Infrared data panoramic stitching method and device, electronic equipment and storage medium

Abstract

This application discloses a method, apparatus, electronic device, and storage medium for panoramic stitching of infrared data, relating to the field of infrared data stitching technology. The method includes: acquiring an original data matrix and a previous stitched data matrix; performing kernel mapping processing on the radiometric measurements in the matrices using kernel functions to obtain an original kernel feature matrix and a previous stitched kernel feature matrix; constructing a radial attenuation model centered on the optical axis of the infrared detector for radiometric consistency correction to obtain an original correction matrix and a previous stitched correction matrix; determining a fusion weight matrix based on the kernel feature similarity at each pixel position in the spatially overlapping region of the original correction matrix and the previous stitched correction matrix; and performing pixel-by-pixel weighted fusion of the data in the spatially overlapping region according to the fusion weight matrix, and stitching it with pixels in the non-overlapping region to obtain the current stitched data matrix for the current round. The above technical solution improves the accuracy and overall consistency of infrared panoramic data fusion.

Description

Technical Field

[0001] This application relates to the field of infrared data processing technology, and more particularly to the field of infrared data stitching technology, specifically to a panoramic stitching method, apparatus, electronic device, and storage medium for infrared data. Background Technology

[0002] Infrared thermal imaging technology achieves non-contact temperature measurement by detecting the radiant energy on the surface of an object, and is widely used in fields such as power line inspection and industrial equipment testing. However, due to the limited field of view of infrared detectors, a single infrared image acquisition is usually insufficient to cover the entire target area, requiring stitching technology to generate a panoramic thermal image.

[0003] Existing technologies typically employ two types of methods to achieve infrared panoramic stitching: one type involves converting raw infrared data into images and then performing weighted averaging for image stitching. This method achieves image fusion by processing the pixel layer, but the process of converting raw infrared data into images leads to a loss of accuracy in temperature measurement data. Furthermore, processing only at the pixel level ignores the actual infrared data, resulting in inaccurate temperature measurement information at the stitching points. The other type is based on Poisson fusion technology, which uses gradient fields to solve Poisson partial differential equations to achieve smooth image transitions and improve the continuity of stitching edges. However, this process requires dynamic range compression of the data, thereby compromising the physical authenticity of the temperature measurement data and making it difficult to meet the requirements of high-precision temperature measurement.

[0004] Therefore, existing solutions struggle to maintain the physical authenticity and accuracy of infrared temperature measurement data while ensuring the smoothness of panoramic images. Summary of the Invention

[0005] This application provides a method, apparatus, electronic device, and storage medium for panoramic stitching of infrared data, which improves the accuracy and overall consistency of infrared panoramic data fusion.

[0006] According to one aspect of this application, a panoramic stitching method for infrared data is provided, comprising: Obtain the original data matrix collected by the infrared detector in the current round and the previous stitched data matrix obtained from the previous round; Kernel functions are used to perform kernel mapping processing on the radiometric measurement values ​​corresponding to each pixel position in the original data matrix and the previous concatenated data matrix to obtain the original kernel feature matrix and the previous concatenated kernel feature matrix; A radial attenuation model centered on the infrared detector optical axis is constructed based on the radial distance of the pixel position relative to the infrared detector optical axis. The original kernel feature matrix and the previous stitched kernel feature matrix are then subjected to radiometric consistency correction according to the radial attenuation model to obtain the original correction matrix and the previous stitched correction matrix. The kernel feature similarity of each pixel is determined based on the kernel feature correction values ​​at each pixel position in the spatially overlapping region of the original correction matrix and the previous concatenated correction matrix, and the fusion weight matrix of the spatially overlapping region is determined based on the kernel feature similarity. Based on the fusion weight matrix, the spatial overlapping areas of the original correction matrix and the previous stitching correction matrix are fused pixel by pixel and stitched with pixels in the non-overlapping areas to obtain the current stitching data matrix for the current round.

[0007] According to another aspect of this application, a panoramic stitching device for infrared data is provided, comprising: The matrix to be stitched module is used to acquire the original data matrix collected by the infrared detector in the current round and the previous stitched data matrix obtained in the previous round. The matrix mapping module is used to perform kernel mapping processing on the radiometric measurement values ​​corresponding to each pixel position in the original data matrix and the previous spliced ​​data matrix using kernel functions to obtain the original kernel feature matrix and the previous spliced ​​kernel feature matrix. The kernel feature matrix correction module is used to construct a radial attenuation model centered on the infrared detector optical axis based on the radial distance of the pixel position relative to the infrared detector optical axis, and to perform radiometric consistency correction on the original kernel feature matrix and the previous spliced ​​kernel feature matrix according to the radial attenuation model to obtain the original correction matrix and the previous spliced ​​correction matrix. The fusion weight matrix generation module is used to determine the kernel feature similarity corresponding to each pixel based on the kernel feature correction values ​​at each pixel position in the spatially overlapping region of the original correction matrix and the previous concatenated correction matrix, and to determine the fusion weight matrix of the spatially overlapping region based on the kernel feature similarity. The current round matrix stitching module is used to perform pixel-by-pixel weighted fusion of the spatial overlapping area of ​​the original correction matrix and the previous stitching correction matrix according to the fusion weight matrix, and stitch it with the pixels in the non-overlapping area to obtain the current stitching data matrix of the current round.

[0008] According to another aspect of this application, an electronic device is provided, the electronic device comprising: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, which enables the at least one processor to perform the panoramic stitching method for infrared data according to any embodiment of this application.

[0009] According to another aspect of this application, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to execute and implement the panoramic stitching method of infrared data according to any embodiment of this application.

[0010] According to another aspect of this application, a computer program product is provided, the computer program product including a computer program that, when executed by a processor, implements any of the panoramic stitching methods for infrared data provided in the embodiments of this application.

[0011] The technical solution of this application embodiment is based on the raw data collected by the infrared detector throughout the stitching process. Kernel mapping is used to map data from different rounds to a unified feature space. Based on this, a radial attenuation model centered on the optical axis is used to correct the radial radiation deviation introduced by the optical system. The similarity of kernel features between data from different rounds within the spatially overlapping area is evaluated at the corrected feature level. The fusion weights used for compressing the spatially overlapping area are determined based on the kernel feature similarity, achieving smooth weighted fusion during the round-by-round expansion of the field of view. By introducing a feature-based processing and radial consistency correction mechanism based on raw data during multi-round infrared panoramic stitching, the problems of radiation information distortion and discontinuous stitching transitions caused by directly stitching image-level data in existing technologies are solved. Without introducing data compression or nonlinear image enhancement, the infrared panoramic stitching results are effectively improved in terms of radiation consistency, spatial continuity, and overall accuracy.

[0012] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent from the following description. Attached Figure Description

[0013] Figure 1 This is a flowchart of a panoramic stitching method for infrared data according to Embodiment 1 of this application; Figure 2 This is a flowchart of another panoramic stitching method for infrared data provided according to Embodiment 2 of this application; Figure 3 This is a schematic diagram of the structure of a panoramic stitching device for infrared data according to Embodiment 3 of this application; Figure 4 This is a schematic diagram of the structure of an electronic device that implements the panoramic stitching method for infrared data according to the embodiments of this application. Detailed Implementation

[0014] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present application.

[0015] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0016] Example 1 Figure 1 This is a flowchart of a panoramic stitching method for infrared data according to Embodiment 1 of this application. This embodiment is applicable to application scenarios involving panoramic stitching of multiple frames of acquired infrared data. The method can be executed by a panoramic stitching device for infrared data, which can be implemented in hardware and / or software and can be configured in a computer device. Figure 1 As shown, the method includes: S101. Obtain the original data matrix collected by the infrared detector in the current round and the previous spliced ​​data matrix obtained from the previous round.

[0017] In this embodiment, the raw data matrix refers to the raw infrared pixel data organized according to the pixel array output by the infrared detector in a single acquisition round. Each matrix element corresponds to the radiometric measurement value of an infrared detector pixel. The acquired data is in 14-bit or 16-bit quantization format and has not undergone pseudo-color mapping or other nonlinear compression processing. The previous stitching data matrix refers to the stitching result data matrix generated during multiple rounds of recursive stitching operations in the infrared panoramic data stitching process. This data matrix is ​​not the final panoramic stitching result, but rather an intermediate stitching result used to participate in subsequent rounds of stitching processing. That is, after completing all predetermined rounds of stitching operations, this intermediate stitching result is gradually expanded to form the complete infrared panoramic data.

[0018] Specifically, at the beginning of each stitching round, the original data matrix collected by the infrared detector in the current round and the previous stitching data matrix formed by the previous stitching operation are acquired and used as input objects for the current stitching process, thereby achieving effective connection between the currently acquired data and the existing stitching results. By simultaneously introducing the currently acquired data and the stitching results of the previous round in each round, the stitching process can progressively expand the data coverage in a recursive manner, avoiding the problem of stitching breaks or discontinuities due to the lack of historical stitching information, which is beneficial to improving the stability and overall consistency of the panoramic data generation process. In addition, this application processes the raw signal output by the infrared detector directly without performing compression or nonlinear transformation operations on the data, so that the final panoramic stitched data can more realistically reflect the original radiation information, thereby improving the accuracy and reliability of the panoramic data results.

[0019] It should be noted that the panoramic stitching operation of infrared data in this application is performed in rounds. In each round, two frames of infrared data with overlapping fields of view are stitched together. When it is the first round of stitching operation, the data matrix for stitching operation is the original data matrix acquired in the current round and another original data matrix in the same scene but with a different field of view. When it is not the first round of stitching operation, the data matrix for stitching operation is the original data matrix acquired in the current round and the previous stitched data matrix obtained in the previous round.

[0020] S102. The kernel function is used to perform kernel mapping processing on the radiometric measurement values ​​corresponding to each pixel position in the original data matrix and the previous concatenated data matrix to obtain the original kernel feature matrix and the previous concatenated kernel feature matrix.

[0021] In this embodiment, the radiation measurement value refers to the measurement result of the target radiation energy at the corresponding pixel position by the infrared detector. It is stored in the original data matrix or the previous stitched data matrix in digital quantization form and is used to characterize the infrared radiation intensity information at that pixel position. The original kernel feature matrix refers to the feature representation matrix formed in the kernel feature space after mapping the radiation measurement values ​​corresponding to each pixel position in the original data matrix through a kernel function. Its matrix structure is consistent with the original data matrix at the pixel position and is used to characterize the distribution characteristics of the original radiation data in the kernel feature space. The previous stitched kernel feature matrix refers to the feature representation matrix obtained after mapping the radiation measurement values ​​corresponding to each pixel position in the previous stitched data matrix to the kernel feature space through the same kernel function used in the kernel mapping process of the original data matrix. This matrix serves as the expression form of the previous round of stitching results in the kernel feature space and is used for subsequent stitching operations with the original kernel feature matrix of the current round.

[0022] Specifically, after obtaining the original data matrix of the current round and the previous stitched data matrix, a kernel function is used to perform kernel mapping processing on the radiometric measurements corresponding to each pixel position in the original data matrix and the previous stitched data matrix. This transforms the original radiometric data from the original feature space to a unified kernel feature space, forming the corresponding original kernel feature matrix and the previous stitched kernel feature matrix, which serve as the feature data for subsequent stitching processing. After mapping the radiometric measurements to the kernel feature space through the kernel function, the nonlinear relationships implicit in the original feature space are expanded into linear relationships in the kernel feature space. This allows the differences and consistency of the data matrix at the feature level to be presented in a linearly evaluable manner, thus providing a more stable and continuous feature foundation for subsequent stitching operations. This improves the adaptability to complex radiometric changes during infrared panoramic stitching and reduces stitching errors caused by nonlinear radiometric data.

[0023] For example, the kernel function used can be the Hilbert kernel function, which maps the radiometric measurements corresponding to each pixel position in the original data matrix and the previous concatenated data matrix from Euclidean space to a high-dimensional Hilbert space, thereby capturing the complex nonlinear dependencies between pixel radiometric characteristics.

[0024] S103. Construct a radial attenuation model centered on the infrared detector optical axis based on the radial distance of the pixel position relative to the infrared detector optical axis, and perform radiometric consistency correction on the original kernel feature matrix and the previous spliced ​​kernel feature matrix according to the radial attenuation model to obtain the original correction matrix and the previous spliced ​​correction matrix.

[0025] In this embodiment, the infrared detector optical axis refers to the reference axis that passes through the optical center of the infrared detector and extends along the detector's field of view, used to determine the radial distance of the pixel position relative to the center of the optical system. The radial attenuation model is a mathematical model constructed with the infrared detector optical axis as the center, used to simulate the attenuation law of the radiation response intensity of each pixel in the infrared imaging system as the radial distance from the pixel to the optical axis changes, thus providing a basis for radiometric consistency correction. The original correction matrix is ​​the matrix obtained after performing radiometric consistency correction on the kernel feature values ​​of each pixel position in the original kernel feature matrix according to the radial attenuation model. This matrix eliminates the radiation intensity deviation caused by differences in the radial position of the pixels while maintaining the original kernel feature space representation. The previous stitching correction matrix is ​​the matrix obtained after performing the same radiometric consistency correction on the kernel feature values ​​of each pixel position in the previous stitching kernel feature matrix according to the radial attenuation model.

[0026] Specifically, a radial attenuation model centered on the optical axis of the infrared detector is constructed to perform radiometric consistency correction on the original kernel feature matrix and the previous stitched kernel feature matrix. This model simulates the variation of pixel radiometric response with its radial distance relative to the optical axis in an infrared imaging system. This allows the radial attenuation caused by pixel deviation from the optical axis to be compensated at the feature level, thereby standardizing the original data and historical stitched data within the kernel feature space, resulting in more accurate radiometric measurements. Through correction based on the radial attenuation model, the uneven brightness problem (bright center and dark edges) caused by radial deviation in the infrared optical system is effectively eliminated. Simultaneously, the stability and overall accuracy of data fusion during panoramic stitching are improved, enabling the final generated infrared panoramic data to more realistically and reliably reflect the radiometric distribution characteristics of the panoramic scene.

[0027] S104. Determine the kernel feature similarity for each pixel based on the kernel feature correction values ​​at each pixel position in the spatially overlapping region of the original correction matrix and the previous spliced ​​correction matrix, and determine the fusion weight matrix of the spatially overlapping region based on the kernel feature similarity.

[0028] In this embodiment, the spatially overlapping region refers to the area in the world space coordinate system where pixel positions coincide in the original correction matrix of the current round and the stitched correction matrix of the previous round, and is the target processing area for data stitching. The kernel feature correction value refers to the feature value corresponding to each pixel position in the original kernel feature matrix or the previous stitched kernel feature matrix after radial attenuation model correction. Kernel feature similarity refers to the degree of consistency between the kernel feature correction values ​​of corresponding pixels in the original correction matrix and the previous stitched correction matrix in the spatially overlapping region. The fusion weight matrix is ​​a weighted coefficient matrix assigned to each pixel in the spatially overlapping region based on the kernel feature similarity, serving as a reference for stitching and fusion of the overlapping regions in the data matrix.

[0029] Specifically, in the spatially overlapping region, the kernel feature correction values ​​corresponding to the original correction matrix and the previous stitching correction matrix represent the feature expressions of the same spatial location in different rounds. By calculating the kernel feature similarity between the two, the degree of feature consistency between the pixel in the current round of data and the historical stitching results can be quantified. Based on the kernel feature similarity, it is mapped to the corresponding fusion weight matrix of the spatially overlapping region, which is used to guide the pixel-level fusion processing in the spatially overlapping region. This method enables pixels with high feature consistency to have a smoother transition effect during the fusion process, while pixels with large feature differences can reduce the data abruptness caused by direct superposition by adjusting the weights during fusion, thus achieving continuity of the stitching result in both spatial and radial dimensions.

[0030] S105. Based on the fusion weight matrix, perform pixel-by-pixel weighted fusion on the spatial overlapping area of ​​the original correction matrix and the previous stitching correction matrix, and stitch it with the pixels in the non-overlapping area to obtain the current stitching data matrix of the current round.

[0031] In this embodiment, the current stitching data matrix refers to the stitching result data matrix formed after the completion of the current stitching round. This data matrix is ​​used as a historical stitching result in subsequent rounds for further panoramic stitching processing. Specifically, after obtaining the fusion weight matrix, the original correction matrix and the previous stitching correction matrix are weighted and fused pixel-by-pixel according to pixel position in the spatially overlapping area based on the fusion weight matrix. The fused result is then stitched with pixels in the non-overlapping area to form the stitching result of the current round, i.e., the current stitching matrix. By introducing a feature consistency-based fusion weight adjustment mechanism in the overlapping area, data from different rounds can be smoothly connected in the transition area, while maintaining the integrity of the original data in the non-overlapping area, avoiding information loss. This is beneficial for continuously expanding the field of view and maintaining radiation and spatial continuity during multi-round recursive stitching, thereby improving the stability and overall quality of the final panoramic data construction.

[0032] The technical solution of this application embodiment is based on the raw data collected by the infrared detector throughout the stitching process. Kernel mapping is used to map data from different rounds to a unified feature space. Based on this, a radial attenuation model centered on the optical axis is used to correct the radial radiation deviation introduced by the optical system. The similarity of kernel features between data from different rounds within the spatially overlapping area is evaluated at the corrected feature level. The fusion weights used for compressing the spatially overlapping area are determined based on the kernel feature similarity, achieving smooth weighted fusion during the round-by-round expansion of the field of view. By introducing a feature-based processing and radial consistency correction mechanism based on raw data during multi-round infrared panoramic stitching, the problems of radiation information distortion and discontinuous stitching transitions caused by directly stitching image-level data in existing technologies are solved. Without introducing data compression or nonlinear image enhancement, the infrared panoramic stitching results are effectively improved in terms of radiation consistency, spatial continuity, and overall accuracy.

[0033] Example 2 Figure 2 This is a flowchart of another panoramic stitching method for infrared data provided in Embodiment 2 of this application. The technical solution of this embodiment further refines the method for determining the fusion weight matrix based on the technical solution of the above embodiments. For example... Figure 2 As shown, the method includes: S201. Obtain the original data matrix collected by the infrared detector in the current round and the previous spliced ​​data matrix obtained from the previous round.

[0034] S202. Kernel functions are used to perform kernel mapping on the radiometric measurements corresponding to each pixel position in the original data matrix and the previous concatenated data matrix to obtain the original kernel feature matrix and the previous concatenated kernel feature matrix.

[0035] S203. Construct a radial attenuation model centered on the infrared detector optical axis based on the radial distance of the pixel position relative to the infrared detector optical axis, and perform radiometric consistency correction on the original kernel feature matrix and the previous spliced ​​kernel feature matrix according to the radial attenuation model to obtain the original correction matrix and the previous spliced ​​correction matrix.

[0036] S204. Obtain the original correction value and the previous stitching correction value corresponding to each pixel position in the spatially overlapping area of ​​the original correction matrix and the previous stitching correction matrix.

[0037] S205. For each pixel position within the spatially overlapping region, calculate the kernel feature similarity between the original correction value and the previous stitched correction value at the same pixel position to obtain the initial weight matrix.

[0038] Kernel feature similarity is used to characterize the degree of consistency between the kernel feature representations of two frames of data at that pixel location.

[0039] S206. Normalize the initial weight matrix to obtain the fused weight matrix.

[0040] The fusion weight matrix is ​​used to control the relative contribution ratio of the original correction matrix and the previous stitching correction matrix in the pixel stitching process.

[0041] In this embodiment, the original correction value and the previous stitching correction value are the correction values ​​corresponding to each pixel position in the original correction matrix and the previous stitching correction matrix, respectively. These original correction values ​​and the previous stitching correction values ​​are the feature representation results after correction based on the original data through feature mapping and radial attenuation model, reflecting the standardized feature results of each corresponding pixel in the current round. The initial weight matrix refers to the initial weight distribution established for each pixel in the spatially overlapping region based on the consistency relationship between the original correction value and the previous stitching correction value in the kernel feature space. Its function is to introduce a continuous weight reference for pixel fusion in the spatially overlapping region.

[0042] Specifically, after completing the radiometric consistency correction, the correction values ​​of the original correction matrix and the previous stitched correction matrix at the corresponding pixel positions within the spatially overlapping region are extracted; these are the original correction values ​​and the previous stitched correction values. Based on this, kernel feature similarity is calculated between the correction values ​​at the same pixel position to obtain an initial weight matrix reflecting the degree of pixel similarity within the spatially overlapping region. This initial weight matrix is ​​then normalized to ensure that the weights corresponding to each pixel have a uniform scale and operability, resulting in a fusion weight matrix that can be used for pixel-by-pixel weighted fusion. By constructing the initial weight matrix using kernel feature similarity as the basis for weight adjustment and normalizing it, a fusion weight matrix that can be directly used for pixel-level weighted fusion is generated. This improves the stability of the fusion weights in terms of numerical scale and spatial distribution, avoids excessive adjustment of the stitching result due to local feature fluctuations, and provides a smoother and more controllable weight foundation for subsequent pixel fusion processes.

[0043] S207. Based on the fusion weight matrix, perform pixel-by-pixel weighted fusion on the spatial overlapping area of ​​the original correction matrix and the previous stitching correction matrix, and stitch it with the pixels in the non-overlapping area to obtain the current stitching data matrix of the current round.

[0044] The technical solution in this embodiment further refines the method for determining the fusion weight matrix. By refining the fusion weight determination process based on kernel feature similarity into two stages—initial weight generation and normalization processing—kernel feature similarity is used as an adjustment basis reflecting pixel feature consistency in weight construction, rather than being directly used as a fusion decision factor. This results in a fusion weight distribution constrained by the overall system. This approach not only improves the stability of the fusion weights on a numerical scale and suppresses the amplified impact of local feature fluctuations on the stitching result, but also enhances the continuity and smoothness of weight transfer during multi-round recursive stitching.

[0045] In one optional implementation, after acquiring the original data matrix collected by the infrared detector in the current round and the previous stitched data matrix obtained in the previous round, the method further includes: mapping the original data matrix and the previous stitched data matrix to the same spatial coordinate system to obtain the original spatial position coordinates and stitched spatial position coordinates corresponding to each pixel position; determining the spatial overlap range based on the original spatial position coordinates and stitched spatial position coordinates, and determining the overlapping pixel range corresponding to the spatial overlap range based on the correspondence between the spatial position coordinates and pixel positions; and determining the corresponding matrix regions in the original data matrix and the previous stitched data matrix as the spatial overlap region based on the overlapping pixel range.

[0046] In this embodiment, the original spatial coordinates and the stitched spatial coordinates represent the spatial positions of each pixel in the original data matrix and the previous stitched data matrix, respectively, under the same spatial coordinate system. The overlapping pixel range refers to the range of pixels determined in the corresponding matrix based on the spatial overlap relationship determined by the original and stitched spatial coordinates and the mapping relationship between spatial positions and pixel positions. Specifically, the original data matrix and the previous stitched data matrix are mapped to the same spatial coordinate system. By identifying and analyzing the position coordinates of each pixel in the same space, the spatial overlap range is determined. Based on the correspondence between spatial coordinates and pixel positions, the overlapping pixel range corresponding to the overlapping spatial range is determined. Finally, the overlapping pixel range is determined to correspond to the matrix region of the original data matrix and the previous stitched data matrix, which is then used as the spatial overlap region. By aligning the positional relationships of each pixel in a unified space, the overlapping areas corresponding to the spatial positions of two frames of data are identified, and the overlapping pixel areas participating in the stitching are accurately extracted from the matrix level. This allows subsequent data fusion operations to be carried out under the constraint of spatial consistency, solving the problem of inaccurate overlapping areas caused by differences in field of view or spatial misalignment when stitching directly based on pixel indexes. This makes the source of the data participating in the fusion clear and the spatial correspondence clear.

[0047] In one optional implementation, a radial attenuation model centered on the infrared detector optical axis is constructed based on the radial distance of the pixel position relative to the infrared detector optical axis. This includes: constructing an initial radial attenuation model with the radial distance as the independent variable, wherein the initial radial attenuation model adopts an even-degree polynomial attenuation model to characterize the attenuation characteristics of the radiation measurement value as the radial distance changes during imaging; extracting the kernel feature values ​​of corresponding pixel positions at the same spatial location from the overlapping area of ​​the original kernel feature matrix and the stitched kernel feature matrix to form spatial point pairs; constructing an observation equation based on the spatial point pairs and the initial radial attenuation model, and solving the polynomial coefficients in the observation equation using the least squares method to obtain the solved radial attenuation model.

[0048] In this embodiment, the initial radial attenuation model is a function model established with the radial distance of the pixel relative to the optical axis of the infrared detector as the independent variable. It is used to describe the attenuation law of the radiation measurement value in the imaging system as the radial position changes. The observation equation is a function relationship equation established with the kernel feature value of the corresponding pixel in the spatially overlapping region as the observation and the initial radial attenuation model. It is used to solve the model parameters so that the model can accurately characterize the actual measured radiation attenuation characteristics.

[0049] Specifically, to correct the radial radiation deviation caused by the optical structure of the infrared imaging system, an initial radial attenuation model is constructed based on the radial distance of the pixel position relative to the optical axis of the infrared detector. An even-degree polynomial fitting method is used to describe the attenuation characteristics of the radiation measurement values ​​with radial position. Kernel feature values ​​corresponding to the spatial positions are extracted from the overlapping region of the original kernel feature matrix and the previous stitching kernel feature matrix to form spatial point pairs. Based on the spatial point pairs and the initial radial attenuation model, observation equations are established and solved to obtain the solved radial attenuation model. By constructing a radial attenuation model based on the radiation characteristics in the radial direction of the optical axis for radiation consistency correction, the radiation deviation problem caused by the inherent non-uniformity of the optical system is avoided. This provides more reliable data support for subsequent pixel-level weighted fusion, enhancing the radiation consistency and spatial continuity of the panoramic stitching results.

[0050] For example, obtain the original kernel feature matrix and the stitched kernel feature matrix of two frames with overlapping fields of view. Construct a radiation attenuation approximation with respect to radial distance. even-degree polynomials As the initial radial attenuation model, where, These are the parameters to be solved. Selected in the spatially overlapping region. For each pair of spatial points, the observation equations are constructed as follows: ; in, The original kernel feature matrix in the overlapping region The eigenvalues ​​at each pair of spatial points To splice the kernel feature matrix of the first The eigenvalues ​​at each pair of spatial points The original kernel feature matrix in the overlapping region The radial distance between two points To splice the kernel feature matrix in the overlapping region The radial distance between pairs of points. Solved using the least squares method. The parameters of the initial radial decay model are updated to obtain the solved radial decay model.

[0051] In one optional implementation, a radial attenuation model centered on the infrared detector optical axis is constructed based on the radial distance of the pixel position relative to the infrared detector optical axis. This includes: constructing an initial radial attenuation model centered on the infrared detector optical axis using a Gaussian function with the radial distance of the pixel relative to the optical axis as the independent variable; calculating the radial distance of each pixel position relative to the infrared detector optical axis, and determining the kernel width value at the corresponding pixel position in the Gaussian function model based on the radial distance and a preset kernel width mapping rule; and introducing the determined kernel width value into the initial radial attenuation model to obtain the solved radial attenuation model.

[0052] In this embodiment, the kernel width mapping rule refers to establishing a correspondence between the radial position of a pixel relative to the optical axis and the Gaussian kernel width based on the radial vignetting characteristic of the infrared optical system's radiation response gradually changing from the center to the edge within the field of view. This allows the kernel width to adaptively adjust with the pixel's position within the field of view. The kernel width value is a Gaussian function scale parameter determined for each pixel position under the kernel width mapping rule constraint. It controls the range of influence of the radial attenuation model on the neighborhood radiation information at that pixel position, and its magnitude reflects the balance required between radiation smoothing intensity and detail preservation in that pixel region.

[0053] Specifically, with the optical axis of the infrared detector as the center, the radial distance of each pixel relative to the optical axis is used as the independent variable. A Gaussian function is introduced to construct an initial radial attenuation model. Based on this, the radial distance from each pixel position to the optical axis is calculated. Combined with a preset kernel width mapping relationship, the scale of the Gaussian function can be adaptively adjusted according to the radial position. Thus, the corresponding kernel width parameter is introduced into the radial attenuation model to complete the model solution. This results in a radial attenuation model that can maintain a sensitive characterization of subtle radiation changes in the central region of the field of view, while effectively smoothing the nonlinear radiation decrease caused by vignetting in the edge region of the field of view. By introducing a Gaussian function to construct the radial attenuation model, the radial attenuation correction process balances detail preservation and radiation smoothing across the entire field of view, avoiding the problems of over-smoothing at the center or insufficient correction at the edges of traditional fixed-scale models. This improves the physical rationality and overall stability of infrared radiation consistency correction, providing a more reliable radiation basis for subsequent stitching and fusion.

[0054] In one optional embodiment, the method further includes: calculating the radial distance from each pixel position to the optical axis of the infrared detector, and determining the kernel width value at the corresponding pixel position in the Gaussian function model based on the radial distance and a preset kernel width mapping rule; wherein the kernel width mapping rule includes: when the radial distance of the pixel is less than a first threshold, selecting a first kernel width value; when the radial distance of the pixel is between the first threshold and a second threshold, adjusting the kernel width value linearly or nonlinearly according to the increase of the radial distance; and when the radial distance of the pixel is greater than the second threshold, selecting a second kernel width value.

[0055] Specifically, the first threshold defines the core region within the field of view where the radiation response is most stable and the vignetting effect is weakest. Radiation changes within this region, with a radial distance less than the first threshold, primarily reflect the high-frequency thermal characteristics of the real target. The corresponding first kernel width value emphasizes scale convergence, avoiding weakening of detailed information in the central region. The second threshold identifies the starting position of the edge region where the vignetting effect significantly intensifies. Within this region, with a radial distance greater than the second threshold, the radiance exhibits a significant non-linear decrease with increasing field of view. The corresponding second kernel width value emphasizes scale expansion, suppressing radiation jumps and grayscale discontinuities by amplifying the smoothing effect range. The second kernel width value is greater than the first kernel width value. In the transition region between the two, the smoothing intensity is continuously adjusted by increasing the kernel width with radial distance, ensuring that the kernel width change aligns with the physical evolution of infrared vignetting from weak to strong, thus avoiding the introduction of new radiation discontinuities due to abrupt scale changes. By setting a collaborative design that links the threshold to the kernel width, the radial attenuation correction process can effectively protect high-frequency thermally significant details at the center of the field of view and fully smooth out radiation anomalies caused by vignetting at the edge of the field of view. This achieves continuity, stability, and physical rationality of radiation correction across the entire field of view, providing a consistent and reliable radiation basis for subsequent stitching and fusion.

[0056] Example 3 Figure 3 This is a structural schematic diagram of a panoramic stitching device for infrared data according to Embodiment 3 of this application. This embodiment is applicable to application scenarios involving panoramic stitching of multiple frames of acquired infrared data. The panoramic stitching device for infrared data can be implemented in hardware and / or software, and can be configured in a computer device. Figure 3 As shown, the panoramic stitching device 300 for infrared data includes: The matrix acquisition module 310 is used to acquire the original data matrix collected by the infrared detector in the current round and the previous spliced ​​data matrix obtained in the previous round. The matrix mapping module 320 is used to perform kernel mapping processing on the radiometric measurement values ​​corresponding to each pixel position in the original data matrix and the previous spliced ​​data matrix using kernel functions to obtain the original kernel feature matrix and the previous spliced ​​kernel feature matrix. The kernel feature matrix correction module 330 is used to construct a radial attenuation model centered on the infrared detector optical axis based on the radial distance of the pixel position relative to the infrared detector optical axis, and to perform radiometric consistency correction on the original kernel feature matrix and the previous spliced ​​kernel feature matrix according to the radial attenuation model to obtain the original correction matrix and the previous spliced ​​correction matrix. The fusion weight matrix generation module 340 is used to determine the kernel feature similarity corresponding to each pixel based on the kernel feature correction values ​​at each pixel position in the spatially overlapping region of the original correction matrix and the previous spliced ​​correction matrix, and to determine the fusion weight matrix of the spatially overlapping region based on the kernel feature similarity. The current round matrix stitching module 350 is used to perform pixel-by-pixel weighted fusion of the spatial overlapping area of ​​the original correction matrix and the previous stitching correction matrix according to the fusion weight matrix, and stitch it with the pixels in the non-overlapping area to obtain the current stitching data matrix of the current round.

[0057] In one optional implementation, the fusion weight matrix generation module 340 is specifically used for: Obtain the original correction value and the previous stitching correction value corresponding to each pixel position in the spatially overlapping area of ​​the original correction matrix and the previous stitching correction matrix; For each pixel position within the spatially overlapping region, the kernel feature similarity between the original correction value and the previous stitched correction value at the same pixel position is calculated to obtain an initial weight matrix; wherein, the kernel feature similarity is used to characterize the degree of consistency of the kernel feature representation of the two frames of data at that pixel position; The initial weight matrix is ​​normalized to obtain the fused weight matrix; wherein the fused weight matrix is ​​used to control the relative contribution ratio of the original correction matrix and the previous stitching correction matrix in the pixel stitching process.

[0058] In one optional embodiment, the panoramic stitching device 300 for infrared data further includes a spatial overlap region determination module, which is specifically used for: The original data matrix and the previous spliced ​​data matrix are mapped to the same spatial coordinate system to obtain the original spatial position coordinates and spliced ​​spatial position coordinates corresponding to each pixel position; Based on the original spatial coordinates and the spliced ​​spatial coordinates, the spatial overlap range is determined, and the overlapping pixel range corresponding to the spatial overlap range is determined based on the correspondence between the spatial coordinates and the pixel positions. Based on the overlapping pixel range, the corresponding matrix regions in the original data matrix and the previous spliced ​​data matrix are determined as spatial overlapping regions.

[0059] In one optional embodiment, the kernel feature matrix correction module 330 further includes a radial attenuation model construction unit, which is specifically used for: Based on the radial distance of the pixel position relative to the optical axis of the infrared detector, an initial radial attenuation model is constructed with the radial distance as the independent variable. The initial radial attenuation model adopts an even-degree polynomial attenuation model to characterize the attenuation characteristics of the radiation measurement value as the radial distance changes during the imaging process. Kernel feature values ​​of corresponding pixel positions at the same spatial location are extracted from the overlapping area of ​​the original kernel feature matrix and the spliced ​​kernel feature matrix to form spatial point pairs; The observation equation is constructed based on the spatial point pair and the initial radial decay model, and the polynomial coefficients in the observation equation are solved by the least squares method to obtain the solved radial decay model.

[0060] In one optional embodiment, the kernel feature matrix correction module 330 further includes a radial attenuation model construction unit, which is specifically used for: An initial radial attenuation model is constructed using a Gaussian function with the radial distance of the pixel relative to the optical axis as the independent variable, centered on the optical axis of the infrared detector. Calculate the radial distance of each pixel position relative to the optical axis of the infrared detector, and determine the kernel width value at the corresponding pixel position in the Gaussian function model based on the radial distance and the preset kernel width mapping rule; The determined kernel width value is introduced into the initial radial decay model to obtain the solved radial decay model.

[0061] In one optional embodiment, the kernel feature matrix correction module 330 further includes a radial attenuation model construction unit, which is specifically used for: Calculate the radial distance from each pixel position to the optical axis of the infrared detector, and determine the kernel width value at the corresponding pixel position in the Gaussian function model based on the radial distance and the preset kernel width mapping rule; The kernel width mapping rules include: When the radial distance of a pixel is less than the first threshold, the first kernel width value is selected; When the radial distance of a pixel is between the first threshold and the second threshold, the kernel width value is adjusted linearly or non-linearly as the radial distance increases. When the radial distance of a pixel is greater than the second threshold, the second kernel width value is selected.

[0062] The panoramic stitching device for infrared data provided in this application can execute the panoramic stitching method for infrared data provided in any embodiment of this application, and has the corresponding functional modules and beneficial effects of executing the method.

[0063] This application also provides an electronic device, a readable storage medium, and a computer program product. The computer-readable storage medium stores a computer program that, when executed by a processor, implements the panoramic stitching method for arbitrary infrared data as described in this application.

[0064] Example 4 Figure 4 This is a schematic diagram of the structure of an electronic device that implements the panoramic stitching method for infrared data according to the embodiments of this application. Figure 4 A schematic diagram of an electronic device 410 that can be used to implement embodiments of this application is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (such as helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the application described and / or claimed herein.

[0065] like Figure 4 As shown, the electronic device 410 includes at least one processor 411 and a memory, such as a read-only memory (ROM) 412 or a random access memory (RAM) 413, communicatively connected to the at least one processor 411. The memory stores computer programs executable by the at least one processor. The processor 411 can perform various appropriate actions and processes based on the computer program stored in the ROM 412 or loaded from storage unit 418 into the RAM 413. The RAM 413 may also store various programs and data required for the operation of the electronic device 410. The processor 411, ROM 412, and RAM 413 are interconnected via a bus 414. An input / output (I / O) interface 415 is also connected to the bus 414.

[0066] Multiple components in electronic device 410 are connected to I / O interface 415, including: input unit 416, such as keyboard, mouse, etc.; output unit 417, such as various types of displays, speakers, etc.; storage unit 418, such as disk, optical disk, etc.; and communication unit 419, such as network card, modem, wireless transceiver, etc. Communication unit 419 allows electronic device 410 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0067] Processor 411 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 411 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 411 performs the various methods and processes described above, such as panoramic stitching methods for infrared data.

[0068] In some embodiments, the panoramic stitching method for infrared data can be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 418. In some embodiments, part or all of the computer program can be loaded and / or installed on electronic device 410 via ROM 412 and / or communication unit 419. When the computer program is loaded into RAM 413 and executed by processor 411, one or more steps of the panoramic stitching method for infrared data described above can be performed. Alternatively, in other embodiments, processor 411 can be configured to perform the panoramic stitching method for infrared data by any other suitable means (e.g., by means of firmware).

[0069] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0070] Computer programs used to implement the methods of this application may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0071] In the context of this application, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium can be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0072] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0073] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0074] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0075] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this application can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this application can be achieved, and this is not limited herein.

[0076] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A panoramic stitching method for infrared data, characterized in that, include: Obtain the original data matrix collected by the infrared detector in the current round and the previous stitched data matrix obtained from the previous round; Kernel functions are used to perform kernel mapping processing on the radiometric measurement values ​​corresponding to each pixel position in the original data matrix and the previous concatenated data matrix to obtain the original kernel feature matrix and the previous concatenated kernel feature matrix; A radial attenuation model centered on the infrared detector optical axis is constructed based on the radial distance of the pixel position relative to the infrared detector optical axis. The original kernel feature matrix and the previous stitched kernel feature matrix are then subjected to radiometric consistency correction according to the radial attenuation model to obtain the original correction matrix and the previous stitched correction matrix. The kernel feature similarity of each pixel is determined based on the kernel feature correction values ​​at each pixel position in the spatially overlapping region of the original correction matrix and the previous concatenated correction matrix, and the fusion weight matrix of the spatially overlapping region is determined based on the kernel feature similarity. Based on the fusion weight matrix, the spatial overlapping areas of the original correction matrix and the previous stitching correction matrix are fused pixel by pixel and stitched with pixels in the non-overlapping areas to obtain the current stitching data matrix for the current round.

2. The method according to claim 1, characterized in that, The step of determining the kernel feature similarity for each pixel based on the kernel feature correction values ​​at each pixel position in the spatially overlapping region of the original correction matrix and the previous concatenated correction matrix, and determining the fusion weight matrix for the spatially overlapping region based on the kernel feature similarity, includes: Obtain the original correction value and the previous stitching correction value corresponding to each pixel position in the spatially overlapping area of ​​the original correction matrix and the previous stitching correction matrix; For each pixel position within the spatially overlapping region, the kernel feature similarity between the original correction value and the previous stitched correction value at the same pixel position is calculated to obtain an initial weight matrix; wherein, the kernel feature similarity is used to characterize the degree of consistency of the kernel feature representation of the two frames of data at that pixel position; The initial weight matrix is ​​normalized to obtain the fused weight matrix; wherein the fused weight matrix is ​​used to control the relative contribution ratio of the original correction matrix and the previous stitching correction matrix in the pixel stitching process.

3. The method according to claim 1, characterized in that, After obtaining the raw data matrix collected by the infrared detector in the current round and the stitched data matrix obtained in the previous round, the following is also included: The original data matrix and the previous spliced ​​data matrix are mapped to the same spatial coordinate system to obtain the original spatial position coordinates and spliced ​​spatial position coordinates corresponding to each pixel position; Based on the original spatial coordinates and the spliced ​​spatial coordinates, the spatial overlap range is determined, and the overlapping pixel range corresponding to the spatial overlap range is determined based on the correspondence between the spatial coordinates and the pixel positions. Based on the overlapping pixel range, the corresponding matrix regions in the original data matrix and the previous spliced ​​data matrix are determined as spatial overlapping regions.

4. The method according to claim 1, characterized in that, The radial attenuation model, constructed based on the radial distance of the pixel position relative to the infrared detector's optical axis and centered on the infrared detector's optical axis, includes: Based on the radial distance of the pixel position relative to the optical axis of the infrared detector, an initial radial attenuation model is constructed with the radial distance as the independent variable. The initial radial attenuation model adopts an even-degree polynomial attenuation model to characterize the attenuation characteristics of the radiation measurement value as the radial distance changes during the imaging process. Kernel feature values ​​of corresponding pixel positions at the same spatial location are extracted from the overlapping area of ​​the original kernel feature matrix and the spliced ​​kernel feature matrix to form spatial point pairs; The observation equation is constructed based on the spatial point pair and the initial radial decay model, and the polynomial coefficients in the observation equation are solved by the least squares method to obtain the solved radial decay model.

5. The method according to claim 1, characterized in that, The radial attenuation model, constructed based on the radial distance of the pixel position relative to the infrared detector's optical axis and centered on the infrared detector's optical axis, includes: An initial radial attenuation model is constructed using a Gaussian function with the radial distance of the pixel relative to the optical axis as the independent variable, centered on the optical axis of the infrared detector. Calculate the radial distance of each pixel position relative to the optical axis of the infrared detector, and determine the kernel width value at the corresponding pixel position in the Gaussian function model based on the radial distance and the preset kernel width mapping rule; The determined kernel width value is introduced into the initial radial decay model to obtain the solved radial decay model.

6. The method according to claim 5, characterized in that, The method further includes: Calculate the radial distance from each pixel position to the optical axis of the infrared detector, and determine the kernel width value at the corresponding pixel position in the Gaussian function model based on the radial distance and the preset kernel width mapping rule; The kernel width mapping rules include: When the radial distance of a pixel is less than the first threshold, the first kernel width value is selected; When the radial distance of a pixel is between the first threshold and the second threshold, the kernel width value is adjusted linearly or non-linearly as the radial distance increases. When the radial distance of a pixel is greater than the second threshold, the second kernel width value is selected.

7. A panoramic stitching device for infrared data, characterized in that, include: The matrix to be stitched module is used to acquire the original data matrix collected by the infrared detector in the current round and the previous stitched data matrix obtained in the previous round. The matrix mapping module is used to perform kernel mapping processing on the radiometric measurement values ​​corresponding to each pixel position in the original data matrix and the previous spliced ​​data matrix using kernel functions to obtain the original kernel feature matrix and the previous spliced ​​kernel feature matrix. The kernel feature matrix correction module is used to construct a radial attenuation model centered on the infrared detector optical axis based on the radial distance of the pixel position relative to the infrared detector optical axis, and to perform radiometric consistency correction on the original kernel feature matrix and the previous spliced ​​kernel feature matrix according to the radial attenuation model to obtain the original correction matrix and the previous spliced ​​correction matrix. The fusion weight matrix generation module is used to determine the kernel feature similarity corresponding to each pixel based on the kernel feature correction values ​​at each pixel position in the spatially overlapping region of the original correction matrix and the previous concatenated correction matrix, and to determine the fusion weight matrix of the spatially overlapping region based on the kernel feature similarity. The current round matrix stitching module is used to perform pixel-by-pixel weighted fusion of the spatial overlapping area of ​​the original correction matrix and the previous stitching correction matrix according to the fusion weight matrix, and stitch it with the pixels in the non-overlapping area to obtain the current stitching data matrix of the current round.

8. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the panoramic stitching method of infrared data according to any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the panoramic stitching method for infrared data as described in any one of claims 1-6.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the panoramic stitching method for infrared data according to any one of claims 1-6.

Images