Data encryption transmission method and system for remote monitoring equipment

Abstract

The invention relates to the technical field of remote monitoring data transmission, and discloses a data encryption transmission method and system for remote monitoring equipment. The method comprises the following steps: acquiring an original data stream, and extracting boundary coordinates of an image key region; separating a core area from a background area and determining the priority of the core area; encrypting the core region and compressing the background region to generate a differentiated data segment; dynamically adjusting the transmission proportion to adapt the bandwidth and the computing power; an integrity verification value is generated in transmission to verify tamper-proofing; and fusing the core and the background region to reconstruct a picture. According to the method, core data security and transmission efficiency are considered, a remote security monitoring result is output, and transmission reliability and practicability are improved.

Description

Technical Field

[0001] This invention relates to the field of remote monitoring data transmission technology, and in particular to a data encryption transmission method and system for remote monitoring equipment. Background Technology

[0002] In existing technologies, remote monitoring equipment has been widely used in various scenarios such as public safety, industrial production, and park management, resulting in an explosive growth in the generated monitoring data streams. Current mainstream data transmission solutions mostly adopt a uniform processing model, failing to differentiate between core sensitive data containing people and objects and background redundant data such as open environments, applying the same level of encryption or compression strategies to both. However, in practical applications, core sensitive data lacks targeted, high-strength protection, making it difficult to effectively guarantee data security. On the other hand, background redundant data consumes a large amount of transmission bandwidth and terminal computing resources, easily leading to transmission stutters and delays, affecting the real-time performance of monitoring.

[0003] In summary, existing technologies cannot achieve differentiated processing of core data and background data, and it is difficult to balance transmission efficiency and resource utilization while ensuring data security. Summary of the Invention

[0004] This invention provides a data encryption transmission method and system for remote monitoring equipment, which solves the problem that existing technologies cannot achieve differentiated processing of core data and background data, and are difficult to balance transmission efficiency and resource utilization while ensuring data security.

[0005] In a first aspect, to solve the above-mentioned technical problems, the present invention provides a data encryption transmission method for remote monitoring equipment, comprising:

[0006] The raw data stream is collected by a remote monitoring device, and the image frames in the raw data stream are feature extracted to obtain the region boundary coordinates.

[0007] Based on the region boundary coordinates, the core region pixel matrix and the background region pixel matrix are separated, and the core region pixel matrix and the background region pixel matrix are weighted and fused to calculate the attention score and determine the priority label of the core region pixel matrix.

[0008] If the priority label meets the preset conditions, the core region pixel matrix is ​​subjected to strength encryption processing to obtain encrypted core ciphertext, the background region pixel matrix is ​​subjected to strength compression processing to obtain compressed background bitstream, and the encrypted core ciphertext and the compressed background bitstream are recombined to output differentiated data segments.

[0009] The current real-time transmission bandwidth is obtained through the network interface, the data segment bitrate requirement of the differentiated data segment is analyzed, and the allocation ratio of the differentiated data segment is dynamically adjusted according to the real-time transmission bandwidth and the data segment bitrate requirement to obtain the adjusted data segment.

[0010] The adjusted data segment is transmitted through a network channel, and an integrity check value is generated during the transmission process.

[0011] The integrity check value is used to verify whether the adjusted data segment at the receiving end has been tampered with; if not, the verified data segment is obtained; if tampering exists, the adjusted data segment is re-acquired and the verification is performed again until the verified data segment is obtained.

[0012] The monitoring screen is reconstructed based on the verified data segment, and the pixel matrix of the core area and the pixel matrix of the background area are fused using local contrast stretching and edge sharpening to obtain the remote security monitoring output.

[0013] In a second aspect, the present invention provides a data encryption transmission system for remote monitoring equipment, comprising:

[0014] The key region extraction module is used to collect raw data streams through remote monitoring equipment, extract features from image frames in the raw data streams, and obtain the region boundary coordinates.

[0015] The data separation and grading module is used to separate the core region pixel matrix and the background region pixel matrix according to the region boundary coordinates, weightedly fuse the core region pixel matrix and the background region pixel matrix, calculate the attention score and determine the priority label of the core region pixel matrix;

[0016] The differentiation processing module is used to perform strength encryption processing on the pixel matrix of the core region to obtain encrypted core ciphertext, perform strength compression processing on the pixel matrix of the background region to obtain compressed background bitstream, and reassemble the encrypted core ciphertext and the compressed background bitstream to output differentiated data segments if the priority label meets the preset conditions.

[0017] The resource adaptation and adjustment module is used to obtain the current real-time transmission bandwidth through the network interface, parse the data segment bitrate requirement of the differentiated data segment, and dynamically adjust the allocation ratio of the differentiated data segment according to the real-time transmission bandwidth and the data segment bitrate requirement to obtain the adjusted data segment.

[0018] The transmission verification module is used to transmit the adjusted data segment through the network channel and generate an integrity verification value during the transmission process.

[0019] The tamper verification module is used to verify whether the adjusted data segment of the receiving end has been tampered with based on the integrity check value; if not, the verified data segment is obtained; if tampering exists, the adjusted data segment is re-acquired and the verification is performed again until the verified data segment is obtained.

[0020] The image fusion module is used to reconstruct the monitoring image based on the verified data segment, and to fuse the pixel matrix of the core area and the pixel matrix of the background area using local contrast stretching and edge sharpening to obtain the remote security monitoring output.

[0021] Compared with the prior art, the present invention has the following beneficial effects:

[0022] (1) This invention extracts the boundary coordinates of key regions of monitoring image frames through convolutional neural networks, separates the core region from the background region, performs dynamic key encryption on the core data, and performs low-intensity compression on the background data to achieve differentiated security processing of data, thus solving the problems of insufficient protection of core data and excessive resource consumption of background data caused by the unified processing of existing technologies.

[0023] (2) This invention combines real-time transmission bandwidth and terminal computing power to dynamically adjust the allocation ratio of core ciphertext and background code stream, adapts to resource-constrained scenarios, and solves the problem that fixed transmission strategies are difficult to balance bandwidth usage and transmission stability, and are prone to stuttering and delay.

[0024] (3) This invention generates an integrity verification value by forward error correction coding and hash tree verification to verify the data anti-tampering status. Then, the image is merged by local contrast stretching and edge sharpening. This solves the problem of difficulty in ensuring authenticity and blurring of key details in the reconstructed image during data transmission, and provides a safe and efficient transmission solution for remote monitoring. Attached Figure Description

[0025] Figure 1 This is a schematic flowchart of a data encryption transmission method for remote monitoring equipment provided in the first embodiment of the present invention;

[0026] Figure 2 This is a schematic diagram of the data encryption transmission system for remote monitoring equipment provided in the second embodiment of the present invention. Detailed Implementation

[0027] 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.

[0028] Reference Figure 1 The first embodiment of the present invention provides a data encryption transmission method for remote monitoring equipment, comprising the following steps:

[0029] S101, the raw data stream is collected through a remote monitoring device, and the image frames in the raw data stream are extracted to obtain the region boundary coordinates;

[0030] S102, Separate the core region pixel matrix and the background region pixel matrix according to the region boundary coordinates, weightedly fuse the core region pixel matrix and the background region pixel matrix, calculate the attention score and determine the priority label of the core region pixel matrix;

[0031] S103, if the priority tag meets the preset conditions, perform strength encryption processing on the core region pixel matrix to obtain encrypted core ciphertext, perform strength compression processing on the background region pixel matrix to obtain compressed background bitstream, and reassemble the encrypted core ciphertext and the compressed background bitstream to output differentiated data segments.

[0032] S104, obtain the current real-time transmission bandwidth through the network interface, parse the data segment bit rate requirement of the differentiated data segment, and dynamically adjust the allocation ratio of the differentiated data segment according to the real-time transmission bandwidth and the data segment bit rate requirement to obtain the adjusted data segment;

[0033] S105, the adjusted data segment is transmitted through the network channel, and an integrity check value is generated during the transmission process;

[0034] S106, verify whether the adjusted data segment of the receiving end has been tampered with according to the integrity check value; if not, obtain the verified data segment; if tampering exists, re-acquire the adjusted data segment and perform verification again until the verified data segment is obtained.

[0035] S107, reconstruct the monitoring screen based on the verified data segment, and fuse the pixel matrix of the core area and the pixel matrix of the background area using local contrast stretching and edge sharpening to obtain the remote security monitoring output.

[0036] In step S101, the raw data stream is acquired through a remote monitoring device, and feature extraction is performed on the image frames in the raw data stream to obtain the region boundary coordinates, including:

[0037] The original data stream is decoded into a single-frame image sequence, and the single-frame image sequence is input into a preset convolutional neural network to output a deep feature map.

[0038] Based on the deep feature map, a region proposal network is used to predict multiple candidate anchor boxes of different scales and proportions, forming a candidate anchor box set.

[0039] The region proposal network performs binary classification prediction on the deep features corresponding to each candidate anchor box to calculate the foreground probability that each anchor box contains the target.

[0040] Filter candidate anchor boxes whose foreground probabilities exceed a preset probability threshold to obtain specific anchor boxes;

[0041] Calculate the coordinate offset of the specific anchor frame, correct the original coordinates of the specific anchor frame based on the coordinate offset, and obtain the region boundary coordinates.

[0042] It's worth noting that the training data for the pre-defined convolutional neural network covers six types of monitoring scenarios, including factory workshops, park roads, and public places. The annotations include boundary coordinates and category labels for targets such as people, equipment, and vehicles. The training data is divided into training, validation, and test sets in an 8:1:1 ratio. The training method uses stochastic gradient descent with a momentum parameter of 0.9, a weight decay coefficient of 1e-4, an initial learning rate of 0.001, and a learning rate that decays to 0.1 every 10 epochs. The total number of training iterations is 50 epochs. The loss function is a weighted sum of classification and regression losses in a 1:1 weighting ratio. An early stopping strategy is used during training; training stops when the validation set loss shows no decrease for five consecutive epochs. The pre-defined convolutional neural network uses a VGG-16 network architecture with 13 convolutional layers and 3 fully connected layers as the feature extraction backbone. After removing the terminal fully connected layers, a single-frame image sequence is input into the network, and its convolutional layers perform feature transformation and downsampling to output the corresponding deep feature maps.

[0043] It should be noted that the Region Proposal Network predicts multiple candidate anchor boxes of different scales and ratios based on deep feature maps. These scales and ratios can be configured according to the actual application scenario and target characteristics. As an exemplary implementation, three scales are preset: 128×128, 256×256, and 512×512; and three ratios are 1:1, 1:2, and 2:1, forming a total of nine anchor box types. Each feature map pixel corresponds to nine candidate anchor boxes, and approximately 9000 candidate anchor boxes can be generated per frame, forming a candidate anchor box set. Binary classification prediction uses the feature vector output from the fully connected layer of the Region Proposal Network. After processing by the Sigmoid activation function, the probability that each candidate anchor box belongs to the foreground or background is obtained, i.e., the foreground probability.

[0044] It is worth noting that the preset probability threshold was determined through experimental statistics. Fifty sets of surveillance image samples from different scenarios were selected, and the false negative and false positive rates were statistically analyzed under different thresholds. When the threshold was set to 0.7, the false negative rate was less than 3%, and the false positive rate was less than 5%, balancing the completeness and accuracy of detection. During the screening process, the candidate anchor box set was traversed one by one, and the foreground probability of each anchor box was compared with the preset probability threshold. Anchor boxes with foreground probabilities exceeding the threshold were retained as specific anchor boxes.

[0045] It should be noted that the coordinate offset is calculated through the regression branch of the region proposal network. This branch outputs four offset parameters, corresponding to the anchor box's x-direction offset, y-direction offset, width offset, and height offset, respectively. During coordinate correction, the coordinates of the top-left and bottom-right corners of the original anchor box are adjusted based on these offset parameters. The region boundary coordinates can be in floating-point form to guide pixel extraction. If the coordinate information itself needs to be recorded or transmitted, it can be converted to integer format for storage. The coordinates and size of the original anchor box are predicted by the region proposal network based on deep feature maps, in pixels. The offset parameters are normalized values ​​learned through the training process, adapting to anchor box correction requirements of different scales and proportions.

[0046] In step S102, the core region pixel matrix and the background region pixel matrix are separated according to the region boundary coordinates, and the core region pixel matrix and the background region pixel matrix are weighted and fused to calculate the attention score and determine the priority label of the core region pixel matrix, including:

[0047] The original data stream is spatially segmented based on the region boundary coordinates, and the pixel matrix of the core region and the pixel matrix of the background region are extracted.

[0048] The core region pixel matrix is ​​feature-encoded to obtain a semantic feature vector, the gray-level variance of the background region pixel matrix is ​​calculated, and the gray-level variance is quantized to obtain the background redundancy.

[0049] The semantic feature vector is matched with a pre-set object category library to obtain a category importance coefficient. The category importance coefficient is then weighted and fused with the background redundancy to obtain an attention score.

[0050] If the attention score meets the preset priority conditions, a priority label is generated and associated with the core region pixel matrix.

[0051] It should be noted that spatial segmentation is performed based on the pixel positions of the top-left and bottom-right corners of the region boundary coordinates. In a single frame image, all pixels within the area defined by the boundary coordinates are extracted to form the core region pixel matrix; the set of pixels in the remaining area of ​​the image forms the background region pixel matrix. The pixel matrices are stored as two-dimensional arrays, with each array element recording the RGB values ​​of the corresponding pixel, ensuring complete preservation of pixel information. The size of the core region pixel matrix is ​​determined by the difference in boundary coordinates, and the size of the background region pixel matrix is ​​the difference between the size of the single frame image and the size of the core region pixel matrix.

[0052] It's worth noting that feature encoding utilizes a pre-trained feature extraction model. This model includes convolutional layers, pooling layers, and fully connected layers. After inputting the pixel matrix of the core region, it outputs a fixed-dimensional semantic feature vector through layer-by-layer computation. The semantic feature vector is set to 1024 dimensions, with each dimension's value ranging from 0 to 1, used to represent the deep semantic information of the core region. The calculation of background grayscale variance first involves converting the background region pixel matrix into a grayscale image. The grayscale value of each pixel is calculated using a weighted formula based on the RGB values. Then, the variance of all grayscale values ​​is calculated. A larger variance indicates more complex background information and lower redundancy; a smaller variance indicates simpler background information and higher redundancy.

[0053] It should be noted that the pre-set object category library includes common monitoring target categories such as people, motors, vehicles, and conveyor belts. Each category corresponds to a unique feature template and a fixed category importance coefficient, ranging from 0 to 1. During matching, the semantic feature vector is compared one by one with the feature templates of each category in the category library, and the coefficient corresponding to the category with the highest similarity is selected as the category importance coefficient. The weighted fusion sets the weight ratio of the category importance coefficient to 7 and the weight ratio of background redundancy to 3. The two are summed proportionally to obtain the attention score, ranging from 0 to 1. The weight ratios are derived from statistical analysis of multiple experimental data.

[0054] It's worth noting that the preset priority conditions were determined through experimental statistics. Sixty sets of sample data covering different monitoring scenarios and target categories were selected, and the importance feedback of the core area corresponding to different scores was statistically analyzed. When the score reaches 0.8, the probability that the core area is a key monitoring target meets the needs of practical application; therefore, 0.8 was set as the priority threshold. When the attention score exceeds this threshold, a high-priority label is generated. The label is stored numerically as 0.9 and directly written into the attribute field of the core area pixel matrix, realizing the association between the label and the core area pixel matrix.

[0055] For example, in a factory workshop monitoring scene, the area boundary coordinates obtained through step S101 are (490, 305) at the top left corner and (690, 455) at the bottom right corner. Based on these coordinates, spatial segmentation is performed on a single frame image of 1920×1080 pixels, extracting a 200×150 pixel core region pixel matrix and a 1720×930 pixel background region pixel matrix. Feature encoding is performed on the core region pixel matrix, outputting a 1024-dimensional semantic feature vector. After converting the background region pixel matrix into a grayscale image, the grayscale variance is calculated to be 10.5. Based on a pre-defined threshold range established through statistical analysis of a large number of monitoring scene background images, this variance value is quantified as a low background redundancy level. For example, a variance below 50 indicates low background redundancy, a variance between 50 and 200 indicates medium redundancy, and a variance above 200 indicates high redundancy. The semantic feature vector is matched against a pre-set object category library, and the motor category is matched, with a corresponding category importance coefficient of 0.9. The scores are weighted and merged at a ratio of 7:3, resulting in a focus score of 0.85. This score exceeds a preset priority threshold of 0.8, generating a high-priority tag with a value of 0.9. The value of this high-priority tag can be set according to the actual system design requirements. As an example implementation, the tag value can be set to 0.9 to distinguish it from other possible priority levels. This value is written into the attribute field of the core region pixel matrix, completing the association of the priority tag.

[0056] It's worth noting that the training data for the feature extraction model includes 80,000 labeled images of different monitored targets, covering 10 common categories such as people, equipment, and vehicles. The data is divided into training, validation, and test sets in an 8:1:1 ratio. Training employs stochastic gradient descent, with an initial learning rate of 0.001, which decays to 0.5 every 15 epochs, for a total of 40 epochs, ensuring the model can stably output effective semantic feature vectors. The feature templates of the pre-defined object category library are calculated by averaging the semantic feature vectors of samples from each category in the training set. The category importance coefficient is determined based on the criticality of each category in the monitoring scenario.

[0057] In step S103, if the priority tag meets a preset condition, strength encryption processing is performed on the core region pixel matrix to obtain encrypted core ciphertext, and strength compression processing is performed on the background region pixel matrix to obtain compressed background bitstream, including:

[0058] Extract the fingerprint texture features of the pixel matrix in the core region, convert the fingerprint texture features into a digital feature vector, and generate a dynamic key sequence.

[0059] Perform multiple rounds of scrambling and diffusion operations on the pixel matrix of the core region according to the dynamic key sequence to obtain the encrypted core ciphertext;

[0060] Downsampling is performed on the background region pixel matrix to obtain a dimension-reduced pixel matrix. Quantization encoding is then performed on the dimension-reduced pixel matrix to obtain a compressed background bitstream.

[0061] It is worth noting that the aforementioned strength encryption processing can specifically employ custom scrambling and diffusion operations. As an example, the scrambling and diffusion operations are set to execute in three rounds. The scrambling operation determines the pixel shifting rules based on the byte values ​​in the dynamic key sequence, randomly shifting the pixels of the core region pixel matrix along the row and column directions. The number of shift steps for each row of pixels is determined by the value of the corresponding key byte, and the shifting direction is determined by the parity of the key byte. The diffusion operation adjusts the pixel values ​​based on the dynamic key sequence, adding or subtracting the RGB values ​​of each pixel from the key byte value, with the result controlled between 0 and 255. Each round of scrambling and diffusion operations uses a different segment of the dynamic key sequence. After three rounds of operations, the core region pixel matrix is ​​converted into an encrypted core ciphertext that is not directly recognizable, and the encrypted core ciphertext is stored in binary stream form.

[0062] It should be noted that downsampling is performed on the background pixel matrix at a fixed ratio, reducing both the width and height of the background pixel matrix to half of their original values. During downsampling, the neighborhood averaging method is used to calculate the new pixel value. The average RGB values ​​of a 2×2 pixel block in the original matrix are selected as the pixel value at the corresponding position after dimensionality reduction, ensuring that the dimensionality-reduced pixel matrix retains the basic outline information of the background. Quantization encoding uses a preset 8-bit quantization table containing 256 quantization levels, each level corresponding to a grayscale range. The RGB values ​​of the dimensionality-reduced pixel matrix are converted to grayscale values. This process discards the background's color information while retaining its brightness outline. Then, the grayscale value is mapped to the corresponding quantization level encoding based on its range. The quantization encoding of each pixel occupies 1 byte, significantly reducing data storage and forming a compressed background bitstream.

[0063] It is worth noting that when presetting the reassembly format, the identifier field is written first, and then the encrypted core ciphertext, background stream length field, and compressed background stream are concatenated in sequence to form a complete differentiated data segment. The data segment is stored in binary file format. The identifier field is a fixed 4-byte value used by the receiving end to identify the data segment format; the background stream length field is 4 bytes and records the number of bytes in the compressed background stream.

[0064] For example, in a factory workshop monitoring scenario, the core area pixel matrix size is 200×150 pixels, with a priority label of 0.9, meeting preset conditions. First, the fingerprint texture features of the core area pixel matrix are extracted using a local binary mode algorithm, and this feature data is converted into an initial intermediate key. Then, this intermediate key is used as input, and a hash-based message authentication code algorithm is used to generate a 256-bit feature vector. Finally, a 128-byte dynamic key sequence is generated based on this feature vector. After three rounds of scrambling and diffusion operations, the encrypted core ciphertext in binary form is obtained, with a ciphertext size of 90KB (200×150×3 bytes). The background area pixel matrix size is 1720×930 pixels. After downsampling, a dimensionality-reduced pixel matrix of 860×465 pixels is obtained. After quantization encoding, a 158KB compressed background bitstream is generated. When reassembling according to the preset format, a 4-byte identifier field is first written, followed by a 90KB encrypted core ciphertext, a 4-byte background stream length field (recording 158KB), and a 158KB compressed background stream, finally outputting a 248KB differentiated data segment.

[0065] It is worth noting that the number of rounds for scrambling and diffusion operations was determined through experimental statistics. Fifty sets of pixel matrix samples from different core regions were selected, and multiple rounds of operations were performed to test the difficulty of cracking the encrypted data and the computation time. When the number of rounds was three, the cracking difficulty met security requirements, and the encryption time for a single frame image was controlled within 10 milliseconds, balancing security and efficiency. The downsampling ratio was determined based on the requirement to retain background information; both the width and height were reduced to half, which significantly reduced the amount of data while preserving the basic outline of the background area, avoiding excessive loss of background information that would affect subsequent image fusion. The quantization table was designed based on the background grayscale distribution characteristics of common monitoring scenarios. By statistically analyzing the grayscale value distribution of a large number of background images, 256 quantization levels were divided to ensure that the quantized and encoded background bitstream could restore the basic appearance of the background.

[0066] In step S104, the allocation ratio of the differentiated data segments is dynamically adjusted according to the real-time transmission bandwidth and the data segment bit rate requirement to obtain the adjusted data segments, including:

[0067] Calculate the difference between the real-time transmission bandwidth and the data segment bit rate requirement to determine the size of the remaining available transmission window;

[0068] Based on the available transmission window size, the encrypted core ciphertext and the compressed background bitstream in the differentiated data segment are weighted and assigned to generate a sequence of data frames to be transmitted;

[0069] The sequence of data frames to be transmitted is parsed and a preliminary estimate is performed to obtain an estimated resource consumption value. If the estimated resource consumption value is less than the preset maximum processing power threshold, the adjusted data segment is output.

[0070] It should be noted that the raw bitrate statistics for differentiated data segments are performed in two parts. The raw bitrate refers to the amount of data to be transmitted per unit time, and its calculation is based on the video frame rate. Assuming the video frame rate is F frames / second, the raw bitrate of the encrypted core ciphertext can be calculated by multiplying the data size (in bits) of the core ciphertext in a single frame by the frame rate F. The raw bitrate of the compressed background stream is calculated in the same way. The two raw bitrates are added together to obtain the total bitrate requirement, with the unit consistent with the real-time transmission bandwidth, such as Mbps.

[0071] It's worth noting that the remaining available transmission window size is calculated by subtracting the total bitrate requirement from the real-time transmission bandwidth. A positive result indicates the window is available, requiring no significant adjustment to the allocation ratio; a negative result indicates insufficient window space, necessitating data compression through ratio adjustments; and a result of zero indicates allocation according to the initial ratio. The window size directly determines the adjustment range of the allocation ratio; a smaller value results in a larger adjustment range.

[0072] It should be noted that the dynamic adjustment is based on a preset initial target data allocation ratio, for example, allocating 60% of the single-frame data to the core region and 40% to the background region. When the available transmission window is sufficient, the processing strategy corresponding to this target ratio (i.e., the compression quality of the background region and the encryption strength of the core region) is maintained. When the window is insufficient, the data volume proportion of the background region is reduced by decreasing its compression quality (e.g., increasing the quantization step size), with each adjustment reducing its target data volume weight by approximately 5%, until it reaches a minimum of 30%. If the window is still insufficient, the data volume proportion of the core region is reduced by appropriately decreasing its encryption strength (e.g., reducing the number of scrambling diffusion rounds), with each adjustment reducing its target data volume weight by approximately 5%, until it reaches a minimum of 50%, to ensure that core data transmission has a higher priority than background data. The processing strategy parameters are adjusted, thus affecting the final generated single-frame data volume and its proportion.

[0073] It is worth noting that when the data frame sequence to be transmitted is generated, each frame is a fixed size of 1400 bytes, arranged in the order of encrypting the core ciphertext frame first, followed by compressing the background bitstream frame. A 1-byte frame type identifier and a 2-byte frame sequence number are added to the beginning of each frame. The frame type identifier 01 indicates a core ciphertext frame, and 02 indicates a background bitstream frame; the frame sequence number increments from 0001. After the sequence is generated, it is stored in binary stream format, awaiting network transmission.

[0074] It should be noted that the estimated resource consumption is derived by statistically analyzing the processing requirements of each frame. The computational resources required for decrypting the encrypted core ciphertext frame and decompressing the compressed background bitstream frame can be estimated using quantifiable metrics such as millions of instructions per second (MIPS) or billions of floating-point operations per second (GFLOPS) based on instruction set simulation or benchmark testing of the target processor platform. The estimated computational resources required per second are obtained by summing the processing resource requirements of all frames and dividing by the transmission duration.

[0075] It is worth noting that the preset maximum processing power threshold was determined by statistically analyzing the hardware performance of common receiving devices. Three main types of receiving devices were selected: edge computing devices, monitoring terminals, and ordinary computers. Ten different models of each type were chosen, and their maximum processing capabilities were tested. The average of all test results, expressed in MIPS or GFLOPS, was multiplied by 0.8 to determine the preset maximum processing power threshold. This ensures that the devices have performance redundancy during operation and avoids overload.

[0076] For example, in a factory workshop monitoring scenario, the real-time transmission bandwidth of the network interface is 12Mbps. Assuming a video frame rate of F frames / second, under the initial processing strategy, the bitrate requirement corresponding to the total data volume of a single frame is 18Mbps, which is higher than the real-time bandwidth. At this point, the processing strategy needs to be adjusted to reduce the data volume per frame. Prioritize reducing the compression quality of the background area (equivalent to reducing its data volume weight). If this is still insufficient, appropriately reduce the encryption strength of the core area (equivalent to reducing its data volume weight) until the bitrate requirement corresponding to the estimated new single-frame data volume matches the real-time bandwidth. For example, through adjustment, the single-frame data volume is reduced, and its corresponding bitrate requirement drops to 12Mbps, matching the real-time bandwidth. At this point, the data volume ratio between the core area and the background area is approximately 55% to 45%. Based on this target, a data frame sequence with a fixed size of 1400 bytes per frame is generated for transmission. After parsing, based on benchmark test data of the target processor platform, the estimated resource consumption is approximately 3200 MIPS. The preset maximum processing power threshold is determined to be 4800 MIPS. Since the estimated value is less than the threshold, the adjusted data segment is output.

[0077] In step S105, the adjusted data segment is transmitted through the network channel, and an integrity check value is generated during the transmission process, including:

[0078] The adjusted data segment is subjected to forward error correction coding to generate an error correction coding block, and the error correction coding block is sent through the physical network link;

[0079] The error correction coding block is sent through the physical network link, and the data packet sequence during transmission is collected;

[0080] The data packet sequence is parsed, an independent hash value is calculated for each data packet, and an integrity hash tree is constructed layer by layer. The root node value of the hash tree is used as the initial verification feature value.

[0081] Obtain the dynamic verification window parameters fed back by the receiving end, and use the dynamic verification window parameters to perform secondary verification on the preliminary verification feature value to obtain the integrity verification value.

[0082] It should be noted that the forward error correction coding adjusts the coding strength according to the current channel packet loss rate. Packet loss on the physical network link is monitored in real time, and the average packet loss rate is calculated every 5 seconds. A higher redundancy coding method is used for the encrypted core ciphertext portion of the adjusted data segment; a lower redundancy coding method is used for the compressed background bitstream portion, balancing transmission security and bandwidth usage. After coding is complete, an error correction coding block containing the original data and error correction symbols is generated, with the block size adjusted according to the network transmission MTU value.

[0083] Independent hash value calculation is performed on the valid data portion of each data packet, which is the actual content after the packet header. After calculation, a 256-bit hash value is obtained, and each hash value is associated with and stored one-to-one with the sequence number of the corresponding data packet. The integrity hash tree is constructed hierarchically, with leaf nodes, intermediate nodes, and the root node. Leaf nodes contain the independent hash values ​​of all data packets. The hash values ​​of two adjacent leaf nodes are concatenated and recalculated to form an intermediate node. This process iterates upwards layer by layer until a unique root node value is obtained, which serves as the initial verification feature value.

[0084] It is worth noting that the dynamic verification window parameters are generated in real time by the receiving end based on its own buffer usage and network jitter status. The parameters include the window length and the maximum tolerable reordering delay. The window length is the number of consecutive data packets currently being processed by the receiving end, and the maximum tolerable reordering delay is the maximum allowed time difference in arrival of the data packets. The receiving end sends the parameters to the sending end through a feedback channel. Upon receiving these parameters, the sending end, based on the window length parameter, extracts the independent hash values ​​of the data packets within the corresponding window range from the locally stored sequence of sent data packets, in the correct sequence number order. It then reconstructs the integrity hash tree corresponding to the window data according to the same rules as before, obtaining a new root node value. The sending end compares this newly calculated root node value with the preliminary verification feature value of the corresponding window calculated and saved during the original transmission. If they match, a second verification is performed to finally determine the integrity verification value; if they do not match, it is marked as a verification anomaly.

[0085] It should be noted that in a factory workshop monitoring scenario, the adjusted data segment size was 248KB, and the average packet loss rate of the real-time detection channel was 3%. High-redundancy encoding was used for the 90KB core encrypted ciphertext, and low-redundancy encoding was used for the 158KB compressed background bitstream, generating a 350KB error correction coding block. This coding block was split into 256 1400-byte data packets. During transmission, a sequence containing these 256 data packets was collected, and a 256-bit independent hash value was calculated for each packet. After constructing an integrity hash tree hierarchically, the root node value was used as the initial verification feature value. The receiving end fed back dynamic verification window parameters, with a window length of 120 data packets and a maximum tolerable reordering delay of 80ms. The sending end verified the initial verification feature value within this window range, confirming that the root node value remained consistent, thus obtaining a passed integrity verification value.

[0086] It's worth noting that the packet fragmentation size is determined based on the network transmission MTU standard. The 1400-byte setting prevents packets from being fragmented due to exceeding the MTU value, improving transmission efficiency. The calculation method for independent hash values ​​was determined through experimental statistics. Multiple hash calculation schemes were tested, and considering both calculation speed and collision resistance, a 256-bit hash value calculation method was ultimately chosen. The dynamic verification window parameter feedback frequency is once every 10 packets. The initial values ​​for the window length and maximum tolerable reordering delay were determined by statistically analyzing transmission data under different network scenarios and dynamically adjusted according to actual transmission conditions.

[0087] In step S106, the integrity check value is used to verify whether the adjusted data segment at the receiving end has been tampered with; if not, the verified data segment is obtained; if tampering exists, the adjusted data segment is reacquired and the verification is performed again until the verified data segment is obtained, including:

[0088] A sliding window mapping of the adjusted data segment is established using the integrity verification value to extract the subset of data to be verified;

[0089] Calculate the local hash fingerprint of the subset of data to be verified, and compare it with the preset standard fingerprint to generate an abnormal offset set;

[0090] Based on the set of abnormal offsets, a tampering feature matrix is ​​constructed. If the similarity between the tampering feature matrix and the preset attack model does not exceed the preset tampering threshold, the verified data segment is output. If the similarity exceeds the preset tampering threshold, the adjusted data segment is reacquired and the verification is performed again until the verified data segment is obtained.

[0091] It should be noted that the window size of the sliding window mapping is dynamically determined based on the total length of the data segment. For example, if the initial baseline value of the window size is set to 10 data packets, the window size increases by 10 data packets for every 1MB increase in the data segment length. For data segment lengths less than 1MB, the window size can use the baseline value of 10 data packets or be adjusted proportionally based on the total number of data packets. The window sliding step is fixed at 5 data packets. When establishing the mapping, the window length is extended backward from the trusted data boundary corresponding to the integrity check value, covering a continuous range of data packets. This ensures that the mapping can dynamically track the data transmission status and fully cover areas that may be tampered with.

[0092] It is worth noting that the data subset extraction prioritizes data packets related to the encrypted core ciphertext, which are distinguished by the type identifier in the header. During the extraction process, core ciphertext data packets within the window mapping are filtered in sequence number order, while a small number of background bitstream data packets are extracted as auxiliary verification. The number of auxiliary data packets is 10% of the number of core ciphertext data packets, ensuring that the subset can focus on key data while also ensuring the integrity of the overall data verification.

[0093] It should be noted that the local hash fingerprint calculation is performed on the valid data portion of each data packet to be verified. The valid data is the actual content after the packet header, and the result is a 256-bit hash value. The default standard fingerprint is the local hash fingerprint calculated by the sender for each data packet before data transmission, and it is stored in the fingerprint database in sequence according to the sequence number. During comparison, the hash fingerprint of the subset of data to be verified is compared one by one with the standard fingerprint of the corresponding sequence number in the fingerprint database. The relative position of the data packets with mismatched fingerprints in the subset is recorded, forming an abnormal offset set.

[0094] It is worth noting that the tampering feature matrix is ​​a two-dimensional matrix. The number of rows in the matrix equals the number of data packets in the subset of data to be verified, and the number of columns is fixed at 3, corresponding to the data packet sequence number, hash deviation degree, and offset position marker, respectively. In the matrix, the hash deviation degree is quantified by the byte difference of mismatched fingerprints, and the offset position marker is 1 or 0, where 1 represents the presence of an anomaly and 0 represents the absence of an anomaly. Data packets without anomalies are filled with 0 at the corresponding positions in the matrix.

[0095] It should be noted that the preset attack model includes feature matrices for various common tampering patterns, such as sudden tampering, interval tampering, and single-point tampering. The construction of these feature matrices is based on statistical analysis of historical tampering data. For each tampering pattern, a large amount of sample data is collected to extract typical patterns from the tampering feature matrix. For example, in the feature matrix of the sudden tampering pattern, the offset position marker often appears as multiple consecutive 1 values, with the hash deviation concentrated in the high value range; the single-point tampering pattern is characterized by isolated positions being marked as 1. Each element value of the feature matrix is ​​determined by the average value or maximum likelihood estimation of the sample data. Similarity comparison is performed by calculating the element matching rate between the tampering feature matrix and the feature matrix of each attack model. The matching rate is calculated by dividing the number of overlapping non-zero elements at the same position by the total number of elements in the matrix; the result is the similarity, with a value ranging from 0 to 1.

[0096] It is worth noting that the preset tampering threshold was determined through experimental statistics. Eighty groups of sample data containing different tampering patterns were selected to test the false positive rate and false negative rate under different thresholds. When the threshold was set to 0.85, the false positive rate was less than 4% and the false negative rate was less than 3%, which can effectively distinguish between normal data fluctuations and malicious tampering. Therefore, this value was determined as the preset tampering threshold.

[0097] It should be noted that when reacquiring the adjusted data segment, the sending end only retransmits the data packets corresponding to the abnormal offset set, not the entire data segment. A retransmission identifier is added to the header of the retransmitted data packets. Upon receiving the retransmitted data packets, the receiving end replaces the original abnormal data packets and re-executes the verification steps. If the similarity still exceeds the threshold after three consecutive retransmissions, a retransmission of the entire data segment is triggered to ensure that the data can complete verification within a reasonable number of attempts.

[0098] For example, in a factory workshop monitoring scenario, the adjusted data segment contains 256 data packets, the sliding window size is determined to be 20 data packets, and the window sliding step is 5 data packets. When extracting the data subset to be verified, 15 core encrypted data packets and 2 background bitstream data packets are selected to form a subset of 17 data packets. After calculating the local hash fingerprint of the subset, it is compared with the preset standard fingerprint. It is found that the fingerprints of data packets No. 8 and No. 12 do not match, and an abnormal offset set [8, 12] is generated. After constructing the tampering feature matrix, it is compared with the preset attack model. The similarity is 0.3, which does not exceed the preset tampering threshold of 0.85, and the verified data segment is output.

[0099] It is worth noting that the calculation method for local hash fingerprints was determined through multiple rounds of testing. Taking into account both computational efficiency and collision resistance, the 256-bit hash value setting effectively avoids misjudgments caused by fingerprint duplication. The limit on the number of retransmissions is based on network transmission stability statistics. The setting of 3 local retransmissions balances verification efficiency and data integrity, avoiding excessive retransmissions that consume bandwidth resources.

[0100] In step S107, the monitoring screen is reconstructed based on the verified data segment, and the pixel matrix of the core region and the pixel matrix of the background region are fused using local contrast stretching and edge sharpening to obtain the remote security monitoring output, including:

[0101] Parse the verified data segment and extract the core pixel block corresponding to the core region pixel matrix and the background redundant frame corresponding to the background region pixel matrix;

[0102] The core pixel block is subjected to local contrast stretching to obtain a clear pixel block, and the clear pixel block is then sharpened at the edges to generate an enhanced texture layer.

[0103] The redundant background frames are matched with a preset static texture library, and the static texture with the highest matching degree is selected as the background base image to be fused.

[0104] Assign weight values ​​to the pixels corresponding to the enhanced texture layer and the background image to be fused, combine the weight values ​​to obtain a fusion mask, and then perform weighted mixing of pixels according to the fusion mask to obtain a complete single-frame image. The complete single-frame image is then used as a remote security monitoring output.

[0105] It should be noted that when parsing and verifying the data segment, the encrypted core ciphertext and compressed background bitstream are separated based on the preset identifier and length fields in the data segment. The encrypted core ciphertext is decrypted to restore the core region pixel matrix, i.e., the core pixel block; the compressed background bitstream is decompressed to restore the background region pixel matrix, i.e., the background redundant frame. The decryption and decompression operations correspond to the encryption and compression logic in step S103, respectively, ensuring the consistency of data restoration. The core pixel block retains its original size and pixel information, and the background redundant frame is restored to its size before downsampling, providing a basis for subsequent fusion.

[0106] It's worth noting that local contrast stretching is performed based on the grayscale distribution of the core pixel block. The grayscale values ​​of all pixels in the core pixel block are statistically analyzed to determine the minimum and maximum values, expanding the grayscale range to 0 to 255. During stretching, the relative grayscale relationships between pixels are maintained, resulting in clearer details in the core area. Edge sharpening employs a gradient enhancement method, calculating the grayscale difference between each pixel in the core pixel block and its neighboring pixels. Pixels with differences exceeding a preset threshold undergo grayscale enhancement to improve edge contour clarity. The preset threshold was determined through experimental statistics. Fifty core pixel block samples were selected to test edge sharpness and noise control effects under different thresholds, ultimately determining a suitable threshold value.

[0107] It should be noted that the preset static texture library stores background texture samples from different monitoring scenarios, including common background types such as workshop walls, park roads, and public ground surfaces. Each sample contains feature information such as grayscale distribution and texture density. During matching, the energy and contrast feature values ​​are calculated based on the grayscale co-occurrence matrix, and the average grayscale value is also calculated. Similarly, the corresponding features of each sample in the static texture library are extracted. The fit calculation can use cosine similarity to measure the similarity between the feature vector of the redundant background frame and the feature vector of the samples in the library. The sample with the highest fit is selected as the background base map to be fused. If the fit of the features of the redundant background frame with all samples in the texture library is lower than the preset standard, the redundant background frame is directly used as the background base map to be fused to ensure the coherence of the fusion.

[0108] It's worth noting that the weights are assigned based on regional importance. The core region of the enhanced texture layer has a weight of 0.9, while the weight of the edge transition region starts at 0.9 at the core region boundary and linearly decreases towards the background region to 0.15. The weight of the background image to be blended and the corresponding region weight of the enhanced texture layer sum to 1, forming a complementary relationship. The blending mask is generated according to this weighting rule, using the boundary of the core pixel block as a reference and expanding outwards to form the transition region. The size of the mask is consistent with the size of a single frame image, and each pixel position corresponds to a weight value. During weighted blending, the RGB values ​​of the corresponding pixels in the enhanced texture layer and the background image to be blended are multiplied by their respective weights and then added together to obtain the blended pixel value.

[0109] It should be noted that during the reconstruction encoding process, a lower quantization parameter (QP) can be set for the image region corresponding to the enhanced texture layer, while a higher quantization parameter (QP) can be set for the region corresponding to the background image to be fused. This achieves the goal of allocating a higher bitrate to the former and a lower bitrate to the latter. The encoding format remains consistent with the original data stream, using the H.264 encoding format. Following the timestamp order of the verified data segments, complete single-frame images are encoded sequentially to form a continuous video stream, i.e., the output for remote security monitoring. Bitrate fluctuations in single-frame images are controlled during the encoding process to ensure smooth video playback.

[0110] For example, in a factory workshop monitoring scenario, after parsing and verifying the data segment, a core pixel block of 200×150 pixels and a background redundant frame of 1720×930 pixels are obtained. Based on the core region coordinate information parsed from the data segment or pre-stored, the core pixel block is positioned at the corresponding location in the background redundant frame. Local contrast stretching is performed on the core pixel block, expanding the grayscale value range from 50-200 to 0-255, and then edge sharpening is used to enhance the motor outline. The background redundant frame is matched with a preset static texture library, and the workshop wall texture with the highest matching degree is selected as the background image to be fused. After generating the fusion mask, pixel-level weighted mixing is performed according to a weight allocation of 0.9 for the core region and gradient attenuation for the transition region, resulting in a complete single-frame image of 1920×1080 pixels. The single-frame image is layered and encoded according to the timestamp sequence, with the enhanced texture layer allocated 65% of the bitrate and the background image allocated 35% of the bitrate, ultimately outputting a smooth remote security monitoring video stream.

[0111] It's worth noting that the samples in the preset static texture library are constructed by collecting a large number of background images from different monitoring scenarios. Each sample is preprocessed and its feature information is stored. Preprocessing includes operations such as grayscale conversion and noise reduction to ensure the accuracy of feature extraction. The width of the transition region of the fusion mask is set to 20 pixels, which was determined through multiple fusion experiments. The bitrate allocation ratio of the layered encoding was determined by testing the video clarity and bandwidth usage under different ratios. The 65%:35% ratio can control the overall bandwidth consumption while ensuring the image quality of the core area, meeting the transmission requirements of remote monitoring.

[0112] In summary, this invention provides a data encryption transmission method for remote monitoring equipment, which can solve the problems of existing technologies being unable to achieve differentiated processing of core data and background data, and being unable to balance transmission efficiency and resource utilization while ensuring data security, thus providing reliable technical support for data encryption transmission of remote monitoring equipment.

[0113] refer to Figure 2 The second embodiment of the present invention provides a data encryption transmission system for remote monitoring equipment, comprising:

[0114] The key region extraction module is used to collect raw data streams through remote monitoring equipment, extract features from image frames in the raw data streams, and obtain the region boundary coordinates.

[0115] The data separation and grading module is used to separate the core region pixel matrix and the background region pixel matrix according to the region boundary coordinates, weightedly fuse the core region pixel matrix and the background region pixel matrix, calculate the attention score and determine the priority label of the core region pixel matrix;

[0116] The differentiation processing module is used to perform strength encryption processing on the pixel matrix of the core region to obtain encrypted core ciphertext, perform strength compression processing on the pixel matrix of the background region to obtain compressed background bitstream, and reassemble the encrypted core ciphertext and the compressed background bitstream to output differentiated data segments if the priority label meets the preset conditions.

[0117] The resource adaptation and adjustment module is used to obtain the current real-time transmission bandwidth through the network interface, parse the data segment bitrate requirement of the differentiated data segment, and dynamically adjust the allocation ratio of the differentiated data segment according to the real-time transmission bandwidth and the data segment bitrate requirement to obtain the adjusted data segment.

[0118] The transmission verification module is used to transmit the adjusted data segment through the network channel and generate an integrity verification value during the transmission process.

[0119] The tamper verification module is used to verify whether the adjusted data segment of the receiving end has been tampered with based on the integrity check value; if not, the verified data segment is obtained; if tampering exists, the adjusted data segment is re-acquired and the verification is performed again until the verified data segment is obtained.

[0120] The image fusion module is used to reconstruct the monitoring image based on the verified data segment, and to fuse the pixel matrix of the core area and the pixel matrix of the background area using local contrast stretching and edge sharpening to obtain the remote security monitoring output.

[0121] It should be noted that the data encryption transmission system for remote monitoring equipment provided in this embodiment of the invention is used to execute all the process steps of the data encryption transmission method for remote monitoring equipment in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0122] It should be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the system embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0123] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for encrypted data transmission in remote monitoring equipment, characterized in that, include: The raw data stream is collected by a remote monitoring device, and the image frames in the raw data stream are feature extracted to obtain the region boundary coordinates. Based on the region boundary coordinates, the core region pixel matrix and the background region pixel matrix are separated, and the core region pixel matrix and the background region pixel matrix are weighted and fused to calculate the attention score and determine the priority label of the core region pixel matrix. If the priority label meets the preset conditions, the core region pixel matrix is ​​subjected to strength encryption processing to obtain encrypted core ciphertext, the background region pixel matrix is ​​subjected to strength compression processing to obtain compressed background bitstream, and the encrypted core ciphertext and the compressed background bitstream are recombined to output differentiated data segments. The current real-time transmission bandwidth is obtained through the network interface, the data segment bitrate requirement of the differentiated data segment is analyzed, and the allocation ratio of the differentiated data segment is dynamically adjusted according to the real-time transmission bandwidth and the data segment bitrate requirement to obtain the adjusted data segment. The adjusted data segment is transmitted through a network channel, and an integrity check value is generated during the transmission process. The integrity check value is used to verify whether the adjusted data segment at the receiving end has been tampered with; if not, the verified data segment is obtained. If tampering is found, the adjusted data segment is retrieved again and the verification is performed again until the verified data segment is obtained. The monitoring screen is reconstructed based on the verified data segment, and the pixel matrix of the core area and the pixel matrix of the background area are fused using local contrast stretching and edge sharpening to obtain the remote security monitoring output.

2. The data encryption transmission method for remote monitoring equipment according to claim 1, characterized in that, The step of acquiring raw data streams through remote monitoring equipment and extracting features from image frames in the raw data streams to obtain region boundary coordinates includes: The original data stream is decoded into a single-frame image sequence, and the single-frame image sequence is input into a preset convolutional neural network to output a deep feature map. Based on the deep feature map, a region proposal network is used to predict multiple candidate anchor boxes of different scales and proportions, forming a candidate anchor box set. The region proposal network performs binary classification prediction on the deep features corresponding to each candidate anchor box to calculate the foreground probability that each anchor box contains the target. Filter candidate anchor boxes whose foreground probabilities exceed a preset probability threshold to obtain specific anchor boxes; Calculate the coordinate offset of the specific anchor frame, correct the original coordinates of the specific anchor frame based on the coordinate offset, and obtain the region boundary coordinates.

3. The data encryption transmission method for remote monitoring equipment according to claim 1, characterized in that, The process of separating the core region pixel matrix and the background region pixel matrix based on the region boundary coordinates, weightedly fusing the core region pixel matrix and the background region pixel matrix, calculating the attention score, and determining the priority label of the core region pixel matrix includes: The original data stream is spatially segmented based on the region boundary coordinates, and the pixel matrix of the core region and the pixel matrix of the background region are extracted. The core region pixel matrix is ​​feature-encoded to obtain a semantic feature vector, the gray-level variance of the background region pixel matrix is ​​calculated, and the gray-level variance is quantized to obtain the background redundancy. The semantic feature vector is matched with a pre-set object category library to obtain a category importance coefficient. The category importance coefficient is then weighted and fused with the background redundancy to obtain an attention score. If the attention score meets the preset priority conditions, a priority label is generated and associated with the core region pixel matrix.

4. The data encryption transmission method for remote monitoring equipment according to claim 1, characterized in that, If the priority tag meets a preset condition, the core region pixel matrix is ​​subjected to strength encryption to obtain encrypted core ciphertext, and the background region pixel matrix is ​​subjected to strength compression to obtain compressed background bitstream, including: Extract the fingerprint texture features of the pixel matrix in the core region, convert the fingerprint texture features into a digital feature vector, and generate a dynamic key sequence. Perform multiple rounds of scrambling and diffusion operations on the pixel matrix of the core region according to the dynamic key sequence to obtain the encrypted core ciphertext; Downsampling is performed on the background region pixel matrix to obtain a dimension-reduced pixel matrix. Quantization encoding is then performed on the dimension-reduced pixel matrix to obtain a compressed background bitstream.

5. The data encryption transmission method for remote monitoring equipment according to claim 1, characterized in that, The step of dynamically adjusting the allocation ratio of the differentiated data segments according to the real-time transmission bandwidth and the data segment bit rate requirements to obtain the adjusted data segments includes: Calculate the difference between the real-time transmission bandwidth and the data segment bit rate requirement to determine the size of the remaining available transmission window; Based on the available transmission window size, the encrypted core ciphertext and the compressed background bitstream in the differentiated data segment are weighted and assigned to generate a sequence of data frames to be transmitted; The sequence of data frames to be transmitted is parsed and a preliminary estimate is performed to obtain an estimated resource consumption value. If the estimated resource consumption value is less than the preset maximum processing power threshold, the adjusted data segment is output.

6. The data encryption transmission method for remote monitoring equipment according to claim 1, characterized in that, The process of transmitting the adjusted data segment through the network channel and generating an integrity check value during transmission includes: The adjusted data segment is subjected to forward error correction coding to generate an error correction coding block, and the error correction coding block is sent through the physical network link; The error correction coding block is sent through the physical network link, and the data packet sequence during transmission is collected; The data packet sequence is parsed, an independent hash value is calculated for each data packet, and an integrity hash tree is constructed layer by layer. The root node value of the hash tree is used as the initial verification feature value. Obtain the dynamic verification window parameters fed back by the receiving end, and use the dynamic verification window parameters to perform secondary verification on the preliminary verification feature value to obtain the integrity verification value.

7. The data encryption transmission method for remote monitoring equipment according to claim 1, characterized in that, The process involves verifying whether the adjusted data segment at the receiving end has been tampered with based on the integrity check value; if not, the verified data segment is obtained. If tampering is found, the adjusted data segment is retrieved again and verification is performed again until the verified data segment is obtained, including: A sliding window mapping of the adjusted data segment is established using the integrity verification value to extract the subset of data to be verified; Calculate the local hash fingerprint of the subset of data to be verified, and compare it with the preset standard fingerprint to generate an abnormal offset set; Based on the set of abnormal offsets, a tampering feature matrix is ​​constructed. If the similarity between the tampering feature matrix and the preset attack model does not exceed the preset tampering threshold, the verified data segment is output. If the similarity exceeds the preset tampering threshold, the adjusted data segment is reacquired and the verification is performed again until the verified data segment is obtained.

8. The data encryption transmission method for remote monitoring equipment according to claim 1, characterized in that, The process of reconstructing the monitoring image based on the verified data segment, and fusing the pixel matrix of the core region and the pixel matrix of the background region using local contrast stretching and edge sharpening to obtain the remote security monitoring output includes: Parse the verified data segment and extract the core pixel block corresponding to the core region pixel matrix and the background redundant frame corresponding to the background region pixel matrix; The core pixel block is subjected to local contrast stretching to obtain a clear pixel block, and the clear pixel block is then sharpened at the edges to generate an enhanced texture layer. The redundant background frames are matched with a preset static texture library, and the static texture with the highest matching degree is selected as the background base image to be fused. Assign weight values ​​to the pixels corresponding to the enhanced texture layer and the background image to be fused, combine the weight values ​​to obtain a fusion mask, and then perform weighted mixing of pixels according to the fusion mask to obtain a complete single-frame image. The complete single-frame image is then used as a remote security monitoring output.

9. A data encryption transmission system for remote monitoring equipment, characterized in that, include: The key region extraction module is used to collect raw data streams through remote monitoring equipment, extract features from image frames in the raw data streams, and obtain the region boundary coordinates. The data separation and grading module is used to separate the core region pixel matrix and the background region pixel matrix according to the region boundary coordinates, weightedly fuse the core region pixel matrix and the background region pixel matrix, calculate the attention score and determine the priority label of the core region pixel matrix; The differentiation processing module is used to perform strength encryption processing on the pixel matrix of the core region to obtain encrypted core ciphertext, perform strength compression processing on the pixel matrix of the background region to obtain compressed background bitstream, and reassemble the encrypted core ciphertext and the compressed background bitstream to output differentiated data segments if the priority label meets the preset conditions. The resource adaptation and adjustment module is used to obtain the current real-time transmission bandwidth through the network interface, parse the data segment bitrate requirement of the differentiated data segment, and dynamically adjust the allocation ratio of the differentiated data segment according to the real-time transmission bandwidth and the data segment bitrate requirement to obtain the adjusted data segment. The transmission verification module is used to transmit the adjusted data segment through the network channel and generate an integrity verification value during the transmission process. The tamper verification module is used to verify whether the adjusted data segment of the receiving end has been tampered with based on the integrity check value; if not, the verified data segment is obtained. If tampering is found, the adjusted data segment is retrieved again and the verification is performed again until the verified data segment is obtained. The image fusion module is used to reconstruct the monitoring image based on the verified data segment, and to fuse the pixel matrix of the core area and the pixel matrix of the background area using local contrast stretching and edge sharpening to obtain the remote security monitoring output.

Images