Survey data cloud multi-dimensional compression and AI super-resolution image reconstruction system

Abstract

This invention belongs to the field of data processing technology. It discloses a cloud-based multi-dimensional compression and AI super-resolution image reconstruction system for surveying and mapping data. The system includes a surveying and mapping parsing module that receives raw surveying and mapping image data, performs format recognition and metadata extraction, and generates a surveying and mapping feature description vector. Based on the surveying and mapping feature description vector, it constructs a surveying and mapping data index structure. A data compression module, based on the surveying and mapping data index structure, performs multi-dimensional collaborative compression processing on the raw surveying and mapping image data, generating a layered compressed data packet containing basic layer compressed data and enhanced layer compressed data. It then performs decompression sampling on the layered compressed data packet and calculates compression quality indicators. A storage management module stores the basic layer compressed data to a hot data storage node and stores the enhanced layer compressed data hierarchically to warm data storage nodes and cold data storage nodes. While ensuring compression efficiency, this system improves spatial structure preservation, cloud storage utilization, and reconstruction clarity.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and more specifically, to a cloud-based multidimensional compression and AI super-resolution image reconstruction system for surveying and mapping data. Background Technology

[0002] With the development of high-resolution Earth observation, UAV aerial surveying, and mobile surveying technologies, the scale of surveying and mapping imagery data continues to grow, exhibiting characteristics of high precision, multi-source data, and massive data volume. Centralized storage, compression management, and on-demand reconstruction of surveying and mapping images in the cloud have become mainstream, making efficient compression and high-quality reconstruction technologies for cloud-based architectures an important research direction.

[0003] However, existing cloud-based multidimensional compression and AI super-resolution image reconstruction systems for surveying and mapping data still have the following problems:

[0004] Existing image compression methods follow a general coding framework, uniformly quantizing and allocating bitrates based on pixel statistical characteristics. Their compression control mechanisms fail to reflect differences in ground features, treating high- and low-value areas equally. As the compression ratio increases, details in critical areas are easily lost, leading to decreased interpretation and reconstruction accuracy. Furthermore, compression strategies are weakly correlated with data type and do not comprehensively consider spatial hierarchy characteristics, easily resulting in over-compression of high-precision data and excessive bitrate allocation for low-value data, thus becoming disconnected from actual business needs.

[0005] In layered coding, existing technologies focus on the hierarchical division of bitstream structure, but lack semantic connections between layers. When only basic layer data is acquired, key semantic details are difficult to recover, limiting the ability to distribute large-scale mapping images over networks.

[0006] In terms of compression quality evaluation, existing technologies mostly use numerical indicators such as pixel-level errors, which only reflect brightness differences and cannot distinguish the spatial distribution of errors. A large amount of information in surveying and mapping images exists in the form of boundaries, and traditional indicators are difficult to detect structural problems in key areas, making it difficult to ensure the authenticity and interpretability of spatial geometry.

[0007] Furthermore, existing coding control strategies have simple quantization parameter adjustments, lack unified mathematical constraints on bitrate allocation and content structure complexity, resulting in an imbalance in rate-distortion distribution between complex and simple regions, making it difficult to balance compression efficiency and structural integrity. Existing compression quality indicators are only used for internal coding control and are not continuously used in subsequent stages, leading to inconsistent evaluation criteria across stages and a lack of linkage at the system level, making it difficult to achieve cross-stage collaborative optimization.

[0008] In view of this, the present invention proposes a cloud-based multidimensional compression and AI super-resolution image reconstruction system for surveying and mapping data to solve the above problems. Summary of the Invention

[0009] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution: a cloud-based multidimensional compression and AI super-resolution image reconstruction system for surveying and mapping data, comprising: The mapping analysis module receives raw mapping image data, performs format recognition and metadata extraction, and generates mapping feature description vectors; based on the mapping feature description vectors, it constructs a mapping data index structure. The data compression module, based on the surveying and mapping data index structure, performs multi-dimensional collaborative compression processing on the original surveying and mapping image data, generates a layered compressed data packet containing basic layer compressed data and enhanced layer compressed data, performs decompression sampling on the layered compressed data packet and calculates the compression quality index. The storage management module stores basic layer compressed data to hot data storage nodes, stores enhanced layer compressed data hierarchically to warm data storage nodes and cold data storage nodes, and establishes a bidirectional mapping table between hierarchical compressed data packets and mapping data index structures. When the image reconstruction module receives an image reconstruction request, it retrieves the corresponding layered compressed data packet and performs decompression processing to obtain the low-resolution image data to be reconstructed; it then inputs the low-resolution image data into the super-resolution reconstruction network model and outputs the super-resolution reconstructed image. The image output module performs display adaptation processing on the super-resolution reconstructed image according to the compression quality index, and outputs it to the visualization terminal for display in a block stream manner.

[0010] Preferably, the method for generating mapping feature description vectors includes: The system receives raw surveying and mapping image data through a cloud data interface. It then identifies the format of the raw surveying and mapping image data by reading the file header information, band organization method markers, and encoding structure, thereby determining the storage format, encoding method, and band organization form of the raw surveying and mapping image data. Based on the format recognition results, the corresponding parsing process is configured to extract spatial geometric metadata, spectral radiation feature metadata, and data structure quality metadata related to the original surveying and mapping image data, and the extracted metadata is normalized. Spatial geometric metadata includes coordinate reference system information, projection parameters, spatial resolution, image coverage, and row and column size information; spectral radiometric metadata includes the number of bands, band parameters, quantization depth, and grayscale value range information; data structure quality metadata includes data organization method identifier, compression method identifier, invalid value marker, acquisition time information, and sensor source information; The spatial geometric metadata, spectral radiometric metadata, and data structure quality metadata are concatenated and encoded according to a preset feature arrangement order to generate a mapping feature description vector that represents the spatial attributes, spectral attributes, and data structure attributes of the original mapping image.

[0011] Preferably, the method for constructing the mapping data index structure includes: After obtaining the mapping feature description vector, the mapping feature description vector is split into different feature sub-vectors according to the preset feature dimension division rules; Each feature subvector is subjected to feature value interval partitioning, discretization encoding, and hierarchical mapping processing to convert continuous physical parameters into sortable index codes. Specifically, spatial resolution, image coverage, and coordinate reference system information are mapped to corresponding spatial hierarchical codes. The number of bands and band type are mapped to spectral category codes, and the data organization method identifier and data block size are mapped to structure type codes. Invalid value markers and acquisition time information are mapped to quality level codes. Using various index codes as index fields, construct a multidimensional composite index structure in the index database. Each index substructure is associated with a preset image unique identifier. The image unique identifier corresponding to the mapping feature description vector is written into the corresponding node position of each index substructure, thereby forming a mapping data index structure.

[0012] Preferably, the method for generating the layered compressed data packet includes: Based on the spatial level code, spectral category code, structural type code, and quality level code recorded in the mapping data index structure, the compression strategy parameters of the corresponding mapping image are determined. The compression strategy parameters include the granularity of code block division, the band joint or separate coding method, and the target code rate range. The original surveying and mapping image data is divided into blocks to construct image coding units with the same granularity as the spatial index; semantic feature analysis is performed on the surveying and mapping images to calculate the statistical mean and standard deviation of the semantic weight distribution; based on the basic quantization parameters selected according to the index structure, a quantization modulation amplitude coefficient is introduced to perform adaptive modulation on the quantization parameters at the position within the coding unit. Transform processing is performed on the image coding unit to convert the original mapping image data into transform domain coefficients. Then, quantization processing is performed on the transform domain coefficients according to the modulated quantization parameters, so that the transform coefficients at different spatial locations are discretized according to the corresponding quantization step size, forming the quantization coefficient data of the corresponding image coding unit. The modulated quantization parameters corresponding to the quantization coefficient data are compared with the preset quantization parameter threshold, and the quantization coefficient data corresponding to each image coding unit is written into the layer identifier field. The layer identifier field includes the base layer identifier and the enhancement layer identifier. When the modulated quantization parameter is greater than or equal to the preset quantization parameter threshold, it is recorded as the base layer identifier, and the corresponding quantization coefficient data is divided into base layer compressed data. When the modulated quantization parameter is less than the preset quantization parameter threshold, it is marked as an enhancement layer identifier, and the corresponding quantization coefficient data is divided into enhancement layer compressed data; the base layer compressed data and the enhancement layer compressed data are encapsulated according to the preset hierarchical relationship to generate a layered compressed data packet.

[0013] Preferably, the method for performing decompression sampling and calculating compression quality indicators on layered compressed data packets includes: Image coding units are selected from the base layer compressed data and enhancement layer compressed data according to a preset sampling ratio and decompressed and reconstructed to obtain the reconstructed image blocks of the image coding units; the reconstructed image blocks are aligned and compared with the corresponding original mapping image blocks, the reconstruction error variance of each image coding unit is calculated, and the actual coding rate and structural complexity index of the corresponding image coding unit are obtained. A structure-aware rate-distortion function is constructed, and statistics are performed on all image coding units. The structure-aware distortion value of the image coding unit is calculated using the structure-aware rate-distortion function, and the structure-aware distortion value is used as a compression quality index.

[0014] Preferably, the method for storing the enhanced layer compressed data hierarchically to warm data storage nodes and cold data storage nodes includes: After generating the hierarchical compressed data packet, the hierarchical identifier field, the image unique identifier, and the index field associated with the mapping data index structure are written to the basic layer compressed data and the enhanced layer compressed data, respectively. The storage scheduling module reads the hierarchical identifier field. When the hierarchical identifier field is the basic layer identifier, the corresponding compressed data is directly determined as high access priority data and written to the hot data storage node. The data is further classified according to the corresponding semantic weights of the enhanced layer compressed data. If the semantic weight is greater than or equal to the preset semantic weight threshold, the corresponding enhanced layer data is written to the warm data storage node, and the remaining enhanced layer data is written to the cold data storage node.

[0015] Preferably, the method for establishing the bidirectional mapping table includes: After generating the layered compressed data packet, a unique image identifier is constructed for the same original mapping image data, and a layer identifier field and a data packet sub-identifier field are written for the base layer compressed data and each enhancement layer compressed data respectively. The combination of the image unique identifier, the layer identifier field and the data packet sub-identifier field is used as the logical data identifier of the layered compressed data packet. In the mapping data index structure, a hierarchical data reference field is set for the index entries to record the corresponding logical data identifiers; when the hierarchical compressed data packets are written to each storage node, the storage management module records the correspondence between the logical data identifiers and the physical storage location information, forming the first mapping relationship from the logical data identifiers to the physical storage locations; Simultaneously, the primary key of the index entry is associated with the corresponding logical data identifier, forming a second mapping relationship from the index entry to the logical data identifier. The logical data identifier is used as an intermediate association key to achieve bidirectional accessibility from the index entry to the physical storage location via the logical data identifier, and vice versa, forming a bidirectional mapping relationship table between the hierarchical compressed data package and the surveying and mapping data index structure.

[0016] Preferably, the method for obtaining the low-resolution image data to be reconstructed includes: Upon receiving an image reconstruction request, at least one of the following is parsed in the request: image unique identifier, spatial range identifier, time identifier, or resolution requirement identifier, and used as a search condition; based on the search condition, a matching query is performed in the mapping data index structure to obtain the logical data identifier corresponding to the target image; Based on the logical data identifier, the corresponding physical storage node and storage path are found in the bidirectional mapping table, and the basic layer compressed data and the enhancement layer compressed data are retrieved. Based on the resolution requirement identifier in the request, the required data level is determined. When the resolution requirement is less than or equal to the preset resolution requirement threshold, the basic layer compressed data is retrieved first for reconstruction. When the resolution requirement is greater than the preset resolution requirement threshold, the enhancement layer compressed data is retrieved for reconstruction. The corresponding decoding parameter set is called according to the hierarchical identifier field to perform hierarchical decompression of the hierarchical compressed data packet. The basic layer compressed data is then subjected to inverse quantization, inverse transform and prediction reconstruction operations to obtain the basic layer reconstructed image block. Simultaneously, the enhanced layer compressed data is retrieved, and residual information superposition is performed based on the spatial alignment of the reconstructed image blocks in the base layer; finally, low-resolution image data that maintains the consistency of the original image's spatial geometry is formed in the preset reconstruction buffer.

[0017] Preferably, the method for outputting super-resolution reconstructed images includes: Low-resolution image data is input into a preset super-resolution reconstruction network model, and super-resolution reconstructed images are output; the super-resolution reconstruction network model is a convolutional neural network model.

[0018] Preferably, the method for display adaptation processing of the super-resolution reconstructed image includes: After obtaining the super-resolution reconstructed image, the spatial index information of the coding unit generated during the compression stage and stored together with the hierarchical compression data packet is called; based on the spatial index information of the coding unit, the spatial region corresponding to each coding unit in the super-resolution reconstructed image is determined, and each corresponding spatial region is defined as a display block unit; the structure-aware distortion value corresponding to each display block unit during the compression stage is read. The structure-aware distortion value is compared with the preset structure distortion risk threshold. When the compression quality index is less than or equal to the preset structure distortion risk threshold, the display block unit is determined to be a structurally reliable block and regular display adaptation processing is performed. When the perceived structural distortion value is greater than the preset structural distortion risk threshold, the display block unit is determined to be a structural risk block, and compensatory display adaptation processing is performed. After the display adaptation processing is completed, a block stream output queue is constructed with the display block unit as the smallest scheduling granularity, and the adaptation results of each display block unit and the spatial location identifier of the display block unit are encapsulated into a block stream data unit. Based on the determination results of the structurally reliable segmentation or the structurally risky segmentation, different transmission priorities are assigned, so that the structurally risky segmentation is transmitted first; the visualization terminal performs block-level splicing at the corresponding position in the display buffer according to the spatial position identifier in the segmented stream data unit and refreshes the display block by block.

[0019] Compared with the prior art, the present invention has the following beneficial effects: This invention constructs a mapping feature description vector and a multi-dimensional index structure, enabling the compression strategy parameters to be driven by spatial, spectral, structural, and quality level encodings, thereby achieving adaptive matching between the compression mode and the mapping data attributes and enhancing the pertinence and rationality of compression parameter configuration. During the encoding stage, a semantic weight distribution and statistical normalization modulation mechanism is introduced to allow the quantization parameters to change dynamically with the importance of spatial semantics, prioritizing the preservation of key ground features and target boundary information, thereby improving the structure preservation capability and information effectiveness of the compressed image. By adopting a hierarchical partitioning method based on quantization parameter thresholds, the basic layer and the enhancement layer carry data with different information retention levels, realizing the correspondence between the hierarchical structure and the primary and secondary information, and improving the flexibility of data utilization; A structure-aware rate-distortion function is constructed, and a structural complexity index is introduced to participate in distortion modulation. This amplifies the pixel error in structurally complex areas, reduces the impact of low-complexity areas, prioritizes the protection of key structural areas, and reduces the risk of boundary deformation of surveying elements and loss of structural information. By incorporating reconstruction error variance, actual coding bitrate, and structural complexity into the same rate-distortion function model, content-characteristic-driven bitrate regulation is achieved, enabling adaptive bitrate allocation and improving the coordination and stability of compression strategies. Structure-aware distortion is defined as a compression quality indicator and passed to subsequent stages, establishing a unified quality measurement mechanism across the entire process and achieving coordinated quality control across multiple processing stages. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the cloud-based multidimensional compression of surveying data and AI super-resolution image reconstruction system of the present invention; Figure 2 This is a schematic diagram of the cloud-based multidimensional compression of surveying data and AI super-resolution image reconstruction method of the present invention. Detailed Implementation

[0021] 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.

[0022] Example 1 Please see Figure 1 As shown, this embodiment provides a cloud-based multidimensional compression and AI super-resolution image reconstruction system for surveying and mapping data, specifically including the following steps: The mapping analysis module receives raw mapping image data, performs format recognition and metadata extraction, and generates mapping feature description vectors; based on the mapping feature description vectors, it constructs a mapping data index structure. The data compression module, based on the surveying and mapping data index structure, performs multi-dimensional collaborative compression processing on the original surveying and mapping image data, generates a layered compressed data packet containing basic layer compressed data and enhanced layer compressed data, performs decompression sampling on the layered compressed data packet and calculates the compression quality index. The storage management module stores basic layer compressed data to hot data storage nodes, stores enhanced layer compressed data hierarchically to warm data storage nodes and cold data storage nodes, and establishes a bidirectional mapping table between hierarchical compressed data packets and mapping data index structures. When the image reconstruction module receives an image reconstruction request, it retrieves the corresponding layered compressed data packet and performs decompression processing to obtain the low-resolution image data to be reconstructed; it then inputs the low-resolution image data into the super-resolution reconstruction network model and outputs the super-resolution reconstructed image. The image output module performs display adaptation processing on the super-resolution reconstructed image according to the compression quality index, and outputs it to the visualization terminal for display in a block stream manner.

[0023] Methods for generating mapping feature description vectors include: The system receives raw surveying and mapping image data through a cloud data interface. The raw surveying and mapping image data originates from surveying and mapping images acquired by aerial photography, UAV aerial photography, satellite remote sensing, or ground mobile surveying equipment. The system reads the file header information, band organization method markers, and encoding structure of the raw surveying and mapping image data to identify the format, storage format, encoding method, and band organization form of the raw surveying and mapping image data. Based on the format recognition results, the corresponding parsing process is configured to extract spatial geometric metadata, spectral radiation feature metadata, and data structure quality metadata related to the original surveying and mapping image data, and the extracted metadata is normalized. Among them, spatial geometric metadata includes coordinate reference system information, projection parameters, spatial resolution, image coverage, and row and column size information; spectral radiometric metadata includes the number of bands, band parameters, quantization bit depth, and grayscale value range information; data structure quality metadata includes data organization method identifier, compression method identifier, invalid value marker, acquisition time information, and sensor source information; The spatial geometric metadata, spectral radiometric metadata, and data structure quality metadata are concatenated and encoded according to a preset feature arrangement order to generate a mapping feature description vector that represents the spatial attributes, spectral attributes, and data structure attributes of the original mapping image.

[0024] Methods for constructing surveying data index structures include: After obtaining the mapping feature description vector, the mapping feature description vector is split into different feature sub-vectors according to the preset feature dimension division rules. The feature sub-vectors include spatial attribute feature sub-vectors, spectral attribute feature sub-vectors, data structure attribute feature sub-vectors, and data quality attribute feature sub-vectors. Each feature subvector is subjected to feature value interval partitioning, discretization encoding, and hierarchical mapping processing to convert continuous physical parameters into sortable index codes. Specifically, spatial resolution, image coverage, and coordinate reference system information are mapped to corresponding spatial hierarchical codes. The number of bands and band type are mapped to spectral category codes, and the data organization method identifier and data block size are mapped to structure type codes. Invalid value markers and acquisition time information are mapped to quality level codes. Using various index codes as index fields, a multidimensional composite index structure is constructed in the index database. The multidimensional composite index structure includes spatial semantic index substructure, spectral feature index substructure, structural complexity index substructure, and quality level index substructure. Each index substructure is associated with a preset image unique identifier. The image unique identifier corresponding to the mapping feature description vector is written into the corresponding node position of each index substructure, thereby forming a mapping data index structure.

[0025] Methods for generating layered compressed data packets include: Based on the spatial level code, spectral category code, structural type code, and quality level code recorded in the mapping data index structure, the compression strategy parameters of the corresponding mapping image are determined. The compression strategy parameters include the granularity of code block division, the band joint or separate coding method, and the target code rate range. The original surveying and mapping image data is divided into blocks to construct image coding units with the same granularity as the spatial index; semantic feature analysis is performed on the surveying and mapping images to obtain the semantic weight distribution function corresponding to the spatial location, and the statistical mean and standard deviation of the semantic weight distribution are calculated; based on the basic quantization parameters selected according to the index structure, a quantization modulation amplitude coefficient is introduced to perform adaptive modulation on the quantization parameters at the location within the coding unit. Adaptive modulation: ;in, Indicates the spatial location within the image coding unit The final quantization parameters used determine the quantization step size of the transform coefficients at the current position, and are control quantities that directly participate in the encoding. The basic quantization parameter is a global reference quantization parameter selected based on the spatial level code, spectral category code, structural type code, and quality level code recorded in the mapping data index structure. This represents the quantization modulation amplitude coefficient, which is a preset constant with a value range of [value range missing]. Preferred This is used to ensure that the quantization accuracy of semantically important regions is increased by no more than one quantization level relative to the basic quantization parameters, while maintaining coding stability. Indicates spatial location The semantic weight value at that location; The statistical mean of the semantic weight distribution of the entire image is the sum of all... The global reference level is obtained by averaging. The standard deviation represents the distribution of semantic weights and is used to measure the spatial dispersion of semantic weights. Transform processing is performed on the image coding unit to convert the original mapping image data into transform domain coefficients. Then, quantization processing is performed on the transform domain coefficients according to the modulated quantization parameters, so that the transform coefficients at different spatial locations are discretized according to the corresponding quantization step size, forming the quantization coefficient data of the corresponding image coding unit. The modulated quantization parameters corresponding to the quantization coefficient data are compared with the preset quantization parameter threshold, and the quantization coefficient data corresponding to each image coding unit is written into the layer identifier field. The layer identifier field includes the base layer identifier and the enhancement layer identifier. When the modulated quantization parameter is greater than or equal to the preset quantization parameter threshold, it is recorded as the base layer identifier, and the corresponding quantization coefficient data is divided into base layer compressed data. When the modulated quantization parameter is less than the preset quantization parameter threshold, it is marked as an enhancement layer identifier, and the corresponding quantization coefficient data is divided into enhancement layer compressed data; the base layer compressed data and the enhancement layer compressed data are encapsulated according to the preset hierarchical relationship to generate a layered compressed data packet.

[0026] Methods for performing decompression sampling and calculating compression quality metrics on layered compressed data packets include: Image coding units are selected from the base layer compressed data and enhancement layer compressed data according to a preset sampling ratio and decompressed and reconstructed to obtain the reconstructed image blocks of the image coding units; the reconstructed image blocks are aligned and compared with the corresponding original mapping image blocks, the reconstruction error variance of each image coding unit is calculated, and the actual coding bit rate and structural complexity index of the corresponding image coding unit are obtained, wherein the structural complexity index is determined by at least one of texture frequency intensity, edge density or semantic target density. It should be noted that, to obtain texture frequency intensity, the image coding unit is first subjected to frequency domain transformation processing, converting spatial domain pixel data into frequency domain coefficients. The proportion of high-frequency component energy to total energy in the frequency domain is then statistically analyzed, or the average amplitude of the high-frequency coefficients is calculated to reflect the spatial frequency level of image grayscale changes. A higher proportion of high-frequency energy indicates more frequent texture changes and richer details in the region, and its texture frequency intensity index increases accordingly. To ensure comparability between different images, the statistical results can be normalized to obtain... Texture frequency intensity value within the range.

[0027] To obtain edge density, gradient calculation or edge detection processing is first performed on the image coding unit to obtain the edge pixel set; the proportion of edge pixels in a unit area is counted, or the proportion of pixels with gradient magnitude exceeding a preset threshold is counted to characterize the complexity of the geometric structure of the region; the denser the edge distribution, the richer the contour information of the region, and the higher its edge density index; the statistical results are also normalized as a component of structural complexity.

[0028] To obtain semantic target density, a preset semantic analysis model is used to perform target recognition or region segmentation on image coding units to identify semantic object regions such as roads, buildings, water bodies, and facilities; the area ratio of regions identified as semantic targets within a unit coding unit is calculated, or the number density of semantic targets is statistically analyzed, as a semantic target density index; when the distribution of semantic targets is dense, it indicates that the region has high information value, and its complexity index is correspondingly increased.

[0029] A structure-aware rate-distortion function is constructed, and statistics are performed on all image coding units. The structure-aware distortion value of the image coding unit is calculated using the structure-aware rate-distortion function, and the structure-aware distortion value is used as a compression quality index.

[0030] The structure-aware rate-distortion function is: ;in, Indicates the first The structural-aware distortion value of each image coding unit under the current bitrate condition; Indicates the image coding unit index number; Indicates the first The actual coding rate of each image coding unit; Indicates the first The reconstruction error variance of each image coding unit; This represents the exponential decay term in rate distortion theory; The complexity modulation coefficient is a preset weighting parameter used to control the strength of distortion amplification caused by structural complexity. Indicates the first Structural complexity index of an image coding unit; It should be noted that, The data is pre-determined through an offline calibration process and can be hierarchically retrieved based on image scene attributes during system operation. Specifically, during the system deployment phase, a representative set of surveying and mapping images covering different landform types is selected. This set includes at least low-texture area images, medium-texture area images, and high-structural-complexity area images. A complete compression coding experiment is performed on each type of sample under different λ values, with the scanning interval of λ set as follows: For each experimental result, the global reconstruction error index and the structure-sensitive quality index are calculated respectively. The global reconstruction error index is used to reflect the overall distortion level, and the structure-sensitive quality index is used to characterize the fidelity of edge, texture or semantic target regions. Based on a preset weighted evaluation function, the encoding results corresponding to different λ values ​​are comprehensively evaluated, and the λ value that makes the comprehensive evaluation function take the optimal value is selected as the default complexity modulation coefficient. The normalized structural complexity index C takes values ​​within the range of... Under the given conditions, the engineering application range of λ is: The preferred value range is: When λ approaches 0, the influence of structural complexity on distortion assessment weakens, and the system behavior is close to the traditional rate-distortion model. When λ increases, the distortion weight of complex regions is enhanced, and the system will allocate more bitrate resources to high-structure regions.

[0031] Methods for storing enhanced layer compressed data hierarchically to warm data storage nodes and cold data storage nodes include: After generating the layered compressed data packets, the layer identifier field, the image unique identifier, and the index field associated with the surveying and mapping data index structure are written to the base layer compressed data and the enhancement layer compressed data, respectively. The storage scheduling module reads the layer identifier field. When the layer identifier field is the base layer identifier, the corresponding compressed data is directly determined as high access priority data and written to the hot data storage node. The hot data storage node is a storage resource node with high read and write performance and low access latency. The data is further graded based on the corresponding semantic weights of the enhanced layer compressed data. If the semantic weight is greater than or equal to a preset semantic weight threshold, the corresponding enhanced layer data is written to a warm data storage node, and the remaining enhanced layer data is written to a cold data storage node. Warm data storage nodes are medium-performance online storage resource nodes, while cold data storage nodes are high-capacity, low-cost archive storage resource nodes.

[0032] Methods for establishing a bidirectional mapping table include: After generating the layered compressed data packet, a unique image identifier is constructed for the same original mapping image data, and a layer identifier field and a data packet sub-identifier field are written for the base layer compressed data and each enhancement layer compressed data respectively. The combination of the image unique identifier, the layer identifier field and the data packet sub-identifier field is used as the logical data identifier of the layered compressed data packet. In the mapping data index structure, a hierarchical data reference field is set for the index entries to record the corresponding logical data identifiers; when the hierarchical compressed data packets are written to each storage node, the storage management module records the correspondence between the logical data identifiers and the physical storage location information; the physical storage location information includes the storage node number, storage path and data block offset address, forming the first mapping relationship from the logical data identifier to the physical storage location; Simultaneously, the primary key of the index entry is associated with the corresponding logical data identifier, forming a second mapping relationship from the index entry to the logical data identifier. The logical data identifier is used as an intermediate association key to achieve bidirectional accessibility from the index entry to the physical storage location via the logical data identifier, and vice versa, forming a bidirectional mapping relationship table between the hierarchical compressed data package and the surveying and mapping data index structure.

[0033] Methods for obtaining low-resolution image data to be reconstructed include: Upon receiving an image reconstruction request, at least one of the following is parsed in the request: image unique identifier, spatial range identifier, time identifier, or resolution requirement identifier, and used as a search condition; based on the search condition, a matching query is performed in the mapping data index structure to obtain the logical data identifier corresponding to the target image; Based on the logical data identifier, the corresponding physical storage node and storage path are found in the bidirectional mapping table, and the basic layer compressed data and the enhancement layer compressed data are retrieved. Based on the resolution requirement identifier in the request, the required data level is determined. When the resolution requirement is less than or equal to the preset resolution requirement threshold, the basic layer compressed data is retrieved first for reconstruction. When the resolution requirement is greater than the preset resolution requirement threshold, the enhancement layer compressed data is retrieved for reconstruction. The corresponding decoding parameter set is called according to the hierarchical identifier field to perform hierarchical decompression of the hierarchical compressed data packet. The basic layer compressed data is then subjected to inverse quantization, inverse transform and prediction reconstruction operations to obtain the basic layer reconstructed image block. Simultaneously, the enhanced layer compressed data is retrieved, and residual information superposition is performed based on the spatial alignment of the reconstructed image blocks in the base layer; finally, low-resolution image data that maintains the consistency of the original image's spatial geometry is formed in the preset reconstruction buffer.

[0034] It should be noted that a layer identifier field is pre-written into the layered compressed data packet to indicate whether the corresponding data belongs to the base layer or the enhancement layer. Simultaneously, a corresponding decoding parameter set is established for different layers during the encoding stage. The decoding parameter set includes, but is not limited to, inverse quantization step size parameters, transform type identifiers, block partitioning mode parameters, prediction mode parameters, and inter-layer alignment offset parameters. After retrieving the layered compressed data packet, the system first reads the layer identifier field, and based on the identifier result, searches for and loads the corresponding layer's decoding parameter set from the decoding parameter management table. This ensures that subsequent decoding operations maintain consistency with the parameter system used in the encoding stage, thereby guaranteeing the consistency of the decoding results in both the numerical and spatial domains.

[0035] For data identified as the base layer, the system performs layered decompression processing according to the loaded base layer decoding parameter set. Specifically, firstly, the base layer quantization coefficient data is dequantized to restore the discretized quantization values ​​to approximately the original transform domain coefficients; then, an inverse transform operation is performed based on the transform type identifier to restore the transform domain coefficients to spatial domain pixel blocks; based on this, a prediction reconstruction operation is performed in conjunction with the prediction mode parameters to compensate for the prediction residuals within or between blocks, resulting in a base layer reconstructed image block. The base layer reconstructed image block retains the geometric structure, main contours, and low- to mid-frequency texture information of the original image, forming the spatial basis for subsequent enhancement layer detail restoration.

[0036] For data identified as enhancement layers, after completing the reconstruction of the base layer image blocks, the system calls the enhancement layer decoding parameter set to decode the compressed enhancement layer data and obtain the corresponding enhancement layer residual information. The enhancement layer residual information has already established a spatial correspondence with the base layer coding units during the encoding stage. Therefore, during the decoding stage, the enhancement layer residual blocks are spatially aligned with the corresponding base layer reconstructed image blocks using inter-layer alignment offset parameters. After alignment, a residual information overlay operation is performed, superimposing the enhancement layer residual values ​​onto the base layer reconstructed pixel values, thereby restoring high-frequency texture details and edge information, achieving layered progressive reconstruction.

[0037] Finally, the image data, after being fused with the base layer reconstruction results and the enhancement layer residual information, is written into a preset reconstruction buffer. This buffer is used to uniformly store image blocks that have undergone spatial alignment and hierarchical fusion, and stitches them together according to the spatial arrangement order of the original mapping images, thereby forming low-resolution or progressively enhanced resolution image data that maintains the consistency of the original image's spatial geometry. Through the above-mentioned layered decoding and residual overlay mechanism, on-demand information recovery under different reconstruction needs is achieved, while ensuring spatial structure stability and hierarchical decoding consistency.

[0038] Methods for outputting super-resolution reconstructed images include: Low-resolution image data is input into a pre-defined super-resolution reconstruction network model, and super-resolution reconstructed images are output. The training process of the super-resolution reconstruction network model includes dividing the dataset into training, validation and test sets, training the model and evaluating its performance. A super-resolution reconstruction network model is constructed. The sample set is a subset of the dataset, and each sample set includes historical low-resolution image data and the corresponding super-resolution reconstructed image. The super-resolution reconstruction network model includes an input layer, a convolutional layer, and an output layer. The super-resolution reconstruction network model is a convolutional neural network model. The model's input layer is used to input historical low-resolution image data, the convolutional layer is used to extract multi-scale texture features and edge structure features from the low-resolution images, and detail compensation is performed through feature recombination; the output layer is used to output super-resolution reconstructed images. Mean squared error is used as the loss function to measure the error between the model's predicted values ​​and the actual values; the super-resolution reconstruction network model is trained using the training set, and the model parameters are updated using the backpropagation algorithm and gradient descent method to minimize the loss function; the performance of the super-resolution reconstruction network model is evaluated by calculating the coefficient of determination using the validation set, and the hyperparameters of the model are tuned. The Adam optimization algorithm was selected as the optimizer, and the hyperparameters of the model were adjusted until the performance no longer improved or the preset number of iterations was reached. The performance of the model in the prediction task was evaluated through the test set. The trained super-resolution reconstruction network model was used to predict the current low-resolution image data to obtain the super-resolution reconstructed image.

[0039] Methods for display adaptation processing of super-resolution reconstructed images include: After obtaining the super-resolution reconstructed image, the spatial index information of the coding unit generated during the compression stage and stored together with the hierarchical compression data packet is called. The spatial index information of the coding unit records the spatial starting coordinates, width and height of each coding unit in the original image. Based on the spatial index information of coding units, the spatial regions corresponding to each coding unit are determined in the super-resolution reconstructed image, and each corresponding spatial region is defined as a display block unit; the structure-aware distortion value corresponding to each display block unit in the compression stage is read; The structure-aware distortion value is compared with a preset structure distortion risk threshold. When the compression quality index is less than or equal to the preset structure distortion risk threshold, the display block unit is determined to be a structurally reliable block, and regular display adaptation processing is performed. Regular display adaptation processing includes standard brightness mapping, default sharpening intensity adjustment, and regular resolution display mapping. When the perceived structural distortion value is greater than the preset structural distortion risk threshold, the display block is determined to be a structural risk block, and a compensatory display adaptation process is performed. The compensatory display adaptation process includes at least one of the following methods to compensate for structural details: increasing the local high-frequency enhancement weight, increasing the edge contrast modulation coefficient, or calling the detail enhancement operator. After completing the display adaptation process, a block stream output queue is constructed with the display block unit as the smallest scheduling granularity. The adaptation results of each display block unit and the spatial location identifier of the display block unit are encapsulated into a block stream data unit. Based on the determination results of the structurally reliable segmentation or the structurally risky segmentation, different transmission priorities are assigned, so that the structurally risky segmentation is transmitted first; the visualization terminal performs block-level splicing at the corresponding position in the display buffer according to the spatial position identifier in the segmented stream data unit and refreshes the display block by block.

[0040] The preset quantization parameter thresholds are set by staff. By collecting different quantization parameters, the average value of multiple quantization parameters is taken as the preset quantization parameter threshold. Similarly, preset semantic weight thresholds, preset resolution requirement thresholds, and preset structural distortion risk thresholds are set.

[0041] In this embodiment, by constructing a mapping feature description vector and a multi-dimensional index structure, the compression strategy parameters are driven by spatial, spectral, structural, and quality level encoding, achieving adaptive matching between the compression mode and the mapping data attributes, and enhancing the pertinence and rationality of the compression parameter configuration. During the encoding stage, a semantic weight distribution and statistical normalization modulation mechanism is introduced to allow the quantization parameters to change dynamically with the importance of spatial semantics, prioritizing the preservation of key ground features and target boundary information, thereby improving the structure preservation capability and information effectiveness of the compressed image. By adopting a hierarchical partitioning method based on quantization parameter thresholds, the basic layer and the enhancement layer carry data with different information retention levels, realizing the correspondence between the hierarchical structure and the primary and secondary information, and improving the flexibility of data utilization; A structure-aware rate-distortion function is constructed, and a structural complexity index is introduced to participate in distortion modulation. This amplifies the pixel error in structurally complex areas, reduces the impact of low-complexity areas, prioritizes the protection of key structural areas, and reduces the risk of boundary deformation of surveying elements and loss of structural information. By incorporating reconstruction error variance, actual coding bitrate, and structural complexity into the same rate-distortion function model, content-characteristic-driven bitrate regulation is achieved, enabling adaptive bitrate allocation and improving the coordination and stability of compression strategies. Structure-aware distortion is defined as a compression quality indicator and passed to subsequent stages, establishing a unified quality measurement mechanism across the entire process and achieving coordinated quality control across multiple processing stages.

[0042] Example 2 Please see Figure 2 As shown, for parts not described in detail in this embodiment, please refer to the description in Embodiment 1. A method for cloud-based multidimensional compression of surveying data and AI super-resolution image reconstruction is provided, including: S1. Receive raw surveying and mapping image data, perform format recognition and metadata extraction, and generate surveying and mapping feature description vectors; construct a surveying and mapping data index structure based on the surveying and mapping feature description vectors; S2. Based on the mapping data index structure, perform multi-dimensional collaborative compression processing on the original mapping image data to generate a layered compressed data packet containing basic layer compressed data and enhanced layer compressed data. Perform decompression sampling on the layered compressed data packet and calculate the compression quality index. S3. Store the basic layer compressed data to the hot data storage node, store the enhanced layer compressed data in layers to the warm data storage node and the cold data storage node, and establish a bidirectional mapping relationship table between the layered compressed data packets and the mapping data index structure. S4. Upon receiving an image reconstruction request, retrieve the corresponding hierarchical compressed data packet and perform decompression processing to obtain the low-resolution image data to be reconstructed; input the low-resolution image data into the super-resolution reconstruction network model and output the super-resolution reconstructed image. S5. Based on the compression quality index, perform display adaptation processing on the super-resolution reconstructed image and output it to the visualization terminal for display in a block stream manner.

[0043] Since the electronic device described in this embodiment is the one used in implementing the cloud-based multidimensional compression and AI super-resolution image reconstruction system for surveying data in this application embodiment, those skilled in the art can understand the specific implementation methods and various variations of the electronic device in this embodiment based on the cloud-based multidimensional compression and AI super-resolution image reconstruction system for surveying data in this application embodiment. Therefore, how the electronic device implements the method in this application embodiment will not be described in detail here. As long as those skilled in the art implement the electronic device used in the cloud-based multidimensional compression and AI super-resolution image reconstruction system for surveying data in this application embodiment, it falls within the scope of protection of this application.

[0044] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0045] The above description is merely a preferred embodiment of the present invention, and the scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for users of ordinary technical skills, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A cloud-based multidimensional compression and AI super-resolution image reconstruction system for surveying and mapping data, characterized in that: include: The mapping analysis module receives raw mapping image data, performs format recognition and metadata extraction, and generates mapping feature description vectors. Construct a surveying data index structure based on surveying feature description vectors; The data compression module, based on the surveying and mapping data index structure, performs multi-dimensional collaborative compression processing on the original surveying and mapping image data, generates a layered compressed data packet containing basic layer compressed data and enhanced layer compressed data, performs decompression sampling on the layered compressed data packet and calculates the compression quality index. The storage management module stores basic layer compressed data to hot data storage nodes, stores enhanced layer compressed data hierarchically to warm data storage nodes and cold data storage nodes, and establishes a bidirectional mapping table between hierarchical compressed data packets and mapping data index structures. When the image reconstruction module receives an image reconstruction request, it retrieves the corresponding layered compressed data packet and performs decompression processing to obtain the low-resolution image data to be reconstructed; it then inputs the low-resolution image data into the super-resolution reconstruction network model and outputs the super-resolution reconstructed image. The image output module performs display adaptation processing on the super-resolution reconstructed image according to the compression quality index, and outputs it to the visualization terminal for display in a block stream manner.

2. The cloud-based multidimensional compression and AI super-resolution image reconstruction system for surveying and mapping data according to claim 1, characterized in that, The method for generating mapping feature description vectors includes: The system receives raw surveying and mapping image data through a cloud data interface. It then identifies the format of the raw surveying and mapping image data by reading the file header information, band organization method markers, and encoding structure, thereby determining the storage format, encoding method, and band organization form of the raw surveying and mapping image data. Based on the format recognition results, the corresponding parsing process is configured to extract spatial geometric metadata, spectral radiation feature metadata, and data structure quality metadata related to the original surveying and mapping image data, and the extracted metadata is normalized. Spatial geometric metadata includes coordinate reference system information, projection parameters, spatial resolution, image coverage, and row and column size information; spectral radiometric metadata includes the number of bands, band parameters, quantization depth, and grayscale value range information; data structure quality metadata includes data organization method identifier, compression method identifier, invalid value marker, acquisition time information, and sensor source information; The spatial geometric metadata, spectral radiometric metadata, and data structure quality metadata are concatenated and encoded according to a preset feature arrangement order to generate a mapping feature description vector that represents the spatial attributes, spectral attributes, and data structure attributes of the original mapping image.

3. The cloud-based multidimensional compression and AI super-resolution image reconstruction system for surveying and mapping data according to claim 2, characterized in that, The method for constructing the mapping data index structure includes: After obtaining the mapping feature description vector, the mapping feature description vector is split into different feature sub-vectors according to the preset feature dimension division rules; Each feature subvector is subjected to feature value interval partitioning, discretization encoding, and hierarchical mapping processing to convert continuous physical parameters into sortable index codes. Specifically, spatial resolution, image coverage, and coordinate reference system information are mapped to corresponding spatial hierarchical codes. The number of bands and band type are mapped to spectral category codes, and the data organization method identifier and data block size are mapped to structure type codes. Invalid value markers and acquisition time information are mapped to quality level codes. Using various index codes as index fields, construct a multidimensional composite index structure in the index database. Each index substructure is associated with a preset image unique identifier. The image unique identifier corresponding to the mapping feature description vector is written into the corresponding node position of each index substructure, thereby forming a mapping data index structure.

4. The cloud-based multidimensional compression and AI super-resolution image reconstruction system for surveying and mapping data according to claim 3, characterized in that, The method for generating the layered compressed data packet includes: Based on the spatial level code, spectral category code, structural type code, and quality level code recorded in the mapping data index structure, the compression strategy parameters of the corresponding mapping image are determined. The compression strategy parameters include the granularity of code block division, the band joint or separate coding method, and the target code rate range. The original surveying and mapping image data is divided into blocks to construct image coding units with the same granularity as the spatial index; semantic feature analysis is performed on the surveying and mapping images to calculate the statistical mean and standard deviation of the semantic weight distribution; based on the basic quantization parameters selected according to the index structure, a quantization modulation amplitude coefficient is introduced to perform adaptive modulation on the quantization parameters at the position within the coding unit. Transform processing is performed on the image coding unit to convert the original mapping image data into transform domain coefficients. Then, quantization processing is performed on the transform domain coefficients according to the modulated quantization parameters, so that the transform coefficients at different spatial locations are discretized according to the corresponding quantization step size, forming the quantization coefficient data of the corresponding image coding unit. The modulated quantization parameters corresponding to the quantization coefficient data are compared with the preset quantization parameter threshold, and the quantization coefficient data corresponding to each image coding unit is written into the layer identifier field. The layer identifier field includes the base layer identifier and the enhancement layer identifier. When the modulated quantization parameter is greater than or equal to the preset quantization parameter threshold, it is recorded as the base layer identifier, and the corresponding quantization coefficient data is divided into base layer compressed data. When the modulated quantization parameter is less than the preset quantization parameter threshold, it is marked as an enhancement layer identifier, and the corresponding quantization coefficient data is divided into enhancement layer compressed data; the base layer compressed data and the enhancement layer compressed data are encapsulated according to the preset hierarchical relationship to generate a layered compressed data packet.

5. The cloud-based multidimensional compression and AI super-resolution image reconstruction system for surveying and mapping data according to claim 4, characterized in that, The method for performing decompression sampling and calculating compression quality indicators on layered compressed data packets includes: Image coding units are selected from the base layer compressed data and enhancement layer compressed data according to a preset sampling ratio and decompressed and reconstructed to obtain the reconstructed image blocks of the image coding units; the reconstructed image blocks are aligned and compared with the corresponding original mapping image blocks, the reconstruction error variance of each image coding unit is calculated, and the actual coding rate and structural complexity index of the corresponding image coding unit are obtained. A structure-aware rate-distortion function is constructed, and statistics are performed on all image coding units. The structure-aware distortion value of the image coding unit is calculated using the structure-aware rate-distortion function, and the structure-aware distortion value is used as a compression quality index.

6. The cloud-based multidimensional compression and AI super-resolution image reconstruction system for surveying and mapping data according to claim 5, characterized in that, The method for storing the enhanced layer compressed data hierarchically to warm data storage nodes and cold data storage nodes includes: After generating the hierarchical compressed data packet, the hierarchical identifier field, the image unique identifier, and the index field associated with the mapping data index structure are written to the basic layer compressed data and the enhanced layer compressed data, respectively. The storage scheduling module reads the hierarchical identifier field. When the hierarchical identifier field is the basic layer identifier, the corresponding compressed data is directly determined as high access priority data and written to the hot data storage node. The data is further classified according to the corresponding semantic weights of the enhanced layer compressed data. If the semantic weight is greater than or equal to the preset semantic weight threshold, the corresponding enhanced layer data is written to the warm data storage node, and the remaining enhanced layer data is written to the cold data storage node.

7. The cloud-based multidimensional compression and AI super-resolution image reconstruction system for surveying and mapping data according to claim 6, characterized in that, The method for establishing the bidirectional mapping table includes: After generating the layered compressed data packet, a unique image identifier is constructed for the same original mapping image data, and a layer identifier field and a data packet sub-identifier field are written for the base layer compressed data and each enhancement layer compressed data respectively. The combination of the image unique identifier, the layer identifier field and the data packet sub-identifier field is used as the logical data identifier of the layered compressed data packet. In the mapping data index structure, a hierarchical data reference field is set for the index entries to record the corresponding logical data identifiers; when the hierarchical compressed data packets are written to each storage node, the storage management module records the correspondence between the logical data identifiers and the physical storage location information, forming the first mapping relationship from the logical data identifiers to the physical storage locations; Simultaneously, the primary key of the index entry is associated with the corresponding logical data identifier, forming a second mapping relationship from the index entry to the logical data identifier. The logical data identifier is used as an intermediate association key to achieve bidirectional accessibility from the index entry to the physical storage location via the logical data identifier, and vice versa, forming a bidirectional mapping relationship table between the hierarchical compressed data package and the surveying and mapping data index structure.

8. The cloud-based multidimensional compression and AI super-resolution image reconstruction system for surveying and mapping data according to claim 7, characterized in that, The method for obtaining the low-resolution image data to be reconstructed includes: Upon receiving an image reconstruction request, at least one of the following is parsed in the request: image unique identifier, spatial range identifier, time identifier, or resolution requirement identifier, and used as a search condition; based on the search condition, a matching query is performed in the mapping data index structure to obtain the logical data identifier corresponding to the target image; Based on the logical data identifier, the corresponding physical storage node and storage path are found in the bidirectional mapping table, and the basic layer compressed data and the enhancement layer compressed data are retrieved. Based on the resolution requirement identifier in the request, the required data level is determined. When the resolution requirement is less than or equal to the preset resolution requirement threshold, the basic layer compressed data is retrieved first for reconstruction. When the resolution requirement is greater than the preset resolution requirement threshold, the enhancement layer compressed data is retrieved for reconstruction. The corresponding decoding parameter set is called according to the hierarchical identifier field to perform hierarchical decompression of the hierarchical compressed data packet. The basic layer compressed data is then subjected to inverse quantization, inverse transform and prediction reconstruction operations to obtain the basic layer reconstructed image block. Simultaneously, the enhanced layer compressed data is retrieved, and residual information superposition is performed based on the spatial alignment of the reconstructed image blocks in the base layer; finally, low-resolution image data that maintains the consistency of the original image's spatial geometry is formed in the preset reconstruction buffer.

9. The cloud-based multidimensional compression and AI super-resolution image reconstruction system for surveying and mapping data according to claim 8, characterized in that, The method for outputting super-resolution reconstructed images includes: Low-resolution image data is input into a preset super-resolution reconstruction network model, and super-resolution reconstructed images are output; the super-resolution reconstruction network model is a convolutional neural network model.

10. The cloud-based multidimensional compression and AI super-resolution image reconstruction system for surveying and mapping data according to claim 9, characterized in that, The method for display adaptation processing of super-resolution reconstructed images includes: After obtaining the super-resolution reconstructed image, the spatial index information of the coding unit generated during the compression stage and stored together with the hierarchical compression data packet is called; based on the spatial index information of the coding unit, the spatial region corresponding to each coding unit in the super-resolution reconstructed image is determined, and each corresponding spatial region is defined as a display block unit; the structure-aware distortion value corresponding to each display block unit during the compression stage is read. The structure-aware distortion value is compared with the preset structure distortion risk threshold. When the compression quality index is less than or equal to the preset structure distortion risk threshold, the display block unit is determined to be a structurally reliable block and regular display adaptation processing is performed. When the perceived structural distortion value is greater than the preset structural distortion risk threshold, the display block unit is determined to be a structural risk block, and compensatory display adaptation processing is performed. After the display adaptation processing is completed, a block stream output queue is constructed with the display block unit as the smallest scheduling granularity, and the adaptation results of each display block unit and the spatial location identifier of the display block unit are encapsulated into a block stream data unit. Based on the determination results of the structurally reliable segmentation or the structurally risky segmentation, different transmission priorities are assigned, so that the structurally risky segmentation is transmitted first; the visualization terminal performs block-level splicing at the corresponding position in the display buffer according to the spatial position identifier in the segmented stream data unit and refreshes the display block by block.

Images