Infrared and visible light image multistage fusion method and device based on fractional order differential and intuitionistic fuzzy set

Abstract

The invention discloses an infrared and visible light image multi-level fusion method and device based on fractional differential and an intuitive fuzzy set, and relates to the field of computer vision. The method comprises the following steps: respectively decomposing an infrared image and a visible light image based on an MDLatLRR decomposition method to obtain a corresponding basic component and a series of multi-scale detail components; fusing the detail components by using a rule based on an intuitionistic fuzzy set to obtain a detail fusion component; fusing the basic components by adopting a fractional order differential rule to obtain a basic fusion component; and obtaining a reconstructed image of the target object according to the detail fusion component and the basic fusion component. According to the method, the local texture richness is enhanced and the key feature extraction effect is improved through the rule of an intuitionistic fuzzy set, meanwhile, the overall feature of the image is effectively reserved through the global and memory features of fractional differential, and finally, the high-quality fusion effect is achieved through recombination of a detail fusion component and a basic fusion component.

Description

Technical Field

[0001] This application relates to the field of computer vision technology, and in particular to a method and apparatus for multi-level fusion of infrared and visible light images based on fractional derivatives and intuitionistic fuzzy sets. Background Technology

[0002] In the field of multimodal image processing, infrared and visible light image fusion can generate comprehensive visual representations by integrating complementary imaging modalities. Infrared images reflect the thermal radiation information of a scene and can effectively highlight important targets even in adverse environments such as nighttime, rain, and fog. However, they cannot clearly present detailed features such as texture. Visible light images, derived from reflected light from the target scene, are sensitive to changes in lighting and inclement weather, but can carry rich edge and texture information. The core objective of infrared and visible light image fusion is to integrate these complementary data streams to generate a synthetic image that combines rich detail with clear targets, supporting subsequent advanced vision tasks such as semantic segmentation and object detection.

[0003] Current image fusion methods are mainly divided into two categories: methods based on traditional mathematical derivation and methods based on deep learning techniques. The latter employs complex network architectures and sophisticated loss functions, enabling it to autonomously learn effective fusion strategies and exhibiting powerful deep feature extraction capabilities.

[0004] However, the high complexity of existing deep learning models requires significant computational resources, and limited training data can lead to poor model fit and weak ensemble effects. Furthermore, the inherent "black box" nature of deep learning models (i.e., lack of interpretability) means their working principles remain poorly understood and inefficient. Simultaneously, some methods overemphasize deep feature extraction while neglecting the interactions between key features, resulting in insufficient preservation of detailed structural and brightness features. Summary of the Invention

[0005] Therefore, it is necessary to provide a method and apparatus for multi-level fusion of infrared and visible light images based on fractional derivatives and intuitionistic fuzzy sets to address the above-mentioned technical problems. This method can fully preserve detailed features while achieving effective feature interaction.

[0006] Firstly, this application provides a multi-level fusion method for infrared and visible light images based on fractional-order differentials and intuitionistic fuzzy sets. This method includes:

[0007] Acquire infrared and visible light images of the target object;

[0008] The infrared and visible light images are decomposed based on the MDLatLRR decomposition method to obtain a basic component and a series of multi-scale detail components.

[0009] The detail components of infrared and visible light images are fused using rules based on intuitionistic fuzzy sets to obtain detail fused components; the basic components of infrared and visible light images are fused using fractional differential rules to obtain basic fused components.

[0010] The reconstructed image of the target object is obtained based on the detail fusion component and the basic fusion component.

[0011] In one embodiment, the detail components corresponding to the infrared image and the visible light image are fused using rules based on intuitionistic fuzzy sets to obtain the detail fused components, including:

[0012] The membership degree function is used to represent the detail components in a fuzzy manner, and the membership degree and non-membership degree of each detail component are obtained.

[0013] Based on membership and non-membership, obtain the degree of hesitation of each detailed component;

[0014] Based on the membership degree, non-membership degree, and hesitation degree of each detail component, obtain the local fuzzy entropy;

[0015] The detail components corresponding to the infrared and visible light images are fused based on the local fuzzy entropy to obtain the fused detail components.

[0016] In one embodiment, fusing the detail components corresponding to the infrared image and the visible light image based on the local fuzzy entropy includes:

[0017] By comparing the local fuzzy entropy corresponding to the infrared image and the local fuzzy entropy corresponding to the visible light image, the target component is selected from the detail components corresponding to the infrared image and the detail components corresponding to the visible light image based on the comparison results.

[0018] The infrared image and the visible light image are fused based on the target component to achieve the fusion of the corresponding detail components.

[0019] In one embodiment, a fractional-order differential rule is used to fuse the fundamental components corresponding to the infrared image and the visible light image to obtain the fundamental fused components, including:

[0020] The fractional order of different regions in infrared and visible light images is adaptively extracted respectively;

[0021] Obtain the negative fractional-order differential operators in the horizontal and vertical directions in infrared and visible light images based on the fractional order;

[0022] Obtain the fractional metric operator from the fractional differential operator;

[0023] Based on the fractional order metric operator, the weights are obtained, and the corresponding basic components of the infrared image and the visible light image are weighted and summed to obtain the basic fusion component.

[0024] In one embodiment, the fractional order of different regions in infrared and visible light images is adaptively extracted, including:

[0025] Obtain the average gradient and spatial frequency, and then determine the fractional order based on the average gradient and spatial frequency.

[0026] In one embodiment, obtaining a reconstructed image of the target object based on detail fusion components and fundamental fusion components includes:

[0027] The detail fusion component and the basic fusion component are added together to obtain the reconstructed image.

[0028] Secondly, this application also provides a multi-level fusion device for infrared and visible light images based on fractional-order differentials and intuitionistic fuzzy sets. The device includes:

[0029] The data acquisition module is used to acquire infrared and visible light images of the target object;

[0030] The multi-level decomposition module is used to decompose infrared and visible light images based on the MDLatLRR decomposition method to obtain a basic component and a series of multi-scale detail components.

[0031] The feature fusion module is used to fuse the detail components corresponding to the infrared image and the visible light image using rules based on intuitionistic fuzzy sets to obtain the detail fusion components; and to fuse the basic components corresponding to the infrared image and the visible light image using fractional differential rules to obtain the basic fusion components.

[0032] The image reconstruction module is used to obtain a reconstructed image of the target object based on the detail fusion component and the basic fusion component.

[0033] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method for multi-level fusion of infrared and visible light images based on fractional derivatives and intuitionistic fuzzy sets.

[0034] Fourthly, this application also provides a computer-readable storage medium. This computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the steps in the above-described method for multi-level fusion of infrared and visible light images based on fractional derivatives and intuitionistic fuzzy sets.

[0035] Fifthly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, implements the steps in the above-described method for multi-level fusion of infrared and visible light images based on fractional-order differentials and intuitionistic fuzzy sets.

[0036] The aforementioned method and apparatus for multi-level fusion of infrared and visible light images based on fractional derivatives and intuitionistic fuzzy sets acquires infrared and visible light images of the target object. It then decomposes the infrared and visible light images using the MDLatLRR decomposition method to obtain a basic component and a series of multi-scale detail components. Next, it fuses the detail components of the infrared and visible light images using rules based on intuitionistic fuzzy sets to obtain a detail fusion component. Finally, it fuses the basic components of the infrared and visible light images using fractional derivative rules to obtain a basic fusion component. The reconstructed image of the target object is then obtained based on the detail fusion component and the basic fusion component. This method enhances local texture richness and improves key feature extraction through rules based on intuitionistic fuzzy sets. Simultaneously, it effectively preserves the overall characteristics of the image through the global and memorable features of fractional derivatives. Finally, it achieves high-quality fusion results through the recombination of the detail fusion component and the basic fusion component. Attached Figure Description

[0037] Figure 1 This is a flowchart illustrating a multi-level fusion method for infrared and visible light images based on fractional derivatives and intuitionistic fuzzy sets in one embodiment.

[0038] Figure 2 This is a comparative illustration of different methods for image fusion in one embodiment. Detailed Implementation

[0039] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0040] This application provides a method for multi-level fusion of infrared and visible light images based on fractional-order differentials and intuitionistic fuzzy sets, such as... Figure 1 As shown, it includes the following steps:

[0041] S101, acquire infrared and visible light images of the target object.

[0042] Infrared (IR) and visible (VI) images are acquired through an image acquisition device, which typically integrates both an infrared imaging module and a visible imaging module.

[0043] Infrared imaging modules generate images using the infrared radiation emitted by the target object itself. Objects at different temperatures exhibit different radiation intensities in the infrared band. The infrared detector in the device can sensitively capture these infrared signals, convert them into electrical signals, and then, after a series of complex data processing and algorithm calculations, finally present an infrared image reflecting the temperature distribution characteristics of the target object.

[0044] The visible light imaging module is based on traditional optical imaging principles. It collects visible light reflected from the target object through a lens, focuses and adjusts the light through an optical system, and projects the light onto an image sensor. The image sensor converts the light signal into a digital signal, and after image processing steps such as color correction and sharpening, generates a visible light image similar to the scene seen by the naked eye. Visible light images can clearly display the appearance, shape, color, and surface details of the target object, and can be used to identify the specific form and characteristics of the target object.

[0045] S102, based on the MDLatLRR decomposition method, decompose the infrared image and the visible light image respectively to obtain a basic component and a series of multi-scale detail components.

[0046] The MDLatLRR (Multi-level Decomposition Based on Latent Low-Rank Representation) method is primarily used to process high-dimensional data. Based on low-rank representation theory, it aims to extract latent low-rank structures from high-dimensional data. A low-rank structure means that a data matrix can be approximated as a low-rank matrix; this structure often contains the essential characteristics and inherent patterns of the data. When processing multi-dimensional data, this method can effectively separate the low-rank components and sparse components, thereby achieving effective data decomposition and feature extraction. The low-rank components are the fundamental components, and the sparse components are the detail components.

[0047] In images, fundamental components contain the global structure and main patterns of the image data. They represent parts of the image with overall consistency. For example, in an image containing a landscape, areas with uniform characteristics, such as the sky and large areas of grass, will be well represented in the fundamental components, typically appearing as the background. Detail components contain noise points, foreground objects, and other details in the image data. Detail components help identify anomalies and special events in the data during data analysis. By analyzing detail components, tasks such as fault detection and target recognition can be performed. For instance, in medical images, detail components may contain information about lesion areas, which doctors can use to aid in disease diagnosis.

[0048] This embodiment uses MDLatLRR decomposition to separate the fundamental components and detail components, which can effectively distinguish the global structure and local anomaly information in the data so that they can be fused separately.

[0049] S103, use rules based on intuitionistic fuzzy sets to fuse the detail components corresponding to the infrared image and the visible light image to obtain the detail fusion component; use fractional differential rules to fuse the basic components corresponding to the infrared image and the visible light image to obtain the basic fusion component.

[0050] Detail components reflect fine image features such as texture and edges. To preserve these features more comprehensively, this embodiment introduces a fusion method based on Intuitionistic Fuzzy Set (IFS). This method has unique advantages over traditional fuzzy sets when dealing with the inherent uncertainty in detail component fusion. IFS theory includes three important concepts: membership, non-membership, and hesitation. Membership refers to the degree to which an element belongs to a set, non-membership refers to the degree to which an element does not belong to a set, and hesitation reflects the degree of hesitation about whether an element belongs to a set. Hesitation captures the uncertainty between membership and non-membership, providing a more refined mechanism for representing subtle differences and relationships between pixels. This three-dimensional modeling capability of membership, non-membership, and hesitation makes the fusion process more adaptive and context-aware, especially in areas where infrared and visible light features conflict or complement each other. By dynamically balancing the contributions of the two modalities according to the level of local uncertainty, the IFS-based method can better simulate human subjective reasoning, conforming to natural cognitive laws during image fusion and effectively preserving subtle details that traditional methods might lose.

[0051] The fundamental components of an image contain its main contour and brightness information. Fractional differentiation can more effectively reflect the sharp areas and energy information of the input image, thereby optimizing the final fusion result. Based on the characteristics of fractional differentiation, this embodiment applies it to the field of image fusion, enhancing edge sharpness while maintaining the smoothness of uniform areas.

[0052] S104, Obtain the reconstructed image of the target object based on the detail fusion component and the basic fusion component.

[0053] The detail fusion component combines detail components from both infrared and visible light images, while the fundamental fusion component combines fundamental components from both. Finally, a specific fusion algorithm is used to combine them to obtain the reconstructed image. This process ensures that the global structural information provided by the fundamental components is accurately represented in the reconstructed image, while also ensuring that the specific information in the detail components is appropriately integrated. This results in a reconstructed image that retains the overall features of the target object while highlighting its local details and unique characteristics. Through this fusion operation, a clearer, more complete, and accurate reconstructed image of the target object is obtained, providing a solid foundation for subsequent tasks such as image recognition, object detection, and image understanding.

[0054] In one embodiment, the detail components corresponding to the infrared image and the visible light image are fused using rules based on intuitionistic fuzzy sets to obtain the detail fused components. This includes: representing the detail components with fuzzy representations through membership functions to obtain the membership degree and non-membership degree of each detail component; obtaining the hesitation degree of each detail component based on the membership degree and non-membership degree; obtaining the local fuzzy entropy based on the membership degree, non-membership degree, and hesitation degree of each detail component; and fusing the detail components corresponding to the infrared image and the visible light image based on the local fuzzy entropy to obtain the detail fused components.

[0055] Based on fuzzy set theory, membership functions can be used to represent image details in a fuzzy manner. The specific membership function definition used in this embodiment is as follows:

[0056]

[0057] in, This represents the feature of the detail component at pixel (m, n). , , mean represents the mean, min represents the minimum, and L represents the maximum number of decomposition layers, which is set to 2 here.

[0058] According to membership function The membership degree of an IFS is determined by the following formula:

[0059]

[0060] Where α takes the value of 1. The non-membership degree is determined by the following formula:

[0061]

[0062] The degree of hesitation is determined by the following formula:

[0063]

[0064] Fuzzy entropy is widely considered a core concept in fuzzy set theory, where an increase in entropy corresponds to greater information richness. This embodiment uses a sliding window approach to select the target region, and calculates the local fuzzy entropy for the center pixel of each target region, defined by the following formula:

[0065]

[0066] Where U represents membership, V represents non-membership, W represents hesitation, M×N represents the sliding window size (3×3 in this case), the sliding step size is 1, and (i, j) represents the pixels in the target region centered at pixel (m, n).

[0067] The blending rules applied to the detail components are as follows:

[0068]

[0069] Where l represents the decomposition level; R(·) represents the reconstruction operator, which is used to restore the image from the fused detail component feature matrix.

[0070] In one embodiment, a fractional-order differential rule is used to fuse the basic components corresponding to the infrared image and the visible light image to obtain the basic fused components. This includes: adaptively extracting the fractional order of different regions in the infrared image and the visible light image respectively; obtaining fractional-order differential operators in the negative horizontal and vertical directions of the infrared image and the visible light image based on the fractional order; obtaining fractional-order metric operators based on the fractional-order differential operators; obtaining weights based on the fractional-order metric operators; and weighted summing the basic components corresponding to the infrared image and the visible light image to obtain the basic fused components.

[0071] In this embodiment, the fractional metric operator is defined as follows:

[0072]

[0073] in, 1 and 2 represent infrared and visible light images, respectively; G1 and G2 represent the negative fractional-order differential operators in the x-direction and y-direction, respectively. Furthermore, to adaptively extract fractional-order features from different regions of the image, this embodiment uses the following method to calculate the fractional orders of G1 and G2:

[0074]

[0075] Here, the values ​​of a and b can be 0.6 and 0.5 respectively, and SF and AG represent the spatial frequency and average gradient of the image.

[0076] In this embodiment, the weights of the basic components are calculated using a fractional differential measure operator. The fusion result is then obtained by a weighted sum of the basic components corresponding to the infrared and visible light images. The weights of the basic components are calculated as follows:

[0077]

[0078] Weighted summation yields:

[0079]

[0080] In summary, incorporating fractional differentiation into basic component fusion techniques offers significant advantages over integer differentiation methods. Integer differentiation focuses on local features, while fractional differentiation possesses global correlation and memory properties, effectively preserving long-range dependencies in brightness and structural information. Furthermore, by adaptively calculating the fractional order, it is possible to respond based on local image features—applying stronger differentiation operations in textured regions while maintaining stability in uniform regions, thus achieving an optimal balance between feature enhancement and natural appearance preservation.

[0081] Finally, by fusing the basic fusion component and the detail fusion component, the complete reconstructed image is obtained, as shown in the following formula:

[0082]

[0083] To evaluate the performance of the method provided in this invention, comparative experiments were conducted on the classic TNO dataset. The method of this invention was compared horizontally with nine advanced fusion techniques, including DCHWT, RFN-Nest, PMGI, FusionGAN, U2Fusion, IFCNN, SDNet, SwinFusion, and CrossFuse. Six objective quality metrics (Information Entropy (EN), Standard Deviation (SD), Mutual Information (MI), Visual Information Fidelity (VIF), Average Gradient (AG), and Spatial Frequency (SF)) were selected to evaluate the comparative methods and the fusion algorithm of this invention. Higher values ​​indicate more ideal fusion results.

[0084] Figure 2 The figure simultaneously displays the results of the fusion of the comparison method and the method of the present invention. The figure uses red and yellow highlighted rectangles to emphasize the subtle differences in each image.

[0085] like Figure 2The results show that the fused images generated by DCHWT, PMGI, FusionGAN, and SDNet generally suffer from significant blurring and loss of texture details (marked by yellow boxes). In contrast, RFN-Nest, U2Fusion, and CrossFuse perform better in preserving texture details, but their brightness information is unreasonable (e.g., insufficient visibility of people marked by red boxes). IFCNN, SwinFusion, and the fusion method proposed in this invention exhibit better image quality and more reasonable presentation of brightness information. However, IFCNN and SwinFusion may lack detailed textures (e.g., missing streetlight features in the yellow box). In contrast, this invention effectively preserves edge features while avoiding artifacts, thus achieving a superior visual effect.

[0086] It is evident that traditional methods often struggle to strike a balance between texture preservation and natural brightness, frequently having to compromise on one over the other. This invention overcomes this limitation with a dual-branch architecture—a fractional-differential branch handles structural integrity and energy conservation, while an intuitionistic fuzzy set branch adaptively enhances texture details in uncertain regions. This integration strategy generates an optimally balanced fusion effect, as evidenced by the clear rendering of the figure in the red box and the streetlight details in the yellow box. The ability to preserve these complementary features fully demonstrates the practical advantages of this invention in complex scenes requiring joint analysis of thermal and textural content.

[0087] Furthermore, Table 1 summarizes the average quantitative performance metrics of different fusion techniques evaluated based on the TNO dataset. The results show that the present invention achieves the best values ​​among the four metrics (standard deviation, visual information fidelity, average gradient, and spatial frequency), and ranks second among the two metrics (information entropy and mutual information). These data fully demonstrate that the method of the present invention outperforms most traditional methods in preserving texture information and target intensity, making it a highly competitive solution for infrared and visible light image fusion tasks.

[0088] Table 1. Average quantitative performance of different fusion techniques evaluated based on the TNO dataset.

[0089] method EN↑ SD↑ MI↑ VIF↑ AG↑ SF↑ DCHWT 6.5678 64.9789 13.1355 0.5056 3.6539 9.7467 RFN-Nest 6.8414 71.9019 13.6827 0.6577 2.7338 6.1276 PMGI 6.9339 71.5486 13.8679 0.7928 3.5725 8.7099 FusionGAN 6.3629 54.3579 12.7257 0.4536 2.2052 5.7910 U2Fusion 6.7571 64.9116 13.5142 0.8543 4.7724 11.3013 IFCNN 6.5953 66.8657 13.1906 0.5903 4.3371 11.4928 SDNet 6.4906 55.5529 12.9812 0.6114 4.4676 11.4048 SwinFusion 6.6812 80.4150 13.3623 0.6449 3.9275 10.2725 CrossFuse 6.8229 80.7646 13.6458 0.6211 3.7175 9.9942 This invention 6.9054 81.8209 13.8109 1.1649 5.2411 13.7638

[0090] The quantitative advantages of this invention stem from its meticulously designed fusion strategy. Excellent results in standard deviation and visual information fidelity metrics demonstrate that the method effectively maintains image contrast and visual naturalness, directly benefiting from an effective fundamental component fusion technique based on fractional derivatives. This technique not only preserves the global structural information and energy distribution characteristics of the source image but also excels in average gradient and spatial frequency metrics, fully showcasing the superior performance of detail component fusion techniques based on intuitionistic fuzzy sets in handling uncertainty in texture-rich regions—a technique that accurately captures subtle texture details and sharp edge features. Although this invention ranks second in EN and MI metrics, these values ​​still indicate strong competitiveness. While some specialized methods may have a slight edge in information entropy or mutual information, this often comes at the expense of other key visual qualities. This invention demonstrates a comprehensive quantitative advantage in the multi-dimensional balance of fusion quality, rather than optimizing specific parameters at the expense of other metrics. Overall, this invention achieves excellent fusion results, generating clearer content and higher visual information quality.

[0091] In summary, the fusion framework proposed in this invention effectively addresses the challenges of target enhancement and detail extraction in complex scenes by integrating fractional differential fusion and intuitionistic fuzzy set techniques. Based on the "foundation-detail" framework design, fractional differential fusion rules are used in the foundational component to ensure the complete preservation of key contour information, while intuitionistic fuzzy set methods are employed in the detail component to improve the accuracy of detail extraction. Experimental results show that this framework can effectively preserve rich details and texture features in images, demonstrating broad application prospects in scenarios such as military reconnaissance and night driving.

[0092] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0093] Based on the same inventive concept, this application also provides an infrared and visible light image multi-level fusion device based on fractional-order differential and intuitionistic fuzzy sets for implementing the aforementioned method for multi-level fusion of infrared and visible light images based on fractional-order differential and intuitionistic fuzzy sets. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the infrared and visible light image multi-level fusion device based on fractional-order differential and intuitionistic fuzzy sets provided below can be found in the limitations of the infrared and visible light image multi-level fusion method based on fractional-order differential and intuitionistic fuzzy sets described above, and will not be repeated here.

[0094] In one embodiment, an infrared and visible light image multi-level fusion device based on fractional derivative and intuitionistic fuzzy set is provided, including: a data acquisition module for acquiring infrared and visible light images of a target object;

[0095] The multi-level decomposition module is used to decompose infrared and visible light images based on the MDLatLRR decomposition method to obtain a basic component and a series of multi-scale detail components.

[0096] The feature fusion module is used to fuse the detail components corresponding to the infrared image and the visible light image using rules based on intuitionistic fuzzy sets to obtain the detail fusion components; and to fuse the basic components corresponding to the infrared image and the visible light image using fractional differential rules to obtain the basic fusion components.

[0097] The image reconstruction module is used to obtain a reconstructed image of the target object based on the detail fusion component and the basic fusion component.

[0098] The modules in the aforementioned infrared and visible light image multi-level fusion device based on fractional derivatives and intuitionistic fuzzy sets can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in hardware or independently of the processor in a computer device, or stored in software in the memory of a computer device, so that the processor can call and execute the corresponding operations of each module.

[0099] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in all of the above method embodiments.

[0100] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in all of the above method embodiments.

[0101] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in all of the above method embodiments.

[0102] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0103] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0104] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0105] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A multi-level fusion method for infrared and visible light images based on fractional-order differentials and intuitionistic fuzzy sets, characterized in that, The method includes: Acquire infrared and visible light images of the target object; The infrared image and the visible light image are decomposed based on the MDLatLRR decomposition method to obtain a basic component and a series of multi-scale detail components. The detail components corresponding to the infrared image and the visible light image are fused using rules based on intuitionistic fuzzy sets to obtain detail fused components; the basic components corresponding to the infrared image and the visible light image are fused using fractional differential rules to obtain basic fused components. The reconstructed image of the target object is obtained based on the detailed fusion component and the basic fusion component.

2. The method according to claim 1, characterized in that, The detail components corresponding to the infrared image and the visible light image are fused using rules based on intuitionistic fuzzy sets to obtain detail fusion components, including: The membership degree and non-membership degree of each detail component are obtained by performing a fuzzy representation of the detail components using a membership degree function. Based on the membership degree and non-membership degree, obtain the hesitation degree of each of the detailed components; Based on the membership degree, non-membership degree, and hesitation degree of each detailed component, obtain the local fuzzy entropy; The detail components corresponding to the infrared image and the visible light image are fused according to the local fuzzy entropy to obtain the fused detail components.

3. The method according to claim 2, characterized in that, The fusion of the detail components corresponding to the infrared image and the visible light image based on the local fuzzy entropy includes: By comparing the local blur entropy corresponding to the infrared image and the local blur entropy corresponding to the visible light image, a target component is selected from the detail components corresponding to the infrared image and the detail components corresponding to the visible light image based on the comparison result. The infrared image and the visible light image are fused based on the target component to achieve the fusion of the corresponding detail components.

4. The method according to claim 1, characterized in that, The fundamental components corresponding to the infrared image and the visible light image are fused using a fractional differential rule to obtain fundamental fused components, including: The fractional order of different regions in the infrared and visible light images is adaptively extracted respectively; Based on the fractional order, obtain the negative fractional differential operators in the horizontal and vertical directions of the infrared and visible light images; Obtain the fractional-order metric operator based on the fractional-order differential operator; Based on the fractional-order metric operator, weights are obtained, and the corresponding basic components of the infrared image and the visible light image are weighted and summed to obtain the basic fusion component.

5. The method according to claim 4, characterized in that, The fractional order of different regions in the infrared and visible light images is adaptively extracted, including: Obtain the average gradient and spatial frequency, and obtain the fractional order based on the average gradient and spatial frequency.

6. The method according to claim 1, characterized in that, Obtaining a reconstructed image of the target object based on the detail fusion component and the basic fusion component includes: The detailed fusion component and the basic fusion component are added together to obtain the reconstructed image.

7. A multi-level fusion device for infrared and visible light images based on fractional-order differentials and intuitionistic fuzzy sets, characterized in that, The device includes: The data acquisition module is used to acquire infrared and visible light images of the target object; The multi-level decomposition module is used to decompose the infrared image and the visible light image respectively based on the MDLatLRR decomposition method to obtain a corresponding basic component and a series of multi-scale detail components. The feature fusion module is used to fuse the detail components corresponding to the infrared image and the visible light image using rules based on intuitionistic fuzzy sets to obtain detail fusion components; and to fuse the basic components corresponding to the infrared image and the visible light image using fractional differential rules to obtain basic fusion components. The image reconstruction module is used to obtain a reconstructed image of the target object based on the detail fusion component and the basic fusion component.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images