Radar data compression and transmission method based on wavelet lifting-DCT (Discrete Cosine Transformation)-mixed entropy coding

Abstract

The invention discloses a radar data compression and transmission method based on wavelet lifting-DCT (Discrete Cosine Transformation)-mixed entropy coding. The method comprises the following steps: processing data into a two-dimensional matrix format; the network condition is monitored in real time, and the network is selected according to the availability and bandwidth characteristics of the network; setting a dynamic compression ratio according to a network environment; performing biorthogonal wavelet transformation on the two-dimensional data in the row direction, performing biorthogonal wavelet transformation on the transformed result in the column direction, and removing redundant information; a block-shaped DCT method is adopted, DCT is further conducted on the low-frequency part and the high-frequency part after wavelet transformation, coefficients are effectively concentrated to the upper left corner of a matrix, energy distribution is concentrated, the compression efficiency is improved, and meanwhile the block size is dynamically determined by monitoring the network state according to the parameters and a dynamic compression ratio setting rule; a mixed entropy coding method is adopted, and lossless compression is carried out by using data statistical characteristics; and transmitting the compressed data.

Description

Technical Field

[0001] This invention relates to the field of rain-measuring radar, and in particular to a radar data compression and transmission method based on wavelet lifting-DCT transform-hybrid entropy coding. Background Technology

[0002] Current rain-measuring radar data processing faces multiple challenges, including large data volumes, heterogeneous networks, limited resources, and extreme scenarios. There is an urgent need to design end-to-end solutions that balance high compression ratios, low complexity, dynamic cross-network adaptation, and error resilience to support all-weather, all-terrain meteorological monitoring operations. During rain-measuring radar data transmission, different communication methods exhibit varying transmission rates. Conventional communication methods such as 4G / 5G, GSM / GPRS, and CDMA networks differ from newer methods like BeiDou short message service in terms of transmission speed. In modern meteorological monitoring systems, radar is a key tool for collecting atmospheric conditions (such as rainfall and wind speed). Radar information, especially extreme weather warnings, needs to be transmitted quickly and accurately. However, radar data is characterized by high resolution and large data volumes, posing challenges to data storage and transmission. Particularly in special application scenarios, such as remote areas or emergency situations, conventional communication methods may be unavailable or limited. This necessitates utilizing multiple communication methods, including BeiDou short message service, to ensure effective data transmission.

[0003] Modern radar technologies (such as dual-polarization radar and phased array radar) generate data with high dynamic range (e.g., reflectivity factors exceeding 60 dB), multiple modes (intensity, velocity, spectral width, and other parameters), and strong spatiotemporal correlation (high similarity of echoes from adjacent regions). This results in an exponential increase in data volume, which conventional transmission bandwidth cannot handle. Faced with massive amounts of radar data, effective compression and transmission are essential to realize the value of radar information. Traditional compression methods are insufficient to meet the demands of cross-network transmission.

[0004] Data transmission needs to be adapted to traditional networks such as 4G / 5G (high bandwidth but limited coverage), GSM / GPRS (wide coverage but low speed), and CDMA (strong anti-interference), as well as new communication methods such as BeiDou short messages (no terrestrial network dependence, but bandwidth is only in the hundreds of bytes). Different networks have significant differences in bandwidth, latency, and reliability, making it difficult for traditional compression methods to meet the requirements of multi-network adaptation. Furthermore, data loss is severe under extreme weather conditions. In extreme weather events such as rainstorms, typhoons, and earthquakes, conventional communication networks may be interrupted, and only BeiDou short messages can be relied upon for early warning information transmission. The single transmission length of a BeiDou short message is limited (generally 120 bytes), and existing radar data compression methods cannot achieve effective transmission under this limitation. It is necessary to solve the problem of high-fidelity compression and priority transmission of critical early warning information under extremely low bandwidth conditions. The bandwidth differences between different communication networks are enormous, and traditional fixed-rate compression strategies cannot dynamically adjust the compression intensity according to channel quality. The lack of "on-demand compression" capability leads to wasted bandwidth resources when network conditions are good, while failing to guarantee the transmission of core data when bandwidth is limited. Summary of the Invention

[0005] To address the problems existing in the prior art, this invention provides a radar data compression and transmission method that can realize a hybrid compression architecture that integrates wavelet lifting-DCT transform-hybrid entropy coding, a dynamic bit rate control mechanism, and a multi-mode communication switching engine.

[0006] The objective of this invention is achieved through the following technical solutions.

[0007] A method for radar data compression and transmission using wavelet lifting-DCT transform-hybrid entropy coding includes the following steps:

[0008] Step S1: Process the input rain radar data into a two-dimensional matrix format;

[0009] Step S2: Monitor the network status in real time and select a suitable network based on network availability and bandwidth characteristics;

[0010] Step S3: Set the dynamic compression ratio according to the network environment, and quantize the DCT coefficients accordingly;

[0011] Step S4: First, perform a biorthogonal wavelet transform on the two-dimensional data in the row direction, and then perform a biorthogonal wavelet transform on the transformed result in the column direction to remove redundant information;

[0012] Step S5: Using the block DCT method, further DCT transformation is performed on the low-frequency and high-frequency parts after wavelet transform, and the coefficients are effectively concentrated in the upper left corner of the matrix to concentrate the energy distribution and improve the compression efficiency. At the same time, the block size is dynamically determined by monitoring the network status and based on these parameters and the dynamic setting compression ratio rules.

[0013] Step S6: Use the hybrid entropy coding method to perform lossless compression by utilizing the statistical characteristics of the data;

[0014] Step S7: Transmit the compressed data.

[0015] Furthermore, step S2 includes an intelligent switching module, which specifically includes:

[0016] (1) Dynamic network awareness: Real-time monitoring of network availability, bandwidth, latency and stability;

[0017] (2) Transmission protocol selection module: Selects the most suitable transmission protocol or mechanism based on the network type;

[0018] (3) Seamless switching: Minimize data interruption and loss when switching communication methods;

[0019] (4) Adaptive compression: Dynamically adjust the compression strategy according to the current communication method;

[0020] (5) Priority management: When Beidou short message is used as a backup link, critical data is transmitted first.

[0021] Furthermore, step S3 specifically includes: implementing dynamic bitrate adjustment based on network parameters; using lossless compression when CQI ≥ 0.8; using lossy compression when 0.5 ≤ CQI < 0.8; and using emergency compression when CQI < 0.5.

[0022] The mapping relationship between quantization step size Q and channel capacity C is as follows:

[0023] ;

[0024] Where α = 0.15, β = 0.08, The channel capacity threshold is set to 0.5. , The minimum and maximum quantization step size is set to 1~100;

[0025] The specific rules are as follows: For high-bandwidth, low-latency 4G / 5G networks, multi-level wavelet decomposition and fine quantization are used to reduce high-frequency information loss; for low-bandwidth, high-latency GSM / GPRS networks, a medium wavelet layer is set, and if the data complexity is high, 8×8 is still used to prevent distortion, and the quantization step size is appropriately increased to 30-50; for medium-bandwidth, medium-latency CDMA networks, standard block partitioning is used to balance computational overhead and compression efficiency, and quantization is moderately set to 20-30 to retain key low-frequency information; BeiDou short messages have extremely low bandwidth, and sampling only retains low-frequency component subbands, using large blocks or performing DCT on the entire data block, setting a larger quantization step size of 50-100 to simplify coding.

[0026] Furthermore, the specific implementation process of biorthogonal wavelet transform to compress radar data in step S4 is as follows:

[0027] (1) Splitting: The original rain radar data is split into two disjoint subsets, and the data is split into even subsets and odd subsets, that is:

[0028] ;

[0029] The original signal is , , ;

[0030] (2) Prediction: Prediction is made by the correlation between even and odd subsets. Generally, even subsets are used to predict odd subsets, and the difference between even and odd subsets reflects the degree of correlation between them. This approximation is called wavelet coefficient, which corresponds to the high-frequency part of the data. The higher the correlation, the smaller the amplitude of the wavelet coefficient. The prediction process is as follows:

[0031] ;

[0032] in, For predicted values, The difference between the actual value and the predicted value.

[0033] (3) Update: The subsets after the first step of splitting may have different overall characteristics from the radar dataset. Updating ensures that the existing subsets maintain consistency with the original radar dataset in terms of overall characteristics. The update process is as follows:

[0034] ;

[0035] Where U is the update operator. for The low-frequency portion;

[0036] (4) Row and column iteration: After completing the above operations in the row direction, repeat steps 1-3 in the column direction of the result matrix to finally obtain the low-frequency sub-band: a large-scale structure for capturing radar echoes; and the high-frequency sub-band: retaining local mutations.

[0037] Furthermore, step S5 specifically includes:

[0038] The DCT transform converts important information in radar data into a set of coefficient matrices concentrated in the upper left corner of the matrix. For the low-frequency region of radar data, the DCT block size is set to 8×8 to balance computational efficiency and detail preservation. For complex regions, 4×4 blocks are used, and for simple regions, 16×16 blocks are used.

[0039] DCT is applied to each row and each column of radar data separately. The specific steps are as follows:

[0040] (1) Apply DCT to each row: Apply DCT to each row to obtain the corresponding one-dimensional transformation coefficients.

[0041] (2) Apply DCT to each column: Treat each column of the coefficient matrix obtained in the previous step as a one-dimensional signal and apply DCT again.

[0042] (3) Finally, the result of this two-dimensional process is a corresponding transformation coefficient matrix, which contains the corresponding frequency information.

[0043] Furthermore, the formula for the two-dimensional DCT forward transform is as follows:

[0044] ;

[0045] Among them, x,y,u,v=0,1,…,N-1, This is the original two-dimensional signal.

[0046] ;

[0047] The formula for the inverse two-dimensional DCT is as follows:

[0048] ;

[0049] Among them, x,y,u,v=0,1,…,N-1, This is the original two-dimensional signal.

[0050] ;

[0051] Furthermore, step S6 specifically includes: for the data after wavelet transform and DCT processing, according to the obtained high-frequency and low-frequency sub-band characteristics, different entropy coding compression is selected. For the high-frequency sub-band with 95% coefficients of 0, run-length coding is used to compress the data, effectively reducing the amount of data. For the low-frequency sub-band, Huffman coding is used to compress the data.

[0052] Furthermore,

[0053] The run-length encoding data compression process is as follows:

[0054] (1) For a two-dimensional matrix of radar data, convert the 2D matrix into a zigzag sequence to make the zero values ​​more continuous;

[0055] (2) Scan the data of each high-frequency subband line by line from left to right and from top to bottom. For each position, check the current element and its subsequent consecutive identical elements.

[0056] (3) Whenever a non-zero value is encountered or the end of the sequence is reached, record the number of consecutive identical values ​​and the value itself. For consecutive zero values, only record the number.

[0057] (4) For each consecutive identical value found, an encoding pair is generated. Finally, all the encoding pairs are arranged in order into a new sequence as the final encoding output.

[0058] The Huffman coding data compression process is as follows:

[0059] (1) Traverse all coefficients in the low-frequency subband LL of the radar data and count the frequency of each coefficient value;

[0060] (2) Construct a frequency table;

[0061] (3) Based on the symbol frequency, use a priority queue to merge low-frequency nodes and generate the optimal binary tree;

[0062] (4) Replace the original data values ​​with the corresponding Huffman codes.

[0063] Furthermore, step S7 specifically includes: transmitting the compressed data; if the transmission is successful, logging is recorded and the cache is cleared; if the transmission fails, attempting to reconnect and switch networks until the maximum number of reconnections is reached; after successfully switching networks, continuing to transmit the compressed data.

[0064] Compared with the prior art, the advantages of this invention are:

[0065] This invention performs two-dimensional format conversion on rain-measuring radar data, and applies wavelet lifting transform and DCT transform to the two-dimensional data. It dynamically adjusts the wavelet transform layer and quantization step size to find a suitable compression ratio and adjust error parameters, taking into account the characteristics of various network types, including conventional communication methods such as 4G / 5G, GSM / GPRS, and CDMA networks, as well as emerging communication methods such as BeiDou short message service. This achieves automatic network switching, lossy compression, lossless compression, and emergency compression in multi-network environments. Based on the compressed data, run-length encoding is used for high-frequency regions, and Huffman coding is used to further compress the data for low-frequency regions, adapting to data transmission across various networks and ensuring stable transmission of early warning information under extreme conditions and rapid data transmission under normal conditions. Attached Figure Description

[0066] Figure 1 This is a flowchart of the wavelet lifting-DCT transform-hybrid entropy coding compression and transmission algorithm.

[0067] Figure 2 Table showing the relationship between compression efficiency and reconstruction quality (Reconstruction Quality (PSNR, SSIM)).

[0068] Figure 3This is a test table for the extreme compression of BeiDou short messages. Detailed Implementation

[0069] To meet the transmission requirements of rain-measuring radar data in multi-network environments, including conventional communication methods such as 4G / 5G, GSM / GPRS, and CDMA networks, as well as emerging communication methods such as BeiDou short message service, this paper studies rain-measuring radar data compression methods and develops an adaptive switching module between network and BeiDou short message communication methods. Radar data exhibits high dynamic range, strong spatiotemporal correlation, and multimodal characteristics. Faced with massive amounts of radar data, effective compression and transmission are essential to realize the value of radar information. Traditional compression methods are insufficient to meet cross-network transmission requirements. This paper proposes a compression architecture based on wavelet lifting-DCT transform-entropy coding and dynamic rate control, combined with a multi-mode communication switching engine, to achieve end-to-end optimization from data compression to channel adaptation. This aims to overcome the challenges of stable transmission of early warning information under extreme conditions and rapid data transmission under normal conditions. The main technical key points are as follows:

[0070] 1. Hybrid compression architecture integrating wavelet lifting, DCT transform, and hybrid entropy coding

[0071] Rainfall radar data typically has high resolution and contains a wealth of detailed information. Parameters such as reflectivity and velocity vary widely in rainfall radar data, with significant differences between flat areas and extreme weather events (such as the core area of ​​a rainstorm). First, biorthogonal wavelet transform is used to decompose the radar data into multiple scales, preserving important details. Then, block-based DCT transform is applied to the processed data to improve energy concentration in local areas. Finally, based on the sparsity characteristics of radar data, zero values ​​are compressed using rodcode programming in the high-frequency domain, and non-zero coefficients are compressed using Huffman coding in the low-frequency domain. This hybrid compression method—wavelet lifting, DCT transform, and hybrid entropy coding—significantly reduces data volume while maintaining data quality, thus adapting to the bandwidth limitations of different communication methods.

[0072] 2. Dynamic bitrate control mechanism

[0073] Due to significant differences in network bandwidth, fixed-rate compression cannot guarantee effective data compression and transmission across all networks. This invention dynamically adjusts parameters such as quantization step size, wavelet decomposition level, and DCT block size based on real-time channel quality index (CQI), network bandwidth, and data priority. This allows for low compression intensity under normal network conditions to ensure data integrity, while switching to high compression mode during extreme weather or when bandwidth is limited, prioritizing the transmission of data from critical areas (such as the core area of ​​a rainstorm). This achieves flexibility, integrity, and efficiency in data compression and transmission across multiple networks.

[0074] 3. Multi-mode communication switching engine

[0075] Design a communication scheduling module that can simultaneously access multiple communication methods (4G / 5G, BeiDou short message service, GSM, etc.). Introduce a communication status awareness mechanism to monitor the bandwidth, latency, packet loss rate, and other indicators of each channel in real time. Based on the task type (e.g., early warning information, routine data) and channel quality, automatically select the optimal communication link and seamlessly switch in case of link failure to ensure reliable transmission of radar data.

[0076] like Figure 1 As shown, the technical solution and specific process of the wavelet lifting-DCT transform-hybrid entropy coding compression and transmission algorithm proposed in this invention are as follows:

[0077] 1. Monitor the network environment and adaptively switch networks.

[0078] To achieve adaptive switching between different communication methods (such as 4G / 5G, GSM / GPRS, CDMA, and BeiDou short message service) and ensure reliable radar data transmission, an intelligent switching module is designed. The following is a layered design and implementation scheme:

[0079] (1) Dynamic network awareness: Real-time monitoring of network availability, bandwidth, latency and stability, etc.;

[0080] (2) Transmission protocol selection module: Selects the most suitable transmission protocol or mechanism based on the network type;

[0081] (3) Seamless switching: Minimize data interruption and loss when switching communication methods;

[0082] (4) Adaptive compression: Dynamically adjust the compression strategy (such as DCT block division and quantization step size) according to the current communication method.

[0083] (5) Priority management: When Beidou short message is used as a backup link, critical data (such as abnormal alarms) is transmitted first.

[0084] 2. Process the input rain radar data into a two-dimensional matrix (N rows × M columns) format.

[0085] 3. Perform a biorthogonal wavelet transform on the two-dimensional data in the row direction first, and then perform a biorthogonal wavelet transform on the transformed result in the column direction to remove redundant information.

[0086] principle:

[0087] Rainfall radar data typically has high resolution and contains a wealth of detailed information. Parameters such as reflectivity and velocity vary widely in rainfall radar data, with significant differences between flat areas and extreme weather events (such as the core of a rainstorm). Bioorthogonal wavelet transform (BOR) is a signal processing method based on wavelet analysis. BOR effectively concentrates the signal energy onto a small number of wavelet coefficients, especially the low-frequency subband. This means that most of the important information (such as the core of extreme weather events) can be well preserved, while the high-frequency components can be compressed or discarded to varying degrees as needed. By adjusting parameters such as the number of decomposition levels and the quantization step size, BOR can provide different levels of compression. Furthermore, its filter design allows for separate analysis and synthesis filters, making it more flexible in processing different types of signals.

[0088] Two-dimensional transformation can be achieved using the separation of variables method, first performing a one-dimensional transformation on the rows of the two-dimensional data, and then performing a one-dimensional transformation on the columns. The lifting steps of bioorthogonal wavelet include decomposition, prediction, and update. For the input two-dimensional rain radar data, a one-dimensional bioorthogonal lifting transformation is first performed on each row to obtain the high and low frequency components in the row direction; then, the same one-dimensional bioorthogonal lifting transformation is performed on each column (based on the result of the row transformation) to obtain the final four subbands: LL (low frequency to low frequency), LH (low frequency to high frequency), HL (high frequency to low frequency), and HH (high frequency to high frequency).

[0089] Specifically, the implementation process of biorthogonal wavelet transform for compressing radar data is as follows:

[0090] (1) Splitting. The original rain radar data is split into two disjoint subsets. Usually, the data is split into an even subset and an odd subset, i.e.:

[0091]

[0092] The original signal is , , .

[0093] (2) Prediction. Prediction is made based on the correlation between even and odd subsets. Generally, even subsets are used to predict odd subsets, and the difference between even and odd subsets reflects the degree of correlation between them. This approximation is called the wavelet coefficient, which corresponds to the high-frequency part of the data. The higher the correlation, the smaller the amplitude of the wavelet coefficient. The prediction process is as follows:

[0094]

[0095] in, For predicted values, This represents the difference between the actual value and the predicted value.

[0096] (3) Update. The subsets after the first step of splitting may have different overall characteristics from the radar dataset. Updating ensures that the existing subsets maintain consistency with the original radar dataset in terms of overall characteristics. The update process is as follows:

[0097]

[0098] Where U is the update operator. for The low-frequency portion.

[0099] (4) Row and column iteration. After performing the above operations in the row direction, repeat steps 1-3 in the column direction of the resulting matrix to finally obtain the low-frequency subband (LL): capturing the large-scale structure of radar echoes (such as the outline of precipitation systems). High-frequency subband (LH / HL / HH): preserving local abrupt changes (such as the gradient at the edge of a storm).

[0100] Biorthogonal wavelets, through multi-resolution analysis, symmetric boundary processing, high vanishing moment characteristics, and energy concentration capabilities, effectively adapt to the strong correlation, local abrupt changes, and multi-scale structure of rain-measuring radar data. Combined with adaptive quantization and entropy coding, they can achieve a balance between priority protection of key areas and efficient compression of overall data.

[0101] 4. A block-based DCT method is adopted, for example, using 8×8 blocks to further perform DCT transformation on the low-frequency and high-frequency components after wavelet transform, and effectively concentrates the coefficients in the upper left corner of the matrix to concentrate energy distribution and improve compression efficiency. At the same time, the block size is dynamically determined by monitoring network status (such as bandwidth, latency, etc.) and dynamically setting compression ratio rules based on these parameters.

[0102] principle:

[0103] Based on the strong spatial correlation and local abrupt change characteristics of rainfall radar data, the two-dimensional discrete cosine transform (DCT) can convert spatial correlation into frequency domain energy concentration (low-frequency coefficients carry >90% of the energy), requiring only a small number of low-frequency coefficients to reconstruct the main features. DCT can effectively concentrate energy, support lossless and lossy compression, and has high computational efficiency. Throughout the compression process, by performing reasonable segmentation, transformation, quantization, and encoding operations on the radar data, the data volume can be significantly reduced without significantly losing key information. Its processing flow combines segmentation strategies, adaptive quantization, and entropy coding, ensuring accuracy in key areas (such as the center of a rainstorm) while drastically reducing the data volume, meeting the transmission requirements of different network environments.

[0104] The DCT transform converts important information from radar data into a set of coefficient matrices mainly concentrated in the upper left corner of the matrix, thus facilitating subsequent operations. For low-frequency regions of radar data, to balance computational efficiency and detail preservation, the DCT block size can be set to 8×8; for complex regions, use 4×4 blocks; and for simple regions (such as stratus clouds), use 16×16 blocks.

[0105] For example, based on an 8×8 pixel block, DCT can be applied to each row and each column of radar data separately. The specific steps are as follows:

[0106] (1) Apply DCT to each row: Apply DCT to each row of the 8×8 block to obtain 8 one-dimensional transformation coefficients.

[0107] (2) Apply DCT to each column: Treat each column of the coefficient matrix obtained in the previous step as a one-dimensional signal and apply DCT again.

[0108] (3) Finally, the result of this two-dimensional process is an 8×8 transformation coefficient matrix, which contains the frequency information of the 8x8 blocks of the image.

[0109] The formula for the two-dimensional DCT forward transform is as follows:

[0110]

[0111] Among them, x,y,u,v=0,1,…,N-1, It is the original two-dimensional signal (such as an 8×8 pixel block).

[0112]

[0113] The formula for the inverse two-dimensional DCT is as follows:

[0114]

[0115] Among them, x,y,u,v=0,1,…,N-1, It is the original two-dimensional signal (such as an 8×8 pixel block).

[0116]

[0117] 5. Set the dynamic compression ratio according to the network environment, and quantize the DCT coefficient accordingly.

[0118] principle:

[0119] To adapt to different network environments (such as 4G / 5G, GSM / GPRS, CDMA, and BeiDou short message service), dynamic rate control is implemented. Different compression parameters can be used for regions or features of varying importance to ensure priority transmission of critical information. Coding parameters (such as wavelet layer number, DCT block size, and quantization step size) are dynamically adjusted based on channel conditions and service requirements to achieve the goals of bandwidth adaptation, controllable quality, and priority transmission of critical information.

[0120] Wavelet transform layer adjustment: In rain radar data compression, the compression ratio can be controlled by adjusting the number of layers in the DB9 / 7 wavelet transform. For networks with limited bandwidth (such as BeiDou short message networks), fewer wavelet decomposition layers can be selected to reduce the retention of high-frequency information, thereby reducing the amount of data.

[0121] Quantization step size adjustment: After DCT transformation, changing the quantization step size can significantly affect the final compression ratio. A larger quantization step size leads to a higher compression ratio but may sacrifice some quality; a smaller quantization step size maintains higher fidelity.

[0122] Dynamic bit rate adjustment is implemented based on current network parameters (Channel Quality Index, etc.). When CQI ≥ 0.8, lossless compression (wavelet + DCT full precision) is used; when 0.5 ≤ CQI < 0.8, lossy compression (dynamic quantization) is used; and when CQI < 0.5, emergency compression (DCT coarse quantization) is used.

[0123] The mapping relationship between quantization step size Q and channel capacity C is as follows:

[0124]

[0125] Where α = 0.15, β = 0.08, Channel capacity threshold (set to 0.5). , This is the minimum and maximum quantization step size (set to 1~100).

[0126] The specific rules are as follows:

[0127] (1) 4G / 5G has the characteristics of high bandwidth and low latency. It uses multi-level wavelet decomposition (8×8) + fine quantization (10-20) to reduce the loss of high frequency information;

[0128] (2) The characteristics of CDMA network are medium bandwidth and medium latency. It adopts standard block (8×8) to balance computing overhead and compression efficiency, and moderately quantizes (20-30) to retain key low frequency information;

[0129] (3) GSM / GPRS is characterized by low bandwidth and high latency. Sampling is used to reduce the number of wavelet decomposition layers. If the data complexity is low, try 16×16 blocks to improve the compression rate; for complex data, still use 8×8 to prevent distortion. Increase the quantization step size (30-50).

[0130] (4) Beidou short messages have extremely low bandwidth (≤256 bytes), only retain low frequency component subbands, use large blocks or perform DCT (16×16) on the whole block of data to maximize energy concentration, greatly increase the quantization step size (50-100), and simplify coding.

[0131] 6. A hybrid entropy coding method is adopted to perform lossless compression by utilizing the statistical characteristics of data.

[0132] Approximately 95% of the coefficients in the high-frequency domain LH / HL / HH after wavelet transform and DCT transform are zero, making them highly suitable for run-length encoding. Run-length encoding can effectively compress consecutive identical values ​​(mainly 0 in this case), thereby significantly reducing the amount of data.

[0133] For data in the low-frequency domain (LL), the optimal entropy coding method, Huffman coding, is selected. Huffman coding allocates variable-length codes according to the probability distribution of the data, so that symbols that occur more frequently are represented by shorter codes, while symbols that occur less frequently are represented by longer codes, thereby effectively reducing the average code length and improving compression efficiency.

[0134] Using hybrid entropy coding organically combines the two types of entropy coding in the low-frequency and high-frequency domains to achieve the best compression effect, which can significantly improve the compression efficiency of radar data while ensuring that important information is not distorted.

[0135] Calculate the sparsity (proportion of zero values) for each subband of radar data and select the optimal encoding method.

[0136] 1. Run-length encoding process:

[0137] (1) For a two-dimensional matrix of radar data, convert the 2D matrix into a zigzag sequence to make the zero values ​​more continuous;

[0138] (2) Scan the data of each high-frequency subband line by line from left to right and from top to bottom. For each position, check the current element and its subsequent consecutive identical elements.

[0139] (3) Whenever a non-zero value (such as reflectivity value) is encountered or the end of the sequence is reached, the number of consecutive identical values ​​and the value itself are recorded. For consecutive zero values ​​(such as strong echo regions), only the number is recorded.

[0140] (4) For each consecutive identical value found, an encoding pair is generated. Finally, all the encoding pairs are arranged in order into a new sequence as the final encoded output.

[0141] 2. Huffman coding process:

[0142] (1) Traverse all coefficients in the low-frequency subband LL of the radar data and count the frequency of each coefficient value;

[0143] (2) Construct a frequency table (e.g., value 256 appears 1000 times, value 187 appears 800 times, ...);

[0144] (3) Based on the symbol frequency, use a priority queue (min-heap) to merge low-frequency nodes and generate the optimal binary tree;

[0145] (4) Replace the original data values ​​with the corresponding Huffman codes.

[0146] Example

[0147] 1. Data Processing

[0148] The one-dimensional rain-measuring radar data is processed into a two-dimensional format with N rows and M columns for subsequent processing.

[0149] 2. Adaptive network switching:

[0150] Real-time monitoring of the current network status, selection of appropriate networks based on network availability, bandwidth and other characteristics, and adaptive switching between 4G / 5G, GSM / GPRS, CDMA and BeiDou short message services ensure reliable transmission of radar data.

[0151] 3. Dynamic compression ratio adjustment

[0152] The compression strategy is dynamically adjusted according to the current communication method. For high-bandwidth, low-latency 4G / 5G networks, multi-level wavelet decomposition (8×8) + fine quantization (10-20) is used to reduce the loss of high-frequency information. For low-bandwidth, high-latency GSM / GPRS networks, the wavelet layer number is set to medium level (16×16). If the data complexity is high, 8×8 is still used to prevent distortion, and the quantization step size is appropriately increased to (30-50). For medium-bandwidth, medium-latency CDMA networks, standard block division (8×8) is used to balance computational overhead and compression efficiency, and moderate quantization (20-30) is used to retain key low-frequency information. Beidou short messages have extremely low bandwidth (≤256 bytes). Sampling only retains low-frequency component subbands. Large blocks are used or DCT is performed on the entire data block (16×16). A larger quantization step size (50-100) is set to simplify the coding.

[0153] 4. Hybrid Entropy Coding

[0154] For the data processed by wavelet transform and DCT, different entropy coding compression methods are selected based on the characteristics of the high-frequency and low-frequency subbands. For the high-frequency subbands where 95% of the coefficients are 0, run-length encoding is used to compress the data, effectively reducing the data volume. For the low-frequency subbands, Huffman coding is used for compression. Using a hybrid entropy coding method can achieve the best compression effect, significantly improving compression efficiency while ensuring the criticality of the data.

[0155] 5. Data transmission

[0156] The compressed data is transmitted. If the transmission is successful, a log is logged and the cache is cleared. If the transmission fails, attempts are made to reconnect and switch networks until the maximum number of reconnections is reached. After successfully switching networks, the compressed data transmission continues.

[0157] This invention uses measured heavy rainfall data from an X-band dual-polarization phased array rain-measuring radar as an example. The data features a 30-second scanning period, 20 layers, and include parameters such as reflectivity factor (Z), radial velocity (V), spectral width (SW), differential reflectivity (ZDR), differential phase shift rate (KDP), and correlation coefficient (CC), forming a multi-dimensional data cube. The data size is approximately 10.2 GB. The data compression result is as follows:

[0158] Experimental platform:

[0159] 1. Hardware platform: Server: AMD EPYC 7763, 256GB RAM, NVIDIA A100 GPU; Embedded terminal: Huawei Atlas 500 smart station (ARM architecture, 4-core CPU / 8GB RAM).

[0160] 2. Comparison benchmark: Fixed parameters, quantization step size of 25, block size of 8×8.

[0161] Dynamic parameter adjustment:

[0162] 1. 4G / 5G: Broadband estimation > 5Mbps, wavelet block setting 8×8, quantization step size setting: 10;

[0163] 2. CDMA: Broadband estimated at 60~100kbps, wavelet block size set to 8×8, quantization step size set to 20;

[0164] 3. GSM / GPRS: Broadband estimated at 30~50kbps, wavelet block size set to 16×16, quantization step size set to 10;

[0165] 4. Beidou short message has extremely low bandwidth: estimated bandwidth ≤ 256 bytes, wavelet block is set to 16×16, quantization step size is set to 80, lossless in key areas + lossy in peripheral areas.

[0166] Experimental results:

[0167] like Figure 2 and 3 As shown, in the BeiDou scenario, by using 4×4 block compression and lossless compression in key areas, the data was successfully compressed to within a 560-byte limit. Lossless encoding was performed by dynamically selecting the core area of ​​the rainstorm (Z>50 dBZ), while lossy compression (QP=40) was used in the outer areas. The positioning error was <1 km, and the intensity error was <2 dBZ.

[0168] By dynamically adjusting the quantization step size and DCT block division, the compression ratio can cover the range of 12x to 35x, adapting to all scenarios from 5G to BeiDou. High accuracy is retained under 4G / 5G (PSNR>38 dB), and key information is prioritized under BeiDou (error <1 km).

Claims

1. A method for radar data compression and transmission using wavelet lifting-DCT transform-hybrid entropy coding, characterized in that, Includes the following steps: Step S1: Process the input rain radar data into a two-dimensional matrix format; Step S2: Monitor the network status in real time and select a suitable network based on network availability and bandwidth characteristics; Step S3: Set the dynamic compression ratio according to the network environment, and quantize the DCT coefficients accordingly; Step S4: First, perform a biorthogonal wavelet transform on the two-dimensional data in the row direction, and then perform a biorthogonal wavelet transform on the transformed result in the column direction to remove redundant information; Step S5: Using the block DCT method, further DCT transformation is performed on the low-frequency and high-frequency parts after wavelet transform, and the coefficients are effectively concentrated in the upper left corner of the matrix to concentrate the energy distribution and improve the compression efficiency. At the same time, the block size is dynamically determined by monitoring the network status and based on these parameters and the dynamic setting compression ratio rules. Step S6: Use the hybrid entropy coding method to perform lossless compression by utilizing the statistical characteristics of the data; Step S7: Transmit the compressed data.

2. The radar data compression and transmission method based on wavelet lifting-DCT transform-hybrid entropy coding according to claim 1, characterized in that, Step S2 includes an intelligent switching module, which specifically includes: Dynamic network awareness: Real-time monitoring of network availability, bandwidth, latency, and stability; Transmission protocol selection module: Selects the most suitable transmission protocol or mechanism based on the network type; Seamless switching: Minimizes data interruption and loss when switching communication methods; Adaptive compression: Dynamically adjusts the compression strategy based on the current communication method; Priority Management: When BeiDou short messages are used as a backup link, critical data is transmitted first.

3. The radar data compression and transmission method based on wavelet lifting-DCT transform-hybrid entropy coding according to claim 1, characterized in that, Step S3 specifically includes: dynamically adjusting the bitrate based on network parameters; using lossless compression when CQI ≥ 0.8; using lossy compression when 0.5 ≤ CQI < 0.8; and using emergency compression when CQI < 0.

5. The mapping relationship between quantization step size Q and channel capacity C is as follows: ； Where α = 0.15, β = 0.08, The channel capacity threshold is set to 0.

5. , The minimum and maximum quantization step size is set to 1~100; The specific rules are as follows: For high-bandwidth, low-latency 4G / 5G networks, multi-level wavelet decomposition and fine quantization are used to reduce high-frequency information loss; for low-bandwidth, high-latency GSM / GPRS networks, a medium wavelet layer is set, and if the data complexity is high, 8×8 is still used to prevent distortion, and the quantization step size is appropriately increased to 30-50; for medium-bandwidth, medium-latency CDMA networks, standard block partitioning is used to balance computational overhead and compression efficiency, and quantization is moderately set to 20-30 to retain key low-frequency information; BeiDou short messages have extremely low bandwidth, and sampling only retains low-frequency component subbands, using large blocks or performing DCT on the entire data block, setting a larger quantization step size of 50-100 to simplify coding.

4. The radar data compression and transmission method based on wavelet lifting-DCT transform-hybrid entropy coding according to claim 1, characterized in that, The specific implementation process of biorthogonal wavelet transform to compress radar data in step S4 is as follows: Splitting: The original rainfall radar data is split into two disjoint subsets, and the data is further split into even subsets and odd subsets, i.e.: ； The original signal is , , ; Prediction: Prediction is made based on the correlation between even and odd subsets, using even subsets to predict odd subsets, and the difference between even and odd subsets reflects the degree of correlation between them; the prediction process is as follows: ； in, For predicted values, The difference between the actual value and the predicted value. Update: The subsets resulting from the first split may have different overall characteristics than the radar dataset. The update process ensures that the existing subsets maintain consistency with the original radar dataset in terms of overall characteristics. The update process is as follows: ； Where U is the update operator. for The low-frequency portion; Row and column iteration: After performing the above operations in the row direction, repeat steps 1-3 in the column direction of the resulting matrix to finally obtain the low-frequency sub-band: a large-scale structure for capturing radar echoes; and the high-frequency sub-band: retaining local mutations.

5. The radar data compression and transmission method based on wavelet lifting-DCT transform-hybrid entropy coding according to claim 1, characterized in that, Step S5 specifically includes: The DCT transform converts important information in radar data into a set of coefficient matrices concentrated in the upper left corner of the matrix. For the low-frequency region of radar data, the DCT block size is set to 8×8 to balance computational efficiency and detail preservation. For complex regions, 4×4 blocks are used, and for simple regions, 16×16 blocks are used. The DCT is applied to each row and each column of radar data separately, and the specific steps are as follows: Apply DCT to each row: Apply DCT to each row to obtain the corresponding one-dimensional transform coefficients; Apply DCT to each column: Treat each column of the coefficient matrix obtained in the previous step as a one-dimensional signal and apply DCT again; Ultimately, the result of this two-dimensional process is a corresponding transformation coefficient matrix, which contains the corresponding frequency information.

6. The radar data compression and transmission method based on wavelet lifting-DCT transform-hybrid entropy coding according to claim 5, characterized in that, The formula for the two-dimensional DCT forward transform is as follows: ； Among them, x,y,u,v=0,1,…,N-1, The original two-dimensional signal; ； The formula for the inverse two-dimensional DCT is as follows: ； Among them, x,y,u,v=0,1,…,N-1, The original two-dimensional signal; 。 7. The radar data compression and transmission method based on wavelet lifting-DCT transform-hybrid entropy coding according to claim 1, characterized in that, Step S6 specifically includes: for the data after wavelet transform and DCT processing, according to the characteristics of the obtained high-frequency and low-frequency sub-bands, different entropy coding compression is selected. For the high-frequency sub-bands with 95% coefficients of 0, run-length coding is used to compress the data, effectively reducing the amount of data. For the low-frequency sub-bands, Huffman coding is used to compress the data.

8. The radar data compression and transmission method based on wavelet lifting-DCT transform-hybrid entropy coding according to claim 7, characterized in that, The run-length encoding data compression process is as follows: For two-dimensional matrix radar data, the 2D matrix is ​​converted into a 1D sequence in a zigzag order to make the zero values ​​more continuous; Scan the data of each high-frequency sub-band line by line from left to right and from top to bottom. For each position, check the current element and its subsequent consecutive identical elements. Whenever a non-zero value is encountered or the end of the sequence is reached, record the number of consecutive identical values ​​and the value itself. For consecutive zero values, only record their count. For each consecutive identical value found, an encoding pair is generated. Finally, all the encoding pairs are arranged in order into a new sequence as the final encoded output. The Huffman coding data compression process is as follows: Traverse all coefficients in the low-frequency subband LL of the radar data and count the frequency of each coefficient value. Construct a frequency table; Based on the symbol frequency, a priority queue is used to merge low-frequency nodes to generate the optimal binary tree; Replace the original data values ​​with the corresponding Huffman codes.

9. The radar data compression and transmission method based on wavelet lifting-DCT transform-hybrid entropy coding according to claim 1, characterized in that, Step S7 specifically includes: transmitting the compressed data; if the transmission is successful, logging is recorded and the cache is cleared; if the transmission fails, attempting to reconnect and switch networks until the maximum number of reconnections is reached; after successfully switching networks, continuing to transmit the compressed data.

Images