Cloud-edge collaborative airborne particle detection data visualization real-time processing and analysis method

Abstract

This invention relates to the field of airborne detection data processing technology, and more particularly to a cloud-edge collaborative real-time visualization processing and analysis method for airborne particle detection data. This method preprocesses and spatiotemporally groups and aggregates the raw data stream at edge nodes, determines a task allocation strategy based on resource and data feature evaluation, and executes edge-side visualization rendering and cloud-based processing in parallel, ultimately fusing them to generate a visualization result. This achieves intelligent dynamic allocation of data processing load, improving the real-time performance of visualization processing and system resource utilization.

Description

Technical Field

[0001] This invention relates to the field of airborne detection data processing technology, and in particular to a cloud-edge collaborative method for real-time visualization, processing and analysis of airborne particle detection data. Background Technology

[0002] Airborne particle detection technology has significant applications in meteorological observation, environmental monitoring, and scientific research. This technology collects physicochemical property data of various particles in the atmosphere using specialized equipment mounted on a flight platform, forming a raw data stream containing spatiotemporal coordinates and multidimensional characteristic parameters. Traditional processing methods typically employ a centralized architecture, transmitting the raw data collected by the airborne equipment entirely via communication links to a remote data center or high-performance computing platform for unified processing and analysis.

[0003] Current conventional practices rely on centralized processing capabilities in cloud or ground-based computing centers. Airborne detection equipment continuously generates massive amounts of particle detection data during flight operations, which is transmitted back to a remote server via wireless communication networks. On the server side, the received data undergoes preprocessing operations such as decoding, quality verification, and spatiotemporal alignment. Subsequently, statistical analysis, feature extraction, and visualization rendering are performed according to preset algorithms, ultimately generating user-interpretable charts or analysis reports. This centralized processing model requires stable and high-bandwidth communication links to support end-to-end data transmission.

[0004] However, this conventional centralized processing method faces significant limitations in practical applications. Wireless communication conditions in airborne detection environments are often unstable, especially in offshore areas, mountainous regions, or under severe weather conditions, where data transmission links are prone to interruptions, delays, or severely limited bandwidth. The end-to-end transmission of massive amounts of raw data consumes a large amount of communication resources and may cause serious delays in data processing results due to transmission latency, failing to meet the requirements of monitoring tasks with high real-time requirements.

[0005] Furthermore, centralized processing architectures struggle to flexibly adapt to the dynamically changing computational demands and resource constraints of airborne exploration missions. The data characteristics, spatial coverage, and real-time requirements of airborne exploration missions may dynamically change with flight paths and scientific objectives, while a strategy of consistently transmitting all data back to the cloud for processing lacks flexibility. This can lead to the loss of critical data due to data transmission congestion when communication resources are strained, or slow response times when handling sudden, high-priority, detailed local analysis needs. Existing solutions typically lack mechanisms for adaptive task scheduling based on real-time network conditions, edge node computing capabilities, and the characteristics of the data itself, impacting overall processing efficiency and the timeliness of results. Summary of the Invention

[0006] This invention provides a cloud-edge collaborative method for real-time visualization, processing, and analysis of airborne particle detection data, which can solve the problems in the prior art.

[0007] A first aspect of this invention provides a method for real-time visualization, processing, and analysis of airborne particle detection data through cloud-edge collaboration, comprising:

[0008] The raw particle detection data stream collected by the airborne particle detection equipment is acquired, and the raw particle detection data stream is preprocessed at the edge computing node to obtain verified particle detection data;

[0009] Based on the spatiotemporal identifiers in the verified particle detection data, a spatiotemporal index structure is determined at the edge computing node, and the verified particle detection data is grouped and aggregated according to the spatiotemporal index structure to obtain a particle detection data set;

[0010] The computing resource status of the edge computing nodes and the data characteristics of the particle detection dataset are evaluated to determine the allocation strategy for data processing tasks. The allocation strategy represents the boundary between edge-side processing tasks and cloud-side processing tasks.

[0011] According to the allocation strategy, the first part of the particle detection dataset is visualized and rendered at the edge computing node to generate an edge-side visualization result, and the second part of the data is transmitted to the cloud computing platform for processing to generate a cloud-based visualization result;

[0012] The edge computing node merges the edge-side visualization results with the cloud-based visualization results to generate a visualization output result.

[0013] The raw particle detection data stream collected by the airborne particle detection equipment is acquired, and the raw particle detection data stream is preprocessed at the edge computing node to obtain verified particle detection data, including:

[0014] The raw particle detection data stream is obtained from the airborne particle detection device via a data interface. The raw particle detection data stream carries timestamp information and detection location information.

[0015] The edge computing node performs data integrity checks on the original particle detection data stream, identifies the start and end identifiers of data frames based on data frame structure rules, and determines whether the data frames in the original particle detection data stream are complete.

[0016] For data frames determined to be incomplete during the data integrity check, the edge computing node generates a data frame repair request based on the timestamp information and the detection location information, and triggers a data retransmission mechanism based on the data frame repair request to obtain the missing data;

[0017] The data frames determined to be complete in the data integrity check are merged with the missing data obtained through the data retransmission mechanism to obtain a complete data frame sequence;

[0018] The rationality of the particle physics feature values ​​in the complete data frame sequence is judged, and abnormal data frames that do not conform to the particle detection physics constraints are removed to obtain the verified particle detection data.

[0019] Based on the spatiotemporal identifiers in the verified particle detection data, a spatiotemporal index structure is determined at the edge computing node, and the verified particle detection data is grouped and aggregated according to the spatiotemporal index structure to obtain a particle detection data set, including:

[0020] Extract time and spatial location identifiers from the verified particle detection data;

[0021] The edge computing node determines the granularity of the time dimension based on the time identifier and the granularity of the spatial dimension based on the spatial location identifier, and determines the spatiotemporal index structure based on the granularity of the time dimension and the granularity of the spatial dimension.

[0022] The verified particle detection data is traversed, and each data point is mapped to the corresponding index unit in the spatiotemporal index structure based on its time identifier and spatial location identifier.

[0023] For each index unit in the spatiotemporal index structure, the verified particle detection data mapped to the same index unit are grouped to obtain data groups with the same spatiotemporal range identifier;

[0024] The particle detection data in the data group is aggregated and calculated to obtain the statistical feature information of the data group. The data group and its corresponding spatiotemporal range identifier and statistical feature information are combined to obtain the particle detection data set.

[0025] The edge computing node determines the granularity of the time dimension based on the time identifier and the granularity of the spatial dimension based on the spatial location identifier. Based on the granularity of the time dimension and the granularity of the spatial dimension, it determines the spatiotemporal index structure, including:

[0026] The edge computing node calculates the time span of the time marker in the verified particle detection data and counts the data distribution frequency of the verified particle detection data within the time span to obtain the time dimension data density;

[0027] Based on the time span and the time dimension data density, time granularity selection constraints are determined, and the time dimension partitioning granularity is determined according to the time granularity selection constraints.

[0028] The edge computing node calculates the spatial coverage area of ​​the spatial location identifier in the verified particle detection data and statistically analyzes the spatial distribution density of the verified particle detection data within the spatial coverage area.

[0029] Based on the spatial coverage and spatial distribution density, spatial granularity selection constraints are determined, and the granularity of spatial dimension division is determined according to the spatial granularity selection constraints.

[0030] The time span is uniformly divided based on the time dimension granularity to generate a time dimension index sequence, and the spatial coverage is gridded based on the spatial dimension granularity to generate a spatial dimension index grid.

[0031] The spatiotemporal index structure is obtained by combining and mapping the time dimension index sequence with the spatial dimension index grid.

[0032] The computational resource status of the edge computing nodes and the data characteristics of the particle detection dataset are evaluated to determine the data processing task allocation strategy, including:

[0033] Resource status parameters of the edge computing node are collected, and data volume and distribution characteristics are extracted from the particle detection dataset.

[0034] The processing capability of the edge computing node is evaluated based on the resource status parameters to determine the maximum data processing volume that the edge computing node can handle. Furthermore, the processing requirements of the particle detection dataset are evaluated based on the data volume and data distribution characteristics to determine the minimum data processing volume required to complete the visualization processing of the particle detection dataset.

[0035] The maximum data processing capacity that the edge computing node can handle is compared with the minimum data processing capacity required to complete the visualization processing of the particle detection dataset, and the comparison result is obtained.

[0036] Based on the comparison results, the cloud-edge processing ratio is calculated. The particle detection data set is then divided according to the cloud-edge processing ratio and the data distribution characteristics to generate an allocation strategy for the data processing tasks.

[0037] The processing capacity of the edge computing node is evaluated based on the resource status parameters to determine the maximum data processing volume that the edge computing node can handle. Furthermore, the processing requirements of the particle detection dataset are evaluated based on the data volume and data distribution characteristics to determine the minimum data processing volume required to complete the visualization processing of the particle detection dataset, including:

[0038] The available computing resources, storage resources, and communication resources of the edge computing node are extracted from the resource status parameters.

[0039] The overall processing capacity of the edge computing node is obtained by comprehensively calculating the available computing resources, the available storage resources, and the available communication resources.

[0040] Based on the comprehensive processing capability value and historical data processing records, determine the maximum data processing volume that the edge computing node can handle under the current resource conditions;

[0041] Calculate the basic processing requirements of the particle detection dataset based on the data volume.

[0042] Based on the data distribution characteristics, the degree of dispersion of the particle detection data set in the spatiotemporal index structure is analyzed, and the index access overhead is calculated based on the degree of dispersion.

[0043] Based on the basic processing requirements and the index access overhead, the minimum data processing amount required to complete the visualization processing of the particle detection dataset is obtained.

[0044] According to the allocation strategy, the first part of the particle detection dataset is visualized and rendered on the edge computing node to generate an edge-side visualization result, and the second part of the data is transmitted to the cloud computing platform for processing to generate a cloud-based visualization result, including:

[0045] According to the allocation strategy, the first part of the data and the second part of the data are extracted from the particle detection dataset, where the first part of the data corresponds to the edge-side processing range and the second part of the data corresponds to the cloud-side processing range.

[0046] The edge computing node performs spatial mapping on the first part of the data based on the spatiotemporal index structure to determine the coordinate position of each data point in the first part of the data in the visualization space;

[0047] The edge computing node performs graphical rendering on the first part of the data according to the coordinate position to generate the edge-side visualization result;

[0048] The second part of the data is transmitted from the edge computing node to the cloud computing platform via a communication interface;

[0049] The cloud computing platform performs spatial mapping on the second part of the data based on the spatiotemporal index structure to determine the coordinate position of each data point in the second part of the data in the visualization space.

[0050] The cloud computing platform performs graphical rendering on the second part of the data based on the coordinate position to generate the cloud visualization result.

[0051] A second aspect of this invention provides a cloud-edge collaborative airborne particle detection data visualization and real-time processing analysis system, comprising:

[0052] The data preprocessing unit is used to acquire the raw particle detection data stream collected by the airborne particle detection equipment, and to preprocess the raw particle detection data stream at the edge computing node to obtain verified particle detection data.

[0053] The spatiotemporal indexing unit is used to determine the spatiotemporal index structure at the edge computing node based on the spatiotemporal identifier in the verified particle detection data, and to group and aggregate the verified particle detection data according to the spatiotemporal index structure to obtain a particle detection data set.

[0054] The task allocation unit is used to evaluate the computing resource status of the edge computing node and the data characteristics of the particle detection dataset, and determine the allocation strategy for data processing tasks. The allocation strategy represents the boundary between edge-side processing tasks and cloud-side processing tasks.

[0055] The visualization generation unit is used to perform visualization rendering of the first part of the particle detection data set on the edge computing node to generate edge-side visualization results according to the allocation strategy, and to transmit the second part of the data to the cloud computing platform for processing to generate cloud-based visualization results.

[0056] The result fusion unit is used to fuse the edge-side visualization results with the cloud-based visualization results at the edge computing node to generate a visualization output result.

[0057] A third aspect of the present invention provides an electronic device, comprising:

[0058] processor;

[0059] Memory used to store processor-executable instructions;

[0060] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0061] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0062] This invention performs real-time preprocessing and verification of the raw particle detection data stream at edge computing nodes, effectively filtering out noise and abnormal data. This ensures the accuracy and reliability of the data source upon which subsequent analysis relies, laying a solid foundation for high-quality visualization. By constructing an index structure based on spatiotemporal identifiers and grouping and aggregating the data, the efficiency of organizing and managing massive amounts of particle detection data is significantly improved. This allows the data to be efficiently retrieved and called according to its inherent spatiotemporal attributes, supporting the smooth execution of subsequent processing flows.

[0063] This invention formulates task allocation strategies based on dynamic evaluation of edge node resource status and dataset characteristics, achieving intelligent collaboration and elastic scheduling of edge and cloud computing resources. This method retains real-time-critical and computationally-light visualization rendering tasks at the edge, ensuring low-latency initial result output; simultaneously, it offloads computationally intensive complex analysis tasks to the cloud, fully utilizing the cloud's powerful computing capabilities, thereby optimizing overall processing capacity and response speed.

[0064] This invention fuses locally generated rendering results with analysis results returned from the cloud at edge nodes, generating a unified and comprehensive visualization output. This cloud-edge collaborative processing architecture not only significantly reduces the bandwidth pressure and latency of transmitting raw data to the cloud, but also achieves multi-level, integrated visualization from macroscopic distribution patterns to microscopic particle characteristics through the combination of real-time rendering at the edge and in-depth analysis in the cloud, greatly improving the real-time processing efficiency and context awareness capabilities of airborne particle detection data. Attached Figure Description

[0065] Figure 1 This is a flowchart illustrating the real-time visualization and analysis method for airborne particle detection data using cloud-edge collaboration, as described in an embodiment of the present invention.

[0066] Figure 2 This is a flowchart illustrating the process of determining the spatiotemporal index structure according to an embodiment of the present invention. Detailed Implementation

[0067] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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.

[0068] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0069] Figure 1 This is a flowchart illustrating the real-time processing and analysis method for cloud-edge collaborative airborne particle detection data visualization according to an embodiment of the present invention. Figure 1 As shown, the cloud-edge collaborative method for real-time visualization and analysis of airborne particle detection data includes:

[0070] The raw particle detection data stream collected by the airborne particle detection equipment is acquired, and the raw particle detection data stream is preprocessed at the edge computing node to obtain verified particle detection data;

[0071] Based on the spatiotemporal identifiers in the verified particle detection data, a spatiotemporal index structure is determined at the edge computing node, and the verified particle detection data is grouped and aggregated according to the spatiotemporal index structure to obtain a particle detection data set;

[0072] The computing resource status of the edge computing nodes and the data characteristics of the particle detection dataset are evaluated to determine the allocation strategy for data processing tasks. The allocation strategy represents the boundary between edge-side processing tasks and cloud-side processing tasks.

[0073] According to the allocation strategy, the first part of the particle detection dataset is visualized and rendered at the edge computing node to generate an edge-side visualization result, and the second part of the data is transmitted to the cloud computing platform for processing to generate a cloud-based visualization result;

[0074] The edge computing node merges the edge-side visualization results with the cloud-based visualization results to generate a visualization output result.

[0075] The raw particle detection data stream collected by the airborne particle detection equipment is acquired, and the raw particle detection data stream is preprocessed at the edge computing node to obtain verified particle detection data, including:

[0076] The raw particle detection data stream is obtained from the airborne particle detection device via a data interface. The raw particle detection data stream carries timestamp information and detection location information.

[0077] The edge computing node performs data integrity checks on the original particle detection data stream, identifies the start and end identifiers of data frames based on data frame structure rules, and determines whether the data frames in the original particle detection data stream are complete.

[0078] For data frames determined to be incomplete during the data integrity check, the edge computing node generates a data frame repair request based on the timestamp information and the detection location information, and triggers a data retransmission mechanism based on the data frame repair request to obtain the missing data;

[0079] The data frames determined to be complete in the data integrity check are merged with the missing data obtained through the data retransmission mechanism to obtain a complete data frame sequence;

[0080] The rationality of the particle physics feature values ​​in the complete data frame sequence is judged, and abnormal data frames that do not conform to the particle detection physics constraints are removed to obtain the verified particle detection data.

[0081] Raw particle detection data streams are acquired through a dedicated data interface on the airborne particle detection equipment. This interface supports high-speed serial communication protocols, with data transmission rates ranging from 10 Mbps to 100 Mbps. The raw particle detection data streams are organized in the form of data frames. Each data frame contains three parts: a frame header, a payload, and a frame trailer. The frame header carries timestamp information in GPS-synchronized UTC time format with microsecond-level accuracy. The detection location information includes geographic coordinates in three dimensions: longitude, latitude, and altitude, with a positioning accuracy better than 10 meters. The payload portion of the data frame records the core physical quantities of particle detection, including particle energy, particle type identifier, detector channel number, and signal amplitude.

[0082] Upon receiving the raw particle detection data stream, the edge computing node immediately initiates a data integrity check process. This check is based on predefined data frame structure rules. These rules stipulate that the frame header uses a specific byte sequence as the start identifier, such as the hexadecimal sequence 0xAA55, and the frame tail uses another specific byte sequence as the end identifier, such as the hexadecimal sequence 0x55AA. The check process scans the received data stream, locates the start and end identifiers of each data frame, calculates the byte length between the two identifiers, and compares this length with the standard frame length defined in the data frame structure rules. If the actual byte length matches the standard frame length and the intra-frame checksum verification passes, the data frame is considered complete; otherwise, if the byte length is insufficient, the checksum is incorrect, or the start or end identifier is missing, the data frame is considered incomplete.

[0083] For data frames deemed incomplete, the corresponding timestamp and detection location information are extracted to generate a data frame repair request. This request includes the time range, geographical location range, and sequence number of the missing data frame, and is sent to the airborne particle detection device via a wireless communication link. The data retransmission mechanism maintains a historical data cache within a certain time window on the airborne device, typically with a cache depth of 5 to 30 seconds of data. Upon receiving the repair request, the airborne device retrieves the data frame with the corresponding timestamp and location from the historical data cache, repackages it, and transmits it to the edge computing node via the data interface. The edge computing node sets a retransmission timeout; if the missing data is not received within the timeout period, the data frame is recorded as permanently lost, and interpolation compensation is performed in subsequent processing.

[0084] Data frames deemed complete during data integrity checks are arranged in timestamp order to form a preliminary data frame sequence. Simultaneously, missing data obtained through the data retransmission mechanism is inserted into the corresponding positions in the preliminary data frame sequence based on its timestamp information, ensuring the entire sequence is strictly ordered in ascending order of time. The merging process employs a merge sort algorithm with a logarithmic linear time complexity, capable of efficiently handling large-scale data streams. After merging, a complete data frame sequence is formed, which is continuous in time and contains no time gaps due to data loss.

[0085] The rationality of particle physical characteristic values ​​in the complete data frame sequence is judged based on particle detection physical constraints, which include energy range constraints, particle type-energy correlation constraints, and signal amplitude threshold constraints. The energy range constraint stipulates that particle energy must be within a physically reasonable range; for example, the energy of cosmic ray particles is typically between 1 MeV and 10 GeV, and values ​​outside this range are considered anomalous. The particle type-energy correlation constraint sets corresponding energy distribution expectations based on the physical characteristics of different particle types. For example, the energy spectrum of electrons and protons have different distribution characteristics; if a data frame identifies the particle type as electron but the energy value conforms to the typical distribution of protons, it is judged as anomalous. The signal amplitude threshold constraint checks whether the amplitude of the electrical signal output by the detector is within the instrument's dynamic range, avoiding erroneous data caused by detector saturation or noise interference. Each frame in the complete data frame sequence is traversed, and the particle energy, particle type identification, and signal amplitude are checked one by one to see if they meet the above physical constraints. Data frames that do not meet any of the constraints are marked as anomalous data frames and removed from the sequence.

[0086] The data sequence obtained after removing anomalous data frames is the verified particle detection data. This dataset has passed verification in both data integrity and physical plausibility dimensions, and can serve as reliable input for subsequent spatiotemporal index construction and visualization processing. The verified particle detection data retains the original timestamp information and detection location information, while adding data quality identifiers and recording the verification steps and results of each data frame, providing a basis for data traceability and quality assessment. The entire preprocessing process is executed in real time on edge computing nodes, with processing latency controlled in the range of milliseconds to hundreds of milliseconds, ensuring that it can keep up with the data output rate of the airborne particle detection equipment and achieving efficient data verification in streaming processing mode.

[0087] Based on the spatiotemporal identifiers in the verified particle detection data, a spatiotemporal index structure is determined at the edge computing node, and the verified particle detection data is grouped and aggregated according to the spatiotemporal index structure to obtain a particle detection data set, including:

[0088] Extract time and spatial location identifiers from the verified particle detection data;

[0089] The edge computing node determines the granularity of the time dimension based on the time identifier and the granularity of the spatial dimension based on the spatial location identifier, and determines the spatiotemporal index structure based on the granularity of the time dimension and the granularity of the spatial dimension.

[0090] The verified particle detection data is traversed, and each data point is mapped to the corresponding index unit in the spatiotemporal index structure based on its time identifier and spatial location identifier.

[0091] For each index unit in the spatiotemporal index structure, the verified particle detection data mapped to the same index unit are grouped to obtain data groups with the same spatiotemporal range identifier;

[0092] The particle detection data in the data group is aggregated and calculated to obtain the statistical feature information of the data group. The data group and its corresponding spatiotemporal range identifier and statistical feature information are combined to obtain the particle detection data set.

[0093] In practical applications of airborne particle detection, because the aircraft continuously moves in space and collects particle information in real time, the verified particle detection data exhibits significant spatiotemporal distribution characteristics. To efficiently organize and manage this data, edge computing nodes need to establish a reasonable spatiotemporal index structure. Specifically, after receiving verified particle detection data, the edge computing node first extracts the time identifier and spatial location identifier from each particle detection record. The time identifier typically contains the timestamp information of the acquisition time, accurate to the millisecond level, used to mark the time attribute of the particle detection data. The spatial location identifier contains the three-dimensional geographic coordinates of the acquisition point, usually composed of longitude, latitude, and altitude information provided by the airborne navigation system, used to mark the spatial attribute of the particle detection data.

[0094] After extracting spatiotemporal identifiers, edge computing nodes need to determine an appropriate spatiotemporal partitioning granularity to construct a spatiotemporal index structure. For the time dimension, edge computing nodes analyze the distribution characteristics of time identifiers and determine the partitioning granularity of the time dimension based on the data stream acquisition frequency and subsequent processing requirements. When airborne particle detection equipment acquires data at a high frequency, the partitioning granularity of the time dimension can be set to a smaller time interval, such as 10 seconds or 30 seconds, to achieve fine temporal resolution. When the acquisition frequency is relatively low or processing resources are limited, the partitioning granularity of the time dimension can be set to a larger time interval, such as 1 minute or 5 minutes. For the spatial dimension, edge computing nodes analyze the distribution range of spatial location identifiers and determine the partitioning granularity of the spatial dimension based on the spatial span and data density of the detection area. In practical applications, a grid partitioning method can be used to decompose three-dimensional space into several regular spatial units, and the size of each spatial unit is the partitioning granularity of the spatial dimension. For example, the longitude direction can be partitioned at 0.1-degree intervals, the latitude direction at 0.1-degree intervals, and the altitude direction at 500-meter intervals to form a three-dimensional grid structure.

[0095] After determining the granularity of the time and spatial dimensions, edge computing nodes establish a spatiotemporal index structure based on these granularities. This spatiotemporal index structure employs a multidimensional indexing approach, combining the time and spatial dimensions to form a four-dimensional index space, which includes one time dimension and three spatial dimensions. Each index unit in the spatiotemporal index structure corresponds to a specific spatiotemporal range, defined by a specific time period and a specific spatial region. For example, an index unit might correspond to a time range from 12:00:00 to 12:00:10 on a certain day, and a spatial range of 110.0 to 110.1 degrees longitude, 30.0 to 30.1 degrees latitude, and 5000 to 5500 meters altitude—a four-dimensional spatiotemporal region. Edge computing nodes can use data structures such as hash tables, tree structures, or multidimensional arrays to implement the spatiotemporal index structure, ensuring rapid location and access to index units corresponding to any spatiotemporal range.

[0096] After establishing the spatiotemporal index structure, the edge computing nodes traverse all verified particle detection data, performing a mapping operation on each data record. Specifically, for the currently traversed particle detection data record, the edge computing node reads the record's time identifier and spatial location identifier, and then calculates the index unit to which the record should belong based on these identifiers. The calculation of the time identifier involves mapping a specific timestamp to a corresponding time interval; for example, mapping a time 12:03:25 to the time interval from 12:03:20 to 12:03:30. The calculation of the spatial location identifier involves mapping three-dimensional coordinates to corresponding spatial grid units; for example, mapping a location with longitude 110.05 degrees, latitude 30.08 degrees, and altitude 5200 meters to spatial units with longitude from 110.0 degrees to 110.1 degrees, latitude from 30.0 degrees to 30.1 degrees, and altitude from 5000 meters to 5500 meters. After completing the mapping calculation, the edge computing node adds the reference to the particle detection data record or the data itself to the data list of the corresponding index unit. By traversing all the data and performing mapping operations, all verified particle detection data are assigned to the various index units of the spatiotemporal index structure.

[0097] After data mapping is completed, the edge computing nodes perform grouping operations on each index cell in the spatiotemporal index structure. For each index cell, all particle detection data mapped to that cell have the same spatiotemporal range attribute, and the edge computing nodes organize these data into a data group. Each data group has a clear spatiotemporal range identifier, which is composed of the time range and spatial range of the corresponding index cell. For example, the spatiotemporal range identifier of a certain data group is time 12:00:00 to 12:00:10, longitude 110.0 degrees to 110.1 degrees, latitude 30.0 degrees to 30.1 degrees, and altitude 5000 meters to 5500 meters. In this way, the originally discretely distributed particle detection data is organized into several data groups with clear spatiotemporal attributes, which facilitates subsequent regional processing and analysis.

[0098] For each data group, the edge computing node further performs aggregation computation to extract statistical feature information. Aggregation computation performs statistical analysis based on the specific content of the particle detection data in the data group, generating statistical indicators that summarize the particle distribution characteristics of that spatiotemporal region. Common statistical features include the total number of particle detection records in the data group, representing the amount of effective data collected within that spatiotemporal range; the average, maximum, and minimum particle concentrations, representing the concentration trend and range of variation of particle concentration within that spatiotemporal region; statistical characteristics of particle size distribution, such as the median and standard deviation of particle size; and other statistical quantities related to particle physical properties. The edge computing node can select which statistical features to calculate based on actual application needs, controlling computational overhead while ensuring information integrity.

[0099] After the aggregation calculation is completed, the edge computing node combines and encapsulates each data group, its corresponding spatiotemporal range identifier, and the calculated statistical feature information. The combined data unit contains the original set of particle detection data records, clear spatiotemporal positioning information, and a general statistical description, constituting an element in the particle detection dataset.

[0100] Figure 2 This is a flowchart illustrating the process of determining the spatiotemporal index structure according to an embodiment of the present invention, as follows: Figure 2 As shown, the edge computing node determines the granularity of the time dimension based on the time identifier and the granularity of the spatial dimension based on the spatial location identifier. Based on the granularity of the time dimension and the granularity of the spatial dimension, it determines the spatiotemporal index structure, including:

[0101] The edge computing node calculates the time span of the time marker in the verified particle detection data and counts the data distribution frequency of the verified particle detection data within the time span to obtain the time dimension data density;

[0102] Based on the time span and the time dimension data density, time granularity selection constraints are determined, and the time dimension partitioning granularity is determined according to the time granularity selection constraints.

[0103] The edge computing node calculates the spatial coverage area of ​​the spatial location identifier in the verified particle detection data and statistically analyzes the spatial distribution density of the verified particle detection data within the spatial coverage area.

[0104] Based on the spatial coverage and spatial distribution density, spatial granularity selection constraints are determined, and the granularity of spatial dimension division is determined according to the spatial granularity selection constraints.

[0105] The time span is uniformly divided based on the time dimension granularity to generate a time dimension index sequence, and the spatial coverage is gridded based on the spatial dimension granularity to generate a spatial dimension index grid.

[0106] The spatiotemporal index structure is obtained by combining and mapping the time dimension index sequence with the spatial dimension index grid.

[0107] Airborne particle detection equipment continuously collects particle data during flight. Each data record includes a time stamp and a spatial location identifier. The time stamp is typically a UTC timestamp, accurate to the millisecond level; the spatial location identifier includes longitude, latitude, and altitude information, acquired in real time via the airborne GPS positioning system. After receiving the verified particle detection data, the edge computing node first extracts the time stamps from all data records and determines the earliest time by traversing the timestamp sequence. and latest time The time span was calculated. Simultaneously, the time span is divided into several fixed-length time windows, and the number of data records within each time window is counted. The time dimension data density is obtained by calculating the average amount of data per unit time. When the data density is high, it indicates that particle detection activity is frequent within that time period, requiring finer time granularity for differentiation; when the data density is low, coarser time granularity can be used for aggregation.

[0108] The selection of time granularity takes into account the relationship between the total length of the time span and the data density. If the total length of the time span is short and the data density is high, then seconds or minutes are chosen as the granularity for dividing the time dimension. If the total length of the time span is long and the data density distribution is uneven, an adaptive strategy is adopted, using a smaller time granularity in data-dense areas and a larger time granularity in data-sparse areas. In specific implementation, a data density threshold is set. When the data density over a certain period of time When the time dimension of this period is divided into granularities, it is set to 1 minute; when At this time, the granularity of the time dimension is set to 5 minutes or 10 minutes. This dynamic adjustment mechanism ensures that a reasonable granularity of the time dimension can be obtained under different data density scenarios.

[0109] When determining the granularity of spatial dimension division, the spatial location markers in the verified particle detection data are first traversed to extract all latitude and longitude coordinates and calculate the longitude range. and latitude range and height range The spatial coverage area is obtained. This area is then divided into a preliminary grid, and the number of data points within each grid cell is counted to calculate the spatial distribution density. Spatial distribution density reflects the degree of aggregation of particle detection data in three-dimensional space, and is usually expressed as the number of data points per unit volume. In airborne particle detection applications, flight trajectories are relatively concentrated, with higher data density near the flight path and lower density in areas deviating from the path.

[0110] The spatial granularity selection constraint needs to balance spatial resolution and data aggregation efficiency. When the spatial coverage is large, using a coarser spatial granularity can reduce the complexity of the index structure; when the spatial distribution density is uneven, a finer spatial granularity should be used in high-density areas and a coarser spatial granularity in low-density areas. In practice, the base grid size is determined based on the latitude and longitude span of the spatial coverage. For example, when the latitude and longitude span exceeds 10 degrees, the base grid is set to 0.1 degrees multiplied by 0.1 degrees; when the latitude and longitude span is between 1 and 10 degrees, the base grid is set to 0.01 degrees multiplied by 0.01 degrees. For the altitude dimension, the altitude layer division is determined based on the flight altitude range, typically every 500 meters is one altitude layer. Based on the base grid, dynamic refinement is performed according to the spatial distribution density of each grid cell. High-density grid cells are further subdivided into smaller subgrids, while low-density grid cells maintain the base granularity or are merged.

[0111] After determining the granularity of the time dimension, the time span is uniformly divided according to this granularity. Let's assume the granularity of the time dimension is... Then the time span Classified as Generate a time-dimensional index sequence for each time period. , where each time index Corresponding time interval The value of i ranges from [1, ..., ... The time dimension index sequence is stored in a sequential structure, which facilitates quick location of the time index to which a specific moment belongs.

[0112] After determining the granularity of spatial dimension division, the spatial coverage area is divided into grids according to this granularity. On the latitude and longitude plane, the granularity is divided according to longitude. Granularity based on latitude Divide the longitude range into Each longitude unit divides the latitude range into 100 latitude units. Each dimensional unit forms a two-dimensional planar grid. In the height dimension, the granularity is divided according to height. Divide the height range into A height layer is generated. Through three-dimensional combination, a spatial dimension index grid is produced. Each grid cell is uniquely identified by a longitude index, a latitude index, and a height index, denoted as […]. , where j is the longitude index, k is the latitude index, and l is the altitude index.

[0113] The computational resource status of the edge computing nodes and the data characteristics of the particle detection dataset are evaluated to determine the data processing task allocation strategy, including:

[0114] Resource status parameters of the edge computing node are collected, and data volume and distribution characteristics are extracted from the particle detection dataset.

[0115] The processing capability of the edge computing node is evaluated based on the resource status parameters to determine the maximum data processing volume that the edge computing node can handle. Furthermore, the processing requirements of the particle detection dataset are evaluated based on the data volume and data distribution characteristics to determine the minimum data processing volume required to complete the visualization processing of the particle detection dataset.

[0116] The maximum data processing capacity that the edge computing node can handle is compared with the minimum data processing capacity required to complete the visualization processing of the particle detection dataset, and the comparison result is obtained.

[0117] Based on the comparison results, the cloud-edge processing ratio is calculated. The particle detection data set is then divided according to the cloud-edge processing ratio and the data distribution characteristics to generate an allocation strategy for the data processing tasks.

[0118] When evaluating the computing resource status of edge computing nodes and the data characteristics of particle detection datasets, it is necessary to comprehensively consider the matching relationship between the hardware capabilities of the edge computing environment and the data processing requirements. A resource monitoring module runs on the edge computing nodes to collect real-time resource status parameters, including CPU utilization, available memory capacity, GPU load, storage device I / O rates, and network bandwidth availability. CPU utilization reflects the current workload of the processor cores; the instantaneous load of each core is obtained by reading the processor monitoring interface provided by the operating system. Available memory capacity is obtained by querying the system memory management unit, including the amount of free physical memory and cache usage. GPU load is particularly critical for visualization rendering tasks; metrics such as video memory usage, compute unit utilization, and temperature are queried through the GPU driver interface. Storage device I / O rates reflect data read / write throughput, measured by monitoring the disk controller's read / write operation frequency and data transfer rate. Network bandwidth availability is determined through network interface card traffic statistics and link quality detection, including the real-time availability of uplink and downlink bandwidth.

[0119] When extracting the data volume scale from the particle detection dataset, the total number of particle records, the storage space occupied by the data, and the arrival rate of the data stream are statistically analyzed. The total number of particle records directly reflects the number of data entries to be processed. The storage space occupied by the data is obtained by summing the number of bytes in each data block. The arrival rate of the data stream is determined by calculating the amount of data received per unit time. Extracting data distribution characteristics involves analyzing the distribution patterns of particle detection data in both time and spatial dimensions. In the time dimension, the distribution of data arrival time intervals is calculated to identify any sudden peaks or periodic fluctuations in the data stream. In the spatial dimension, the geographical distribution density of particle detection locations is analyzed to determine the degree of data aggregation in different spatial regions. By constructing a statistical histogram of data distribution, the spatial region is divided into multiple grid cells, and the number of particles in each grid cell is counted to form a spatial density distribution map. Simultaneously, the correlation characteristics of the data are extracted to determine whether there is a strong correlation between particle detection data from adjacent spatiotemporal locations. This correlation directly affects the computational complexity of subsequent visualization rendering.

[0120] When evaluating the processing capabilities of edge computing nodes based on collected resource status parameters, a processing capability evaluation model is established. This model comprehensively considers the computing power of the central processing unit (CPU), the rendering capability of the graphics processing unit (GPU), and the caching capability of memory. The CPU's computing power is calculated using its number of cores, clock speed, and current utilization rate to determine the available computing resources. The GPU's rendering capability is determined by its number of stream processors, memory bandwidth, and current load to determine the intensity of rendering tasks it can handle. Memory caching capability depends on available memory capacity and data access speed. These hardware resource capabilities are mapped to a unified processing capability metric and comprehensively evaluated using a weighted summation method. The weighting coefficients are set according to the characteristics of the visualization task; for the visualization rendering of particle detection data, the GPU capability has a higher weight. Based on this, considering the real-time load of the edge computing node, the resource occupancy currently allocated to other tasks is subtracted from the evaluated theoretical processing capability to obtain the maximum data processing capacity that the edge computing node can handle. This maximum data processing capacity is expressed as the number of particle records that can be processed or the number of visualization units that can be rendered per unit time.

[0121] When assessing the processing requirements of the particle detection dataset, the computational resource consumption required to complete the visualization process is calculated based on the data volume and distribution characteristics. Different visualization rendering methods correspond to different computational complexities. For scattered point visualization, the computational complexity is proportional to the number of particles; for density field visualization, the computational complexity is related to the resolution of the spatial grid and the accuracy of the interpolation algorithm. Based on the preset visualization quality requirements, rendering parameter configurations are determined, including image resolution, color depth, and anti-aliasing level. These rendering parameters are combined with the data volume to estimate the number of floating-point operations, memory accesses, and image rendering calls required to complete the visualization of the entire dataset. Historical task execution records stored in the performance benchmark database are used to find historical cases with similar characteristics to the current task, and the actual execution performance data of these cases is used to calibrate the theoretical estimation results. Finally, the minimum data processing volume required to complete the visualization of the particle detection dataset is determined, which represents the total amount of computational resources required to complete the task.

[0122] The maximum data processing capacity of an edge computing node is compared with the minimum data processing capacity required to complete the visualization of the particle detection dataset. A comparison result is generated. When the maximum data processing capacity is greater than the minimum, it indicates that the edge computing node is capable of independently completing the entire visualization task, and the comparison result is marked as "edge sufficient." When the maximum data processing capacity is less than the minimum, it indicates that the edge computing node has insufficient resources and requires cloud assistance; the comparison result is marked as "edge insufficient." The difference between the two is calculated, representing either resource surplus or resource gap. Resource surplus reflects the idle capacity of the edge computing node, which can be used to process more complex visualization effects or higher rendering quality. Resource gap reflects the scale of computing resources that need to be supplemented by the cloud and is a key basis for determining the cloud-edge processing ratio.

[0123] When calculating the cloud-edge processing ratio based on the comparison results, under conditions of sufficient edge computing, all data is processed first on edge computing nodes, with the cloud-edge processing ratio set to 0:100 (zero cloud processing, one hundred edge processing). Under conditions of insufficient edge computing, the proportion of tasks that need to be offloaded to the cloud is determined based on the ratio of resource gap to minimum data processing volume. The cloud processing ratio equals the resource gap divided by the minimum data processing volume, and the edge processing ratio equals 1 minus the cloud processing ratio. A dynamic adjustment mechanism is introduced to consider the impact of network transmission latency on overall processing performance.

[0124] The processing capacity of the edge computing node is evaluated based on the resource status parameters to determine the maximum data processing volume that the edge computing node can handle. Furthermore, the processing requirements of the particle detection dataset are evaluated based on the data volume and data distribution characteristics to determine the minimum data processing volume required to complete the visualization processing of the particle detection dataset, including:

[0125] The available computing resources, storage resources, and communication resources of the edge computing node are extracted from the resource status parameters.

[0126] The overall processing capacity of the edge computing node is obtained by comprehensively calculating the available computing resources, the available storage resources, and the available communication resources.

[0127] Based on the comprehensive processing capability value and historical data processing records, determine the maximum data processing volume that the edge computing node can handle under the current resource conditions;

[0128] Calculate the basic processing requirements of the particle detection dataset based on the data volume.

[0129] Based on the data distribution characteristics, the degree of dispersion of the particle detection data set in the spatiotemporal index structure is analyzed, and the index access overhead is calculated based on the degree of dispersion.

[0130] Based on the basic processing requirements and the index access overhead, the minimum data processing amount required to complete the visualization processing of the particle detection dataset is obtained.

[0131] After obtaining the resource status parameters of the edge computing nodes and the data characteristics of the particle detection dataset, it is necessary to accurately assess the processing capabilities and data processing requirements of the edge computing nodes in order to provide a scientific basis for task allocation.

[0132] The available computing resources of edge computing nodes are extracted from resource status parameters. This availability specifically reflects the number of CPU cores, current CPU utilization, and available memory capacity. For airborne particle detection scenarios, the extraction process includes reading the system call interfaces provided by the edge computing node's operating system to obtain key indicators such as current CPU idle rate, memory free volume, and processor cache hit rate. Specifically, by querying the system process table, the number of currently running processes and their occupied computing resources are counted, thereby calculating the remaining computing resources available for new task allocation. For multi-core processor architectures, the load of each core is statistically analyzed separately to avoid resource waste caused by some cores being overloaded while others are idle.

[0133] Extracting the availability of storage resources involves two levels: persistent storage space and memory cache space of edge computing nodes. Persistent storage space is primarily used to temporarily store particle detection data awaiting transmission to the cloud, and the remaining space capacity of disks or solid-state storage devices is obtained by querying the file system interface. Memory cache space is used to store the datasets being processed and intermediate computation results, and the number of allocable memory pages is obtained through the system memory management interface. Considering the impact of vibration and temperature on storage devices in airborne environments, it is also necessary to check the health status of the storage devices, excluding storage areas with bad blocks or degraded performance, to ensure the accuracy and reliability of the extracted availability.

[0134] The availability of communication resources reflects the data transmission capacity between edge computing nodes and the cloud computing platform. The extraction process includes measuring current network bandwidth usage, network latency jitter, and packet loss rate, among other communication quality indicators. In airborne scenarios, communication links are established via satellite communication or air-to-ground data links, and link quality is significantly affected by flight altitude, weather conditions, and electromagnetic interference. By continuously monitoring network throughput and round-trip latency within the most recent time window, the effective bandwidth available for data transmission is calculated. Simultaneously, network connection stability is recorded; for communication links with frequent interruptions or drastic latency fluctuations, their availability assessment value is appropriately reduced to reserve a safety margin for task allocation.

[0135] After obtaining the availability of the three types of resources mentioned above, a comprehensive calculation is performed to obtain the overall processing capacity value of the edge computing node. The comprehensive calculation adopts a weighted summation method, and weight coefficients are set for the availability of computing resources, storage resources, and communication resources. The weight coefficients are determined based on the characteristics of particle detection data processing tasks. For computationally intensive visualization rendering tasks, computing resources have a higher weight; for data aggregation tasks that require a large amount of data caching, storage resources have a higher weight; and for tasks that require frequent interaction with the cloud, communication resources have a higher weight. The availability of each resource is multiplied by its corresponding weight, summed, and then normalized to obtain an overall processing capacity value ranging from 0 to 1. The closer the value is to 1, the stronger the processing capacity of the edge computing node.

[0136] Based on the overall processing capacity and historical data processing records, the maximum data processing volume that an edge computing node can handle under the current resource conditions is determined. Historical data processing records include information such as the data volume, task type, actual resource consumption, and task completion time of past data processing tasks. By analyzing historical records, a mapping relationship between the overall processing capacity and the actual data processing volume is established. Using regression analysis, the overall processing capacity of historical tasks is used as the independent variable, and the actual data processing volume is used as the dependent variable to fit a functional relationship between capacity and processing volume. Considering the differences in resource consumption patterns for different types of tasks, mapping models are established separately for data preprocessing tasks, visualization rendering tasks, and data transmission tasks. For the current overall processing capacity, the corresponding mapping function is substituted to calculate the theoretically maximum data volume that the edge computing node can handle. Adjustments are made appropriately based on real-time load conditions to avoid resource overruns leading to system response delays or task failures.

[0137] For the assessment of processing requirements for particle detection datasets, the basic processing requirements are first calculated based on the data volume. The data volume is obtained by counting the number of data records and the number of bytes per record in the particle detection dataset. The basic processing requirements are positively correlated with the data volume; the larger the data volume, the more computational resources and time are required for data parsing, format conversion, and visualization rendering. The formula for calculating the basic processing requirements is determined based on the standard processing overhead per unit data volume, which is obtained through experimental testing of typical particle detection data samples. Considering the differences in structural complexity among different types of particle detection data, different processing overhead coefficients are set for different data types. The total basic processing requirements are obtained by multiplying the data volume of each type of data by its corresponding coefficient and then summing the results.

[0138] The degree of dispersion in data distribution directly affects data access efficiency. This paper analyzes the dispersion of the particle detection dataset within the spatiotemporal index structure based on data distribution characteristics, specifically calculating the distribution span of each data group within the spatiotemporal index tree. If data records are relatively concentrated in the time dimension, with high data density within consecutive time periods, batch continuous reading can be achieved during time index access, resulting in low access overhead. If data records are scattered across multiple discontinuous geographical areas in the spatial dimension, spatial index access requires frequent jumps to different index nodes, significantly increasing access overhead. The degree of dispersion is quantified by calculating the spatial and temporal variances of the data groups; a larger variance value indicates a higher degree of dispersion.

[0139] Index access overhead is calculated based on the degree of dispersion. Index access overhead includes the number of index node traversals, the number of disk reads due to cache misses, and the computational cost of index maintenance operations. For datasets with high dispersion, index traversal requires accessing more intermediate and leaf nodes, leading to a decrease in cache hit rate and more disk I / O operations.

[0140] According to the allocation strategy, the first part of the particle detection dataset is visualized and rendered on the edge computing node to generate an edge-side visualization result, and the second part of the data is transmitted to the cloud computing platform for processing to generate a cloud-based visualization result, including:

[0141] According to the allocation strategy, the first part of the data and the second part of the data are extracted from the particle detection dataset, where the first part of the data corresponds to the edge-side processing range and the second part of the data corresponds to the cloud-side processing range.

[0142] The edge computing node performs spatial mapping on the first part of the data based on the spatiotemporal index structure to determine the coordinate position of each data point in the first part of the data in the visualization space;

[0143] The edge computing node performs graphical rendering on the first part of the data according to the coordinate position to generate the edge-side visualization result;

[0144] The second part of the data is transmitted from the edge computing node to the cloud computing platform via a communication interface;

[0145] The cloud computing platform performs spatial mapping on the second part of the data based on the spatiotemporal index structure to determine the coordinate position of each data point in the second part of the data in the visualization space.

[0146] The cloud computing platform performs graphical rendering on the second part of the data based on the coordinate position to generate the cloud visualization result.

[0147] Based on the established allocation strategy, the particle detection dataset needs to be segmented and processed in a distributed manner. The allocation strategy defines the criteria for dividing the processing scope into edge-side and cloud-side processing. For example, the edge-side processing scope covers real-time data with timestamps within the last 5 minutes and simple data with particle concentrations below a set threshold, while the cloud-side processing scope covers historical data requiring complex calculations and dense data with high particle concentrations requiring detailed analysis. Based on this strategy, a first part and a second part of the data are extracted from the particle detection dataset. The extraction of the first part involves traversing the particle detection dataset and checking the timestamp, spatial coordinates, particle type, and particle concentration of each data record to determine if it meets the edge-side processing conditions. Records that meet the conditions are marked as the first part of the data and stored in the local storage of the edge computing node. The extraction of the second part of the data uses a similar traversal process; records that do not meet the edge-side processing conditions are marked as the second part of the data, which needs to be transmitted to the cloud computing platform for processing.

[0148] For the processing of the first part of the data, spatial mapping operations are performed at the edge computing nodes using the established spatiotemporal index structure. This structure includes a time dimension index and a spatial dimension index. The time dimension index maps timestamps to discrete time points on the time axis, while the spatial dimension index maps longitude, latitude, and altitude to a three-dimensional spatial grid. The spatial mapping process first obtains the longitude, latitude, and altitude information of each record in the first part of the data. Based on preset visualization space range and resolution parameters, the actual geographic coordinates are converted into normalized coordinates in the visualization space. The visualization space uses a Cartesian coordinate system, with the coordinate range typically set to the interval [-1, 1]. During coordinate transformation, longitude is mapped to the x-axis, latitude to the y-axis, and altitude to the z-axis. The mapping process uses linear transformation, with transformation parameters determined based on the geographic range of the data to ensure that the coordinate positions of all data points fall within the effective range of the visualization space. Simultaneously, based on the time information in the spatiotemporal index structure, a time series number is assigned to each data point for subsequent dynamic visualization.

[0149] After spatial mapping, the first part of the data is rendered on the edge computing nodes. The rendering employs a GPU-accelerated rendering pipeline, fully utilizing the graphics processing capabilities of the edge computing nodes. The rendering process first assigns different visual attributes to different particle types. For example, aerosol particles are represented by gray spheres, ice crystal particles by blue cubes, and water droplets by transparent spheres. Visual attributes include shape, color, transparency, and size, which are dynamically set based on the particle's physical properties. Particle density is represented by color depth or dot size; high-density areas use darker colors or larger dots, while low-density areas use lighter colors or smaller dots. During rendering, a vertex buffer is constructed, storing the coordinates and visual attributes of each data point, and then submitting it to the GPU for parallel rendering. The rendering pipeline consists of two stages: vertex shaders and fragment shaders. Vertex shaders handle coordinate transformations and projections, while fragment shaders handle color calculations and lighting effects. To improve rendering efficiency, an instantiation rendering technique is used, rendering particles of the same type in batches to reduce the number of rendering calls. The rendering output is image data in the frame buffer, which is the edge-side visualization result, stored in the edge computing node in the form of image files or video streams.

[0150] The second part of the data needs to be transmitted from the edge computing node to the cloud computing platform via a communication interface. This interface uses a standard network communication protocol and supports TCP or UDP transmission modes. Before transmission, the second part of the data is encapsulated, packaging the data into data packets according to a predetermined data format. Each data packet contains a header and a body. The header records the packet's sequence number, data type, data length, and checksum information, while the body stores the actual particle detection data. To ensure transmission reliability, data compression and error control mechanisms are employed. Data compression uses a lossless compression algorithm to reduce the amount of data transmitted and lower network bandwidth usage. Error control is achieved by adding a cyclic redundancy check (CRC) code to the data packets. The receiving end verifies data integrity using the checksum and requests retransmission of corrupted data packets. During transmission, network status is monitored, and the transmission rate is dynamically adjusted based on network bandwidth and latency to avoid network congestion. When network conditions are good, the transmission rate is increased to speed up data transmission; when network congestion occurs, the transmission rate is reduced to ensure transmission stability.

[0151] After receiving the second part of the data, the cloud computing platform performs a spatial mapping and graphics rendering process similar to that on the edge side. Based on a spatiotemporal index structure, the cloud platform performs spatial mapping on the second part of the data, converting geographic coordinates into coordinate positions in the visualization space. The spatial mapping in the cloud uses the same coordinate system and transformation parameters as on the edge side, ensuring spatial consistency between the visualization results generated on the edge side and in the cloud. The cloud computing platform has stronger computing power and can handle larger and more complex datasets; therefore, it can employ higher-precision interpolation algorithms and finer grid divisions during the spatial mapping process, improving the accuracy and detail of the visualization results.

[0152] Cloud-based graphics rendering also employs GPU acceleration technology, but due to the higher-performance GPU clusters available on cloud computing platforms, it can achieve more complex rendering effects. The rendering process supports volumetric rendering, representing the particle concentration field as three-dimensional volumetric data and generating volumetric rendering images through ray casting algorithms to visually display the distribution characteristics of particles in space. Multiple visualization effects can also be overlaid during rendering, such as isosurface extraction, streamline visualization, and slice display, helping users understand particle detection data from different perspectives. The output of cloud-based graphics rendering is a cloud-based visualization, presented as a high-resolution image or an interactive 3D scene, supporting user operations such as zooming, rotating, and switching viewpoints. After the cloud-based visualization is generated, it is transmitted back to edge computing nodes via the network for fusion processing with the edge-side visualization results.

[0153] A second aspect of this invention provides a cloud-edge collaborative airborne particle detection data visualization and real-time processing analysis system, comprising:

[0154] The data preprocessing unit is used to acquire the raw particle detection data stream collected by the airborne particle detection equipment, and to preprocess the raw particle detection data stream at the edge computing node to obtain verified particle detection data.

[0155] The spatiotemporal indexing unit is used to determine the spatiotemporal index structure at the edge computing node based on the spatiotemporal identifier in the verified particle detection data, and to group and aggregate the verified particle detection data according to the spatiotemporal index structure to obtain a particle detection data set.

[0156] The task allocation unit is used to evaluate the computing resource status of the edge computing node and the data characteristics of the particle detection dataset, and determine the allocation strategy for data processing tasks. The allocation strategy represents the boundary between edge-side processing tasks and cloud-side processing tasks.

[0157] The visualization generation unit is used to perform visualization rendering of the first part of the particle detection data set on the edge computing node to generate edge-side visualization results according to the allocation strategy, and to transmit the second part of the data to the cloud computing platform for processing to generate cloud-based visualization results.

[0158] The result fusion unit is used to fuse the edge-side visualization results with the cloud-based visualization results at the edge computing node to generate a visualization output result.

[0159] A third aspect of the present invention provides an electronic device, comprising:

[0160] processor;

[0161] Memory used to store processor-executable instructions;

[0162] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0163] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0164] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0165] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A cloud-edge collaborative method for real-time visualization, processing, and analysis of airborne particle detection data, characterized in that: include: The raw particle detection data stream collected by the airborne particle detection equipment is acquired, and the raw particle detection data stream is preprocessed at the edge computing node to obtain verified particle detection data; Based on the spatiotemporal identifiers in the verified particle detection data, a spatiotemporal index structure is determined at the edge computing node, and the verified particle detection data is grouped and aggregated according to the spatiotemporal index structure to obtain a particle detection data set; The computing resource status of the edge computing nodes and the data characteristics of the particle detection dataset are evaluated to determine the allocation strategy for data processing tasks. The allocation strategy represents the boundary between edge-side processing tasks and cloud-side processing tasks. According to the allocation strategy, the first part of the particle detection dataset is visualized and rendered at the edge computing node to generate an edge-side visualization result, and the second part of the data is transmitted to the cloud computing platform for processing to generate a cloud-based visualization result; The edge computing node merges the edge-side visualization results with the cloud-based visualization results to generate a visualization output result.

2. The method according to claim 1, characterized in that, The raw particle detection data stream collected by the airborne particle detection equipment is acquired, and the raw particle detection data stream is preprocessed at the edge computing node to obtain verified particle detection data, including: The raw particle detection data stream is obtained from the airborne particle detection device via a data interface. The raw particle detection data stream carries timestamp information and detection location information. The edge computing node performs data integrity checks on the original particle detection data stream, identifies the start and end identifiers of data frames based on data frame structure rules, and determines whether the data frames in the original particle detection data stream are complete. For data frames determined to be incomplete during the data integrity check, the edge computing node generates a data frame repair request based on the timestamp information and the detection location information, and triggers a data retransmission mechanism based on the data frame repair request to obtain the missing data; The data frames determined to be complete in the data integrity check are merged with the missing data obtained through the data retransmission mechanism to obtain a complete data frame sequence; The rationality of the particle physics feature values ​​in the complete data frame sequence is judged, and abnormal data frames that do not conform to the particle detection physics constraints are removed to obtain the verified particle detection data.

3. The method according to claim 1, characterized in that, Based on the spatiotemporal identifiers in the verified particle detection data, a spatiotemporal index structure is determined at the edge computing node, and the verified particle detection data is grouped and aggregated according to the spatiotemporal index structure to obtain a particle detection data set, including: Extract time and spatial location identifiers from the verified particle detection data; The edge computing node determines the granularity of the time dimension based on the time identifier and the granularity of the spatial dimension based on the spatial location identifier, and determines the spatiotemporal index structure based on the granularity of the time dimension and the granularity of the spatial dimension. The verified particle detection data is traversed, and each data point is mapped to the corresponding index unit in the spatiotemporal index structure based on its time identifier and spatial location identifier. For each index unit in the spatiotemporal index structure, the verified particle detection data mapped to the same index unit are grouped to obtain data groups with the same spatiotemporal range identifier; The particle detection data in the data group is aggregated and calculated to obtain the statistical feature information of the data group. The data group and its corresponding spatiotemporal range identifier and statistical feature information are combined to obtain the particle detection data set.

4. The method according to claim 3, characterized in that, The edge computing node determines the granularity of the time dimension based on the time identifier and the granularity of the spatial dimension based on the spatial location identifier. Based on the granularity of the time dimension and the granularity of the spatial dimension, it determines the spatiotemporal index structure, including: The edge computing node calculates the time span of the time marker in the verified particle detection data and counts the data distribution frequency of the verified particle detection data within the time span to obtain the time dimension data density; Based on the time span and the time dimension data density, time granularity selection constraints are determined, and the time dimension partitioning granularity is determined according to the time granularity selection constraints. The edge computing node calculates the spatial coverage area of ​​the spatial location identifier in the verified particle detection data and statistically analyzes the spatial distribution density of the verified particle detection data within the spatial coverage area. Based on the spatial coverage and spatial distribution density, spatial granularity selection constraints are determined, and the granularity of spatial dimension division is determined according to the spatial granularity selection constraints. The time span is uniformly divided based on the time dimension granularity to generate a time dimension index sequence, and the spatial coverage is gridded based on the spatial dimension granularity to generate a spatial dimension index grid. The spatiotemporal index structure is obtained by combining and mapping the time dimension index sequence with the spatial dimension index grid.

5. The method according to claim 1, characterized in that, The computational resource status of the edge computing nodes and the data characteristics of the particle detection dataset are evaluated to determine the data processing task allocation strategy, including: Resource status parameters of the edge computing node are collected, and data volume and distribution characteristics are extracted from the particle detection dataset. The processing capability of the edge computing node is evaluated based on the resource status parameters to determine the maximum data processing volume that the edge computing node can handle. Furthermore, the processing requirements of the particle detection dataset are evaluated based on the data volume and data distribution characteristics to determine the minimum data processing volume required to complete the visualization processing of the particle detection dataset. The maximum data processing capacity that the edge computing node can handle is compared with the minimum data processing capacity required to complete the visualization processing of the particle detection dataset, and the comparison result is obtained. Based on the comparison results, the cloud-edge processing ratio is calculated. The particle detection data set is then divided according to the cloud-edge processing ratio and the data distribution characteristics to generate an allocation strategy for the data processing tasks.

6. The method according to claim 5, characterized in that, The processing capacity of the edge computing node is evaluated based on the resource status parameters to determine the maximum data processing volume that the edge computing node can handle. Furthermore, the processing requirements of the particle detection dataset are evaluated based on the data volume and data distribution characteristics to determine the minimum data processing volume required to complete the visualization processing of the particle detection dataset, including: The available computing resources, storage resources, and communication resources of the edge computing node are extracted from the resource status parameters. The overall processing capacity of the edge computing node is obtained by comprehensively calculating the available computing resources, the available storage resources, and the available communication resources. Based on the comprehensive processing capability value and historical data processing records, determine the maximum data processing volume that the edge computing node can handle under the current resource conditions; Calculate the basic processing requirements of the particle detection dataset based on the data volume. Based on the data distribution characteristics, the degree of dispersion of the particle detection data set in the spatiotemporal index structure is analyzed, and the index access overhead is calculated based on the degree of dispersion. Based on the basic processing requirements and the index access overhead, the minimum data processing amount required to complete the visualization processing of the particle detection dataset is obtained.

7. The method according to claim 1, characterized in that, According to the allocation strategy, the first part of the particle detection dataset is visualized and rendered on the edge computing node to generate an edge-side visualization result, and the second part of the data is transmitted to the cloud computing platform for processing to generate a cloud-based visualization result, including: According to the allocation strategy, the first part of the data and the second part of the data are extracted from the particle detection dataset, where the first part of the data corresponds to the edge-side processing range and the second part of the data corresponds to the cloud-side processing range. The edge computing node performs spatial mapping on the first part of the data based on the spatiotemporal index structure to determine the coordinate position of each data point in the first part of the data in the visualization space; The edge computing node performs graphical rendering on the first part of the data according to the coordinate position to generate the edge-side visualization result; The second part of the data is transmitted from the edge computing node to the cloud computing platform via a communication interface; The cloud computing platform performs spatial mapping on the second part of the data based on the spatiotemporal index structure to determine the coordinate position of each data point in the second part of the data in the visualization space. The cloud computing platform performs graphical rendering on the second part of the data based on the coordinate position to generate the cloud visualization result.

8. A cloud-edge collaborative airborne particle detection data visualization and real-time processing analysis system, used to implement the method as described in any one of claims 1-7, characterized in that, include: The data preprocessing unit is used to acquire the raw particle detection data stream collected by the airborne particle detection equipment, and to preprocess the raw particle detection data stream at the edge computing node to obtain verified particle detection data. The spatiotemporal indexing unit is used to determine the spatiotemporal index structure at the edge computing node based on the spatiotemporal identifier in the verified particle detection data, and to group and aggregate the verified particle detection data according to the spatiotemporal index structure to obtain a particle detection data set. The task allocation unit is used to evaluate the computing resource status of the edge computing node and the data characteristics of the particle detection dataset, and determine the allocation strategy for data processing tasks. The allocation strategy represents the boundary between edge-side processing tasks and cloud-side processing tasks. The visualization generation unit is used to perform visualization rendering of the first part of the particle detection data set on the edge computing node to generate edge-side visualization results according to the allocation strategy, and to transmit the second part of the data to the cloud computing platform for processing to generate cloud-based visualization results. The result fusion unit is used to fuse the edge-side visualization results with the cloud-based visualization results at the edge computing node to generate a visualization output result.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images