A method for safe transmission of images of a drone

Abstract

This application provides a method for secure transmission of UAV images, comprising the following steps: extracting texture feature statistics and full information hash digests from the original UAV image; performing mixed operations on the texture feature statistics, hash digests, and a preset master key to generate a dynamic chaotic seed for initializing a chaotic system; using the dynamic chaotic seed as an initial value, inputting it into a finite-field fixed-point integer chaotic mapping system, and generating a pseudo-random sequence through an iterative process without floating-point operations; constructing a measurement matrix that dynamically changes with the image content using the pseudo-random sequence, performing compressed sensing linear projection observations on the sparse coefficients of the image after sparse transformation to obtain compressed observation data; quantizing the compressed observation data into an integer data stream, dynamically selecting encoding rules from a preset DNA encoding rule pool based on the real-time state value of the chaotic system to encode the data stream, and simultaneously introducing a mask generated by the output feedback OFB mode to XOR obfuscate the encoded data to obtain the final ciphertext.

Description

Technical Field

[0001] This invention relates to the field of secure image transmission technology, and more specifically to a method for secure image transmission from a drone. Background Technology

[0002] In modern industrial inspection, UAVs equipped with hyperspectral and infrared imaging devices face severe challenges such as limited bandwidth, open channels, and limited onboard computing power / energy consumption. They must ensure both the confidentiality of image transmission and real-time performance. Existing technologies suffer from structural flaws: the traditional "compression-encryption" serial architecture separates security from compression, resulting in high computational overhead, vulnerability to side-channel attacks, and failure to reduce data volume to adapt to low-bandwidth environments. While traditional compressed sensing (CS)-based solutions can reduce the dimensionality of observations, their linear transformations are vulnerable to known-plaintext attacks, and their reliance on floating-point operations makes them susceptible to precision drift due to hardware heterogeneity, leading to decryption failures.

[0003] Furthermore, existing encryption schemes mostly use static keys, lacking dynamic association with plaintext content, making them vulnerable to chosen-plaintext attacks. Chaotic encryption schemes, on the other hand, suffer from synchronization failures due to the heterogeneity of floating-point operations in hardware. Additionally, existing solutions cannot provide differentiated protection based on image texture features, and their decryption strategies are rigid, failing to achieve a flexible balance between rapid preview and high-precision analysis in resource-constrained industrial scenarios. Therefore, there is an urgent need for a lightweight secure transmission solution that integrates compression and encryption, adapts to heterogeneous hardware, and possesses content-aware capabilities. Summary of the Invention

[0004] The purpose of this invention is to solve the technical problems mentioned above and to propose a method for secure transmission of UAV images, comprising the following steps: S1. Dynamic chaotic seed generation: Extract the texture feature statistics and full information hash digest of the original UAV image, and perform mixed operation on the texture feature statistics, hash digest and preset master key to generate a dynamic chaotic seed for initializing the chaotic system. S2. Establishment of Integer Chaotic Basis: The dynamic chaotic seed is used as the initial value and input into the finite field fixed-point integer chaotic mapping system. A pseudo-random sequence is generated through an iterative process without floating-point operations. S3. Chaos-driven compressed sensing encryption uses pseudo-random sequences to construct a measurement matrix that dynamically changes with the image content. Compressed sensing linear projection observation is performed on the sparse coefficients of the image after sparse transformation to obtain compressed observation data. S4, Cascaded Diffusion Obfuscation Encryption, quantizes compressed observation data into an integer data stream. Based on the real-time state value of the chaotic system, it dynamically selects encoding rules from a pre-set DNA encoding rule pool to encode the data stream. At the same time, it introduces a mask generated by the output feedback OFB mode to XOR obfuscate the encoded data to obtain the final ciphertext.

[0005] In a preferred embodiment, in step S1, the texture feature statistics are obtained from the image derived data through edge or texture operators, and the hash digest is calculated from the image derived data through a cryptographic hash algorithm; the texture feature statistics, hash digest, and preset key material are mixed and operated to generate a dynamic chaotic seed or to update the initial chaotic value and / or control parameters.

[0006] In the preferred embodiment, in step S2, the finite field fixed-point integer chaotic mapping system is a preset bit-width integer field improved Logistic chaotic mapping system. Its state evolution equation includes nonlinear iterative terms and linear XOR perturbation terms. Through modular arithmetic, the state evolution trajectory of the chaotic system is strictly constrained within the preset bit-width integer field, ensuring the bit-level output consistency of the chaotic sequence on the heterogeneous hardware platform and eliminating the floating-point precision drift problem.

[0007] In the preferred scheme, in step S3, the sparse transformation adopts the 3rd level Haar discrete wavelet transform (DWT) to decompose the original UAV image at multiple scales, concentrate the image energy in the low-frequency sub-band, and preserve the high-frequency edge details to construct a sparse representation of the image. Compressed sensing linear projection observation is performed synchronously on the high-frequency and low-frequency sub-band coefficients after the DWT transformation.

[0008] In the preferred embodiment, in step S3, the construction process of the dynamic measurement matrix is ​​as follows: the integer pseudo-random sequence generated in step S2 is mapped through a Gaussian distribution to generate a measurement matrix that is strongly correlated with the content of the current frame image; the measurement matrix is ​​updated in real time with the texture features and hash features of each frame of UAV image, and serves as a one-time key for the compressed sensing process, realizing the deep integration of compression and encryption.

[0009] In the preferred embodiment, in step S4, the preset DNA coding rule pool contains at least 128 coding rules based on base pairing and base operations. The selection result of the current coding rule is uniquely determined by the real-time state value of the chaotic system. The key space is expanded by dynamically switching coding rules to resist brute-force attacks.

[0010] In the preferred scheme, in step S4, the specific execution method of the output feedback OFB mode is as follows: the output feedback information of the previous round is used to participate in the current round mask generation / parameter update, participate in the generation of the current round XOR mask, establish a chain diffusion relationship between ciphertext data blocks, so that the small changes in plaintext or key are diffused to all ciphertext data, and smooth out the original statistical features of the image.

[0011] The preferred solution also includes: S5. Ciphertext Integrity Verification: After obtaining the final ciphertext in step S4, calculate the message authentication code of the ciphertext data packet and encapsulate it together with the ciphertext and the parameters required for recovery. The receiving end performs message authentication code verification before decryption. If the verification fails, the data packet is discarded directly without performing subsequent decryption and reconstruction operations.

[0012] The preferred solution also includes: S6. Adaptive decryption and reconstruction: The receiving end selects a reconstruction algorithm to perform decryption and reconstruction based on a preset mode, user selection, or application scenario requirements.

[0013] In the preferred scheme, the reconstruction algorithm includes three types: the pseudo-inverse matrix FAST algorithm for real-time fast preview of the scene, the fast iterative shrinking threshold FISTA algorithm for standard transmission scene, and the orthogonal matching pursuit OMP algorithm for post-processing high-precision image analysis scene. The receiving end achieves a dynamic balance between image reconstruction quality and processing delay by switching algorithms.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: 1) By combining image texture / edge features with hash digests, the system's sensitivity to plaintext changes is improved, and the diffusion effect is enhanced.

[0015] 2) By introducing a preset bit-width integer field to improve the Logistic chaotic mapping system, the precision error of floating-point operations between DSP, FPGA and GPU is eliminated from the algorithm bottom layer. This helps to reduce the synchronization risk caused by precision drift of heterogeneous platforms and improve cross-platform consistency and engineering implementation stability.

[0016] 3) The four stages of DWT sparsification, CS dynamic observation, DNA encoding, and OFB obfuscation are interconnected and mutually causal. If attackers cannot accurately obtain the initial chaotic trajectory, they cannot deduce the CS measurement matrix, let alone parse the subsequent DNA encoding rules, thus achieving an "onion-like" multi-layered security barrier.

[0017] 4) The system natively supports multiple reconstruction algorithms such as FISTA, FAST, and OMP, allowing the receiving end to dynamically adjust the decryption strategy according to the computing power constraints of the job scenario, thereby achieving optimal resource utilization. Attached Figure Description

[0018] Figure 1 This is a flowchart of the end-to-end cascaded transmission process proposed in this invention.

[0019] Figure 2 This is a comparison chart of platform consistency between a 30-bit integer chaotic system and a traditional floating-point chaotic system.

[0020] Figure 3This is a comparison chart of plaintext sensitivity tests before and after the introduction of Sobel+SHA3.

[0021] Figure 4 This is a comparison chart showing the decryption results and time consumption of different reconstruction algorithms (FISTA / FAST / OMP).

[0022] Figure 5 This is a histogram uniformization effect resulting from DNA dynamic coding and OFB feedback. Detailed Implementation

[0023] Example 1 This invention proposes a method for secure transmission of images from unmanned aerial vehicles (UAVs), comprising the following steps: S1. Dynamic chaotic seed generation: Extract the texture feature statistics and full information hash digest of the original UAV image, and perform mixed operation on the texture feature statistics, hash digest and preset master key to generate a dynamic chaotic seed for initializing the chaotic system. S2. Establishment of Integer Chaotic Basis: The dynamic chaotic seed is used as the initial value and input into the finite field fixed-point integer chaotic mapping system. A pseudo-random sequence is generated through an iterative process without floating-point operations. S3. Chaos-driven compressed sensing encryption uses pseudo-random sequences to construct a measurement matrix that dynamically changes with the image content. Compressed sensing linear projection observation is performed on the sparse coefficients of the image after sparse transformation to obtain compressed observation data. S4, Cascaded Diffusion Obfuscation Encryption, quantizes compressed observation data into an integer data stream, dynamically selects encoding rules from a pre-set DNA encoding rule pool based on the real-time state value of the chaotic system to encode the data stream, and introduces a mask generated by the output feedback OFB mode to XOR obfuscate the encoded data to obtain the final ciphertext. S5. Ciphertext Integrity Verification: After obtaining the final ciphertext in step S4, the message authentication code of the ciphertext data packet is calculated and encapsulated together with the ciphertext and the parameters required for recovery. The receiving end performs message authentication code verification before decryption. If the verification fails, the data packet is discarded directly without performing subsequent decryption and reconstruction operations. S6. Adaptive decryption and reconstruction: The receiving end selects a reconstruction algorithm to perform decryption and reconstruction based on a preset mode, user selection, or application scenario requirements.

[0024] In step S1, the texture feature statistics are extracted from the image derived data using edge or texture operators, and the hash digest is calculated from the image derived data using a cryptographic hash algorithm. The texture feature statistics, hash digest, and preset key material are mixed and operated to generate a dynamic chaotic seed or to update the initial chaotic value and / or control parameters.

[0025] Preferably, Sobel edge enhancement is used for image preprocessing; texture feature statistics can be obtained from the statistical analysis of multiple convolutional kernel responses; and the hash perturbation factor can be calculated from the scrambled observation data using SHA3-256.

[0026] In step S2, the finite field fixed-point integer chaotic mapping system is a preset bit-width integer field improved Logistic chaotic mapping system. Its state evolution equation includes nonlinear iterative terms and linear XOR perturbation terms. Through modular arithmetic, the state evolution trajectory of the chaotic system is strictly constrained within the preset bit-width integer field, ensuring the bit-level output consistency of the chaotic sequence on the heterogeneous hardware platform and eliminating the floating-point precision drift problem.

[0027] Preferably, the finite field fixed-point integer chaotic mapping system is a 30-bit integer field improved Logistic chaotic mapping system. Its state evolution equation includes nonlinear iterative terms and linear XOR perturbation terms. Through modular arithmetic, the state evolution trajectory of the chaotic system is strictly constrained within the 30-bit integer field, ensuring the bit-level output consistency of the chaotic sequence on heterogeneous hardware platforms and eliminating the floating-point precision drift problem.

[0028] Preferably, in step S3, the sparse transformation adopts the 3rd level Haar discrete wavelet transform (DWT) to decompose the original UAV image at multiple scales, concentrate the image energy in the low-frequency sub-band, and preserve the high-frequency edge details to construct a sparse representation of the image. Compressed sensing linear projection observation is performed synchronously on the high-frequency and low-frequency sub-band coefficients after the DWT transformation.

[0029] Preferably, in step S3, the construction process of the dynamic measurement matrix is ​​as follows: the integer pseudo-random sequence generated in step S2 is mapped through a Gaussian distribution to generate a measurement matrix that is strongly correlated with the content of the current frame image; the measurement matrix is ​​updated in real time with the texture features and hash features of each frame of UAV image, and serves as a one-time key for the compressed sensing process, realizing the deep integration of compression and encryption.

[0030] Preferably, in step S4, the preset DNA coding rule pool contains at least 128 coding rules based on base pairing and base operation. The selection result of the current coding rule is uniquely determined by the real-time state value of the chaotic system. The key space is expanded by dynamically switching coding rules to resist exhaustive attacks.

[0031] In step S4, the specific execution method of the Output Feedback (OFB) mode is as follows: The output feedback information from the previous round is used to participate in the generation of the mask / parameter update in the current round, and to participate in the generation of the XOR mask in the current round, establishing a chain-like diffusion relationship between ciphertext data blocks. This allows minor changes in plaintext or the key to spread to all ciphertext data, smoothing out the original statistical features of the image. Preferably, the output feedback information from the previous round is used to participate in the generation of the XOR mask in the current round, establishing a chain-like diffusion relationship between ciphertext data blocks, allowing minor changes in plaintext or the key to spread to all ciphertext data, smoothing out the original statistical features of the image.

[0032] Preferably, the message authentication code calculated in step S5 is an HMAC-SHA256 message authentication code, which is appended to the header or footer of the data packet.

[0033] Preferably, in step S6, after the HMAC-SHA256 message authentication code is verified, the receiving end detects the current hardware node's computing load, remaining power, and real-time requirements in real time, and dynamically selects the appropriate algorithm from a variety of preset image reconstruction algorithms to complete the image decryption and reconstruction.

[0034] Preferably, the reconstruction algorithms include three types: the pseudo-inverse matrix FAST algorithm for real-time fast preview of the scene, the fast iterative shrinking threshold FISTA algorithm for standard transmission scenarios, and the orthogonal matching pursuit OMP algorithm for post-processing high-precision image analysis scenarios. The receiving end achieves a dynamic balance between image reconstruction quality and processing latency by switching algorithms.

[0035] Example 2 This invention constructs a deep cascaded system driven by plaintext features. Its core logic lies in establishing a "parameter dependency" and "state deadlock" mechanism: the output state of the previous stage directly determines the operating parameters of the next stage, and any attempt to decouple any stage will lead to the failure of the entire chain. The logical deduction of the system flow is as follows: Plaintext Sensitive Injection (Sobel + SHA3): By extracting the Sobel texture features (high-frequency component statistics) and SHA-3 hash digest (full information fingerprint) from the original image, and XORing them with the static master key, a dynamic chaotic seed is generated. This step establishes the cryptographic benchmark of "one-time pad".

[0036] Cross-platform Chaos Benchmark (30-bit Integer Logistic): Utilizes a dynamic seed to launch an improved 30-bit integer logistic chaotic system. This system simulates nonlinear dynamics through fixed-point arithmetic, ensuring absolute consistency of chaotic sequences across heterogeneous hardware platforms and eliminating the risk of precision drift.

[0037] Sparse Transform and Dynamic Compression (DWT + CS): Chaotic sequences are mapped to dynamic measurement matrices, and the sparse coefficients after the DWT transformation are compressed and sampled. Since the measurement matrix changes dynamically with the plaintext features, the compression process itself constitutes the first layer of encryption.

[0038] Biological dynamic diffusion (DNA + OFB + XOR): The compressed data stream enters the 128-rule DNA dynamic coding pool and undergoes XOR obfuscation using the OFB output feedback mode. This step achieves strong diffusion of the ciphertext, completely smoothing out local statistical features.

[0039] Integrity Closed Loop (HMAC): The final ciphertext is appended with an HMAC signature, and multiple reconstruction algorithms (FISTA / Fast / OMP) are supported at the receiving end to adapt to changing engineering needs.

[0040] The specific implementation steps are as follows: Step 1: Construct a plaintext-sensitive 30-bit integer chaotic system.

[0041] During system initialization, the direct use of a pre-set static key is strictly prohibited. Instead, a strong coupling relationship must be established between plaintext features and chaotic initial values. Specifically, assume the original image is... The system first calculates the mean feature of its Sobel edge detection operator. and SHA-3 256-bit hash digest Dynamic Chaos Seeds The formula for generating it is as follows:

[0042] Subsequently, the improved Logistic mapping is initialized using this seed. To avoid floating-point operation errors and eliminate the short-period effect caused by finite precision, this invention constructs a 30-bit integer Logistic mapping with linear perturbation, whose state evolution equation consists of two parts: nonlinear iteration and XOR perturbation.

[0043]

[0044] in For modulus, This indicates a bitwise XOR operation. These are control parameters derived from the master key. This refers to the current iteration round. The technical effect of this step is twofold: firstly, it ensures strict cross-platform consistency through integer operations; secondly, it addresses the linear perturbation term. This effectively disrupts the short-period property of finite field mappings, allowing the generated pseudo-random sequences to maintain highly complex dynamic behavior over long periods. Simultaneously, thanks to the avalanche effect of SHA-3, small image variations can lead to… and Dramatic changes, leading to chaotic trajectories The complete reconstruction fundamentally blocked the path of differential attacks.

[0045] Step 2: Chaos-driven DWT-CS joint compression.

[0046] Multi-scale sparsity: A 3-level Haar wavelet transform (DWT) is performed on the image to concentrate image energy in the low-frequency subband while preserving high-frequency edge details, thus establishing an efficient sparse representation of the image. .

[0047] Dynamic measurement matrix construction: Using the integer chaotic sequence generated in step 1, a measurement matrix is ​​constructed through Gaussian distribution mapping. It should be emphasized that this matrix is ​​not a fixed random matrix, but a dedicated matrix determined by the image content.

[0048] Compressed sensing observation: Performing linear projection observations .

[0049] Technical Effect Analysis: Based on the finite isometry property (RIP) of compressed sensing, the original signal can be reconstructed with high probability from the low-dimensional projection. In this scheme, the measurement matrix... It acts as a "one-time key." If an attacker cannot obtain the accurate chaotic trajectory, they cannot construct the sensing matrix required for decryption, thus achieving a deep integration of compression and encryption at the physical level.

[0050] Step 3: Dynamic DNA coding and obfuscation in OFB mode.

[0051] 128. Rule-based dynamic pool mapping: compressing data. Quantize into a stream of unsigned integers. Utilize the current chaotic state value. The current operation rule is dynamically selected from 128 predefined DNA coding rules (including base pairing and operation rules):

[0052] This mechanism greatly expands the key space, making exhaustive attacks computationally infeasible.

[0053] OFB Output Feedback and XOR Confusion: The Output Feedback (OFB) mode utilizes information from the previous round's output feedback to generate the current round's XOR mask. :

[0054] Technical Effect Analysis: OFB mode endows the system with extremely strong diffusion properties. Small changes in plaintext or key not only affect the current data block, but also affect all subsequent data blocks through feedback loops, resulting in an ideally uniform distribution of the ciphertext histogram, completely masking the statistical characteristics of the image.

[0055] Preferably, during the system initialization phase, the user password and a 16-byte random salt value are input into PBKDF2-HMAC-SHA256, and the master key is derived through 100,000 iterations. Sub-keys are then extracted according to their uses: 32 bytes are used for HMAC integrity authentication, and 16 bytes are used for sensitive parameter protection encryption, thereby achieving separate management of authentication keys and encryption keys.

[0056] Step 4: HMAC authentication and multi-mode decryption architecture.

[0057] Integrity verification: The HMAC-SHA256 digest of the final encrypted data packet is calculated and appended to the header to prevent man-in-the-middle attacks and malicious tampering in the channel.

[0058] Adaptive decryption strategy: After verifying HMAC, the receiver flexibly selects a reconstruction algorithm based on the current computing resource status. FAST pseudo-inverse mode: Based on the least squares method, the pseudo-inverse matrix reconstruction takes only seconds and is suitable for preview scenarios with extremely high real-time requirements.

[0059] FISTA Iterative Mode: A fast iterative shrinking threshold algorithm that provides a balance between image quality and speed within seconds, suitable for standard transmission scenarios.

[0060] OMP Fine-grained Mode: Orthogonal Matching Pursuit Algorithm provides the highest accuracy in image reconstruction, suitable for high-value post-hoc intelligence analysis.

[0061] The optimization objective of this invention is: Under the boundary conditions of limited unmanned aerial vehicle (UAV) computing resources (SWaP), a cascading mechanism is used to achieve a paradigm shift of "Compression-as-Encryption" to maximize covert throughput. Simultaneously, the system ensures robustness against differential attacks (through the Sobel+SHA3 mechanism), imperceptibility to statistical analysis (through the DNA+OFB mechanism), and engineering stability against heterogeneous computing platforms (through the 30-bit Integer mechanism).

[0062] Example 3: Cascaded Transmission of UAV Infrared Images In this embodiment, the core system parameters are set as follows: the wavelet basis is Haar wavelet, the compressed sensing sampling rate is set to 0.65, and the OFB diffusion rounds are set to 3 rounds.

[0063] Feature Coupling Stage: The system inputs an infrared thermal imaging image with a resolution of 512×512. The processor calculates the Sobel edge detection mean and SHA-3 256-bit hash digest of the image, and performs a bitwise XOR operation with the calculated result and a preset master key to generate a unique dynamic chaotic seed. .

[0064] Integer evolution stages: utilizing Start a 30-bit integer Logistic chaotic system to iteratively generate highly random pseudo-random sequences. Upon verification, the sequence output on the PC and DSP sides is consistent.

[0065] Compression and encryption stage: A 3-level Haar DWT transform is performed on the original image to extract sparse coefficients. Subsequently, the sequence... Constructing a dynamic Gaussian measurement matrix Calculate the observed values This process reduces the amount of data to 65% of the original image.

[0066] Cascade diffusion stage: Observations Quantize to a byte stream. Utilize sequence. The state value dynamically controls the switching of 128 DNA coding rules, and introduces the mask generated by the previous round of OFB feedback for XOR operation to complete the final obfuscation of data.

[0067] Authentication output stage: Calculate the HMAC signature of the ciphertext data and append it to the end of the data packet to output the final NPZ format file.

[0068] Decryption and verification phase: After verifying the HMAC signature, the receiving end selects the FISTA algorithm for image reconstruction based on the current computing power load.

[0069] Example 4 like Figure 3 In the sensitivity test shown, changing the grayscale value of only one pixel in the original image from 100 to 101 caused a drastic change in the generated chaotic seed by the Sobel+SHA3 mechanism. The correlation coefficient between the two subsequently generated ciphertext streams was close to 0, and both the non-critical pixel change rate (NPCR) and uniform average change intensity (UACI) were close to the theoretical limit, confirming the avalanche effect of the cascade mechanism.

[0070] like Figure 4In the reconstruction performance test shown, using the FAST pseudo-inverse mode at the receiving end, the system can complete a low-precision preview within 3 seconds; switching to FISTA mode, a high-quality image with a peak signal-to-noise ratio (PSNR) greater than 32dB can be obtained within 10 seconds; and using OMP mode, the highest precision reconstruction details can be obtained. This result verifies that the present invention, while ensuring a high level of security, possesses extremely strong engineering adaptability and flexibility.

[0071] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for secure transmission of images from unmanned aerial vehicles (UAVs), characterized in that, Includes the following steps: S1. Dynamic chaotic seed generation: Extract the texture feature statistics and full information hash digest of the original UAV image, and perform mixed operation on the texture feature statistics, hash digest and preset master key to generate a dynamic chaotic seed for initializing the chaotic system. S2. Establishment of Integer Chaotic Basis: The dynamic chaotic seed is used as the initial value and input into the finite field fixed-point integer chaotic mapping system. A pseudo-random sequence is generated through an iterative process without floating-point operations. S3. Chaos-driven compressed sensing encryption uses pseudo-random sequences to construct a measurement matrix that dynamically changes with the image content. Compressed sensing linear projection observation is performed on the sparse coefficients of the image after sparse transformation to obtain compressed observation data. S4, Cascaded Diffusion Obfuscation Encryption, quantizes compressed observation data into an integer data stream. Based on the real-time state value of the chaotic system, it dynamically selects encoding rules from a pre-set DNA encoding rule pool to encode the data stream. At the same time, it introduces a mask generated by the output feedback OFB mode to XOR obfuscate the encoded data to obtain the final ciphertext.

2. The method for secure transmission of UAV images according to claim 1, characterized in that, In step S1, the texture feature statistics are extracted from the image derived data using edge or texture operators, and the hash digest is calculated from the image derived data using a cryptographic hash algorithm. The texture feature statistics, hash digest, and preset key material are mixed and operated to generate a dynamic chaotic seed or to update the initial chaotic value and / or control parameters.

3. The method for secure transmission of UAV images according to claim 1, characterized in that, In step S2, the finite field fixed-point integer chaotic mapping system is a preset bit-width integer field improved Logistic chaotic mapping system. Its state evolution equation includes nonlinear iterative terms and linear XOR perturbation terms. Through modular arithmetic, the state evolution trajectory of the chaotic system is strictly constrained within the preset bit-width integer field, ensuring the bit-level output consistency of the chaotic sequence on the heterogeneous hardware platform and eliminating the floating-point precision drift problem.

4. The method for secure transmission of UAV images according to claim 1, characterized in that, In step S3, the sparse transformation adopts the 3rd level Haar discrete wavelet transform (DWT) to decompose the original UAV image at multiple scales, concentrate the image energy in the low-frequency subband, and preserve the high-frequency edge details to construct a sparse representation of the image. Compressed sensing linear projection observation is performed synchronously on the high-frequency and low-frequency subband coefficients after the DWT transformation.

5. A method for secure transmission of UAV images according to claim 1, characterized in that, In step S3, the construction process of the dynamic measurement matrix is ​​as follows: the integer pseudo-random sequence generated in step S2 is mapped through a Gaussian distribution to generate a measurement matrix that is strongly correlated with the content of the current frame image; the measurement matrix is ​​updated in real time with the texture features and hash features of each frame of UAV image, and serves as a one-time key for the compressed sensing process, realizing the deep integration of compression and encryption.

6. The method for secure transmission of UAV images according to claim 1, characterized in that, In step S4, the preset DNA coding rule pool contains at least 128 coding rules based on base pairing and base operation. The selection result of the current coding rule is uniquely determined by the real-time state value of the chaotic system. The key space is expanded by dynamically switching coding rules to resist exhaustive attacks.

7. The method for secure transmission of UAV images according to claim 1, characterized in that, In step S4, the specific execution method of the output feedback OFB mode is as follows: the output feedback information of the previous round is used to participate in the current round mask generation / parameter update, participate in the generation of the current round XOR mask, establish a chain diffusion relationship between ciphertext data blocks, so that the small changes in plaintext or key are diffused to all ciphertext data, and smooth out the original statistical features of the image.

8. A method for secure transmission of UAV images according to claim 1, characterized in that, Also includes: S5. Ciphertext integrity verification: After obtaining the final ciphertext in step S4, calculate the message authentication code of the ciphertext data packet and encapsulate it together with the ciphertext and the parameters required for recovery. Before decryption, the receiving end performs message authentication code verification. If the verification fails, the data packet is discarded directly without performing subsequent decryption and reconstruction operations.

9. A method for secure transmission of UAV images according to claim 8, characterized in that, Also includes: S6. Adaptive decryption and reconstruction: The receiving end selects a reconstruction algorithm to perform decryption and reconstruction based on a preset mode, user selection, or application scenario requirements.

10. A method for secure transmission of UAV images according to claim 9, characterized in that, The reconstruction algorithms include three categories: the pseudo-inverse matrix FAST algorithm for real-time fast preview of the scene, the fast iterative shrinking threshold FISTA algorithm for standard transmission scenarios, and the orthogonal matching pursuit OMP algorithm for post-processing high-precision image analysis scenarios. The receiving end achieves a dynamic balance between image reconstruction quality and processing latency by switching algorithms.

Images