A video stream encoding and decoding method and system based on Morse code encryption

Abstract

The application discloses a video stream encoding and decoding method and system based on Morse code encryption, and belongs to the technical field of multimedia information security, and aims to solve the technical problem that existing video sensitive information transmission is easily detected, and has weak anti-interception and anti-decoding capabilities. Its technical scheme comprises the following steps: a sending end separates secret data blocks and carrier videos of an original video, maps expression symbol sequences to the secret data blocks after encryption, hides synchronization information in the carrier videos, and transmits the synchronization information through double-channel separation; a receiving end extracts the synchronization information to reconstruct a mapping password book, analyzes the expression symbol sequences to obtain ciphertext data, and fuses the secret data blocks back to the carrier videos after decryption to reconstruct a complete video. The application improves the anti-interception and anti-decoding capabilities through double-channel separation transmission and dynamic mapping of the password book, the carrier videos keep visual coherence to avoid secret transmission being detected, and is suitable for high-security video transmission scenes.

Description

Technical Field

[0001] This invention relates to the field of multimedia information security technology, and more specifically, to a video stream encoding and decoding method and system based on Morse encoding encryption. Background Technology

[0002] In intelligent substation inspection scenarios, inspection robots or fixed cameras collect real-time video data from substation equipment (such as transformers, insulators, and circuit breakers) and transmit it to the monitoring center for analysis and storage. To ensure the security of sensitive information in the video data, existing technologies mainly employ two approaches:

[0003] The first approach is to encrypt the entire smart inspection video stream using traditional symmetric encryption algorithms (such as AES) before transmission. This approach requires encrypting the entire video stream, but smart inspection videos in substations typically contain a large amount of non-sensitive content (such as open spaces and backgrounds with no defective equipment). Full encryption would lead to a surge in data processing volume and increased transmission latency, making it difficult to meet the needs of real-time identification and alarm of equipment defects in smart inspection scenarios.

[0004] The second approach is to embed sensitive information (such as equipment defect markers) into the carrier video using steganography and then transmit it through a single communication channel (such as substation industrial Ethernet). Although this approach reduces the need for separate encryption, single-channel transmission poses significant security risks. If the transmission channel is intercepted, both the carrier video and the embedded sensitive information will be leaked simultaneously, leading to the illegal acquisition of substation equipment defect information and affecting the safe operation and maintenance of the power grid.

[0005] In view of this, a video stream encoding and decoding method and system based on Morse coding encryption is proposed. Summary of the Invention

[0006] The purpose of this invention is to provide a video stream encoding and decoding method and system based on Morse encoding encryption, so as to solve the technical problems of high transmission delay and high information security risks of existing video encryption encoding and decoding technologies in the context of intelligent substation inspection.

[0007] To solve the above-mentioned technical problems, this invention provides a video stream encoding and decoding method based on Morse coding encryption, comprising the following steps:

[0008] At the sending end, perform the following steps:

[0009] S101. Process the original video to separate at least one secret data block and the carrier video after removing the secret data block;

[0010] S102. Encrypt the secret data block to generate binary ciphertext data, and map the binary ciphertext data into a Morse-like signal sequence composed of first-class units and second-class units.

[0011] S103. Based on the dynamically generated mapping cipherbook, the first type of unit and the second type of unit in the Morse-like signal sequence are mapped to different emojis respectively to generate an emoji sequence.

[0012] S104. Generate synchronization information containing the generation parameters of the mapping cipherbook and the spatiotemporal location information of the secret data block, and embed the synchronization information into the carrier video in a steganographic manner.

[0013] S105. The carrier video with embedded synchronization information and the emoji sequence are transmitted separately through different communication channels;

[0014] At the receiving end, perform the following steps:

[0015] S106. Extract the steganized synchronization information from the received carrier video;

[0016] S107. Reconstruct the mapping cipherbook according to the generation parameters in the synchronization information, and use the reconstructed mapping cipherbook to reverse-parse the received emoji sequence into the Morse-like signal sequence, and then restore it to binary ciphertext data;

[0017] S108. The binary ciphertext data is decrypted and data is restored to obtain a secret data block. Based on the spatiotemporal location information in the synchronization information, the secret data block is fused into the carrier video to reconstruct a complete video.

[0018] As a further improvement to this technical solution, step S101 specifically includes:

[0019] Based on preset sensitive semantic categories, computer vision models are used to identify and locate target regions in the original video as the secret data blocks;

[0020] Using video restoration technology, based on the pixel information and temporal continuity information around the target area, the missing area after the extraction of the secret data block is filled with content-aware information to generate the visually coherent carrier video.

[0021] As a further improvement to this technical solution, in step S103, the dynamically generated mapping codebook is generated through a process that includes:

[0022] A random seed is calculated using a cryptographic hash function based on the pre-shared master key and the current session timestamp.

[0023] Using the random seed to control the pseudo-random number generator, a first emoji mapping table and a second emoji mapping table corresponding to the first type of unit and the second type of unit are randomly drawn from the preset emoji pool; wherein, the session timestamp is included as the generation parameter in the synchronization information described in step S104.

[0024] As a further improvement to this technical solution, step S102, which involves encrypting the secret data block, specifically includes:

[0025] The secret data block is compressed using the WebP image format;

[0026] The compressed data is encrypted using the Advanced Encryption Standard (AES) algorithm.

[0027] As a further improvement to this technical solution, the first type of unit corresponds to the dot signal in Morse code, and the second type of unit corresponds to the dash signal in Morse code.

[0028] As a further improvement to this technical solution, in step S103, during the process of generating the emoji sequence, interference emojis that do not carry secret information are randomly inserted into the sequence according to a preset probability.

[0029] As a further improvement to this technical solution, the step S104 of embedding the synchronization information into the carrier video via steganography specifically involves:

[0030] The synchronization information is encoded into a watermark signal, and the watermark signal is superimposed and embedded into the transform domain coefficients of the chroma channel of the carrier video.

[0031] A video stream encoding and decoding system based on Morse encoding and encryption, wherein the video stream encoding and decoding system based on Morse encoding and encryption is used to implement the above-mentioned video stream encoding and decoding method based on Morse encoding and encryption, comprising:

[0032] Transmitting device, comprising:

[0033] The separation processing module separates at least one secret data block and the carrier video after removing the secret data block;

[0034] The encryption and encoding module is used to encrypt the secret data block to generate binary ciphertext data, then convert it into a Morse code sequence, and map it into an emoji sequence based on a dynamically generated mapping cipherbook.

[0035] A synchronous steganography module is used to generate synchronization information containing the generation parameters of the mapping cipherbook and the spatiotemporal location information of the secret data block, and to steganographically embed the synchronization information into the carrier video.

[0036] The transmission control module is used to separate and transmit the carrier video with embedded synchronization information and the emoji sequence through different communication channels;

[0037] The receiving end device includes:

[0038] The receiving module is used to receive the carrier video and the emoji sequence;

[0039] The information extraction module is used to extract the steganized synchronization information from the received carrier video;

[0040] The decoding and restoration module is used to reconstruct the mapping cipherbook based on the generation parameters in the synchronization information, and to parse the emoji sequence to restore the binary ciphertext data;

[0041] The data decryption module is used to decrypt and restore the binary ciphertext data to obtain the secret data block;

[0042] The video reconstruction module is used to fuse the restored secret data blocks into the carrier video based on the spatiotemporal location information in the synchronization information, and reconstruct the complete video.

[0043] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0044] 1. In this video stream encoding and decoding method and system based on Morse coding encryption, by separating the secret data block from the carrier video, and encrypting only the sensitive secret data block, the amount of encrypted data in the substation intelligent inspection video is reduced, the transmission delay is shortened, and the needs of real-time identification of equipment defects in intelligent inspection scenarios are met; at the same time, dual-channel separate transmission is adopted, with the carrier video and the emoji sequence transmitted through different channels. Even if a single channel is intercepted, complete sensitive information cannot be obtained, thus improving the transmission security of substation intelligent inspection data.

[0045] 2. In this video stream encoding and decoding method and system based on Morse encoding encryption, secret data blocks are compressed using WebP format, which further reduces the amount of data transmission while ensuring the clarity of details of defects in substation equipment.

[0046] 3. In this video stream encoding and decoding method and system based on Morse encoding encryption, a mapping keybook is dynamically generated (based on a pre-shared master key and session timestamp). The emoji mapping rules are different for each smart patrol session, thus preventing fixed mapping rules from being cracked.

[0047] 4. In this video stream encoding and decoding method and system based on Morse encoding encryption, by inserting interference symbols into the emoji sequence, the probability of the emoji sequence being identified as sensitive data can be reduced, further enhancing the anti-interception capability. Attached Figure Description

[0048] Figure 1 This is a flowchart illustrating the overall method of the present invention;

[0049] Figure 2 This is an overall system block diagram of the present invention. Detailed Implementation

[0050] 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.

[0051] Example 1

[0052] Currently, in substation intelligent inspection scenarios, inspection robots or fixed cameras collect video data of equipment (such as transformers, insulators, and circuit breakers) within the substation in real time and transmit it to the monitoring center for analysis and storage. To ensure the security of sensitive information in the video data, existing technologies mainly employ methods such as encrypting the complete intelligent inspection video stream using traditional symmetric encryption algorithms (such as AES) before transmission, or embedding sensitive information (such as equipment defect markers) into the carrier video using steganography technology and transmitting it through a single communication channel (such as substation industrial Ethernet). However, existing technologies suffer from high transmission latency and significant information security risks.

[0053] In view of this, please refer to Figure 1 As shown, one of the objectives of this invention is to provide a video stream encoding and decoding method based on Morse coding encryption. This embodiment describes the application of this method in a real-time video transmission scenario for a substation inspection robot. The implementation process of this method is explained in detail below with specific steps:

[0054] Considering that only equipment defect areas and inspector faces are considered sensitive information in substation intelligent inspection videos, while non-sensitive content (such as the background of defect-free equipment and open areas) does not require encryption, separation processing can reduce the amount of subsequent encrypted data and lower transmission latency; therefore, at the sending end, step S101 is executed to process the original video, separating at least one secret data block and the carrier video after removing the secret data block. The specific implementation method is as follows:

[0055] A lightweight YOLOv8-n object detection model (deployed on the edge computing unit of the inspection robot, with a computing power of 2 TOPS and an inference speed of 35 fps) was adopted. Based on preset sensitive semantic categories (equipment defect areas, inspector faces, and equipment nameplate text), frame-level recognition was performed on the raw video (1920×1080 resolution, 25 fps) collected by the inspection robot. Taking transformer oil leakage defects as an example, the model identified the coordinate range of the oil leakage area in the video frame as (x1=320, y1=450) to (x2=580, y2=620) and marked the frame number of the area. (e.g., frames 100-150) This region of data is used as a secret data block. For the empty regions after removing the secret data blocks, STTN (Spatiotemporal Transformer Network) video inpainting technology is used. This model establishes long-distance dependencies in both time and space dimensions. Based on the pixel information around the empty region (such as the texture and color of the transformer shell) and temporal coherence information (defect-free region data from the first 10 frames and the last 10 frames), content-aware filling is performed to generate a visually coherent carrier video (the texture similarity between the filled empty region and the surrounding environment is ≥95%, with no obvious visual breaks).

[0056] By following the steps above, only the secret data blocks, which account for 15% to 25% of the video data, are processed, reducing the amount of data processing and laying the foundation for real-time transmission.

[0057] Since the secret data block needs to be compressed first to reduce transmission size, then encrypted to ensure confidentiality, and finally mapped to a Morse-like signal sequence to provide a basis for subsequent emoji mapping, step S102 involves encrypting the secret data block to generate binary ciphertext data, and mapping the binary ciphertext data to a Morse-like signal sequence composed of first-type units and second-type units. The specific implementation method is as follows:

[0058] First, the secret data block is compressed using the WebP image format. For defective areas, a lossy compression mode is selected (compression quality parameter Q=75, ensuring clear defect details while compressing the data volume to 40%–50% of the original). For the equipment nameplate text, a lossless compression mode is selected (to avoid blurry text affecting recognition). The compressed data is then encrypted using the AES-256 encryption algorithm. The encryption key is a 32-bit key pre-shared between the substation monitoring center and the inspection robot (imported via offline USB key to avoid network distribution and leakage). The encryption mode is CBC (initialization vector IV is generated by a subsequent random seed to avoid outputting the same ciphertext after encrypting the same plaintext), generating binary ciphertext data. The binary ciphertext data is then mapped to a Morse-like signal sequence according to the rule that "each bit corresponds to one Morse signal unit." A binary "1" maps to a "dot" signal (first type unit) in Morse coding, and a binary "0" maps to a "dash" signal (second type unit) in Morse coding. For example, the binary ciphertext "1011" corresponds to the Morse-like signal sequence "dot-dash-dot-dot".

[0059] Through the above steps, WebP compression reduces the transmission size of secret data blocks, AES-256 encryption ensures data confidentiality, and Morse code mapping provides a standardized format for subsequent emoji conversion.

[0060] Considering that if a fixed emoji mapping rule is used, once the rule is cracked, attackers can easily parse the emoji sequence. Dynamically generating a mapping cipherbook allows for different mapping rules in each smart patrol session, improving resistance to cracking. Therefore, in step S103, based on the dynamically generated mapping cipherbook, the first and second types of units in the Morse code sequence are mapped to different emojis respectively, generating an emoji sequence. The specific implementation is as follows:

[0061] Based on the pre-shared master key (same as the AES encryption key, 32 bits) and the current session timestamp (e.g., "20241001143025123", synchronized with the substation clock server via NTP protocol), a 256-bit random seed is calculated using the SHA-256 hash function. Using the MT19937 pseudo-random number generator, 20 emojis are randomly selected from a pre-set emoji pool (containing 100 emojis unrelated to the substation, preventing attackers from guessing content through emoji associations) as the first emoji mapping table. (Corresponding to the "dot" signal), then extract 20 different emojis as the second emoji mapping table (corresponding to the "swipe" signal), and use the remaining 60 emojis as the interference symbol pool; according to the above mapping table, the Morse code-like signal sequence "dot-swipe-dot-dot" is converted into an emoji sequence; at the same time, according to a preset probability of 8% (determined based on actual measurement of the 4G / 5G channel bandwidth of the substation, when the bandwidth is 10Mbps, 8% interference symbols do not affect the effective data transmission efficiency), interference symbols are randomly selected from the interference symbol pool and inserted into the emoji sequence to generate the final sequence;

[0062] Through the above steps, the emoji mapping rules are different for each session, and attackers cannot crack the current sequence through historical data; the insertion of interference symbols makes the emoji sequence closer to ordinary text data, reducing the probability of it being identified as sensitive information.

[0063] Since the receiving end needs to reconstruct the mapping codebook and the location of the secret data block through synchronization information in order to restore the complete video, embedding the synchronization information into the carrier video in a steganographic manner can prevent the synchronization information from being intercepted when transmitted separately. Therefore, step S104, generating synchronization information containing the generation parameters of the mapping codebook and the spatiotemporal location information of the secret data block, and embedding the synchronization information in a steganographic manner into the carrier video, is specifically implemented as follows:

[0064] The synchronization information includes two parts:

[0065] The first is the generation parameter of the mapping codebook, namely the session timestamp ("20241001143025123").

[0066] Second, the spatiotemporal location information of the secret data block, namely the frame number (frames 100-150) and pixel coordinates (x1=320, y1=450; x2=580, y2=620) where the defective area is located.

[0067] After the synchronization information is converted into an 8-bit wide binary data stream, it is encoded into a watermark signal using BCH encoding (error correction capability t=3, which can correct errors within 3 bits during transmission and adapt to the noisy environment of industrial Ethernet in substations). The carrier video uses the H.264 encoding format (a commonly used format for intelligent substation inspection, with strong compatibility), which is converted into the YUV420 color space. An 8×8 block DCT transformation is performed on the Cb chroma channel (the human eye's visual sensitivity to the Cb channel is lower than that to the Y channel, making the embedded watermark less noticeable). The intermediate frequency coefficients of each DCT block are taken (intermediate frequency coefficients, while ensuring anti-compression capability, are less likely to cause block effects perceptible to the human eye compared to low-frequency coefficients, and are less likely to be lost due to compression compared to high-frequency coefficients). The watermark signal is superimposed on the intermediate frequency coefficients with an embedding strength of 0.1 (determined through subjective visual evaluation, after embedding, the carrier video has no obvious color change). Then, through IDCT transformation and YUV to RGB conversion, a carrier video with embedded synchronization information is generated.

[0068] Through the above steps, the synchronization information is highly concealed, and the peak signal-to-noise ratio of the embedded carrier video does not affect the monitoring center's observation of the equipment background; BCH encoding ensures the correct extraction of synchronization information in noisy environments.

[0069] Considering that the carrier video and the emoji sequence are easily intercepted simultaneously during single-channel transmission, dual-channel separate transmission can reduce the risk of information leakage and adapt to the transmission characteristics of different channels in substations (stable industrial Ethernet, flexible 4G / 5G). Therefore, in step S105, the carrier video with embedded synchronization information and the emoji sequence are transmitted separately through different communication channels; the specific implementation method is as follows:

[0070] The video, which carries the embedded synchronization information, is transmitted via the substation's industrial Ethernet (transmission rate 100Mbps, latency ≤50ms, suitable for transmitting large-volume video data), with the destination address being the video receiving server at the monitoring center (IP address 192.168.1.100, port 554). The emoticon sequence is transmitted via the inspection robot's 4G / 5G module (using TCP protocol to ensure reliable data transmission, transmission rate ≥2Mbps, meeting the small-volume transmission requirements of the emoticon sequence), with the destination address being the encrypted data receiving server at the monitoring center (IP address 192.168.1.101, port 8080).

[0071] Through the above steps, the two channels transmit independently. Even if the industrial Ethernet is intercepted, the attacker can only obtain the carrier video without sensitive information; if the 4G / 5G channel is intercepted, only meaningless emoji sequences can be obtained, and the secret data cannot be recovered.

[0072] Considering that the receiving end needs to obtain synchronization information before it can reconstruct the codebook and the location of the secret data block, the extraction process must ensure the integrity and accuracy of the synchronization information. Therefore, step S106, extracting the steganized synchronization information from the received carrier video, is specifically implemented as follows:

[0073] After receiving the carrier video, the monitoring center receiving server converts it to YUV420 format and extracts the Cb chroma channel data; it performs 8×8 block DCT transformation on the Cb channel and extracts the intermediate frequency coefficients of each DCT block; based on the preset embedding strength (0.1) and the watermark signal length (determined based on the BCH encoding length of the synchronization information), the watermark signal is separated from the intermediate frequency coefficients; BCH decoding (corresponding to the encoding parameter t=3 at the sending end) is used to correct possible transmission errors and restore the synchronization information (session timestamp and spatiotemporal location of secret data block).

[0074] The above steps improve the accuracy of synchronization information extraction, ensuring that synchronization information can still be correctly obtained even if there is slight noise in the carrier video during transmission.

[0075] Since a meaningless emoji sequence can only be parsed into a meaningful Morse-like signal and thus restored to binary ciphertext data by reconstructing the mapping cipherbook through synchronization information, step S107 involves reconstructing the mapping cipherbook based on the generation parameters in the synchronization information, and using the reconstructed mapping cipherbook to reverse-parse the received emoji sequence into the Morse-like signal sequence, thereby restoring it to binary ciphertext data. The specific implementation method is as follows:

[0076] After receiving the emoji sequence, the receiving encrypted data server first filters out interference symbols in the sequence according to a preset interference symbol pool (consistent with the sending end, 60 interference symbols), retaining valid emojis; based on the session timestamp ("20241001143025123") in the synchronization information and the pre-shared master key, it reconstructs the random seed using the SHA-256 hash function; using the MT19937 pseudo-random number generator, it reconstructs the first and second emoji mapping tables based on the random seed; according to the reconstructed mapping tables, it reverse-parses the valid emoji sequence into a Morse code-like signal sequence "dot-dash-dot-dot", then maps "dot" to binary "1" and "dash" to binary "0", restoring the binary ciphertext data "1011";

[0077] The above steps ensure the accuracy of interference symbol filtering and emoji sequence parsing, and can accurately restore binary ciphertext data.

[0078] Since the secret data block obtained after decrypting the binary ciphertext needs to be placed back to its original spatiotemporal location in order to reconstruct the complete intelligent patrol video containing sensitive information for analysis by the monitoring center, step S108 involves decrypting and restoring the binary ciphertext data to obtain the secret data block, and then fusing the secret data block into the carrier video based on the spatiotemporal location information in the synchronization information to reconstruct the complete video. The specific implementation method is as follows:

[0079] The AES-256 decryption algorithm is used, along with a pre-shared master key and initialization vector (extracted from a random seed), to decrypt the binary ciphertext data, obtaining a compressed secret data block. WebP decoding (corresponding to the compression mode at the sending end, lossy / lossless decoding are performed separately) is then used to reconstruct the original secret data block (image of the transformer oil leak area). Based on the spatiotemporal location information in the synchronization information (frames 100-150, coordinates x1=320, y1=450; x2=580, y2=620), the secret data block is used to replace the filling area in the carrier video frame by frame, generating a complete intelligent patrol video. The complete video is then transmitted to the video analysis platform of the monitoring center (such as the State Grid Intelligent Patrol Cloud Platform) for maintenance personnel to view details of equipment defects.

[0080] The above steps ensure the accuracy of secret data block restoration, resulting in low frame synchronization error after complete video reconstruction, and do not affect maintenance personnel's identification and analysis of equipment defects.

[0081] Example 2

[0082] The core steps of this embodiment are the same as those of Embodiment 1, the difference being that it adapts to the non-real-time requirements of historical video playback. The specific adjustments are as follows:

[0083] In step S101, the secret data block selects the substation equipment nameplate text (sensitive information that needs to be archived), and uses the Tesseract 5.0 optical character recognition model to identify the text area. The video restoration uses a more refined StyleGAN restoration technology to make the texture similarity between the filled area and the surrounding environment ≥98%.

[0084] In step S103, the preset probability of interference symbols is increased to 15% (historical video retransmission is not sensitive to latency, and security can be improved by increasing interference symbols).

[0085] In step S105, the carrier video is transmitted through the substation's fiber optic channel (rate 1Gbps, suitable for large-volume historical videos), and the emoji sequence is transmitted through the substation's intranet (IP address 192.168.1.102, port 8081).

[0086] In step S108, the reconstructed complete video is stored on the NAS storage server in the monitoring center (capacity 100TB, supports long-term video archiving) for subsequent operation and maintenance auditing.

[0087] Example 3

[0088] Considering that the sensitive information in the substation intelligent inspection scenario has clear industry attributes, it is necessary to define sensitive semantic categories in a targeted manner to accurately separate secret data blocks and avoid omissions or misjudgments. Therefore, step S101 specifically includes: based on the preset sensitive semantic categories, using a computer vision model to identify and locate the target area in the original video as the secret data block; using video restoration technology, based on the pixel information and temporal continuity information around the target area, performing content-aware filling on the empty area after the extraction of the secret data block to generate the visually coherent carrier video; this embodiment, in conjunction with embodiment 1, supplements the implementation details as follows:

[0089] The pre-defined sensitive semantic categories include three types: The first type is equipment defect areas (such as transformer oil leakage, insulator damage, and circuit breaker contact erosion), which uses the YOLOv8-n object detection model. The model training dataset contains 100,000 images of substation equipment defects (covering equipment from different manufacturers and models); the second type is the faces of inspection personnel, which uses the MTCNN face detection model (lightweight and suitable for edge computing) and can recognize frontal and side faces; the third type is equipment nameplate text (such as equipment model, production date, and rated parameters), which uses the Tesseract 5.0 optical character recognition model and supports common nameplate fonts such as Song and Hei. Through the above steps, the targeted definition of sensitive semantic categories improves the accuracy of secret data block separation, avoids misjudging non-sensitive content (such as substation walls) as secret data blocks, and avoids missing key sensitive information such as equipment defects.

[0090] Considering that the restored video carrier needs to meet the monitoring center's requirements for observing the equipment background, and that distortion in the filled area should not affect operational judgment, it is necessary to clarify the specific parameters of the restoration technology. Therefore, STTN (Spatiotemporal Transformer Network) video restoration technology (based on a deep generative dynamic video restoration algorithm) is adopted. The restoration window size is set to 5 frames (referencing 2 frames of data before and after to ensure temporal continuity), the number of iterations is set to 10 (to balance restoration effect and processing speed), and the loss function is L1 loss (to reduce the blurriness of the restored area). The restoration effect is verified through subjective evaluation (5 maintenance personnel visually score the restored video, with a maximum score of 10 points and an average score ≥ 8.5 points) and objective indicators (peak signal-to-noise ratio PSNR ≥ 38dB, structural similarity SSIM ≥ 0.92). Through the above steps, the restored carrier video has no obvious visual breaks, and maintenance personnel cannot distinguish the filled area with the naked eye, thus not affecting the judgment of the equipment background status.

[0091] Example 4

[0092] The dynamically generated mapping keybook in step S103 includes the following generation process: a random seed is calculated using a cryptographic hash function based on the pre-shared master key and the current session timestamp; the random seed is used to control a pseudo-random number generator to randomly extract and generate a first emoji mapping table and a second emoji mapping table corresponding to the first and second types of units, respectively, from a preset emoji pool; wherein the session timestamp is included as a generation parameter in the synchronization information described in step S104. This embodiment, in conjunction with Embodiment 1, supplements the implementation details as follows:

[0093] Since the pre-shared master key is the foundation of the dynamic mapping keybook, it is necessary to ensure the security of the key and the reliability of its distribution to prevent key leakage from causing the entire encryption system to fail. Therefore, the specific implementation method for generating and distributing the pre-shared master key is as follows:

[0094] The pre-shared master key is generated by the substation safety management system. A random number generator produces a 32-bit binary key (key entropy ≥ 256, meeting cryptographic security requirements). Key distribution is offline, using a dedicated USB key (compliant with the State Cryptography Administration GM / T0027 standard) to import the key into the inspection robot's edge unit and the monitoring center server. Network transmission is prohibited during distribution, and the USB key requires password unlocking (the password is kept by two maintenance personnel, with both unlocking it). The key is updated every three months (adjusted according to the substation's security level; important substations are updated monthly). Through these steps, the generation and distribution of the pre-shared master key comply with power industry security standards, reducing the risk of key leakage and providing a secure foundation for dynamic mapping codebooks.

[0095] Since the session timestamp needs to be strictly synchronized with both the sender and receiver—otherwise the receiver cannot reconstruct the random seed and thus cannot parse the emoji sequence—the synchronization method and format must be clearly defined. Therefore, the specific implementation method for the synchronization and format of the session timestamp is as follows:

[0096] The session timestamp uses the format "year-month-day-hour-minute-second-millisecond" (e.g., "20241001143025123", 17 bits in total). The sending and receiving ends synchronize with the substation clock server (deployed in the substation control room, with a clock accuracy ≤1ms) via the NTP protocol, with a synchronization period of 1 minute (to avoid clock offset caused by long-term operation). If NTP synchronization fails (e.g., clock server failure), the sending end automatically stops transmission and sends an alarm signal to the monitoring center (alarm information is sent to the mobile phones of maintenance personnel via an independent SMS channel). Through the above steps, the timestamp synchronization error between the sending and receiving ends is reduced, ensuring that the receiving end can accurately reconstruct the random seed, and timely alarms can be issued when timestamp synchronization fails, avoiding invalid transmission.

[0097] Example 5

[0098] In step S102, the secret data block is encrypted, specifically including: compressing the secret data block using the WebP image format; encrypting the compressed data using the Advanced Encryption Standard (AES) algorithm; This embodiment, in conjunction with Embodiment 1, provides the following supplementary implementation details:

[0099] Because the types of secret data blocks in substation intelligent inspections are different (images, text), different compression parameters need to be selected to minimize the transmission volume while ensuring information integrity. Therefore, the selection criteria for WebP compression parameters are as follows:

[0100] For equipment defect areas (image-type secret data blocks), WebP lossy compression mode is adopted, with the compression quality parameter Q ranging from 70 to 80 (determined experimentally: when Q=70, the defect area recognition accuracy is 96%, and the data volume is compressed to 45% of the original; when Q=80, the recognition accuracy is 98%, and the data volume is compressed to 55% of the original). In practical applications, the value is adjusted according to the defect type (e.g., Q=75 is selected for insulator damage, and Q=80 is selected for transformer oil leakage). For equipment nameplate text (text-type secret data blocks), WebP lossless compression mode is adopted. After compression, the text clarity is consistent with the original, and the data volume is compressed to 60%-70% of the original. Through the above steps, WebP compression reduces the transmission volume of secret data blocks without affecting equipment defect recognition and text reading, balancing compression efficiency and information integrity.

[0101] Considering that AES encryption requires selecting appropriate modes and parameters to avoid outputting the same ciphertext after encrypting the same plaintext, and to ensure that the encryption speed is compatible with the computing power of edge computing units, the specific modes and parameters of AES encryption are as follows:

[0102] The AES-256 encryption algorithm is adopted, and the encryption mode is CBC (Cryptographic Block Chaining, requiring initialization of vector N). N is generated by bitwise XOR operation between the first 16 bits of the random seed and the sequence number of the current secret data block (ensuring the uniqueness of N for different data blocks under the same timestamp and avoiding the risk of N reuse). The block size is 128 bits. The encryption process is executed in the hardware encryption module of the edge computing unit of the inspection robot (supporting AES hardware acceleration, encryption speed ≥100Mbps), avoiding the occupation of CPU resources and affecting video acquisition. Through the above steps, AES-256 encryption meets the security requirements of the State Cryptography Administration for the encryption of sensitive information. Hardware acceleration reduces encryption latency and does not affect real-time video transmission.

[0103] Example 6

[0104] The first type of unit corresponds to the dot signal in Morse code, and the second type of unit corresponds to the dash signal in Morse code. This embodiment, in conjunction with Embodiment 1, supplements the implementation details as follows:

[0105] To clarify the correspondence between dot / dash signals and binary data, ensure consistency in mapping logic between the transmitting and receiving ends, and avoid parsing errors, the mapping rules between dot / dash signals and binary data are defined as follows:

[0106] In the substation intelligent inspection scenario, binary data "1" corresponds to valid information related to equipment status (such as "equipment has a defect" or "text character is valid"). Therefore, binary "1" is mapped to a "dot" signal (first type of unit) in Morse code, and the signal duration is set to 100ms (referencing the Morse code standard to adapt to the transmission rhythm of emoticon sequences). Binary "0" corresponds to invalid or auxiliary information (such as "equipment has no defect" or "text character interval"). Therefore, binary "0" is mapped to a "dash" signal (second type of unit) in Morse code, and the signal duration is set to 300ms (three times that of the "dot" signal, conforming to the time ratio of Morse code). Through the above steps, the clear mapping rules ensure that the signal conversion logic of the sending end and the receiving end is consistent, avoiding parsing errors caused by ambiguous mapping relationships.

[0107] Example 7

[0108] In step S103, during the generation of the emoji sequence, interference emojis without carrying secret information are randomly inserted into the sequence according to a preset probability. This embodiment supplements the implementation details with Embodiment 1 as follows:

[0109] Since the interfering emojis must have no obvious correlation with the valid emojis, and the insertion logic must be random, in order to effectively confuse attackers without affecting the receiver's filtering, the implementation method for selecting and inserting interfering emojis is as follows:

[0110] The interference emoticons are selected from a preset interference pool (containing 60 everyday emoticons unrelated to substations, avoiding symbols related to power equipment); the insertion logic uses pseudo-random number control, based on a random number sequence generated by the session timestamp (different from the random seed of the mapping codebook to avoid correlation). When the random number falls within the range of [0, preset probability], one interference symbol is inserted; otherwise, a valid symbol is transmitted. The preset probability is adjusted according to the transmission channel bandwidth, set to 5%–10% for industrial Ethernet and 4G / 5G real-time transmission scenarios, and 15%–20% for fiber optic and intranet historical transmission scenarios. Through the above steps, the interference symbols increase the randomness of the emoticon sequence, making it impossible for attackers to determine valid information through symbol frequency, and the receiving end can accurately filter through the interference pool.

[0111] Example 8

[0112] Step S104, which involves embedding the synchronization information into the carrier video, specifically involves encoding the synchronization information into a watermark signal and then embedding the watermark signal into the transform domain coefficients of the chroma channel of the carrier video. This embodiment, in conjunction with Embodiment 1, provides the following supplementary implementation details:

[0113] Because choosing channels and transform domains with low human visual sensitivity is crucial to ensure that embedding synchronization information does not affect the visual quality of the carrier video and avoids detection by attackers, the implementation of the chroma channel and transform domain is as follows:

[0114] The video carrier uses the YUV420 color space, where the Y channel is the luminance channel (the human eye is sensitive to changes in luminance), and Cb and Cr are the chrominance channels (the human eye is less sensitive to changes in chrominance). Therefore, the Cb channel was chosen as the steganography carrier. An 8×8 block DCT transform was performed on the Cb channel. After the DCT transform, the coefficients are divided into mid-frequency coefficients (corresponding to the overall color trend of the video) and high-frequency coefficients (corresponding to color details). High-frequency coefficients are easily lost due to compression, and modifying low-frequency coefficients can easily lead to visual distortion. Therefore, mid-frequency coefficients (DCT block Zig-Zag scan) were selected. The 10th to 25th coefficients in the tracing sequence are used as the embedding positions; the embedding strength is set to 0.1 (determined experimentally: when the strength is <0.05, the accuracy of synchronization information extraction is <95%; when the strength is >0.15, the carrier video shows obvious color distortion; when the strength is =0.1, the extraction accuracy is ≥99.5% and there is no obvious distortion). Through the above steps, after the synchronization information is embedded, the visual quality of the carrier video does not decrease significantly (subjective score ≥8.5 points, PSNR ≥38dB), and the synchronization information can still be accurately extracted after transmission and compression.

[0115] Example 9

[0116] The preset sensitive semantic categories include at least one of faces, text information, and specific objects; the computer vision model includes an object detection model or an optical character recognition model. This embodiment, in conjunction with Embodiment 1, supplements the implementation details as follows:

[0117] Since the "specific objects" in substation intelligent inspection need to be clearly defined as safety-related equipment components, and the model needs to be optimized for the substation scenario to improve recognition accuracy, the specific scope of specific objects and model optimization are disclosed:

[0118] Specific objects include key equipment components such as transformer bushings, insulator strings, circuit breaker contacts, and grounding switches (defects in these components directly affect power grid safety). The target detection model (YOLOv8-n) is used for transfer learning in substation scenarios. Based on pre-training on the COCO dataset, it is fine-tuned using 100,000 substation equipment images, adjusting the anchor frame size of the model (to adapt to the aspect ratio of substation equipment, such as the slender anchor frame of insulator strings). Through the above steps, the specific object definition and model optimization make the separation of secret data blocks more accurate and avoid missing defect information of key equipment components.

[0119] Because the text on substation equipment nameplates may be blurry or tilted, the text processing workflow of the OCR model needs to be optimized to accurately identify text regions. Therefore, the details of the OCR model's text processing are disclosed:

[0120] The Tesseract 5.0 optical character recognition model is used, and the text processing workflow includes three steps: The first step is text region preprocessing, which involves grayscale conversion (to reduce color interference), binarization (threshold set to 127 to distinguish text from background), and tilt correction (based on Hough transform, with a correction angle range of -15° to 15°) of the identified nameplate region; the second step is text segmentation, which uses connected component analysis to segment the text into individual characters; the third step is character recognition, which loads a special character set for substation nameplates (including numbers 0-9, letters A, Z, and common symbols such as "-" and " / ") to improve recognition accuracy; the recognized text region is used as a secret data block, and regions where no text is recognized are excluded. Through the above steps, the preprocessing workflow improves the recognition accuracy of the OCR model for blurry and tilted nameplates, and avoids misjudgment of secret data blocks due to text recognition errors.

[0121] Example 10

[0122] To implement the methods of Embodiments 1 to 9, please refer to... Figure 2 As shown, the purpose of Embodiment 10 is to provide a video stream encoding and decoding system based on Morse coding encryption, comprising:

[0123] The transmitting device (deployed in the edge computing unit of the inspection robot) includes:

[0124] The separation processing module separates at least one secret data block and the carrier video after removing the secret data block;

[0125] The encryption and encoding module is used to encrypt the secret data block to generate binary ciphertext data, then convert it into a Morse code sequence, and map it into an emoji sequence based on a dynamically generated mapping cipherbook.

[0126] A synchronous steganography module is used to generate synchronization information containing the generation parameters of the mapping cipherbook and the spatiotemporal location information of the secret data block, and to steganographically embed the synchronization information into the carrier video.

[0127] The transmission control module is used to separate and transmit the carrier video with embedded synchronization information and the emoji sequence through different communication channels;

[0128] The receiving device (a server cluster deployed in the monitoring center) includes:

[0129] The receiving module is used to receive the carrier video and the emoji sequence;

[0130] The information extraction module is used to extract the steganized synchronization information from the received carrier video;

[0131] The decoding and restoration module is used to reconstruct the mapping cipherbook based on the generation parameters in the synchronization information, and to parse the emoji sequence to restore the binary ciphertext data;

[0132] The data decryption module is used to decrypt and restore the binary ciphertext data to obtain the secret data block;

[0133] The video reconstruction module is used to fuse the restored secret data blocks into the carrier video based on the spatiotemporal location information in the synchronization information, and reconstruct the complete video.

[0134] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A video stream encoding and decoding method based on Morse coding encryption, characterized in that, Includes the following steps: At the sending end, perform the following steps: S101. Process the original video to separate at least one secret data block and the carrier video after removing the secret data block. Using video restoration technology, based on the pixel information and temporal continuity information around the target area, perform content-aware filling on the empty area after the secret data block is extracted to generate the visually coherent carrier video. S102. Encrypt the secret data block to generate binary ciphertext data, and map the binary ciphertext data into a Morse-like signal sequence composed of first-class units and second-class units. S103. Based on the dynamically generated mapping cipherbook, the first type of unit and the second type of unit in the Morse-like signal sequence are mapped to different emojis respectively to generate an emoji sequence. S104. Generate synchronization information containing the generation parameters of the mapping cipherbook and the spatiotemporal location information of the secret data block, and embed the synchronization information into the carrier video in a steganographic manner. S105. The carrier video with embedded synchronization information and the emoji sequence are transmitted separately through different communication channels; At the receiving end, perform the following steps: S106. Extract the steganized synchronization information from the received carrier video; S107. Reconstruct the mapping cipherbook according to the generation parameters in the synchronization information, and use the reconstructed mapping cipherbook to reverse-parse the received emoji sequence into the Morse-like signal sequence, and then restore it to binary ciphertext data; S108. The binary ciphertext data is decrypted and data is restored to obtain a secret data block. Based on the spatiotemporal location information in the synchronization information, the secret data block is fused into the carrier video to reconstruct a complete video.

2. The video stream encoding and decoding method based on Morse coding encryption according to claim 1, characterized in that, Step S101 specifically includes: Based on preset sensitive semantic categories, the target region in the original video is identified and located using a computer vision model as the secret data block; Using video restoration technology, based on the pixel information and temporal continuity information surrounding the target area, content-aware filling is performed on the empty area after the secret data block is extracted to generate the visually coherent carrier video.

3. The video stream encoding and decoding method based on Morse coding encryption according to claim 1, characterized in that, In step S103, the dynamically generated mapping codebook is generated through a process that includes: A random seed is calculated using a cryptographic hash function based on the pre-shared master key and the current session timestamp. Using the random seed to control the pseudo-random number generator, a first emoji mapping table and a second emoji mapping table corresponding to the first type of unit and the second type of unit are randomly drawn from the preset emoji pool; wherein, the session timestamp is included as the generation parameter in the synchronization information described in step S104.

4. The video stream encoding and decoding method based on Morse coding encryption according to claim 1, characterized in that, In step S102, the encryption process for the secret data block specifically includes: The secret data block is compressed using the WebP image format; The compressed data is encrypted using the Advanced Encryption Standard (AES) algorithm.

5. The video stream encoding and decoding method based on Morse coding encryption according to claim 1, characterized in that: The first type of unit corresponds to the dot signal in Morse code, and the second type of unit corresponds to the dash signal in Morse code.

6. The video stream encoding and decoding method based on Morse coding encryption according to claim 1, characterized in that: In step S103, during the process of generating the emoji sequence, interference emojis that do not carry secret information are randomly inserted into the sequence according to a preset probability.

7. The video stream encoding and decoding method based on Morse coding encryption according to claim 1, characterized in that, The step S104, which involves embedding the synchronization information into the carrier video using steganography, specifically involves: The synchronization information is encoded into a watermark signal, and the watermark signal is superimposed and embedded into the transform domain coefficients of the chroma channel of the carrier video.

8. The video stream encoding and decoding method based on Morse coding encryption according to claim 2, characterized in that, The preset sensitive semantic categories include at least one of text information and specific objects; the computer vision model includes an object detection model or an optical character recognition model.

9. A video stream encoding and decoding system based on Morse encoding and encryption, wherein the video stream encoding and decoding system based on Morse encoding and encryption is used to implement the video stream encoding and decoding method based on Morse encoding and encryption as described in any one of claims 1 to 8, characterized in that, include: Transmitting device, comprising: The separation processing module separates at least one secret data block and the carrier video after removing the secret data block; The encryption and encoding module is used to encrypt the secret data block to generate binary ciphertext data, then convert it into a Morse code sequence, and map it into an emoji sequence based on a dynamically generated mapping cipherbook. A synchronous steganography module is used to generate synchronization information containing the generation parameters of the mapping cipherbook and the spatiotemporal location information of the secret data block, and to steganographically embed the synchronization information into the carrier video. The transmission control module is used to separate and transmit the carrier video with embedded synchronization information and the emoji sequence through different communication channels; The receiving end device includes: The receiving module is used to receive the carrier video and the emoji sequence; The information extraction module is used to extract the steganized synchronization information from the received carrier video; The decoding and restoration module is used to reconstruct the mapping cipherbook based on the generation parameters in the synchronization information, and to parse the emoji sequence to restore the binary ciphertext data; The data decryption module is used to decrypt and restore the binary ciphertext data to obtain the secret data block; The video reconstruction module is used to fuse the restored secret data blocks into the carrier video based on the spatiotemporal location information in the synchronization information, and reconstruct the complete video.

Images