparison, and absolute error arithmetic boundary deviation detection control to prevent misjudgment and time-limited process logic. Its core calculation unit module calls local parsing to construct elevation nodes, extract latitude, longitude, and altitude feature data parameters, perform algebraic calculations to correct path diffusion rate features including terrain slope vectors, and compensate for coupling matching and additional friction coefficients. It outputs physical time offset address boundary adjustment parameters and uses address allocation to extract instructions to generate reconstructed memory control offsets, read variable window attributes, inject sequence features, and perform encrypted volume encoding conversion and transmission actions. The cloud server system utilizes a remote, high-security computing power device cluster with a cloud architecture to configure server nodes. Inside the cluster, high-concurrency process middleware is deployed to parse and extract payload files, complete unpacking, decryption, and release of data, extract data, identify and peel off extreme value data of in-phase state segments, perform calculations, comparisons, and judgments, and process task actions. It also uses application layer interfaces to drive external devices to communicate with the network and publish information.

[0063] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A dynamic data integration method for IoT edge computing based on edge-cloud collaboration, applied to agricultural IoT, characterized in that: include: Edge nodes receive agricultural monitoring data uploaded by sensing devices deployed in different farmland locations, package the agricultural monitoring data according to the default time sequence window, and upload it to the cloud server. When the edge node detects a sudden change in the physical state of the source sensing device, it extracts the initial trigger timestamp of the state change event; The edge node acquires the spatial topological relationship between the source sensing device and the downstream sensing device, and deduces the physical propagation vector of the state mutation event by combining the initial trigger timestamp; Calculate the physical delay offset of the state change event reaching different downstream sensing devices based on the physical propagation vector; The edge node applies the physical delay offset to the default timing window to generate a phase-aligned integral window with time axis translation characteristics; The edge nodes extract the agricultural monitoring data from the source sensing device and the downstream sensing device according to the phase-aligned integral window, assemble them into an event evolution spatial sequence package, and upload them to the cloud server.

2. The method for dynamic integration of IoT edge computing data based on edge-cloud collaboration according to claim 1, characterized in that, The process of packaging the agricultural monitoring data according to the default time-series window includes: Extract the absolute physical timestamp attached to the received agricultural monitoring data; The absolute physical timestamps are assigned to a fixed time interval according to the absolute physical timeline to form the default time window; Agricultural monitoring data belonging to the same default time window are concatenated into a time-aligned two-dimensional matrix, and the two-dimensional matrix is ​​encapsulated into the payload field of a regular periodic message.

3. The method for dynamic integration of IoT edge computing data based on edge-cloud collaboration according to claim 1, characterized in that, The edge node detected a sudden change in the physical state of the source sensing device, including: Construct a moving average calculation queue and update the dynamic baseline of the output values ​​of the source sensing device in real time; Calculate the absolute deviation difference between the latest output value and the dynamic baseline base value; When the absolute deviation exceeds the preset fluctuation tolerance limit and the duration exceeds the false alarm filtering window, it is determined that the source sensing device has experienced a sudden change in physical state.

4. The method for dynamic integration of IoT edge computing data based on edge-cloud collaboration according to claim 1, characterized in that, The deduction of the physical propagation vector of the state change event includes: Send a request to the cloud server to retrieve the 3D digital elevation model of the farmland; Locate the three-dimensional spatial nodes corresponding to the source sensing device and the downstream sensing device in the three-dimensional digital elevation model of the farmland; Calculate the straight-line Euclidean distance and elevation difference / slope between the nodes in the three-dimensional space; Extract the default planar diffusion rate corresponding to the category to which the state mutation event belongs, and use the elevation difference slope to perform vector correction on the default planar diffusion rate to obtain the physical propagation vector.

5. The method for dynamic integration of IoT edge computing data based on edge-cloud collaboration according to claim 4, characterized in that, The calculation of the physical delay offset of the state change event reaching different downstream sensing devices based on the physical propagation vector includes: Obtain the ground vegetation friction resistance coefficient between the source sensing device and the downstream sensing device; The basic propagation time is obtained by dividing the linear Euclidean distance by the velocity component in the physical propagation vector. The damping additional time is obtained by multiplying the basic propagation time by the friction resistance coefficient of the ground vegetation; The physical delay offset is obtained by superimposing the basic propagation time with the damping additional time.

6. The method for dynamic integration of IoT edge computing data based on edge-cloud collaboration according to claim 1, characterized in that, The step of applying the physical delay offset to the default timing window to generate a phase-aligned integration window with time axis translation characteristics includes: The start and end time spans of the default timing window corresponding to the source sensing device remain unchanged. For any of the downstream sensing devices, the initial trigger timestamp is added to the corresponding physical delay offset to serve as the starting trigger reference for the downstream dedicated integration window. According to the preset physical evolution observation length, the phase-aligned integration window with time axis translation characteristics is generated by extending backward from the starting trigger reference.

7. The method for dynamic integration of IoT edge computing data based on edge-cloud collaboration according to claim 6, characterized in that, The assembly into an event evolution space sequence package and its uploading to the cloud server includes: Add tags containing spatial topological hierarchy and phase state type to the extracted agricultural monitoring data; According to the spatial topological hierarchy, the agricultural monitoring data with added tags are serialized and concatenated to form the event evolution spatial sequence package that records the entire life cycle of the state mutation event along its evolution path; The event evolution spatial sequence packet is compressed using a lossless data compression algorithm and then pushed to the cloud server through an encrypted tunnel.

8. The method for dynamic integration of IoT edge computing data based on edge-cloud collaboration according to claim 7, characterized in that, The event evolution space sequence package uploaded to the cloud server is used for the cloud server to parse in order to compare the change amplitude of the environmental peak under the same phase state type of the source sensing device and the downstream sensing device. When the environmental peak change amplitude corresponding to the downstream sensing device decreases and falls below the warning attenuation threshold, the cloud server determines that the state change event has been blocked in the physical propagation of the farmland, and generates an on-site investigation work order based on the spatial location of the blocked downstream sensing device and sends it to the associated agricultural operation and maintenance terminal.

9. The method for dynamic integration of IoT edge computing data based on edge-cloud collaboration according to claim 6, characterized in that, After generating the phase-aligned integration window with time-axis translation characteristics, the method further includes: The edge node continuously monitors the agricultural monitoring data output by all downstream sensing devices; When the agricultural monitoring data from all downstream sensing devices have fallen back to the stable baseline level before the physical state abrupt change, the phase alignment integral window assigned to each downstream sensing device is cancelled. Resume execution of the step of packaging the agricultural monitoring data according to the default timing window.

10. A dynamic data integration system for IoT edge computing based on edge-cloud collaboration, applied to agricultural IoT, characterized in that: include: The cloud server is used to receive and process event evolution space sequence packets; Sensing devices are deployed in different locations in farmland to collect and upload agricultural monitoring data; Edge nodes are communicatively connected to both the cloud server and the sensing devices. These edge nodes receive agricultural monitoring data uploaded by sensing devices deployed in different farmland locations, package the data according to a default time window, and upload it to the cloud server. When a physical state change is detected at the source sensing device, the initial trigger timestamp of the state change event is extracted. The spatial topology relationship between the source sensing device and downstream sensing devices is obtained, and the physical propagation vector of the state change event is deduced based on the initial trigger timestamp. The physical delay offset of the state change event arriving at different downstream sensing devices is calculated based on the physical propagation vector; the physical delay offset is applied to the default timing window to generate a phase-aligned integral window with time axis translation characteristics; The agricultural monitoring data from the source sensing device and the downstream sensing device are extracted according to the phase-aligned integral window, assembled into an event evolution space sequence package, and uploaded to the cloud server. The edge node works in conjunction with the cloud server to perform the data processing steps as defined in any one of claims 1 to 9.

Images