Image encryption method in garbage classification process

Abstract

The invention provides an image encryption method in a garbage classification process, and the method comprises the following steps: S1, collecting a garbage image, and carrying out the preprocessing of the garbage image; s2, a deep learning model is adopted to classify the preprocessed junk images; and S3, the classified junk images are encrypted through an encryption system, the confidentiality of the junk images in the transmission process is ensured, and the problem of sensitive information leakage possibly caused in the junk classification process can be effectively solved through the method.

Description

Technical Field

[0001] The present invention relates to the technical field of image encryption, and in particular to an image encryption method in a garbage classification process. Background Art

[0002] With the rapid development of information technology, the security of image data has become increasingly prominent. Especially in the field of garbage classification, file-type garbage may contain sensitive information, such as personal privacy, business secrets, etc. These confidential garbage data need to be encrypted and specially processed during transmission and storage. Traditional image encryption methods have limitations in protecting data security, especially when dealing with large amounts of data and resisting complex attacks. Therefore, a garbage image confidentiality method is urgently needed to solve the problem of sensitive information leakage that may occur during garbage classification. Summary of the invention

[0003] The purpose of the present invention is to solve the technical problems in the above background and propose an image encryption method in the process of garbage classification, comprising the following steps:

[0004] S1, collect garbage images and perform preprocessing;

[0005] S2, using deep learning model to classify the pre-processed garbage images;

[0006] S3. Encrypt the classified junk images through an encryption system to ensure the confidentiality of the junk images during transmission.

[0007] In a preferred solution, the encryption process in step S3 includes the following steps:

[0008] S31, establishing a quantum image representation model, scanning each color channel of the collected garbage image to obtain a one-dimensional vector, and using a quantum sequence to store color information, while using the probability amplitude of a quantum bit to store pixel position information;

[0009] S32. Introduce the design technology of plaintext associated symmetric encryption system into the quantum image encryption framework, integrating the inherent characteristics of chaotic system and quantum mechanics;

[0010] S33, decomposing the original image into a three-dimensional bit matrix, and performing quantization processing using an enhanced quantum image representation model;

[0011] S34, using the random sequence and the plaintext feature to jointly generate a sequence for controlling the quantum shift operation, and realizing the pixel cyclic shift operation of the quantum image by controlling the quantum exchange gate;

[0012] S35, converting the generated pseudo-random sequence into an integer to obtain a new password stream, and applying it to the quantum XOR operation on the image to further change the pixel value;

[0013] S36. In order to break the correlation between adjacent pixels of the image, quantum generalized Arnold transform is added in the final stage of encryption to obtain the final ciphertext.

[0014] In a preferred embodiment, the process of establishing the quantum image representation model includes the following steps:

[0015] First, scan each color channel of the image in a specific order to obtain a one-dimensional vector; add element 0 to make the vector length reach 2 n , n is an integer; 8 quantum sequences are used to store color information

[0016] Then use a quantum bit probability amplitude |g(j)> to store the pixel position information, as shown in the following formula:

[0017] |g(j)>=cosθ j |0>+sinθ j |1>;

[0018] The expression of the quantum image representation method of any dimension is obtained as follows:

[0019]

[0020] Where |g(j)>=cosθ j |0>+sinθ j |1>Store pixel location information, |i n i n-1 …i0> is the ground state of n quantum bits, and is also the original coordinate of the i-th pixel in the image;

[0021] Stores color information.

[0022] In a preferred solution, the encryption system in step S2 specifically refers to a plaintext-associated quantum image encryption system based on quantum shift and quantum XOR operation, using the designed quantum image representation model to verify the validity of the quantum image representation model.

[0023] In a preferred solution, the encryption steps of the plaintext-associated quantum image encryption system based on quantum shift and quantum XOR operation include:

[0024] S31, decomposing the original image P into a three-dimensional bit matrix P 8bit (M×N×8), and then the original image P is quantized using the enhanced quantum image representation model to obtain |P>, assuming that the plaintext image size is 2 n ×2 n , the new enhanced quantum grayscale image is expressed as:

[0025]

[0026] In the formula Used to encode image position information, |y>=|y n-1 y n-2 ...y0> encodes the vertical position information of the grayscale image, |x> = |x n-1 x n-2 …x0> encodes the horizontal position information of the grayscale image, Used to encode grayscale information;

[0027] S32, respectively find the 3D bit matrix P 8bit The sum of the elements at each level:

[0028]

[0029] S33, to resist the chosen plaintext attack, the encryption process is associated with the plaintext, using a random sequence x n and plaintext feature Sum k (k = 0, 1 ... 7) together generate a sequence for controlling quantum shift operations Use the control quantum switching gate U YX To implement the pixel circular shift operation of the quantum image:

[0030]

[0031] byU YX Constructor Operation S YX :

[0032]

[0033] Where S YX satisfy The corresponding matrix is ​​a unitary matrix, YesS YX The Hermitian conjugate matrix of YX Perform 2 on |P> 2n This operation can complete the quantum shift operation of all pixels of the quantum image and obtain |P1>;

[0034] S34, the pseudo-random sequence x generated in point (1) of the research plan n Integerize to get a new password stream And apply it to the quantum XOR operation of the image |P1>, further change the pixel value to obtain |P2>; let the quantum XOR operation be B, and the effect of the quantum XOR operation is:

[0035]

[0036] Among them, B YXis a sub-operation of B, and its expression is:

[0037]

[0038] S35. In order to break the correlation between adjacent pixels of the image, scrambling is a commonly used digital image encryption operation. Therefore, the quantum generalized Arnold transform is added in the final stage of encryption to obtain the final ciphertext C.

[0039] In the preferred solution, the generalized Arno1d transform of the quantum state image is expressed as:

[0040]

[0041] Where (x, y) is the original image position, (x′, y′) is the scrambled pixel position, and A is used to represent the quantum generalized Arnold transform. Then the first quantum generalized Arnold transform is:

[0042]

[0043] Sequence for controlling quantum shift operations The generation of is related to the plaintext.

[0044] In a preferred solution, in step S2, ResNet or Inception is used as a deep learning model, and the deep learning model is trained using a labeled data set. By inputting pictures into the model, the model is gradually learned to extract features of garbage, and then these features are mapped to corresponding garbage classification labels;

[0045] Each type of garbage in the garbage image dataset has a corresponding label, which is used to train the deep learning model. The dataset includes garbage images at different angles, lighting conditions, and occlusion conditions to ensure the generalization ability of the model;

[0046] During the training process, the accuracy and convergence speed of the model can be improved by setting the learning rate, loss function and optimization algorithm;

[0047] The deep learning model predicts and classifies garbage images and outputs a probability distribution, indicating the probability that the image belongs to each type of garbage. Based on the probability distribution, the category with the highest probability is selected as the classification of the garbage.

[0048] In a preferred solution, the preprocessing operation in step S1 includes denoising, enhancement and segmentation processing.

[0049] In a preferred solution, in step S2, the classified junk images include file-type junk images;

[0050] The file-type junk images are further identified, and the file-type junk images with sensitive information identified are encrypted.

[0051] In a preferred solution, the process of identifying sensitive information specifically includes the following steps:

[0052] Use OCR technology to convert text in images into machine-readable text data;

[0053] Apply NLP techniques to analyze the converted text, including language modeling, keyword extraction, and entity recognition, to identify sensitive information;

[0054] Also included are the following steps in parallel with the above steps:

[0055] Analyze spam images to identify specific patterns or signs, including: seals, signatures or specific logos;

[0056] Combine the file's contextual information, including its origin, type, and history, to infer whether the file contains sensitive information.

[0057] The beneficial effects of the present invention are:

[0058] (1) By integrating the encryption system with the garbage sorting process, the risk of sensitive data leakage in the garbage sorting process is avoided;

[0059] (2) Based on the analysis and summary of the characteristics and application scenarios of existing quantum image representation models, in order to more effectively implement quantum image encryption systems, a quantum image representation model suitable for encryption applications is proposed;

[0060] (3) Analyze and summarize the existing quantum image encryption algorithms, use the results of cryptanalysis to guide cipher design, and then, starting from the architecture of the encryption system, integrate the scrambling and diffusion operations into an indivisible whole to avoid being broken one by one by attackers, while improving encryption efficiency;

[0061] (4) For the first time, the plaintext-associated symmetric encryption technology was introduced into the research of quantum image encryption system, which achieved full association between the encryption system and the plaintext image, improved the security of the system, and further condensed the general design principles of quantum image cryptographic algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 It is a flow chart of the overall solution of the present invention.

[0063] Figure 2 This is a schematic diagram of the scanning order of image color channels.

[0064] Figure 3 This is a block diagram of a plaintext-associated quantum image encryption system based on quantum shift and quantum XOR operations.

[0065] Figure 4It is a schematic diagram of the bit matrix of the plaintext image. DETAILED DESCRIPTION

[0066] Example 1

[0067] like Figures 1 to 4 As shown in the figure, an encryption algorithm based on quantum image representation model is proposed. The research process is as follows:

[0068] Establishment of quantum image representation model

[0069] The quantum image representation model is studied in depth from the perspective of quantum information theory, including the following specific contents:

[0070] (1) Analyze the characteristics of existing quantum image representation models, available quantum image processing operations and their applicable application scenarios;

[0071] (2) Based on the definition of quantum images, combined with the representation methods of classical images and the requirements of image encryption, explore the effective representation of quantum images and the best way to represent classical images into quantum images;

[0072] (3) Establish a quantum image representation model that is more suitable for image encryption. This model has richer quantum image processing operations that can be used for quantum image encryption and is easier to implement scrambling and diffusion operations in image encryption.

[0073] Design of Symmetric Quantum Image Encryption System with Plaintext Association

[0074] (1) Study the defects of existing quantum image encryption algorithms, such as insufficient encryption operations, low encryption efficiency, poor encryption effect, inability to resist powerful differential attacks and chosen plaintext attacks, and the entry points of related attack methods;

[0075] (2) Taking full advantage of the advantages of chaotic systems and quantum mechanics in digital image encryption, a symmetric quantum image encryption architecture with scrambling-diffusion interaction of plaintext association is proposed. The scrambling and diffusion operations of the corresponding encryption system are integrated into an inseparable whole, improving the efficiency of image encryption, and certain features of the plaintext image are used to control the image encryption process, thereby achieving plaintext association;

[0076] (3) Conduct a comprehensive and rigorous security performance analysis of the proposed quantum image algorithm, such as key space analysis, information entropy analysis, correlation analysis, differential attack analysis, chosen plaintext attack, noise attack and system complexity analysis, etc.

[0077] Hardware simulation of quantum image encryption system

[0078] Build an FPGA quantum simulator and perform preliminary simulation of the designed quantum image representation model and quantum image encryption system on FPGA to solve related technical problems and verify the efficiency and practicality of the encryption algorithm.

[0079] The overall research plan to be adopted in this project:

[0080] The overall research plan is as follows Figure 1 As shown in the figure, we first analyze several existing quantum image representation models, and establish a new quantum image representation model for image encryption applications. This model enriches the quantum image processing operations that can be used for quantum image encryption, and makes it easier to implement scrambling and diffusion operations in image encryption. Then, based on the summary of the security vulnerabilities of existing quantum image encryption algorithms, we introduce the design technology of plaintext-associated symmetric encryption systems into the quantum image encryption framework, integrate the inherent characteristics of chaotic systems and quantum mechanics, improve the security of quantum image cryptographic systems, and condense a set of universal security criteria for quantum image cryptographic algorithms on this basis; finally, based on the above foundation, we establish an FPGA quantum simulator on the FPGA platform to simulate the proposed quantum image encryption algorithm, solve related technical problems, and promote the practical progress of quantum image cryptography.

[0081] Research Plan for Quantum Image Representation Model

[0082] According to the different ways of encoding the grayscale values ​​of the image, quantum image representation models can be divided into two categories.

[0083] The first category: using a quantum bit that stores angle parameters to encode grayscale value information, such as Qubit Lattice, FRQI, etc. This type of model is suitable for image encryption fields such as random phase encoding and quantum rotation operations, but is not suitable for value-dependent operations such as XOR and circular shift.

[0084] The second category: using a set of quantum bit sequences to store grayscale value information, such as NEQR, GQOR, etc. This category is opposite to the first category and is more suitable for value-dependent operations.

[0085] In order to enrich the calculation methods, this project proposes the following quantum image representation model suitable for encryption applications. First, each color channel of the image is divided into Figure 2 Scan in order to get a one-dimensional vector. Add element 0 to make the vector length 2 n , n is an integer. Use 8 quantum sequences to store color information Then use a quantum bit probability amplitude |g(j)> to store the pixel position information, as shown in the following formula

[0086] |g(j)>=cosθ j |0>+sinθ j|1> (1)

[0087] The expression of arbitrary dimensional quantum image representation (ADQIR: Arbitrary Dimensional Quantum Image Representation) is obtained as follows:

[0088]

[0089] Where |g(j)>=cosθ j |0>+sinθ j |1>Store pixel location information, |i n i n-1 …i0> is the ground state of n quantum bits and is also the original coordinate of the i-th pixel in the image. Stores color information.

[0090] It can be seen that grayscale images or color images of any size can be stored with only n+9 quantum bits, which has great advantages over most quantum image representation models. In addition, this model can easily perform scrambling and diffusion operations commonly used in image encryption, making it suitable for use in the field of image encryption.

[0091] Research on the architecture of quantum image encryption system and analysis and design of encryption system

[0092] (1) The disadvantage of quantum image encryption schemes based on coordinate information perturbation of different quantum transformations is that the histogram remains unchanged and multiple complex transformations are required. This not only increases the complexity of the circuit, but also cannot resist statistical attacks. On the other hand, these encryption systems are either one-time pad encryption systems, or the encryption process is independent of the plaintext and cannot resist chosen plaintext attacks.

[0093] (2) Using the chaotic sequences generated by typical chaotic systems, the pseudo-random sequence performance results of different pseudo-random sequence generation strategies (such as retaining low-order strategy, threshold division strategy, etc.) under the TESTU01 advanced test package are analyzed, and the appropriate chaotic system and its pseudo-random sequence generation strategy are selected.

[0094] (3) In order to solve the series of shortcomings of the chaotic image encryption algorithm and the existing quantum image encryption algorithm mentioned above, this project makes full use of the characteristics of quantum mechanics and chaotic systems, and proposes the following encryption framework and corresponding encryption system scheme, improves the scrambling-diffusion encryption system architecture in the existing literature, and defines the corresponding quantum image processing operations.

[0095] For scheme 1, a plaintext-associated quantum image encryption system based on quantum shift and quantum XOR operation is designed. This system realizes plaintext association in the scrambling stage, with the aim of studying the security and efficiency of plaintext-associated symmetric encryption technology at the bit level. In addition, the system also uses the designed ADQIR quantum image representation model to verify the effectiveness of the model.

[0096] Solution 1: Plaintext-associated quantum image encryption system based on quantum shift and quantum XOR operation

[0097] A digital image can be viewed as a 3-dimensional matrix (M×N×8), and the grayscale value of each pixel can be represented by an 8-bit binary number, such as Figure 4 As shown, the digital image can be converted into a 3-dimensional bit matrix M×N×8.

[0098] The main steps of encryption are as follows:

[0099] Step 1: Decompose the original image P into Figure 3 The three-dimensional bit matrix P shown 8bit (M×N×8),. Next, the original image P is quantized using NEQR (Novel Enhanced Quantum Presentation) to obtain |P>. Assume that the plaintext image size is 2 n ×2 n , the new enhanced quantum grayscale image is expressed as:

[0100]

[0101] In the formula Used to encode image position information, |y>=|y n-1 y n-2 ...y0> encodes the vertical position information of the grayscale image, |x> = |x n-1 xn -2 …x0> encodes the horizontal position information of the grayscale image, Used to encode grayscale information.

[0102] Step 2: Calculate the 3D bit matrix P 8bit The sum of the elements at each level:

[0103]

[0104] Step 3: The encryption process is associated with the plaintext, which is an effective means to make the encryption system effectively resist the chosen plaintext attack. To achieve this goal, a random sequence x is used. n and plaintext feature Sum k (k = 0, 1 ... 7) together generate a sequence for controlling quantum shift operations Here, we use the control quantum switching gate U YX To realize the pixel circular shift operation of quantum image.

[0105]

[0106] byU YX Constructor Operation S YX

[0107]

[0108] Where S YX satisfy The corresponding matrix is ​​a unitary matrix, YesS YX The Hermitian conjugate matrix of . Using S YX Perform 2 on |P> 2n This operation can complete the quantum shift operation of controlling all pixels of the quantum image and obtain |P1>.

[0109] Step 4: Substitute the pseudo-random sequence x generated in point (1) of the research plan n Integerize to get a new password stream And it is applied to the quantum XOR operation of the image |P1>, further changing the pixel value to obtain |P2>. Let the quantum XOR operation be B, and the effect of the quantum XOR operation is shown in the following formula.

[0110]

[0111] Among them, B YX is a sub-operation of B, and its expression is:

[0112]

[0113] Step 5: To break the correlation between adjacent pixels of the image, scrambling is a commonly used digital image encryption operation. Therefore, the quantum generalized Arnold transform is added in the final stage of encryption to obtain the final ciphertext C. The generalized Arnold transform of the quantum state image can be expressed as:

[0114]

[0115] Where (x, y) is the original image position, and (x′, y′) is the scrambled pixel position. A represents the quantum generalized Arnold transform, and a quantum generalized Arnold transform is:

[0116]

[0117] Since the sequence used to control the quantum shift operation The generation of is related to the plaintext, which means that any change in the plaintext image can cause a change in T, resulting in a huge change in the final ciphertext C. That is, the encryption system has a strong plaintext sensitivity and can effectively resist powerful chosen plaintext attacks.

[0118] Implementation of quantum image encryption system on FPGA hardware platform

[0119] Quantum computing provides a great acceleration when performing data encryption. Classical computing devices can be used to model quantum algorithms, but classical computer simulations cannot effectively handle the parallelism present in quantum algorithms. Quantum circuits are a convenient way to describe quantum algorithms. Using the analogy between quantum circuits and digital circuits, we study the use of classical circuits to design quantum circuit simulators. We plan to design simulators for quantum algorithms in FPGAs, and simulate the components in quantum circuits in existing FPGA chips, so that new quantum algorithms can be used for effective experiments. Compared with software simulations, FPGAs can map inherently parallel computing tasks more efficiently. The overall design process is shown in the figure below:

[0120] First, after completing the design of the quantum image encryption algorithm, the quantum circuit for quantum computing in the encryption algorithm will be designed.

[0121] Next, we will construct a quantum circuit using the simple quantum gate components we created.

[0122] Then, the correctness of the circuit is verified through FPGA simulation.

[0123] Finally, the circuit is synthesized on hardware to achieve the required performance and hardware simulation of the quantum image encryption algorithm is performed.

[0124] Example 2

[0125] Based on the garbage classification image confidentiality method in Example 1, the homomorphic encryption algorithm can be integrated to encrypt and decrypt the image before classification, as follows:

[0126] Choose a suitable homomorphic encryption algorithm that supports necessary computational operations on image data, such as addition and multiplication. Currently, commonly used homomorphic encryption algorithms include Paillier encryption and algorithms based on the Learning With Errors (LWE) problem.

[0127] Generate the public key and private key of the homomorphic encryption algorithm. The public key is used to encrypt the image data, while the private key is used to decrypt;

[0128] Before encryption, the collected garbage images are preprocessed, such as resizing and normalizing, to meet the requirements of the homomorphic encryption algorithm;

[0129] The pre-processed image data is encrypted using the public key of the homomorphic encryption algorithm. This step ensures the security of the image data during transmission and storage;

[0130] If the homomorphic encryption algorithm supports it, you can run a deep learning model directly on the encrypted image for classification. This usually requires the model to be able to process homomorphic encrypted data, or to use a special homomorphic encryption-compatible machine learning algorithm;

[0131] For homomorphic encryption algorithms that do not support complex calculations directly on encrypted data, it may be necessary to extract the calculation results (such as the probability distribution of classification labels) from the encrypted state;

[0132] The classification results or extracted calculation results are decrypted using the private key of the homomorphic encryption algorithm to obtain the final classification label; the decrypted data is post-processed, such as converting the classification results into a readable format or further analyzing the accuracy of the classification.

[0133] Through the above steps, the privacy of garbage classification image data can be protected while achieving effective image classification. This method of integrating homomorphic encryption can not only improve data security, but also maintain the flexibility of data processing.

[0134] Example 3

[0135] A garbage intelligent classification system based on deep learning and AI vision, which uses deep learning algorithms and AI vision technology to intelligently classify garbage. It includes the following steps:

[0136] 1. Recognition module design: A large number of labeled garbage image datasets are required, with corresponding labels for each type of garbage, to train the deep learning model. The dataset should contain garbage images at different angles, lighting conditions, and occlusions to ensure the generalization ability of the model. Convolutional neural networks are usually used as deep learning models. You can choose pre-trained models such as ResNet and Inception, or design and train models yourself. Use labeled datasets to train the deep learning model. By inputting pictures into the model, it gradually learns to extract the features of garbage, and then maps these features to the corresponding garbage classification labels. During the training process, the learning rate, loss function, and optimization algorithm need to be reasonably set to improve the accuracy and convergence speed of the model. When the system receives garbage images, the deep learning model predicts and classifies them. The model outputs a probability distribution, indicating the probability that the image belongs to each type of garbage. According to the probability distribution, the system selects the category with the highest probability as the classification of the garbage.

[0137] 2. Classification device design: The classification device rotates the corresponding trash can to the user through the steering gear rotation device. Here, it is necessary to control the rotation of the steering gear according to the output results of the model to achieve intelligent classification.

[0138] 3. WeChat applet design: Develop a WeChat applet with real-time garbage classification function, voice recognition function, text search function, garbage classification question answering function, etc.

[0139] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for image encryption in a garbage classification process, characterized in that: The following steps are involved: S1, collect garbage images and perform preprocessing; S2, using deep learning model to classify the pre-processed garbage images; S3. Encrypt the classified junk images through an encryption system to ensure the confidentiality of the junk images during transmission.

2. According to the image encryption method in the garbage classification process of claim 1, its characteristics are: The encryption process in step S3 includes the following steps: S31, establishing a quantum image representation model, scanning each color channel of the collected garbage image to obtain a one-dimensional vector, and using a quantum sequence to store color information, while using the probability amplitude of a quantum bit to store pixel position information; S32. Introduce the design technology of plaintext associated symmetric encryption system into the quantum image encryption framework, integrating the inherent characteristics of chaotic system and quantum mechanics; S33, decomposing the original image into a three-dimensional bit matrix, and performing quantization processing using an enhanced quantum image representation model; S34, using the random sequence and the plaintext feature to jointly generate a sequence for controlling the quantum shift operation, and realizing the pixel cyclic shift operation of the quantum image by controlling the quantum exchange gate; S35, converting the generated pseudo-random sequence into an integer to obtain a new password stream, and applying it to the quantum XOR operation on the image to further change the pixel value; S36. In order to break the correlation between adjacent pixels of the image, quantum generalized Arnold transform is added in the final stage of encryption to obtain the final ciphertext.

3. According to the image encryption method in the garbage classification process described in claim 2, it is characterized by: quantum The process of building an image representation model includes the following steps: First, scan each color channel of the image in a specific order to obtain a one-dimensional vector; add element 0 to make the vector length reach 2 n , n is an integer; 8 quantum sequences are used to store color information Then use a quantum bit probability amplitude |g(j)> to store the pixel position information, as shown in the following formula: |g(j)>=cosθ j |0>+sinθ j |1>; The expression of the quantum image representation method of any dimension is obtained as follows: Where |g(j)>=cosθ j |0>+sinθ j |1>Store pixel location information, |i n i n-1 …i0> is the ground state of n quantum bits, and is also the original coordinate of the i-th pixel in the image; Stores color information.

4. According to the image encryption method in the garbage classification process of claim 1, its characteristics are: The encryption system in step S2 specifically refers to a plaintext-associated quantum image encryption system based on quantum shift and quantum XOR operation, which uses the designed quantum image representation model to verify the validity of the quantum image representation model.

5. According to claim 4, the image encryption method in the garbage classification process is characterized by: The encryption steps of the plaintext-associated quantum image encryption system based on quantum shift and quantum XOR operation include: S31, decomposing the original image P into a three-dimensional bit matrix P 8bit (M×N×8), and then the original image P is quantized using the enhanced quantum image representation model to obtain |P>, assuming that the plaintext image size is 2 n ×2 n , the new enhanced quantum grayscale image is expressed as: In the formula Used to encode image position information, |y>=|y n-1 y n-2 ...y0> encodes the vertical position information of the grayscale image, |x> = |x n-1 x n-2 …x0> encodes the horizontal position information of the grayscale image, Used to encode grayscale information; S32, respectively find the 3D bit matrix P 8bit The sum of the elements at each level: S33, to resist the chosen plaintext attack, the encryption process is associated with the plaintext, using a random sequence x n and plaintext feature Sum k (k=0,1…7) together generate a sequence for controlling quantum shift operations Use the control quantum switching gate U YX To implement the pixel circular shift operation of the quantum image: byU YX Constructor Operation S YX : Where S YX satisfy The corresponding matrix is ​​a unitary matrix, YesS YX The Hermitian conjugate matrix of YX Perform 2 on |P> 2n This operation can complete the quantum shift operation of all pixels of the quantum image and obtain |P1>; S34, the pseudo-random sequence x generated in point (1) of the research plan n Integerize to get a new password stream And apply it to the quantum XOR operation of the image |P1>, further change the pixel value to obtain |P2>; let the quantum XOR operation be B, and the effect of the quantum XOR operation is: Among them, B YX is a sub-operation of B, and its expression is: S35. In order to break the correlation between adjacent pixels of the image, scrambling is a commonly used digital image encryption operation. Therefore, the quantum generalized Arnold transform is added in the final stage of encryption to obtain the final ciphertext C.

6. According to claim 5, the image encryption method in the garbage classification process is characterized by: The generalized Arnold transform of the quantum state image is expressed as: Where (x, y) is the original image position, (x', y') is the scrambled pixel position, and A represents the quantum generalized Arnold transform. Then a quantum generalized Arnold transform is: Sequence for controlling quantum shift operations The generation of is related to the plaintext.

7. According to claim 1, the image encryption method in the garbage classification process is characterized by: In step S2, ResNet or Inception is used as a deep learning model, and the labeled data set is used to train the deep learning model. By inputting the picture into the model, it gradually learns to extract the features of the garbage, and then maps these features to the corresponding garbage classification labels; Each type of garbage in the garbage image dataset has a corresponding label, which is used to train the deep learning model. The dataset includes garbage images at different angles, lighting conditions, and occlusion conditions to ensure the generalization ability of the model; During the training process, the accuracy and convergence speed of the model can be improved by setting the learning rate, loss function and optimization algorithm; The deep learning model predicts and classifies garbage images and outputs a probability distribution, indicating the probability that the image belongs to each type of garbage. Based on the probability distribution, the category with the highest probability is selected as the classification of the garbage.

8. According to claim 1, the image encryption method in the garbage classification process is characterized by: The preprocessing operation in step S1 includes denoising, enhancement and segmentation processing.

9. The image encryption method in the garbage classification process according to claim 1 is characterized in that: In step S2, the classified junk images include file-type junk images; The file-type junk images are further identified, and the file-type junk images with sensitive information identified are encrypted.

10. The image encryption method in the garbage classification process according to claim 9, characterized in that: The process of identifying sensitive information includes the following steps: Use OCR technology to convert text in images into machine-readable text data; Apply NLP techniques to analyze the converted text, including language modeling, keyword extraction, and entity recognition to identify sensitive information; Also included are the following steps in parallel with the above steps: Analyze spam images to identify specific patterns or signs, including: seals, signatures or specific logos; Combine the file's contextual information, including its origin, type, and history, to infer whether the file contains sensitive information.

Images