Aerial multi-image encryption method based on improved chaotic mapping and regional adaptive encryption

Abstract

This invention discloses an aerial multi-image encryption method based on improved chaotic mapping and regional adaptive encryption, comprising the following steps: acquiring multiple images of arbitrary size in parallel; concatenating and connecting the hash values ​​of all images, converting the composite hash values ​​into a normalized set of numerical parameters through dynamic prime number generation and modulo operation methods to generate a key, and iterating a 2D-HCSCS chaotic system to generate a driving sequence; performing region-adaptive partitioning on the input multiple images of arbitrary size, separating them into foreground target regions and background regions, and then preprocessing and encrypting the separated foreground target regions and background regions; structurally integrating the encrypted foreground target region data stream and the encrypted background region data stream, and reconstructing the one-dimensional encryption sequence into a two-dimensional image matrix with complete spatial topology, finally generating the ciphertext image. The statistical characteristics of the ciphertext of this invention are highly consistent with random distribution; it has excellent resistance to salt-and-pepper noise and other noise.

Description

Technical Field

[0001] This invention belongs to the field of encryption technology and relates to an aerial multi-image encryption method based on improved chaotic mapping and regional adaptive encryption. Background Technology

[0002] In recent years, image encryption algorithms based on chaotic systems have been extensively studied. Although many schemes have been proposed, some core aspects still have limitations.

[0003] Existing chaotic image encryption schemes mostly rely on traditional low-dimensional or complex high-dimensional chaotic systems. The former generally suffers from problems such as limited parameter space and susceptibility to periodic windows, resulting in insufficient key space; the latter is too computationally complex to meet the efficiency requirements of practical applications.

[0004] In multi-image encryption scenarios, current methods typically require the images to be processed to have a fixed or consistent size. This limitation reduces the algorithm's adaptability to images from different sources and at different resolutions, and also restricts its application in concurrent encryption scenarios of multi-source heterogeneous image data.

[0005] Existing encryption technologies suffer from several problems: traditional low-dimensional chaotic systems have narrow parameter ranges and insufficient security due to periodic windows; high-dimensional chaotic systems are computationally complex and difficult to implement; multi-image encryption suffers from efficiency degradation and security vulnerabilities due to the need for extensive padding caused by differences in size and number; traditional static key mechanisms are susceptible to known-plaintext and chosen-plaintext attacks; and traditional single scrambling or diffusion processes are unable to simultaneously and efficiently destroy image spatial redundancy, channel correlation, and pixel value statistical regularities, resulting in a comprehensive security problem. Summary of the Invention

[0006] To address the issues of narrow parameter range and insufficient security due to periodic windows in traditional low-dimensional chaotic systems, as well as the computational complexity and implementation difficulties of high-dimensional chaotic systems; the efficiency degradation and security vulnerabilities caused by the large amount of padding required due to size differences in multi-image encryption; the vulnerability of traditional static key mechanisms to known-plaintext and chosen-plaintext attacks; and the comprehensive security problem that traditional single scrambling or diffusion processes cannot simultaneously and efficiently destroy image spatial redundancy, channel correlation, and pixel value statistical regularities, this invention adopts the following technical solution: an aerial multi-image encryption method based on improved chaotic mapping and regional adaptive encryption, comprising the following steps:

[0007] S1: Acquire multiple images of arbitrary size in parallel; S2: By using dynamic prime number generation and modular arithmetic methods, the unique hash value of each image is converted into a composite hash value and then into a normalized set of numerical parameters to generate a key; S3: Based on the generated key, iterate the 2D-HCSCS chaotic system to generate an index sequence; S4: Perform region-adaptive segmentation on multiple input images of arbitrary size, separating them into foreground target regions and background regions, and then preprocess the separated foreground target regions and background regions. S5: Based on the chaotic sequence and the index sequence after the chaotic sequence is sorted, the preprocessed foreground target region is subjected to multi-level structured processing, multi-dimensional chaotic linkage block scrambling algorithm, spatial position recovery operation, chaotic-driven pixel rearrangement scrambling with adaptive Z-order spatial filling curve and concentric ring double mapping diffusion algorithm to achieve encryption of the foreground target region; The preprocessed background region is first subjected to lossless compression using Huffman coding, and then a lightweight chaotic operation is performed to encrypt the background region. S6: Based on the geometric constraints and region mapping relationships of multiple original images of arbitrary size, the encrypted foreground target region data stream and the encrypted background region data stream are structurally integrated. By reconstructing the one-dimensional encryption sequence into a two-dimensional image matrix with complete spatial topology, the ciphertext image is finally generated.

[0008] Furthermore: the method of generating dynamic prime numbers and modulo operations converts the unique hash value of each image into a composite hash value and then into a normalized set of numerical parameters. The key generation process is as follows: The binary representation of the composite hash value is evenly divided into four segments of equal length; Each segment is mapped to a large prime number within a specific range using a dynamic prime number generation algorithm; By using the large integers corresponding to the composite hash values, modulo operations and normalization are performed on the prime numbers mapped to each field to generate the four initial parameters required for the chaotic system.

[0009] Furthermore: the process of performing region-adaptive segmentation on multiple input images of arbitrary size, separating them into foreground target regions and background regions, and then preprocessing the separated foreground target regions and background regions is as follows: Multiple images to be processed are loaded in parallel; A pre-trained YOLO v8 detection network was used to perform multi-scale feature extraction and instance segmentation on each image; the identified semantically salient target regions were defined as foreground target regions to be encrypted, while the remaining non-target regions in the image were classified as background regions. The preprocessing process for the foreground target region is as follows: while extracting the target image patch, its original spatial coordinate metadata is fully recorded, and the target regions extracted from all images are integrated into a unified global cache queue; The preprocessing process for the background region is as follows: a spatial reorganization algorithm based on a column-first scanning strategy is adopted to rearrange the background pixels in each image into a continuous and regular rectangular data array according to the original two-dimensional spatial adjacency relationship of the discretely distributed background pixels.

[0010] Furthermore, the process of the multi-level structured preprocessing is as follows: Modal determination and data standardization are performed on the preprocessed foreground target region image, and color space distribution characteristics are identified and converted into 8-bit unsigned integer format. Based on the image in integer format, a fixed-size grid segmentation strategy is used to segment the image into blocks. The grayscale image is directly segmented into a two-dimensional grid, while the color image is processed using a channel-independent mechanism, which performs independent segmentation operations on the R, G, and B color channels to obtain all image blocks. After all image blocks are segmented, they are integrated into a unified global data storage pool to form a block set data matrix with structured features.

[0011] Furthermore, the segmentation operation is as follows: the spatial dimension features of each image are calculated using a dynamic size detection algorithm, and boundary expansion and filling methods are used to ensure that all generated image blocks have uniform geometric specifications; at the same time, a complete block metadata management system is constructed to record the multidimensional traceable information of each image block, including the unique identifier index of the source image, the channel's encoded identifier, the row and column coordinates in the original image space, and the standardized size parameters of the block.

[0012] Furthermore: The foreground target region is divided into image blocks, and a multi-dimensional chaotic linked block scrambling algorithm is executed to achieve global reorganization across image blocks. The specific process is as follows: All image blocks in the foreground target region are integrated into a unified two-dimensional grid topology, and a zero-value filling strategy is used to normalize the grid into rectangles; Based on the rectangularized image, multiple sets of dynamic index sequences generated by a chaotic system are used to drive the grid to perform multi-level spatial cyclic permutation operations for global permutation, specifically including cyclic permutations in the row, column, and diagonal directions; the permutation process is completely guided by the randomness of the index sequences. After the global replacement is completed, the pre-established block metadata tracking mechanism is used to accurately remap the replaced block units to the topological framework of each original image, thereby achieving global reorganization across image blocks.

[0013] Furthermore: After globally reconstructing the foreground target region, a spatial location restoration operation is performed on the image. The specific process is as follows: A reverse engineering strategy is used to reassemble the globally arranged block units into a complete image topology; For grayscale images, two-dimensional reconstruction is performed based on the original grid layout; For color images, a channel decoupling reconstruction strategy is adopted to independently reconstruct the block matrix of each color channel, and then the reconstruction results of each channel are merged into a multi-channel image through multi-channel fusion technology.

[0014] Furthermore: the image after the foreground target region is recovered is then scrambled based on the chaotic-driven adaptive Z-order space-filling curve pixel rearrangement, the specific process of which is as follows: Based on the MC size features of the image after the foreground target region is restored, an adaptive Z-order scanning sequence is dynamically generated: the Morton code value of each pixel in the core region of the image is calculated by bit cross-coding, and an accurate mapping from spatial coordinates to a one-dimensional sequence is established. For irregular boundary pixels that exceed the core region, a row-first traversal supplementary scanning strategy is adopted to construct a serial representation covering the entire image domain. Extract segments of the same length as the scan sequence from the index sequence, construct a permutation mapping function through sorting operations, and perform a nonlinear rearrangement operation based on a chaotic driving process on the serial pixel vector to achieve global randomization of pixel positions; Based on the inverse mapping relationship of the original scan sequence, the scrambled one-dimensional pixel sequence is reconstructed into a two-dimensional image matrix ZC, thus obtaining the rearranged and scrambled image.

[0015] Furthermore: Based on the rearranged and scrambled foreground target region image, the concentric ring double mapping diffusion algorithm is used to generate the ciphertext image FE of the foreground target region. The specific process is as follows: First: Construct a complete circular pixel sequence starting from the outermost boundary of the image; Then, a pixel-by-pixel XOR operation is performed on the quantized value of the chaotic sequence and the corresponding ring pixel to achieve initial chaotic modulation. For the inner ring pixels, a dual-channel selective diffusion mechanism is adopted: two outer ring pixel index mappings are constructed in parallel: the first mapping forms an ordered mapping based on the statistical median segmentation principle, and the second mapping generates a random mapping using hash pseudo-random permutation. The mapping method is dynamically selected according to the chaotic bit sequence, and the selected outer ring pixel is XORed with the inner ring pixel. This process iterates layer by layer inward, and the pixel value processed in each layer is immediately used for the diffusion calculation of the next inner ring, forming a chain propagation effect from the outside to the inside.

[0016] Furthermore, the process of first performing Huffman coding lossless compression on the preprocessed background region, and then implementing lightweight chaotic operations to encrypt the background region is as follows: Perform channel separation and image type determination on the preprocessed background image; Then, adaptive Huffman coding compression is performed, the statistical distribution of pixel values ​​in each channel is calculated, the optimal prefix coding dictionary is constructed, the image data is converted into a variable-length coded bitstream, and the coding length feature of each pixel unit is recorded synchronously during the compression process; a chaotic preprocessing mechanism is introduced into the coding layer, and the coded bitstream is nonlinearly interleaved and rearranged using a pre-generated chaotic index sequence; Then, a dynamic filling algorithm based on adjacent pixel copying is used to normalize the data length of each channel in order to achieve a synchronous processing standard. Subsequently, a dual encryption operation is performed: the first stage of encryption extracts a scrambled index from the index sequence based on the composite parameters of the image and channels, and spatially scrambles the compressed data; the second stage of encryption quantizes the chaotic sequence into integer values ​​between 0 and 255 after spatial scrambling, and performs pixel value diffusion encryption through bitwise XOR operation.

[0017] This invention provides an aerial multi-image encryption method based on improved chaotic mapping and regional adaptive encryption. It proposes a new improved chaotic system 2D-HCSCS, which solves the problems of narrow parameter range and insufficient security due to periodic windows in traditional low-dimensional chaotic systems, as well as the computational complexity and implementation difficulties of high-dimensional chaotic systems. It provides a pseudo-random sequence source with excellent dynamic characteristics and practicality for image encryption.

[0018] A cross-plane pixel scrambling mechanism is proposed to address the efficiency degradation and security risks caused by the need for extensive padding due to size and quantity differences in multi-image encryption.

[0019] A method combining the SHA-512 hash algorithm with plaintext image features is proposed, which solves the problem that traditional static key mechanisms are vulnerable to known plaintext and chosen plaintext attacks. It realizes dynamic key generation of multi-image joint keys, making the key space strongly correlated with the plaintext.

[0020] A two-stage encryption mechanism consisting of "adaptive block scrambling + three-channel hybrid scrambling + dynamic selective diffusion" is proposed, which solves the comprehensive security problem that traditional single scrambling or diffusion processes cannot simultaneously and efficiently destroy image spatial redundancy, channel correlation and pixel value statistical regularity.

[0021] Multiple encrypted aerial images can be losslessly recovered into plaintext data identical to the original images through the corresponding decryption process. We conducted systematic security tests and performance evaluations on the encryption algorithm: In terms of key analysis, we verified that the dynamic key generated based on SHA-512 and plaintext features has a large key space and exhibits high sensitivity to the content of the aerial images and initial parameters; in terms of statistical properties, we comprehensively tested the histogram, spatial correlation, NPCR and UACI, and information entropy of the ciphertext images, and the results showed that the statistical properties of the ciphertext are highly consistent with the random distribution; in the robustness test, the algorithm showed excellent resistance to salt-and-pepper noise and different cropping ratios, verifying its reliability and practicality in the secure transmission and storage of aerial images. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a flowchart of the method described in this application; Figure 2 This is the key generation method of this method; Figure 3 It is a multidimensional chaotic linkage block scrambling process; Figure 4 This is a diagram of the chaotic pixel rearrangement scrambling process driven by the adaptive Z-order space-filling curve. Figure 5 It is a concentric ring double-mapping diffusion process; Figure 6 These are histograms of the images before and after encryption; where (a) is the plaintext image histogram and (b) is the ciphertext image histogram. Detailed Implementation

[0024] It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of the present invention can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and embodiments.

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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. The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the present invention or its application or use. 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.

[0026] Figure 1 This is a flowchart of the method described in this application; A method for encrypting multiple aerial images based on improved chaotic mapping and regional adaptive encryption includes the following steps: S1: Acquire multiple images of arbitrary size in parallel; S2: Generate a unique hash value for each image using the SHA-512 hash function, concatenate the hash values ​​of all images, and perform a second SHA-512 operation on the concatenated result to generate a composite hash value; convert the composite hash value into a normalized set of numerical parameters using dynamic prime number generation and modular arithmetic methods to generate a key, which serves as the initial value and control parameter required by the chaotic system. S3: Based on the generated key, iterate the 2D-HCSCS chaotic system to generate a chaotic sequence;

[0027] in: and It is the system's state variable. and These are the system's control parameters.

[0028] S4: Perform region-adaptive segmentation on multiple input images of arbitrary size, separating them into foreground target regions and background regions, and then preprocess the separated foreground target regions and background regions. S5: Based on the chaotic sequence and the index sequence after the chaotic sequence is sorted, the preprocessed foreground target region is subjected to multi-level structured processing, multi-dimensional chaotic linkage block scrambling algorithm, spatial position recovery operation, chaotic-driven pixel rearrangement scrambling with adaptive Z-order (Z-order) space filling curve and concentric ring double mapping diffusion algorithm to realize the encryption of the foreground target region; The preprocessed background region is first subjected to lossless compression using Huffman coding, and then a lightweight chaotic operation is performed to encrypt the background region. S6: Based on the geometric constraints and region mapping relationships of multiple original images of arbitrary size, the encrypted foreground target region data stream and the encrypted background region data stream are structurally integrated. By reconstructing the one-dimensional encryption sequence into a two-dimensional image matrix with complete spatial topology, the ciphertext image is finally generated.

[0029] The steps S1 / S2 / S3 / S4 / S5 / S6 are executed sequentially; Figure 2 This is the key generation method of this method; The expression for the 2D-HCSCS chaotic system is as follows:

[0030] in: and It is the system's state variable. and These are the system's control parameters.

[0031] The key generation process, which involves converting the composite hash value into a normalized set of numerical parameters using dynamic prime number generation and modular arithmetic, is as follows: The binary representation of the composite hash value is evenly divided into four segments of equal length; Each segment is mapped to a large prime number within a specific range using a dynamic prime number generation algorithm; By using the large integers corresponding to the composite hash values, and simultaneously performing modulo operations and normalization on the prime numbers mapped to each field, four initial values ​​required for the chaotic system are generated: , , , These initial values ​​will be used to drive the iteration of the chaotic system, generating the chaotic sequence required for encryption.

[0032] Furthermore, the process of performing region-adaptive segmentation on multiple input images of arbitrary size, separating them into foreground target regions and background regions, and then preprocessing the separated foreground target regions and background regions is as follows: Based on parallel loading of multiple images to be processed; A pre-trained YOLO v8 detection network was used to perform multi-scale feature extraction and instance segmentation on each image; the identified semantically salient target regions were defined as foreground target regions to be encrypted, while the remaining non-target regions in the image were classified as background regions. The preprocessing process for the foreground target region is as follows: while extracting the target image patch, its original spatial coordinate metadata is fully recorded, and the target regions extracted from all images are integrated into a unified global cache queue; The preprocessing process for the background region is as follows: A spatial reorganization algorithm based on a column-first scanning strategy is adopted. According to the original two-dimensional spatial adjacency relationship of the discretely distributed background pixels in each image, the discretely distributed background pixels in each image are rearranged into a continuous and regular rectangular data array.

[0033] Furthermore, for the foreground target region F, perform multi-level structured preprocessing operations: First, modality determination and data standardization are performed on the foreground target area image, and color space distribution characteristics are identified and converted into 8-bit unsigned integer format. Then, a fixed-size grid segmentation strategy is adopted: grayscale images are directly segmented into two-dimensional grids, while color images are processed using a channel-independent mechanism, which performs independent segmentation operations on the R, G, and B color channels to generate image blocks. After all image blocks have been processed, they are integrated into a unified global data storage pool to form a block set data matrix with structured features, providing a standardized data infrastructure for subsequent encryption operations.

[0034] The segmentation process is as follows: the spatial dimension features of each image are calculated using a dynamic size detection algorithm, and boundary expansion and filling methods are used to ensure that all generated image blocks have uniform geometric specifications; at the same time, a complete block metadata management system is constructed to record the multidimensional traceable information of each image block, including the unique identifier index of the source image, the channel encoding identifier, the row and column coordinates in the original image space, and the standardized size parameters of the block.

[0035] Figure 3 It is a multidimensional chaotic linkage block scrambling process; Furthermore: The foreground target region is divided into image blocks, and a multi-dimensional chaotic linked block scrambling algorithm is executed to achieve global reorganization across image blocks. The specific process is as follows: All image blocks in the foreground target region are integrated into a unified two-dimensional grid topology, and a zero-value filling strategy is used to normalize the grid into rectangles; Based on the rectangularized image, an index sequence generated by a chaotic system is used to drive the grid to perform multi-level spatial cyclic permutation operations for global permutation, specifically including cyclic permutations in the row, column, and diagonal directions; the permutation process is completely guided by the randomness of the index sequence. After the global replacement is completed, the pre-established block metadata tracking mechanism is used to accurately remap the replaced block units to the topological framework of each original image, thereby achieving global reorganization across image blocks.

[0036] Furthermore: After globally reconstructing the foreground target region, a spatial location restoration operation is performed on the image. The specific process is as follows: A reverse engineering strategy is used to reassemble the globally arranged block units into a complete image topology; For grayscale images, two-dimensional reconstruction is performed based on the original grid layout; For color images, a channel decoupling reconstruction strategy is adopted to independently reconstruct the block matrix of each color channel, and then the reconstruction results of each channel are merged into a multi-channel image through multi-channel fusion technology.

[0037] Figure 4 This is a diagram of the chaotic pixel rearrangement scrambling process driven by the adaptive Z-order space-filling curve. Furthermore: the image after the foreground target region is recovered is then scrambled based on the chaotic-driven adaptive Z-order space-filling curve pixel rearrangement, the specific process of which is as follows: Based on the MC size features of the image after the foreground target region is restored, an adaptive Z-order scanning sequence is dynamically generated: the Morton code value of each pixel in the core region of the image is calculated by bit cross-coding, and an accurate mapping from spatial coordinates to a one-dimensional sequence is established. For irregular boundary pixels that exceed the core region, a row-first traversal supplementary scanning strategy is adopted to construct a serial representation covering the entire image domain. Extract segments of the same length as the scan sequence from the index sequence, construct a permutation mapping function through sorting operations, and perform a nonlinear rearrangement operation based on a chaotic driving process on the serial pixel vector to achieve global randomization of pixel positions; Based on the inverse mapping relationship of the original scan sequence, the scrambled one-dimensional pixel sequence is reconstructed into a two-dimensional image matrix ZC, thus obtaining the rearranged and scrambled image.

[0038] Figure 5 It is a concentric ring double-mapping diffusion process; Furthermore: Based on the rearranged and scrambled foreground target region image, the concentric ring double mapping diffusion algorithm is used to generate the ciphertext image FE of the foreground target region. The specific process is as follows: First: Construct a complete circular pixel sequence starting from the outermost boundary of the image; Then, a pixel-by-pixel XOR operation is performed on the quantized value of the chaotic sequence and the corresponding ring pixel to achieve initial chaotic modulation. For the inner ring (inner circle pixels), a dual-channel selective diffusion mechanism is adopted: two outer ring pixel index mappings are constructed in parallel: the first mapping forms an ordered mapping based on the statistical median segmentation principle, and the second mapping generates a random mapping using hash pseudo-random permutation. The mapping method is dynamically selected according to the chaotic bit sequence, and the selected outer ring pixel is XORed with the inner ring pixel. This process iterates layer by layer inward, and the pixel value processed in each layer is immediately used for the diffusion calculation of the next inner ring, forming a chain propagation effect from the outside to the inside.

[0039] Furthermore, the process of first performing Huffman coding lossless compression on the preprocessed background region, and then implementing lightweight chaotic operations to encrypt the background region is as follows: Perform channel separation and image type determination (determine whether it is a grayscale or color image) on the preprocessed background image; Then, adaptive Huffman coding compression is performed, the statistical distribution of pixel values ​​in each channel is calculated, the optimal prefix coding dictionary is constructed, the image data is converted into a variable-length coded bitstream, and the coding length feature of each pixel unit is recorded synchronously during the compression process; a chaotic preprocessing mechanism is introduced into the coding layer, and the coded bitstream is non-linearly interleaved and rearranged using a pre-generated index sequence; Then, a dynamic filling algorithm based on adjacent pixel copying is used to normalize the data length of each channel in order to achieve a synchronous processing standard. Subsequently, a dual encryption operation is performed: the first stage of encryption extracts a scrambled index from the index sequence based on the composite parameters of the image and channels, and spatially scrambles the compressed data; the second stage of encryption quantizes the chaotic sequence into integer values ​​between 0 and 255 after spatial scrambling, and performs pixel value diffusion encryption through bitwise XOR operation.

[0040] Example 1: A method for encrypting multiple aerial images based on improved chaotic mapping and regional adaptive encryption includes the following steps: S1: Acquire multiple images of arbitrary size in parallel; S2: Generate a unique hash value for each image using the SHA-512 hash function, concatenate the hash values ​​of all images, and perform a second SHA-512 operation on the concatenated result to generate a composite hash value; convert the composite hash value into a normalized set of numerical parameters using dynamic prime number generation and modular arithmetic methods to generate a key, which serves as the initial value and control parameter required by the chaotic system. S3: Based on the generated key, iterate the 2D-HCSCS chaotic system to generate a chaotic sequence; S4: Perform region-adaptive segmentation on multiple input images of arbitrary size, separating them into foreground target regions and background regions, and then preprocess the separated foreground target regions and background regions. S5: Based on the chaotic sequence and the index sequence after the chaotic sequence is sorted, the preprocessed foreground target region is subjected to multi-level structured processing, multi-dimensional chaotic linkage block scrambling algorithm, spatial position recovery operation, chaotic-driven pixel rearrangement scrambling with adaptive Z-order (Z-order) space filling curve and concentric ring double mapping diffusion algorithm to realize the encryption of the foreground target region; The preprocessed background region is first subjected to lossless compression using Huffman coding, and then a lightweight chaotic operation is performed to encrypt the background region. S6: Based on the geometric constraints and region mapping relationships of multiple original images of arbitrary size, the encrypted foreground target region data stream and the encrypted background region data stream are structurally integrated. By reconstructing the one-dimensional encryption sequence into a two-dimensional image matrix with complete spatial topology, the ciphertext image is finally generated.

[0041] The steps S1 / S2 / S3 / S4 / S5 / S6 are executed sequentially; The key generation process, which involves converting the composite hash value into a normalized set of numerical parameters using dynamic prime number generation and modular arithmetic, is as follows: S21: Divide the binary representation of the composite hash value into four equal segments; S22: Each sub-segment is mapped to a large prime number within a specific range using a dynamic prime number generation algorithm; The specific range refers to the prime number candidate value falling within the interval [50000, 1049999]. S23: Using the large integers corresponding to the composite hash values, perform modulo operations and normalization on the prime numbers mapped to each field simultaneously to generate the four initial values ​​required for the chaotic system: , , , These initial values ​​will be used to drive the iteration of the chaotic system, generating the chaotic sequence required for encryption.

[0042] The formula for generating the key is as follows:

[0043]

[0044] Where: H represents the decimal integer corresponding to the hash value. This represents a dynamically generated prime number parameter. The function represents a conversion function that maps large integers to a specific range of values. M represents a conversion function that maps large integers to a specific range of values. nextprime represents a function that returns the smallest prime number that is greater than or equal to the input value.

[0045] Furthermore, the process of performing region-adaptive segmentation on multiple input images of arbitrary size, separating them into foreground target regions and background regions, and then preprocessing the separated foreground target regions and background regions is as follows: S41: Based on parallel loading of multiple images to be processed; S42: Use a pre-trained YOLO v8 detection network to perform multi-scale feature extraction and instance segmentation on each image; the identified semantically salient target regions are defined as foreground target regions to be encrypted, and the remaining non-target regions in the image are classified as background regions; S43: The preprocessing process for the foreground target region is as follows: while extracting the target image patch, its original spatial coordinate metadata is fully recorded, and the target regions extracted from all images are integrated into a unified global cache queue; The preprocessing process for the background region is as follows: A spatial reorganization algorithm based on a column-first scanning strategy is adopted. According to the original two-dimensional spatial adjacency relationship of the discretely distributed background pixels in each image, the discretely distributed background pixels in each image are rearranged into a continuous and regular rectangular data array.

[0046] Based on the chaotic sequence and the index sequence after sorting the chaotic sequence, the preprocessed foreground target region is subjected to multi-level structuring processing, multi-dimensional chaotic linkage block scrambling algorithm, spatial position recovery operation, chaotic-driven pixel rearrangement scrambling with adaptive Z-order (Z-order) space filling curve, and concentric ring double mapping diffusion algorithm to achieve encryption of the foreground target region, including: S51: For the foreground target region F, perform multi-level structured preprocessing operations; S52: After dividing the foreground target region into image blocks, multidimensional chaotic link block replacement is performed based on the multidimensional chaotic link block replacement algorithm to achieve global recombination across image blocks; S53: Perform spatial location restoration on the globally reconstructed image of the foreground target region. The specific process is as follows; S54: The image after the foreground target region is restored is then scrambled using a chaotic-driven adaptive Z-order space-filling curve pixel rearrangement. S55: Based on the rearranged and scrambled foreground target region image, a concentric ring double mapping diffusion algorithm is used to generate a ciphertext image of the foreground target region.

[0047] Furthermore, for the foreground target region F, perform multi-level structured preprocessing operations: S511: First, perform modal determination and data standardization on the foreground target area image, identify the color space distribution characteristics and convert them into an 8-bit unsigned integer format; S512: Then, a fixed-size grid segmentation strategy is adopted: grayscale images are directly segmented into two-dimensional grids, while color images are processed using a channel-independent mechanism, which performs independent segmentation operations on the three color channels R, G, and B to generate image blocks. S513: After completing the block processing of all images, integrate all image blocks into a unified global data storage pool to form a block set data matrix with structured features, providing a standardized data infrastructure for subsequent encryption operations.

[0048] The segmentation operation process is as follows: The spatial dimension features of each image are calculated using a dynamic size detection algorithm, and boundary expansion and filling methods are used to ensure that all generated image blocks have uniform geometric specifications. At the same time, a complete block metadata management system was built to record multidimensional traceable information for each image block.

[0049] The multidimensional traceability information includes the unique identifier index of the source image, the coded identifier of the channel, the row and column coordinates in the original image space, and the standardized size parameters of the block.

[0050] Furthermore: After dividing the foreground target region into image blocks, a multidimensional chaotic linked block replacement algorithm is used to perform multidimensional chaotic linked block replacement, achieving global reorganization across image blocks. The specific process is as follows: S521: First, all image blocks generated in the preprocessing stage of step S2 are integrated into a unified two-dimensional grid topology, and the grid is normalized into rectangles using a zero-value filling strategy. S522: Subsequently, based on multiple sets of dynamic index sequences (sequences after driving sequence sorting) generated by the chaotic system, the grid is driven to perform multi-level spatial cyclic permutation operations, specifically including cyclic permutations in the row, column and diagonal directions; the permutation process is completely guided by the randomness of the index sequence, ensuring the unpredictability of the permutation pattern; S523: After completing the global replacement, the pre-established block metadata tracking mechanism (covering multi-dimensional identity information such as source image index, channel identifier and spatial coordinates) is used to accurately remap the replaced block units to the topological framework of each original image, thereby realizing global reorganization across image blocks.

[0051] The pseudocode is as follows: Input: F 1. Flatten the block and record information 2. Reshape into a 2D mesh and fill accordingly 3. for each row r = 1 to grid_rows do / / Row Circular Shift 4. Shift ← [r] mod grid_cols 5. if [r] is odd then 6. Shift ← -ShiftEnd if 7.Circularly shift row r of big_grid by shift positions 8 end for 9 for each column c = 1 to grid_cols do / / Column Circular Shift 10.shift ← [grid_rows + c] mod grid_rows 11.if [grid_rows + c] is odd then 12. Shift ← -ShiftEnd if 13.Circularly shift column c of big_grid by shift positions 14. end for 15. for each diagonal offset k do / / Main Diagonal Circular Shift 16.Extract main diagonal Dk; shift ← [k] mod length (Dk) 17.Circularly shift Dk along diagonal direction 18. end for 19. for each anti-diagonal offset k do 20.Extract anti-diagonal ADk; shift ← [k] mod length (ADk) 21.Circularly shift ADk along diagonal direction 22. end for 23. Remove fill and reshape image Output: MC Furthermore: Spatial location restoration is performed on the globally reconstructed image of the foreground target region. The specific process is as follows: A reverse engineering strategy is used to reassemble the globally arranged block units into a complete image topology; For grayscale images, two-dimensional reconstruction is performed based on the original grid layout; For color images, a channel decoupling reconstruction strategy is adopted to independently reconstruct the block matrix of each color channel, and then the reconstruction results of each channel are merged into a multi-channel image through multi-channel fusion technology. During the reconstruction process, a dynamic boundary redundancy cropping algorithm was implemented to restore the original geometric dimensions of the image, so as to accurately remove the edge padding pixels introduced in the block segmentation stage, thereby meeting the block size multiple requirement, restoring the original geometric dimensions and topological integrity of the image, and finally generating the image multichannel (MC).

[0052] Furthermore: For the image after the foreground target region is restored, pixel rearrangement and scrambling based on the chaotic-driven adaptive Z-order space-filling curve is performed. The specific process is as follows: S541: First, an adaptive Z-order scanning sequence is dynamically generated based on the MC size features of the image after the foreground target region is restored: the Morton code value of each pixel in the core region (maximum square region) of the image is calculated using bit cross-coding to establish an accurate mapping from spatial coordinates to a one-dimensional sequence; for irregular boundary pixels beyond the core region, a row-first traversal supplementary scanning strategy is adopted to construct a serial representation covering the entire image domain. S542: Subsequently, segments of equal length to the scan sequence are extracted from the index sequence, a permutation mapping function is constructed through sorting operations, and a nonlinear rearrangement operation based on a chaotic driving process is performed on the serial pixel vector to achieve global randomization of pixel positions. S543: Finally, based on the inverse mapping relationship of the original scan sequence, the scrambled one-dimensional pixel sequence is reconstructed into a two-dimensional image matrix ZC. The pseudocode is as follows: Input: MC 1. Image Size Initialization; Obtain image height H and width W fromchannel 2. Let N = 2^ log2(min(H, W)) / / Adaptive Z-Order Scan Generation 3. if N<1 then 4.scan_order ← linear indices from 1 to H×W 5.else 6.Generate coordinate grids X, Y over [0, N 1] 7.for each pixel coordinate (x, y) in the N × N region do 8.Compute Morton code by interleaving bits of x and y 9. end for 10.Sort Morton codes in ascending order 11. Obtain scan_order for the N × N region accordingly 12.Append remaining pixels outside the N × N region 13. using row-wise scan order 14.end if 15. Extract pixel values ​​of channel according to scan_order / / Z-Order Based Sequence Construction 16. Form a one-dimensional sequence sequence 17. Reorder seq according to permx to obtain seq_shuffled 18. Initialize shuffled image of size H × W with zeros 19. Assign seq_shuffled back to shuffled according to scan_order Output: ZC Furthermore: Based on the rearranged and scrambled foreground target region image, a concentric ring double-mapping diffusion algorithm is used to generate a ciphertext image of the foreground target region. The specific process is as follows: The rearranged and scrambled foreground target region image is regarded as a topological structure composed of multiple concentric rings, and global chaos of pixel values ​​is achieved through a hierarchical progressive mechanism from the outside to the inside. S551: First: Construct a complete circular pixel sequence starting from the outermost boundary of the rearranged and scrambled foreground target region image; S552: Then, a pixel-by-pixel XOR operation is performed on the quantized value of the chaotic sequence and the corresponding ring pixel to achieve initial chaotic modulation. For the inner ring, a dual-channel selective diffusion mechanism is adopted: two outer ring pixel index mapping relationships are constructed in parallel—the first mapping forms an ordered mapping based on the statistical median segmentation principle, and the second mapping generates a random mapping using hash pseudo-random permutation. The mapping method is dynamically selected according to the chaotic bit sequence, and the selected outer ring pixel is XORed with the inner ring pixel. This process iterates layer by layer inward, and the pixel value processed in each layer is immediately used for the diffusion calculation of the next inner ring, forming a chain propagation effect from the outside to the inside.

[0053] This ensures that changes in a single pixel can propagate layer by layer through the circular structure to the entire image area, achieving highly complex global diffusion of pixel values ​​and ultimately generating a ciphertext image (FE). The pseudocode is as follows: Input: ZC 1. while top<= bottom and left<= right do 2.ring_idx = compute_clockwise_indices(top, bottom, left, right) / / Process outer ring 3. if is_outer_ring then 4. len_ring = length(ring_idx) 5.chaos_slice = chaos_seq[chaos_idx : chaos_idx + len_ring - 1] 6. for i = 1 to len_ring do 7.diffused[ring_idx[i]] = diffused[ring_idx[i]]XOR chaos_slice[i] 8. end for 9. chaos_idx += len_ring 10. is_outer_ring = false 11.end if 12. / / Process inner ring 13.inner_top = top + 1; inner_bottom = bottom – 1; inner_left = left+ 1; inner_right = right - 1 14.if inner_top<= inner_bottom and inner_left<= inner_right then 15.inner_ring = compute_clockwise_indices(inner_top, inner_bottom,inner_left, inner_right) 16.if N<2 then 17.Median_Split_Order = ring_idx; Hash_Permutation = ring_idx 18.else 19.outer_values = convert_to_numeric(diffused[ring_idx]) 20.median_val = median(outer_values); above_median = empty list;below_median = empty list 21.for i = 1 to N do 22.if outer_values[i]>= median_val then 23.append ring_idx[i] to above_medianelse 24.append ring_idx[i] to below_medianend if 25.end for 26.Median_Split_Order = concatenate(above_median, below_median) 27.Hash_Permutation = array of size N / / Hash Permutation 28.for i = 1 to N do 29.hash = (outer_values[i] i) mod N + 1; Hash_Permutation[i] = ring_idx[hash]; end for 30.end if 31.for idx = 1 to length(inner_ring) do / / Inner ring diffusion usingchaos bit selection 32.bit_choice = chaos_bit_seq[bit_idx]; bit_idx += 1 33. if bit_choice == 0 then 34.ext_idx = Median_Split_Order[idx]; else 35.ext_idx = Hash_Permutation[idx] 36.end if 37.diffused[inner_ring[idx]] = diffused[inner_ring[idx]]XOR diffused[ext_idx] 38.end for 39.top += 1; bottom -= 1; left += 1; right -= 1; / / Shrink the processing area 40. end while Output: FE Furthermore: For the preprocessed background region B, a multi-level compression and encryption process is performed, the specific process of which is as follows: S561: First, perform channel separation and mode determination on the preprocessed background image; S562: Then, adaptive Huffman coding compression is performed, the statistical distribution of pixel values ​​in each channel is calculated, the optimal prefix coding dictionary is constructed, the image data is converted into a variable-length coded bitstream, and the coding length feature of each pixel unit is recorded synchronously during the compression process; a chaotic preprocessing mechanism is introduced in the coding layer, and the coded bitstream is non-linearly interleaved and rearranged using a pre-generated chaotic index sequence; To address the issue of inconsistent data lengths across multiple channels in color images, a dynamic filling algorithm based on adjacent pixel replication is employed to normalize the data length of each channel, thereby achieving a synchronized processing standard. Subsequently, a double encryption operation is performed: The first stage of encryption extracts a scrambled index from the index sequence based on composite parameters of the image and channels, and spatially perturbs the compressed data; the second stage quantizes the chaotic sequence into integer values ​​between 0 and 255, and performs pixel value diffusion encryption through bitwise XOR operations. The entire encryption process preserves the complete Huffman coding dictionary, compression parameters, and padding metadata, thus ensuring the reversibility of the decryption process.

[0054] Figure 6 These are histograms of the images before and after encryption; where (a) is the plaintext image histogram and (b) is the ciphertext image histogram.

[0055] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; 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 or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for encrypting multiple aerial images based on improved chaotic mapping and regional adaptive encryption, characterized in that: Includes the following steps: S1: Acquire multiple images of arbitrary size in parallel; S2: By using dynamic prime number generation and modular arithmetic methods, the unique hash value of each image is converted into a composite hash value and then into a normalized set of numerical parameters to generate a key; S3: Based on the generated key, iterate the 2D-HCSCS chaotic system to generate a chaotic sequence; S4: Perform region-adaptive segmentation on multiple input images of arbitrary size, separating them into foreground target regions and background regions, and then preprocess the separated foreground target regions and background regions. S5: Based on the chaotic sequence and the index sequence after the chaotic sequence is sorted, the preprocessed foreground target region is subjected to multi-level structured processing, multi-dimensional chaotic linkage block scrambling algorithm, spatial position recovery operation, chaotic-driven pixel rearrangement scrambling with adaptive Z-order spatial filling curve and concentric ring double mapping diffusion algorithm to achieve encryption of the foreground target region; The preprocessed background region is first subjected to lossless compression using Huffman coding, and then a lightweight chaotic operation is performed to encrypt the background region. S6: Based on the geometric constraints and region mapping relationships of multiple original images of arbitrary size, the encrypted foreground target region data stream and the encrypted background region data stream are structurally integrated. By reconstructing the one-dimensional encryption sequence into a two-dimensional image matrix with complete spatial topology, the ciphertext image is finally generated.

2. The aerial multi-image encryption method based on improved chaotic mapping and regional adaptive encryption according to claim 1, characterized in that: The key generation process involves using dynamic prime number generation and modular arithmetic to convert the unique hash value of each image into a composite hash value, which is then converted into a normalized set of numerical parameters. The binary representation of the composite hash value is evenly divided into four segments of equal length; Each segment is mapped to a large prime number within a specific range using a dynamic prime number generation algorithm; By using the large integers corresponding to the composite hash values, modulo operations and normalization are performed on the prime numbers mapped to each field to generate the four initial parameters required for the chaotic system.

3. The aerial multi-image encryption method based on improved chaotic mapping and regional adaptive encryption according to claim 1, characterized in that: The process of performing region-adaptive segmentation on multiple input images of arbitrary size, separating them into foreground target regions and background regions, and then preprocessing the separated foreground target regions and background regions is as follows: Multiple images to be processed are loaded in parallel; A pre-trained YOLO v8 detection network was used to perform multi-scale feature extraction and instance segmentation on each image; the identified semantically salient target regions were defined as foreground target regions to be encrypted, while the remaining non-target regions in the image were classified as background regions. The preprocessing process for the foreground target region is as follows: while extracting the target image patch, its original spatial coordinate metadata is fully recorded, and the target regions extracted from all images are integrated into a unified global cache queue; The preprocessing process for the background region is as follows: a spatial reorganization algorithm based on a column-first scanning strategy is adopted to rearrange the background pixels in each image into a continuous and regular rectangular data array according to the original two-dimensional spatial adjacency relationship of the discretely distributed background pixels.

4. The aerial multi-image encryption method based on improved chaotic mapping and regional adaptive encryption according to claim 1, characterized in that: The multi-level structured preprocessing process is as follows: Modal determination and data standardization are performed on the preprocessed foreground target region image, and color space distribution characteristics are identified and converted into 8-bit unsigned integer format. Based on the image in integer format, a fixed-size grid segmentation strategy is used to segment the image into blocks. The grayscale image is directly segmented into a two-dimensional grid, while the color image is processed using a channel-independent mechanism, which performs independent segmentation operations on the R, G, and B color channels to obtain all image blocks. After all image blocks are segmented, they are integrated into a unified global data storage pool to form a block set data matrix with structured features.

5. The aerial multi-image encryption method based on improved chaotic mapping and regional adaptive encryption according to claim 4, characterized in that: The segmentation process is as follows: the spatial dimension features of each image are calculated using a dynamic size detection algorithm, and boundary expansion and filling methods are used to ensure that all generated image blocks have uniform geometric specifications; at the same time, a complete block metadata management system is constructed to record the multidimensional traceable information of each image block, including the unique identifier index of the source image, the channel encoding identifier, the row and column coordinates in the original image space, and the standardized size parameters of the block.

6. The aerial multi-image encryption method based on improved chaotic mapping and regional adaptive encryption according to claim 1, characterized in that: The foreground target region is divided into image blocks, and a multi-dimensional chaotic linked block scrambling algorithm is executed to achieve global reorganization across image blocks. The specific process is as follows: All image blocks in the foreground target region are integrated into a unified two-dimensional grid topology, and a zero-value filling strategy is used to normalize the grid into rectangles; Based on the rectangularized image, an index sequence generated by a chaotic system is used to drive the grid to perform multi-level spatial cyclic permutation operations for global permutation, specifically including cyclic permutations in the row, column, and diagonal directions; the permutation process is completely guided by the randomness of the index sequence. After the global replacement is completed, the pre-established block metadata tracking mechanism is used to accurately remap the replaced block units to the topological framework of each original image, thereby achieving global reorganization across image blocks.

7. The aerial multi-image encryption method based on improved chaotic mapping and regional adaptive encryption according to claim 1, characterized in that: After globally reconstructing the foreground target region, a spatial location restoration operation is performed on the image. The specific process is as follows: A reverse engineering strategy is used to reassemble the globally arranged block units into a complete image topology; For grayscale images, two-dimensional reconstruction is performed based on the original grid layout; For color images, a channel decoupling reconstruction strategy is adopted to independently reconstruct the block matrix of each color channel, and then the reconstruction results of each channel are merged into a multi-channel image through multi-channel fusion technology.

8. The aerial multi-image encryption method based on improved chaotic mapping and regional adaptive encryption according to claim 1, characterized in that: The image after the foreground target region is restored is then scrambled using a chaotic-driven adaptive Z-order space-filling curve pixel rearrangement process, as follows: Based on the MC size features of the image after the foreground target region is restored, an adaptive Z-order scanning sequence is dynamically generated: the Morton code value of each pixel in the core region of the image is calculated by bit cross-coding, and an accurate mapping from spatial coordinates to a one-dimensional sequence is established. For irregular boundary pixels that exceed the core region, a row-first traversal supplementary scanning strategy is adopted to construct a serial representation covering the entire image domain. Extract segments of the same length as the scan sequence from the index sequence, construct a permutation mapping function through sorting operations, and perform a nonlinear rearrangement operation based on a chaotic driving process on the serial pixel vector to achieve global randomization of pixel positions; Based on the inverse mapping relationship of the original scan sequence, the scrambled one-dimensional pixel sequence is reconstructed into a two-dimensional image matrix ZC, thus obtaining the rearranged and scrambled image.

9. The aerial multi-image encryption method based on improved chaotic mapping and regional adaptive encryption according to claim 1, characterized in that: Based on the rearranged and scrambled foreground target region image, the concentric ring double-mapping diffusion algorithm is used to generate the encrypted foreground target region image FE. The specific process is as follows: First: Construct a complete circular pixel sequence starting from the outermost boundary of the image; Then, a pixel-by-pixel XOR operation is performed on the quantized value of the chaotic sequence and the corresponding ring pixel to achieve initial chaotic modulation. For the inner ring pixels, a dual-channel selective diffusion mechanism is adopted: two outer ring pixel index mappings are constructed in parallel: the first mapping forms an ordered mapping based on the statistical median segmentation principle, and the second mapping generates a random mapping using hash pseudo-random permutation. The mapping method is dynamically selected according to the chaotic bit sequence, and the selected outer ring pixel is XORed with the inner ring pixel. This process iterates layer by layer inward, and the pixel value processed in each layer is immediately used for the diffusion calculation of the next inner ring, forming a chain propagation effect from the outside to the inside.

10. The aerial multi-image encryption method based on improved chaotic mapping and regional adaptive encryption according to claim 1, characterized in that: The process of encrypting the preprocessed background region by first performing lossless compression using Huffman coding and then implementing lightweight chaotic operations is as follows: Perform channel separation and image type determination on the preprocessed background image; Then, adaptive Huffman coding compression is performed, the statistical distribution of pixel values ​​in each channel is calculated, the optimal prefix coding dictionary is constructed, the image data is converted into a variable-length coded bitstream, and the coding length feature of each pixel unit is recorded synchronously during the compression process. A chaotic preprocessing mechanism is introduced at the coding layer, using a pre-generated chaotic index sequence to non-linearly interleave and rearrange the encoded bit stream; Then, a dynamic filling algorithm based on adjacent pixel copying is used to normalize the data length of each channel in order to achieve a synchronous processing standard. Subsequently, a dual encryption operation is performed: the first stage of encryption extracts a scrambled index from the index sequence based on the composite parameters of the image and channels, and spatially scrambles the compressed data; the second stage of encryption quantizes the chaotic sequence into integer values ​​between 0 and 255 after spatial scrambling, and performs pixel value diffusion encryption through bitwise XOR operation.

Images