Unstructured commodity data synchronization transmission method and system based on edge computing

Abstract

This invention discloses a method and system for synchronous transmission of unstructured commodity data based on edge computing, relating to the field of edge computing technology. Its key technical points are: reading the diffuse map stream and normal map stream contained in the unstructured commodity data to generate surface texture features; inputting the geometric topological features and surface texture features into a preset semantic weight evaluation model; outputting a first weight value and a second weight value; obtaining topological micro-slices; obtaining texture macro-slices; establishing a data synchronous transmission channel between edge computing nodes; detecting the real-time available bandwidth between the current edge computing node and the target edge computing node; dividing the real-time available bandwidth into a first sub-band and a second sub-band according to a preset bandwidth allocation ratio; transmitting the topological micro-slices through the first sub-channel and the texture macro-slices through the second sub-channel. This invention improves the transmission efficiency and bandwidth utilization of unstructured commodity data in an edge computing environment.

Description

Technical Field

[0001] This invention relates to the field of edge computing technology, specifically to a method and system for synchronous transmission of unstructured commodity data based on edge computing. Background Technology

[0002] In scenarios such as new retail supported by edge computing, 3D product display, real-time rendering of cross-platform product models, and multi-node digital shelf collaboration, unstructured product data usually exists in the form of 3D models, containing both geometric topological information for constructing the product's outline and surface texture information for presenting the product's appearance and texture. This type of data is characterized by complex structure, large data volume, and high real-time requirements. It needs to be transmitted synchronously and quickly, stably, and in an orderly manner between multiple edge computing nodes to ensure that the product models rendered by different edge nodes are consistent, smooth, and complete. However, the edge environment has inherent characteristics such as dispersed nodes, limited bandwidth, large network status fluctuations, and scarce transmission resources, which place extremely high demands on the reliability and real-time performance of the synchronous transmission of unstructured product data.

[0003] Existing unstructured data transmission technologies suffer from the following core defects that make them unsuitable for edge computing scenarios: They fail to differentiate the importance of geometric topology and surface texture features, employing a uniform transmission strategy for core topology data and large-volume texture data, resulting in a lack of priority for core data transmission; bandwidth resources between edge computing nodes are limited, and bandwidth is not dynamically allocated according to data characteristics and volume, leading to bandwidth congestion caused by large-volume texture data, resulting in delays and packet loss in core topology data transmission; when network congestion occurs, the transmission channels for topology and texture data cannot be differentiated, making the overall transmission link prone to blockage; data segmentation and transmission weights are not adaptively adjusted based on edge network latency jitter and bandwidth fluctuations, resulting in a rigid transmission strategy; and when data transmission is missing, a high-priority retransmission mechanism for core topology data is not set, easily leading to the inability to render product models correctly due to missing topology data. Summary of the Invention

[0004] To achieve the above objectives, the present invention provides the following technical solution: A method for synchronizing and transmitting unstructured commodity data based on edge computing includes: Acquire unstructured product data; parse the vertex coordinate set and face connection relationship contained in the unstructured product data to construct geometric topological features; read the diffuse map flow and normal map flow contained in the unstructured product data to generate surface texture features; Input geometric topological features and surface texture features into a pre-defined semantic weight evaluation model; Output the first weight value and the second weight value; wherein the first weight value is greater than the second weight value; The unstructured commodity data is segmented at the first scale based on the first weight value to obtain topological micro-slices; The unstructured commodity data is segmented at the second scale based on the second weight value to obtain texture macro slices; the data volume of the texture macro slices is greater than that of the topological micro slices. Establish data synchronization and transmission channels between edge computing nodes; Detect the real-time available bandwidth between the current edge computing node and the target edge computing node; The real-time available bandwidth is divided into a first sub-band and a second sub-band according to a preset bandwidth allocation ratio; wherein the bandwidth of the first sub-band is less than that of the second sub-band; a first sub-channel is established on the first sub-band and a second sub-channel is established on the second sub-band. Topological microslices are transmitted via the first sub-channel, and texture macroslices are transmitted via the second sub-channel.

[0005] Furthermore, geometric topological features and surface texture features are input into a pre-defined semantic weight evaluation model, including: Generate the spatial curvature mean of the vertex coordinate set and the variance of the side lengths of the facet connection relationships; The first eigenvector is obtained based on the mean of spatial curvature and the variance of side length; Extract the pixel contrast of the diffuse map stream to generate a second feature vector; The first fully connected layer and the second fully connected layer are activated respectively within the semantic weight evaluation model based on the first feature vector and the second feature vector.

[0006] Furthermore, topological microslices are obtained, including: Extract the first preset segmentation threshold corresponding to the first weight value; The vertex array in the unstructured product data is truncated according to the first preset segmentation threshold; Add a first check header to the truncated vertex array and encapsulate it as a topological microslice.

[0007] Furthermore, texture macro slices are obtained, including: Extract the second preset segmentation threshold corresponding to the second weight value, wherein the second preset segmentation threshold is greater than the first preset segmentation threshold; Based on the second preset segmentation threshold, adjacent pixel blocks in non-structured product data are combined; A second verification header is injected into the merged adjacent pixel blocks and encapsulated into a texture macro slice; Establish a first sub-channel on the first sub-band and a second sub-channel on the second sub-band.

[0008] Furthermore, the transmission of topological micro-slices via the first sub-channel and texture macro-slices via the second sub-channel includes: Monitor the real-time packet loss rate of the data synchronization transmission channel; Determine if the real-time packet loss rate is greater than the preset congestion threshold; If the congestion threshold is greater than the preset congestion threshold, then the first expansion rate corresponding to the first sub-channel and the second expansion rate corresponding to the second sub-channel are extracted, wherein the first expansion rate is greater than the second expansion rate. Adjust the transmission window size of the first sub-channel according to the first expansion rate; The transmission window size of the second sub-channel is reduced according to the second expansion rate.

[0009] Furthermore, adjusting the transmission window size of the first sub-channel according to the first expansion rate includes: Based on the base sending window size and the first expansion rate, generate the first target window increment; Based on the first target window increment and the current transmission window size, the updated transmission window size of the first sub-channel is generated.

[0010] Furthermore, it also includes: Receive the first and second acknowledgment messages from the target edge computing node; Verify the microslice sequence number carried in the first confirmation message; When a missing microslice sequence number is detected; A high-priority retransmission mechanism for topology microslices is triggered within the first sub-channel.

[0011] Furthermore, a high-priority retransmission mechanism for topology microslices is triggered within the first sub-channel, including: Pause the transmission of texture macro slices within the second sub-channel; Extract the topology micro-slice corresponding to the missing sequence number and place it at the beginning of the sending queue; Resend the extracted topology micro-cut to the target edge computing node.

[0012] Furthermore, after activating the first fully connected layer and the second fully connected layer within the semantic weight evaluation model based on the first feature vector and the second feature vector, the model further includes: Obtain the network latency jitter value and bandwidth fluctuation coefficient of the current edge computing node; Construct a network state feature matrix, and then weight and fuse the network state feature matrix with the first feature vector and the second feature vector; The first and second weight values ​​are recalculated based on the fused feature vectors. The ratio of the first weight value to the second weight value is positively correlated with the network state.

[0013] An edge computing-based unstructured commodity data synchronization and transmission system includes: The unstructured product data acquisition module acquires unstructured product data; parses the vertex coordinate set and facet connection relationship contained in the unstructured product data to construct geometric topological features; and reads the diffuse map flow and normal map flow contained in the unstructured product data to generate surface texture features. The product feature deep analysis module inputs geometric topological features and surface texture features into a preset semantic weight evaluation model; The feature weight intelligent evaluation module outputs a first weight value and a second weight value; wherein, the first weight value is greater than the second weight value. The dual-weighted hierarchical output module performs first-scale segmentation on unstructured commodity data based on the first weight value to obtain topological micro-slices; The high-weight data fine-grained slicing module performs second-scale segmentation on unstructured commodity data based on the second weight value to obtain texture macro slices; wherein, the data volume of texture macro slices is greater than that of topological micro slices; A low-weight data high-efficiency macro slicing module establishes a data synchronization and transmission channel between edge computing nodes; The edge node synchronous transmission link construction module detects the real-time available bandwidth between the current edge computing node and the target edge computing node. The bandwidth dynamic allocation and sub-channel construction module divides the real-time available bandwidth into a first sub-band and a second sub-band according to a preset bandwidth allocation ratio; wherein the bandwidth of the first sub-band is less than the bandwidth of the second sub-band; a first sub-channel is established on the first sub-band and a second sub-channel is established on the second sub-band. The slice data differential synchronous transmission module transmits topological micro-slices through the first sub-channel and texture macro-slices through the second sub-channel.

[0014] This invention provides a method and system for synchronous transmission of unstructured commodity data based on edge computing, which has the following beneficial effects: 1. This invention combines edge network latency jitter and bandwidth fluctuations to adaptively adjust feature weights, allowing data segmentation and transmission strategies to adapt to dynamic edge network environments, improving transmission robustness in complex networks, setting a high-priority retransmission mechanism for missing topological micro-slices, pausing texture transmission to ensure core data retransmission, ensuring the integrity of core geometric data of product models, and improving the success rate of edge node data synchronization.

[0015] 2. This invention distinguishes the transmission priority of geometric topology and surface texture by semantic weight, and divides the core geometric topology into small-volume topology micro-slices and the surface texture into large-volume texture macro-slices, so as to prioritize the transmission and processing efficiency of core topology data and avoid non-core texture data from affecting the basic formation of the product model; 3. This invention divides sub-channels according to bandwidth ratio, using small-bandwidth sub-channels to transmit core topology micro-slices and large-bandwidth sub-channels to transmit texture macro-slices, accurately matching data features with edge bandwidth resources, and significantly improving bandwidth utilization between edge computing nodes; 4. This invention differentiates and adjusts the sub-channel transmission window during network congestion, expanding the transmission window for core topology data and shrinking the transmission window for texture data, prioritizing the stable transmission of core topology data and solving the problem of core data interruption under congestion. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the steps of the edge computing-based unstructured commodity data synchronization and transmission method provided in an exemplary embodiment of this application. Figure 2 This is a schematic diagram of an edge computing-based unstructured commodity data synchronization transmission system provided in an exemplary embodiment of this application. Detailed Implementation

[0017] 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 of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1: Please see Figure 1 This embodiment provides a method for synchronous transmission of unstructured commodity data based on edge computing, including: Step 101: Obtain unstructured product data. The unstructured product data referred to here is the .mesh, .obj, and .stl format files collected by the 3D scanning equipment. Specifically, it contains four data components: vertex coordinate set, facet connectivity, diffuse map flow, and normal map flow. The vertex coordinate set is stored in the form of a three-dimensional floating-point array, with each vertex containing three coordinate values: x, y, and z. The facet connectivity uses a vertex index sequence to define the corresponding topological connections between each triangle facet and vertex.

[0019] Step 102: Analyze the vertex coordinate set and facet connection relationships contained in the unstructured commodity data to construct geometric topological features; specifically, firstly, extract the spatial curvature mean of the vertex coordinate set and the side length variance of the facet connection relationships, and combine these two statistics into the first feature vector; the spatial curvature mean refers to the arithmetic mean of the local curvature values ​​of all vertices in the vertex coordinate set, where the local curvature of each vertex is calculated by the spatial position deviation between it and its adjacent vertices, and Gaussian curvature or average curvature can be used as the calculation benchmark; the side length variance is the variance of the Euclidean distance of the three sides of each triangular facet, reflecting the regularity of the triangular mesh on the model surface.

[0020] Step 103: Read the diffuse map stream and normal map stream contained in the unstructured product data to generate surface texture features; specifically, extract the pixel contrast of the diffuse map stream and quantize the pixel contrast into a second feature vector; the pixel contrast mentioned here refers to the color or brightness difference between adjacent pixels or adjacent regions in the diffuse map, which is used to characterize the richness of texture details; the standard deviation method or the local variance method can be used to calculate the pixel contrast.

[0021] Step 104: Input the geometric topological features and surface texture features into the preset semantic weight evaluation model, and output the first weight value and the second weight value. The semantic weight evaluation model here refers to a two-branch neural network, in which the first branch receives the first feature vector and calculates the topological feature weight through the first fully connected layer, and the second branch receives the second feature vector and calculates the texture feature weight through the second fully connected layer. The outputs of the two branches are normalized by the fusion layer to obtain the first weight value and the second weight value, respectively. The first weight value is greater than the second weight value, which means that the topological features are given a higher semantic weight in the subsequent segmentation task.

[0022] Step 105: Perform first-scale segmentation on the unstructured commodity data according to the first weight value to obtain topological micro-slices; specifically, firstly extract the first preset segmentation threshold corresponding to the first weight value, truncate the vertex array according to the first preset segmentation threshold, and encapsulate the truncated vertex array into a topological micro-slice; the first preset segmentation threshold mentioned here is used to control the truncation ratio of the vertex array, and the value range is determined by the first weight value. The truncation operation retains the prefix part with the highest weight contribution in the vertex array.

[0023] Step 106: Perform second-scale segmentation on the unstructured product data according to the second weight value to obtain texture macro slices. The data volume of the texture macro slices is greater than that of the topological micro slices. Specifically, extract the second preset segmentation threshold corresponding to the second weight value. The second preset segmentation threshold is greater than the first preset segmentation threshold. Based on the second preset segmentation threshold, merge adjacent pixel blocks in the unstructured product data and encapsulate the merged pixel blocks into texture macro slices. The adjacent pixel blocks referred to here are the geographically adjacent block sets after the diffuse map stream is pre-divided into several fixed-size pixel blocks. The merging operation stitches multiple adjacent pixel blocks into larger-sized data units to meet the data volume requirements of the second-scale segmentation.

[0024] Step 107: Establish a data synchronization transmission channel between edge computing nodes. This data synchronization transmission channel refers to the network communication link established between the current edge computing node and the target edge computing node. It can be implemented using a TCP long connection or a bidirectional streaming channel based on the QUIC protocol. This channel is used to carry the subsequent transmission tasks of topology micro-tiles and texture macro-tiles.

[0025] Step 108: Probe the real-time available bandwidth between the current edge computing node and the target edge computing node. Specifically, by sending a continuous sequence of probe packets to the target edge computing node, the real-time available bandwidth is calculated based on the packet reception rate reported by the target node. The probe packet sequence referred to here is a set of fixed-size UDP test packets, and the sending interval and number of packets are determined according to a preset probe strategy.

[0026] Step 109: Divide the real-time available bandwidth into a first sub-band and a second sub-band according to the preset bandwidth allocation ratio. Establish a first sub-channel on the first sub-band and a second sub-channel on the second sub-band. The bandwidth allocation ratio mentioned here is a preset parameter based on the difference in data volume between topological micro-slices and texture macro-slices. Since the data volume of texture macro-slices is usually significantly larger than that of topological micro-slices, the bandwidth allocation ratio of the second sub-band is correspondingly larger than that of the first sub-band. The division of sub-bands is achieved through frequency division multiplexing or time division multiplexing technology, and each sub-band carries the establishment of an independent sub-channel.

[0027] Step 110: Transmit the topology micro-slice through the first sub-channel and the texture macro-slice through the second sub-channel; the transmission process here includes writing the topology micro-slice and the texture macro-slice into the transmission buffer of the corresponding sub-channel respectively, and sending them to the target edge computing node through the first sub-channel and the second sub-channel according to the preset transmission sequence.

[0028] The above technical solution calculates differentiated weights for geometric topological features and surface texture features in unstructured commodity data based on a semantic weight evaluation model, thereby achieving semantic layering of data content. Multi-scale segmentation is performed based on dual weight values, separating and transmitting high-weight topological micro-slices and low-weight texture macro-slices in the frequency domain. This fully utilizes the real-time available bandwidth between edge computing nodes, ensuring priority transmission of topological data while maintaining the integrity of texture data, effectively improving the synchronous transmission efficiency and reconstruction quality of unstructured commodity data in an edge computing environment.

[0029] In some embodiments, geometric topological features and surface texture features are input into a preset semantic weight evaluation model, including: Step 201: Generate the spatial curvature mean of the vertex coordinate set and the variance of the side length of the patch connection relationship; the spatial curvature mean is calculated as follows: for each vertex in the vertex coordinate set, search for its K nearest neighbor vertices in the patch connection relationship, construct a local neighborhood based on the K nearest neighbor vertices and fit a local surface, solve for the Gaussian curvature or average curvature of the local surface at the current vertex as the local curvature value of the vertex, and take the arithmetic mean of the local curvature values ​​of all vertices to obtain the spatial curvature mean.

[0030] Step 202: Based on the mean of spatial curvature and the variance of side length, obtain the first feature vector; specifically, concatenate the mean of spatial curvature and the variance of side length into a two-dimensional feature vector, which serves as the input to the first branch of the semantic weight evaluation model; it can be understood that the first feature vector reflects the geometric complexity of unstructured commodity data, where the mean of spatial curvature represents the degree of surface curvature and the variance of side length represents the uniformity of the grid.

[0031] Step 203: Extract the pixel contrast of the diffuse map stream and generate the second feature vector; specifically, after grayscale conversion, the local standard deviation of the diffuse map stream is calculated, the spatial mean of the local standard deviation is quantized into a pixel contrast scalar, and the scalar is expanded into a second feature vector as the input of the second branch of the semantic weight evaluation model.

[0032] Step 204: Based on the first feature vector and the second feature vector, activate the first fully connected layer and the second fully connected layer respectively within the semantic weight evaluation model; specifically, the neurons in the first fully connected layer of the semantic weight evaluation model receive the components of each dimension of the first feature vector and perform weighted summation and nonlinear activation operations, and the neurons in the second fully connected layer receive the components of each dimension of the second feature vector and perform weighted summation and nonlinear activation operations; the output vectors of the two fully connected layers are input into the fusion layer after dimension alignment, the fusion layer performs weighted summation to obtain the fused feature vector, and finally outputs the first weight value and the second weight value after normalization processing.

[0033] The above technical solution constructs a first feature vector based on the mean of the spatial curvature of the vertex coordinate set and the variance of the side length of the patch connectivity relationship, and generates a second feature vector through pixel contrast quantization. The two feature vectors are then fused after being processed by independent fully connected layers, which avoids mutual interference between topological features and texture features in the feature space and improves the accuracy and discriminativeness of semantic weight evaluation.

[0034] In some embodiments, obtaining a topological microslice includes: Step 301: Extract the first preset segmentation threshold corresponding to the first weight value; the first preset segmentation threshold is calculated by the first weight value through a preset mapping function. The mapping function adjusts the truncation ratio according to the size of the first weight value, so that the larger the first weight value, the higher the truncation ratio, thereby retaining more topological key vertices.

[0035] Step 302: Truncate the vertex array in the unstructured commodity data according to the first preset segmentation threshold; specifically, sort the semantic contribution of all vertices in the vertex array in descending order according to the preset sorting rules, and the truncation operation only retains the vertices located in the prefix interval after sorting. The proportion of the number of vertices in the prefix interval to the total number of original vertices is equal to the first preset segmentation threshold; the semantic contribution mentioned here is determined by the first branch activation value calculated by the vertex in the semantic weight evaluation model.

[0036] Step 303: Add a first verification header to the truncated vertex array and encapsulate it into a topological micro-slice; specifically, the first verification header includes a version number field, a slice type identifier, a total number of original vertices field, a truncated vertex count field, a checksum field, and a timestamp field; the encapsulation process concatenates the first verification header and the truncated vertex array in byte order to form a complete topological micro-slice data unit.

[0037] The above technical solution controls the truncation range of the vertex array with a first preset segmentation threshold, prioritizing the retention of topological key vertices with the highest semantic contribution, thus balancing the data compression rate of topological micro-slices with the integrity of subsequent 3D model reconstruction.

[0038] In some embodiments, obtaining a texture macro slice includes: Step 401: Extract the second preset segmentation threshold corresponding to the second weight value; the second preset segmentation threshold mentioned here is greater than the first preset segmentation threshold. The value of this threshold is determined by the second weight value through a preset mapping function. The amplification ratio of the second preset segmentation threshold matches the expected data volume amplification factor of the texture macro-slice relative to the topology micro-slice. Step 402: Merge adjacent pixel blocks in unstructured product data according to the second preset segmentation threshold; specifically, firstly, the diffuse map flow is divided into several fixed-size pixel blocks, each pixel block having the same length and width in pixels; then, the size of the merging window is determined according to the second preset segmentation threshold, and multiple adjacent pixel blocks within the merging window are stitched together into a large-size pixel block group; the adjacent pixel blocks referred to here are a set of blocks that are geographically adjacent in pixel coordinate space and within the size range of the merging window.

[0039] Step 403: Inject a second verification header into the merged adjacent pixel blocks and encapsulate it into a texture macro slice; specifically, the second verification header includes a version number field, a macro slice type identifier, a number of original pixel blocks, a number of merged pixel blocks, a checksum field, and a timestamp field; the encapsulation process concatenates the second verification header and the merged pixel block group in byte order to form a complete texture macro slice data unit.

[0040] By using the above technical solution, the merging granularity of texture macro slices is controlled by the second preset segmentation threshold. Under the premise of ensuring that the data volume of texture macro slices is greater than that of topological micro slices, efficient organization of texture data is achieved by merging and encapsulating adjacent pixel blocks, which takes into account both transmission efficiency and texture reconstruction quality.

[0041] In some embodiments, transmitting topological micro-slices via a first sub-channel and transmitting texture macro-slices via a second sub-channel includes: Step 501: Monitor the real-time packet loss rate of the data synchronization transmission channel; specifically, the real-time packet loss rate is calculated by counting the difference between the total number of sent packets and the number of received acknowledgment packets within a preset time window, and dividing the difference by the total number of sent packets to obtain the packet loss rate percentage; the total number of sent packets mentioned here includes all data packets sent through the first sub-channel and the second sub-channel, and the number of acknowledgment packets is fed back by the receiving end through acknowledgment messages.

[0042] Step 502: Determine whether the real-time packet loss rate is greater than the preset congestion threshold. The preset congestion threshold is used to determine whether the current network has entered a congested state. An initial value can be preset based on historical network performance data. When the real-time packet loss rate is greater than the preset congestion threshold, the network is determined to have entered a congested state, and the window adjustment process is triggered.

[0043] Step 503: Extract the first expansion rate corresponding to the first sub-channel and the second expansion rate corresponding to the second sub-channel; the first expansion rate and the second expansion rate mentioned here are parameters obtained by looking up the current packet loss rate through a preset expansion rate mapping table, wherein the first expansion rate is greater than the second expansion rate; the mapping table is pre-calibrated based on network simulation or actual measurement data, and the higher the packet loss rate, the greater the difference between the first expansion rate and the second expansion rate.

[0044] Step 504: Adjust the transmission window size of the first sub-channel according to the first expansion rate, and shrink the transmission window size of the second sub-channel according to the second expansion rate; specifically, the transmission window size of the first sub-channel is appropriately expanded under congestion to prioritize the transmission bandwidth of the topology microslice, and the transmission window size of the second sub-channel is shrunk under congestion to reduce bandwidth contention for the first sub-channel.

[0045] By employing the above technical solutions, when the network enters a congested state, a differentiated window adjustment strategy is used to prioritize the transmission bandwidth of topology micro-slices, while shrinking the transmission window of texture macro-slices to reduce network load. This achieves reliable transmission and fair resource allocation of unstructured commodity data in congested network environments.

[0046] In some embodiments, adjusting the transmission window size of the first sub-channel according to a first expansion rate includes: Step 601: Multiply the base transmission window size by the first expansion rate to obtain the first target window increment. The base transmission window size refers to the current transmission window size of the first sub-channel before the congestion state is triggered. This value is preset by the system based on the measured bandwidth and round-trip delay. The specific steps for multiplying the base transmission window size by the first expansion rate to obtain the first target window increment are as follows: Obtain the current transmission window size of the first sub-channel before the congestion state is triggered, and use this size as the base transmission window size. This base transmission window size is preset by the system based on the measured bandwidth and round-trip delay to ensure that the initial transmission window is consistent with the current transmission window size. The network's basic transmission capacity is matched; the first expansion rate obtained in step 503 is extracted. The value of the first expansion rate is greater than 1 and less than or equal to 1.5. The higher the packet loss rate, the closer the value of the first expansion rate is to 1.5, which is used to control the magnitude of window expansion; in the third step, a multiplication operation is performed, multiplying the basic sending window size by the first expansion rate. The result obtained by this multiplication operation is the first target window increment. The magnitude of this increment is positively correlated with the basic sending window size and the current network packet loss rate, which can both prioritize the transmission of topology micro-slices and avoid excessive network bandwidth consumption leading to increased congestion.

[0047] Step 602: Add the first target window increment to the current transmission window size to update the transmission window size of the first sub-channel.

[0048] The above technical solution calculates the window increment based on the product of the first expansion rate and the basic transmission window size, and then adds the increment to the current transmission window, thereby realizing the adaptive window expansion of the first sub-channel under congestion conditions and ensuring the priority transmission of topology microslices under bandwidth-constrained conditions.

[0049] In some embodiments, it also includes: Step 701: Receive the first and second acknowledgment messages from the target edge computing node. The first and second acknowledgment messages are data reception credentials returned by the target edge computing node after successfully receiving and verifying the topology micro-tiles and texture macro-tiles. Each acknowledgment message carries the corresponding micro-tile sequence number or macro-tile sequence number and a reception status identifier.

[0050] Step 702: Verify the micro-slice sequence number carried in the first confirmation message; specifically, after receiving the first confirmation message, the current edge computing node parses the micro-slice sequence number field from the message payload, compares the field with the micro-slice sequence numbers already sent in the sending record, and checks whether there is a missing sequence number.

[0051] Step 703: When a microslice sequence number is detected to be missing, a high-priority retransmission mechanism for the topology microslice is triggered in the first sub-channel. Specifically, when the expected microslice sequence number in the transmission record is not continuous with the sequence number fed back by the receiver's acknowledgment message, it is determined that a microslice loss has occurred, and a high-priority retransmission process is then initiated on the first sub-channel. The high-priority retransmission mentioned here means that the topology microslice corresponding to the missing sequence number is extracted to the beginning of the transmission queue and retransmitted with priority over other data to be transmitted.

[0052] The above technical solution monitors the integrity of data transmission in real time by responding to confirmation messages from the target edge computing node. When a topology micro-slice is detected to be lost, a high-priority retransmission is immediately triggered on the first sub-channel to avoid incomplete 3D model reconstruction due to slice loss, thus effectively ensuring the reliability of data transmission.

[0053] In some embodiments, triggering a high-priority retransmission mechanism for topology microslices within a first sub-channel includes: Step 801: Pause the transmission task of texture macro slices in the second sub-channel; specifically, perform a suspension operation on the transmission task of the second sub-channel, stop transmitting the texture macro slice data units that are currently being transmitted, and enter the waiting queue for texture macro slices that have been allocated but not yet written to the transmission buffer; the purpose of the suspension operation is to release the bandwidth resources of the second sub-channel for the retransmission of topology microslices.

[0054] Step 802: Extract the topology microslice corresponding to the missing sequence number to the first position of the transmission queue; specifically, retrieve the topology microslice data unit corresponding to the missing sequence number from the local transmission buffer, remove it from the original queue position and insert it at the front of the transmission queue to ensure that the topology microslice is transmitted first in the next transmission cycle.

[0055] Step 803: Retransmit the extracted topology micro-slice to the target edge computing node; specifically, rewrite the extracted topology micro-slice into the transmission buffer of the first sub-channel and send it to the target edge computing node according to the transmission timing of the first sub-channel; after receiving the retransmitted topology micro-slice, the target node performs sequence number verification. If the verification passes, it updates the reception status and resumes the transmission task of the second sub-channel.

[0056] By employing the aforementioned technical solution, when a topology micro-slice is lost, the transmission of the texture macro-slice is paused and the topology micro-slice is placed at the head of the transmission queue for priority retransmission. This ensures that the topology data is recovered in the first instance, effectively shortening the impact time of slice loss on 3D model reconstruction.

[0057] In some embodiments, after activating the first fully connected layer and the second fully connected layer within the semantic weight evaluation model based on the first feature vector and the second feature vector, the method further includes: Step 901: Obtain the network latency jitter value and bandwidth fluctuation coefficient of the current edge computing node; the network latency jitter value refers to the standard deviation or peak deviation of the round-trip time over a period of time, which is used to characterize the degree of fluctuation of network latency; the bandwidth fluctuation coefficient refers to the change ratio of real-time available bandwidth to historical average bandwidth within a preset time window, which is used to characterize the stability of network bandwidth.

[0058] Step 902: Construct a network state feature matrix and perform weighted fusion of the network state feature matrix with the first feature vector and the second feature vector. Specifically, the network state feature matrix is ​​represented by a two-dimensional vector composed of network latency jitter value and bandwidth fluctuation coefficient. This vector is dimensionally aligned with the first feature vector and the second feature vector and then fused at the feature level through a weighted fusion layer. The fusion weights are preset with initial values ​​based on the sensitivity of each feature dimension to the network state.

[0059] Step 903: Recalculate the first weight value and the second weight value based on the fused feature vector, so that the ratio of the first weight value to the second weight value increases as the network condition deteriorates; specifically, input the fused feature vector into the re-evaluation branch of the semantic weight evaluation model, and the output of this branch is normalized to obtain the updated first weight value and second weight value; it can be understood that when network latency jitter increases or bandwidth fluctuations increase, the ratio of the first weight value to the second weight value increases, which means that the topology micro-slice is given a higher semantic weight in subsequent segmentation tasks to prioritize transmission quality.

[0060] By incorporating network state characteristics into the semantic weight evaluation process, the weight values ​​are adjusted in real time to adapt to the network state. When network conditions deteriorate, the semantic weight of the topology micro-slice is automatically increased, ensuring that key topology data is given priority transmission under limited bandwidth conditions.

[0061] Example 2: Reference Figure 2 Based on Example 1, this embodiment also provides an edge computing-based unstructured commodity data synchronization and transmission system, including: The unstructured product data acquisition module acquires unstructured product data; parses the vertex coordinate set and facet connection relationship contained in the unstructured product data to construct geometric topological features; and reads the diffuse map flow and normal map flow contained in the unstructured product data to generate surface texture features. The product feature deep analysis module inputs geometric topological features and surface texture features into a preset semantic weight evaluation model; The feature weight intelligent evaluation module outputs a first weight value and a second weight value; wherein, the first weight value is greater than the second weight value. The dual-weighted hierarchical output module performs first-scale segmentation on unstructured commodity data based on the first weight value to obtain topological micro-slices; The high-weight data fine-grained slicing module performs second-scale segmentation on unstructured commodity data based on the second weight value to obtain texture macro slices; wherein, the data volume of texture macro slices is greater than that of topological micro slices; A low-weight data high-efficiency macro slicing module establishes a data synchronization and transmission channel between edge computing nodes; The edge node synchronous transmission link construction module detects the real-time available bandwidth between the current edge computing node and the target edge computing node. The bandwidth dynamic allocation and sub-channel construction module divides the real-time available bandwidth into a first sub-band and a second sub-band according to a preset bandwidth allocation ratio; wherein the bandwidth of the first sub-band is less than the bandwidth of the second sub-band; a first sub-channel is established on the first sub-band and a second sub-channel is established on the second sub-band. The slice data differential synchronous transmission module transmits topological micro-slices through the first sub-channel and texture macro-slices through the second sub-channel.

[0062] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0063] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0064] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for synchronous transmission of unstructured commodity data based on edge computing, characterized in that, include: Obtain unstructured product data; parse the vertex coordinate set and face connection relationship contained in the unstructured product data to construct geometric topological features; Read the diffuse map stream and normal map stream contained in the unstructured product data to generate surface texture features; Input geometric topological features and surface texture features into a pre-defined semantic weight evaluation model; Output the first weight value and the second weight value; wherein the first weight value is greater than the second weight value; The unstructured commodity data is segmented at the first scale based on the first weight value to obtain topological micro-slices; The unstructured commodity data is segmented at the second scale based on the second weight value to obtain texture macro slices; the data volume of the texture macro slices is greater than that of the topological micro slices. Establish data synchronization and transmission channels between edge computing nodes; Detect the real-time available bandwidth between the current edge computing node and the target edge computing node; The real-time available bandwidth is divided into a first sub-band and a second sub-band according to a preset bandwidth allocation ratio; wherein the bandwidth of the first sub-band is less than that of the second sub-band; a first sub-channel is established on the first sub-band and a second sub-channel is established on the second sub-band. Topological microslices are transmitted via the first sub-channel, and texture macroslices are transmitted via the second sub-channel.

2. The method for synchronous transmission of unstructured commodity data based on edge computing according to claim 1, characterized in that, Geometric topological features and surface texture features are input into a predefined semantic weight evaluation model, including: Generate the spatial curvature mean of the vertex coordinate set and the variance of the side lengths of the facet connection relationships; The first eigenvector is obtained based on the mean of spatial curvature and the variance of side length; Extract the pixel contrast of the diffuse map stream to generate a second feature vector; The first fully connected layer and the second fully connected layer are activated respectively within the semantic weight evaluation model based on the first feature vector and the second feature vector.

3. The method for synchronous transmission of unstructured commodity data based on edge computing according to claim 1, characterized in that, The topological microslices obtained include: Extract the first preset segmentation threshold corresponding to the first weight value; The vertex array in the unstructured product data is truncated according to the first preset segmentation threshold; Add a first check header to the truncated vertex array and encapsulate it as a topological microslice.

4. The method for synchronous transmission of unstructured commodity data based on edge computing according to claim 1, characterized in that, Obtain texture macro slices, including: Extract the second preset segmentation threshold corresponding to the second weight value, wherein the second preset segmentation threshold is greater than the first preset segmentation threshold; Based on the second preset segmentation threshold, adjacent pixel blocks in non-structured product data are combined; A second verification header is injected into the merged adjacent pixel blocks and encapsulated into a texture macro slice; Establish a first sub-channel on the first sub-band and a second sub-channel on the second sub-band.

5. The method for synchronous transmission of unstructured commodity data based on edge computing according to claim 1, characterized in that, The topology micro-slice is transmitted through the first sub-channel, and the texture macro-slice is transmitted through the second sub-channel, including: Monitor the real-time packet loss rate of the data synchronization transmission channel; Determine if the real-time packet loss rate is greater than the preset congestion threshold; If the congestion threshold is greater than the preset congestion threshold, then the first expansion rate corresponding to the first sub-channel and the second expansion rate corresponding to the second sub-channel are extracted, wherein the first expansion rate is greater than the second expansion rate. Adjust the transmission window size of the first sub-channel according to the first expansion rate; The transmission window size of the second sub-channel is reduced according to the second expansion rate.

6. The method for synchronous transmission of unstructured commodity data based on edge computing according to claim 5, characterized in that, Adjusting the transmission window size of the first sub-channel according to the first expansion rate includes: Based on the base sending window size and the first expansion rate, generate the first target window increment; Based on the first target window increment and the current transmission window size, the updated transmission window size of the first sub-channel is generated.

7. The method for synchronous transmission of unstructured commodity data based on edge computing according to claim 1, characterized in that, Also includes: Receive the first and second acknowledgment messages from the target edge computing node; Verify the microslice sequence number carried in the first confirmation message; When a missing microslice sequence number is detected; A high-priority retransmission mechanism for topology microslices is triggered within the first sub-channel.

8. The method for synchronous transmission of unstructured commodity data based on edge computing according to claim 7, characterized in that, The high-priority retransmission mechanism for topology microslices is triggered within the first sub-channel, including: Pause the transmission of texture macro slices within the second sub-channel; Extract the topology micro-slice corresponding to the missing sequence number and place it at the beginning of the sending queue; Resend the extracted topology micro-cut to the target edge computing node.

9. The method for synchronous transmission of unstructured commodity data based on edge computing according to claim 2, characterized in that, After activating the first fully connected layer and the second fully connected layer within the semantic weight evaluation model based on the first feature vector and the second feature vector, the model further includes: Obtain the network latency jitter value and bandwidth fluctuation coefficient of the current edge computing node; Construct a network state feature matrix, and then weight and fuse the network state feature matrix with the first feature vector and the second feature vector; The first and second weight values ​​are recalculated based on the fused feature vectors. The ratio of the first weight value to the second weight value is positively correlated with the network state.

10. A synchronous transmission system for unstructured commodity data based on edge computing, characterized in that, The system is used to implement the method for synchronous transmission of unstructured commodity data based on edge computing as described in any one of claims 1-9, the system comprising: The unstructured product data acquisition module acquires unstructured product data; parses the vertex coordinate set and facet connection relationship contained in the unstructured product data to construct geometric topological features; and reads the diffuse map flow and normal map flow contained in the unstructured product data to generate surface texture features. The product feature deep analysis module inputs geometric topological features and surface texture features into a preset semantic weight evaluation model; The feature weight intelligent evaluation module outputs a first weight value and a second weight value; wherein, the first weight value is greater than the second weight value. The dual-weighted hierarchical output module performs first-scale segmentation on unstructured commodity data based on the first weight value to obtain topological micro-slices; The high-weight data fine-grained slicing module performs second-scale segmentation on unstructured commodity data based on the second weight value to obtain texture macro slices; wherein, the data volume of texture macro slices is greater than that of topological micro slices; A low-weight data high-efficiency macro slicing module establishes a data synchronization and transmission channel between edge computing nodes; The edge node synchronous transmission link construction module detects the real-time available bandwidth between the current edge computing node and the target edge computing node. The bandwidth dynamic allocation and sub-channel construction module divides the real-time available bandwidth into a first sub-band and a second sub-band according to a preset bandwidth allocation ratio; wherein the bandwidth of the first sub-band is less than the bandwidth of the second sub-band; a first sub-channel is established on the first sub-band and a second sub-channel is established on the second sub-band. The slice data differential synchronous transmission module transmits topological micro-slices through the first sub-channel and texture macro-slices through the second sub-channel.

Images