A method and apparatus for detecting weak infrared targets based on frequency domain data spherization

Abstract

This invention discloses an infrared weak target detection method and apparatus based on frequency domain data spherization. It includes: (1) constructing a three-dimensional spatiotemporal tensor from the image frame to be detected and its preceding and following adjacent image frames; (2) converting from the spatial domain to the frequency domain; (3) designing and applying a high-pass filter to obtain a high-pass filter tensor; (4) constructing a frequency domain spherized space and mapping the high-pass filter tensor to the spherized space to obtain a frequency domain spherized tensor; (5) converting to the spatial domain to obtain a spherized tensor; and (6) designing an information filter and applying it to the central frontal slice to obtain a target detection result matrix T. t The above calculation is repeated for each frame image to be detected, thereby obtaining a target detection sequence as the infrared weak target detection result, thus realizing infrared weak target detection. The infrared weak target detection method based on frequency domain data spherization of the present invention can effectively improve the detection performance of infrared weak targets.

Description

Technical Field

[0001] This invention relates to the field of image processing, and in particular to a method and apparatus for detecting small infrared targets based on frequency domain data spherization. Background Technology

[0002] Benefiting from the "all-weather, strong penetration, and thermal sensitivity" characteristics of infrared imaging technology, infrared small target detection has become a research hotspot. In various land, sea, and air scenarios, infrared small target detection aims to achieve continuous detection of targets such as drones and vehicles at long distances. Currently, infrared small target detection faces several challenges: First, infrared targets are small and weak, typically appearing as point shapes ranging from 2×2 to 9×9 pixels. Under various radiation interferences such as atmospheric clouds, the target signal strength weakens, making feature extraction more difficult. Second, complex backgrounds often contain various background components and clutter interference, including cloud cover, atmospheric changes, and illumination effects. Furthermore, they are affected by various environmental noises and inherent sensor noise, easily leading to false alarms in infrared small target detection algorithms. Finally, the timeliness of infrared small target detection technology is another key factor determining the practicality of infrared detection and tracking systems. How to optimize the computational efficiency of infrared small target detection technology to meet application requirements remains an unsolved problem.

[0003] Currently, infrared weak target detection methods can be broadly categorized into background estimation filtering-based methods, local feature representation-based methods, machine learning-based methods, deep learning-based methods, and component analysis-based methods. Among these, background estimation filtering-based methods are unsuitable for detecting infrared weak targets in complex scenes; methods based on local feature representation are sensitive to complex backgrounds and noise clutter; machine learning-based methods heavily rely on feature design and exhibit poor scene robustness; deep learning-based methods require large amounts of data and have high computational complexity; and component analysis-based methods depend on the accurate representation of the intrinsic features of each component. Therefore, in-depth research into efficient infrared weak target detection methods remains a key research focus. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the present invention aims to provide an infrared weak target detection method and apparatus based on frequency domain data spheroidization. This method constructs a three-dimensional spatiotemporal tensor from the infrared image frame to be detected and its adjacent infrared image frames, fully utilizing the spatiotemporal information of the infrared image sequence and transforming it to the frequency domain. By employing a designed frequency domain high-pass filter and frequency domain data spheroidization method, background signals are effectively suppressed. Furthermore, the data signal is transformed to the spatial domain, and an information filter is designed to further enhance the target detection and background suppression capabilities of the method. Comprehensive verification through embodiments demonstrates the superior performance of the infrared weak target detection method based on frequency domain data spheroidization in improving detection performance and background suppression, providing new ideas and methods for research and application in related fields.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] This invention discloses a method for detecting weak infrared targets based on frequency domain data spherization, characterized by comprising the following steps:

[0007] Step 1): For the original thermal infrared image sequence, take the current infrared image frame D to be detected. t Centered on D t The adjacent preceding and following image frames are constructed into a three-dimensional spatiotemporal tensor.

[0008] Step 2): Convert the three-dimensional spacetime tensor Transforming from the spatial domain to the frequency domain yields the three-dimensional spatiotemporal tensor in the frequency domain.

[0009] Step 3): Design a high-pass filter and filter... The low-frequency information is extracted, while the high-frequency target information is preserved to obtain the high-pass filter tensor.

[0010] Step 4): Construct the frequency domain spherical space Ω FDS and the high-pass filter tensor Mapping to the spherical space yields the frequency domain spherical tensor.

[0011] Step 5): Sphere the frequency domain tensor Transform to the spatial domain to obtain the spherical tensor.

[0012] Step 6): Design an information filter and apply it to the spherical tensor. The center front slice Obtain the target detection result matrix T tThis allows us to obtain the target detection result matrix for each frame of the original thermal infrared image sequence, thus obtaining the target detection sequence T and realizing the detection of weak infrared targets.

[0013] The present invention also discloses an infrared weak target detection device based on spatiotemporal background reconstruction for implementing the method, comprising:

[0014] The 3D spatiotemporal tensor construction module is used to construct a 3D spatiotemporal tensor from the infrared image frame to be detected and its adjacent infrared image frames.

[0015] The spatial domain to frequency domain conversion module is used to convert the three-dimensional spatiotemporal tensor from the spatial domain to the frequency domain, so as to obtain the three-dimensional spatiotemporal tensor in the frequency domain.

[0016] The high-pass filter module is used to design high-pass filters and filter three-dimensional spatiotemporal tensors in the frequency domain;

[0017] The frequency domain data spheroidization module is used to construct the frequency domain spheroidization space and perform frequency domain data spheroidization on the high-pass filter tensor.

[0018] The information filtering module is used to design information filters and apply them to the spherical tensor. The target detection result is obtained from the central frontal slice;

[0019] The target detection result output module is used to output the infrared weak target detection result map of each frame of the infrared image sequence.

[0020] Compared with the prior art, the beneficial effects of the present invention include:

[0021] 1) This invention constructs a three-dimensional spatiotemporal tensor to fully utilize the spatiotemporal information of infrared image sequences, which helps to express background and target signals; and provides an effective background suppression method. By transforming the data to the frequency domain through Fourier transform, target features can be identified and extracted more accurately; through frequency domain high-pass filtering and frequency domain data spheroidization, the high-frequency features of the image can be significantly improved and low-frequency signals can be filtered out.

[0022] 2) The information filter designed in this invention can further improve the background suppression and target detection capabilities of the detector, ensure the accurate identification and extraction of small targets, and complete the detection of weak targets in the image frame to be detected with high quality. Furthermore, this method is applicable to infrared images under various scenarios and conditions and has good versatility. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of constructing a three-dimensional spacetime tensor in this invention;

[0024] Figure 2This is a schematic diagram of the infrared weak target detection device based on frequency domain data spheroidization in this invention;

[0025] Figure 3 Example frame images of a thermal infrared image sequence used for experimental testing;

[0026] Figure 4 The thermal infrared weak target detection results are obtained by using the method proposed in this invention on example frames of thermal infrared image sequences.

[0027] Figure 5 The image shows the original image of a thermal infrared image example frame and the thermal infrared weak target detection results detected by the methods of this invention, GSWLCM, NRAM, STT-TRNR, SSTTC and TWTVR. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this invention clearer, the invention is described in detail below with reference to specific embodiments. Specific embodiments are described below to simplify the invention. However, it should be understood that the invention is not limited to the described embodiments, and various modifications are possible without departing from the basic principles; these equivalent forms also fall within the scope defined by the appended claims.

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0030] The basic steps of the infrared weak target detection method based on frequency domain data spheroidization of the present invention in this embodiment mainly include:

[0031] Step 1: For the original thermal infrared image sequence, take the current infrared image frame D to be detected. t Centered on D t The adjacent preceding and following image frames are constructed into a three-dimensional spatiotemporal tensor.

[0032] Specifically, such as Figure 1 As shown, for the original thermal infrared image sequence, with the current infrared image frame D to be detected... t Centered on the image D of the preceding and following frames, the images adjacent to it are... t-l ,…,D t-2 D t-1 D t+1 D t+2 ,…,D t+l Following the temporal order of the image frames, 2l+1 consecutive images are stacked sequentially to construct a three-dimensional spacetime tensor. Its size is n1×n2×n3, where n1 represents The width, n2 represents The height, n3 represents The thickness, and n3 = 2l + 1; where, D t-1 Indicates that it is located at D t The first frame image before, D t-2 Indicates that it is located at D t The second frame image before, D t-l Indicates that it is located at D t The l-th frame image before, D t+1 Indicates that it is located at D t The first frame image after D t+2 Indicates that it is located at D t The second frame image after D t+l Indicates that it is located at D t The l-th frame of the image. Typically, but not limited to, l can be 1-5.

[0033] In the specific embodiment, l is set to 3, and the constructed three-dimensional spacetime tensor The dimensions are 256×256×7.

[0034] Step 2: Convert the three-dimensional spacetime tensor Transforming from the spatial domain to the frequency domain yields the three-dimensional spatiotemporal tensor in the frequency domain.

[0035] Specifically, the three-dimensional spacetime tensor Each frame of image D t-k Transforming from the spatial domain to the frequency domain yields the frequency domain representation F. t-k :

[0036]

[0037] Where tk represents the tk-th frame in the original thermal infrared image sequence, and -l ≤ k ≤ l, x represents the horizontal coordinate in the spatial domain, y represents the vertical coordinate in the spatial domain, (u, v) represents the frequency domain coordinates, and F t-k (u,v) represents the signal value at coordinates (u,v) in the frequency domain image;

[0038] Thus, the three-dimensional spacetime tensor is realized. Transforming from the spatial domain to the frequency domain yields the three-dimensional spatiotemporal tensor in the frequency domain.

[0039] Step 3: Design a high-pass filter and filter... The low-frequency information is extracted, while the high-frequency target information is preserved to obtain the high-pass filter tensor.

[0040] Specifically, a high-pass filter H(u,v) is designed and defined as follows:

[0041]

[0042] Where d0 is the cutoff frequency, and d(u,v) is the distance from the frequency point to the frequency center, we have:

[0043]

[0044] Using a high-pass filter H(u,v) to filter the three-dimensional spatiotemporal tensor in the frequency domain Low-frequency information is preserved, while high-frequency target information is retained; for the three-dimensional spatiotemporal tensor in the frequency domain... Each frontal slice F t-k ,have

[0045]

[0046] From this, the high-pass filter tensor can be obtained.

[0047] In the specific embodiment, d0 is set to 60.

[0048] Step 4: Construct the frequency domain spherical space Ω FDS and the high-pass filter tensor Mapping to the spherical space yields the frequency domain spherical tensor.

[0049] Specifically, for the high-pass filter tensor The centered high-pass filter tensor is obtained by performing a centered process.

[0050]

[0051] Where m(·) represents the mean calculation operator;

[0052] Center the high-pass filter tensor Expand into a two-dimensional matrix get Where reshape(·) represents the data reconstruction operator, and [n1n2,n3] represents... The dimensions are: width n1n2, height n3;

[0053] Based on two-dimensional matrix Construct the corresponding spheroidized space in the frequency domain; first calculate The covariance matrix ∑:

[0054]

[0055] in, Representation matrix Transpose of;

[0056] Perform eigenvalue decomposition on the covariance matrix ∑:

[0057] ∑=UΛU T (7)

[0058] Where U represents the eigenvector matrix and Λ represents the diagonal matrix;

[0059] Using U and Λ, establish the frequency domain spherical space Ω FDS The projection matrix S corresponding to this frequency domain spherical space is

[0060]

[0061] Two-dimensional matrix Mapped to the frequency domain spherical space Ω FDS In the process, the frequency domain sphericization matrix is ​​obtained. Among them, Ω FDS (·) denotes the frequency domain spherification operator; furthermore, the frequency domain spherification matrix F sphered Reconstructed into a frequency domain spherical tensor of size n1×n2×n3

[0062] Step 5: Spherize the frequency domain tensor Transform to the spatial domain to obtain the spherical tensor.

[0063] Specifically, for the frequency domain spheroidized tensor obtained in step 4 The spherical tensor is obtained by transforming back to the spatial domain using the inverse Fourier transform. For the frequency domain spheric tensor Frontal slice have

[0064]

[0065] Where x represents the abscissa in the spatial domain, y represents the ordinate in the spatial domain, and (u,v) represents the frequency domain coordinates. For frequency domain spherization tensor Frontal slice The signal value with coordinates (u, v) in the middle. Let represent the spherical image corresponding to the tk-th frame in the original infrared image sequence, and -l ≤ k ≤ l, from which the spherical tensor is obtained.

[0066] Step 6: Design an information filter and apply it to the spherical tensor The center front slice Obtain the target detection result matrix Tt This leads to the target detection result matrix of each frame of infrared image in the original thermal infrared image sequence, thus obtaining the target detection sequence T and realizing the detection of weak infrared targets.

[0067] Specifically, for the spherical tensor obtained in step 5 The center front slice An information filter is constructed to filter out low-response pixel values, thereby enhancing high-frequency target information and obtaining the infrared image frame D to be detected. t The corresponding target detection result T t :

[0068]

[0069] in, Indicates will An operator that projects elements whose index value belongs to set ξ onto themselves and projects all other elements to zero. Indicates obtaining The operator for the index set ξ corresponding to the k largest elements in the middle, k = αn1n2, where α is an adjustable positive constant;

[0070] In the specific embodiment, α is set to 0.0001;

[0071] For each infrared image frame to be detected in the original thermal infrared image sequence, repeat the steps 1 to 6 to obtain the target detection result sequence T, which serves as the infrared weak target detection result of the original infrared thermal infrared image sequence, thus realizing infrared weak target detection based on frequency domain data spherization.

[0072] Corresponding to the aforementioned embodiment of an infrared weak target detection method based on frequency domain data spherization, the present invention also provides an embodiment of an infrared weak target detection device based on spatiotemporal background reconstruction.

[0073] Figure 2 This is a block diagram illustrating an infrared weak target detection device based on frequency domain data spheroidization according to an exemplary embodiment, such as... Figure 2 As shown, the device includes:

[0074] The 3D spatiotemporal tensor construction module is used to construct a 3D spatiotemporal tensor from the infrared image frame to be detected and its adjacent infrared image frames.

[0075] The spatial domain to frequency domain conversion module is used to convert the three-dimensional spatiotemporal tensor from the spatial domain to the frequency domain, so as to obtain the three-dimensional spatiotemporal tensor in the frequency domain.

[0076] The high-pass filter module is used to design high-pass filters and filter three-dimensional spatiotemporal tensors in the frequency domain;

[0077] The frequency domain data spheroidization module is used to construct the frequency domain spheroidization space and perform frequency domain data spheroidization on the high-pass filter tensor.

[0078] The information filtering module is used to design information filters and apply them to the spherical tensor. The target detection result is obtained from the central frontal slice;

[0079] The target detection result output module is used to output the infrared weak target detection result map of each frame of the infrared image sequence.

[0080] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0081] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the description of the method embodiments. The device embodiments described above are merely illustrative; the various modules in the device represent a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules may be combined or integrated into another unit. Furthermore, the connections between the displayed or discussed modules may be communication connections through some interfaces, which may be electrical or other forms. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs. Those skilled in the art can understand and implement this without creative effort. The following uses a publicly available real thermal infrared image sequence as an example to illustrate specific implementation methods to demonstrate the technical effects of the present invention; specific steps in the embodiments will not be repeated.

[0082] The accompanying drawings illustrating the embodiments of the present invention will make the objectives, technical solutions, and advantages of the present invention clearer. It should be noted that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention. All equivalent substitutions and improvements made within the framework of the methods and principles provided by the present invention should be included within the protection scope of the present invention.

[0083] In this embodiment, the effectiveness of the infrared weak target detection method based on frequency domain data spherization will be verified using a publicly available infrared image sequence. The publicly available infrared image sequence contains 100 consecutive infrared image frames, each with a size of 256×256. The background includes buildings, mountains, and the sky, and there is significant clutter and noise. A small drone is flying in the air within the sequence, creating a high contrast with the surrounding background. Figure 3 The image shown is an example frame of an infrared image sequence. Figure 4The images shown are example frames from an infrared image sequence, obtained using the infrared weak target detection method based on frequency domain data spheroidization according to this invention. The intuitive detection results demonstrate that this method effectively suppresses complex backgrounds, noise, and clutter in the image, while significantly enhancing small targets. To more accurately and objectively evaluate the effectiveness of the infrared weak target detection method based on frequency domain data spheroidization, a qualitative analysis is conducted comparing and evaluating the target detection results on example frames in an infrared image sequence. Quantitatively, the detection performance of the algorithm is evaluated using the 3D-ROC evaluation index system, as follows:

[0084] (1) Detector effectiveness

[0085] AUC (D,F) ∈[0,1]

[0086] (2) Target detection probability

[0087] AUC ADP ≡AUC (D,τ) ∈[0,1]

[0088] (3) Background suppression probability

[0089] AUC BDP =1-AUC (F,τ) ∈[0,1]

[0090] (4) Joint target detection capability

[0091] AUC JAD =AUC (D,F) +AUC ADP ∈[0,2]

[0092] (5) Combined background suppression capability

[0093] AUC JBS

[0094] AUC JBS =AUC (D,F) +AUC BDP ∈[0,2]

[0095] (6) Combined target detection capability and background suppression capability

[0096] AUC ADBS =AUC (D,τ) -AUC (F,τ) ∈[-1,1]

[0097] (7) Target-to-background signal ratio

[0098]

[0099] (8) Comprehensive target detection capability

[0100] AUC OAD =AUC ADP +AUC BDP ∈[0,2]

[0101] To more objectively verify the effectiveness of the method of this invention, five thermal infrared weak target detection methods were selected for comparison with the method of this invention. The comparison methods include GSWLCM, NRAM, STT-TRNR, SSTTC, and TWTVR. Table 1 shows the 3D-ROC evaluation system of the results of infrared weak target detection using the comparison methods and the method of this invention for this thermal infrared image sequence. The bold and underlined values ​​represent the best and second-best performance, respectively.

[0102] GSWLCM comes from Qiu, Zhaobing, et al. "Global sparsity-weighted local contrastmeasure for infrared small target detection." IEEE Geoscience and RemoteSensing Letters 19(2022):1-5.

[0103] NRAM originated from Zhang, Landan, et al. "Infrared small target detection via non-convex rank approximation minimization joint l 2,1norm." Remote Sensing10.11(2018):1821.

[0104] STT-TRNR comes from Yi, Haiyang, et al. "Spatial-Temporal Tensor Ring NormRegularization for Infrared Small Target Detection." IEEE Geoscience andRemote Sensing Letters 20(2023):1-5.

[0105] SSTTC comes from Xia, Chaoqun, et al. "Separable Spatial-Temporal Patch-TensorPair Completion for Infrared Small Target Detection." IEEE Transactions onGeoscience and Remote Sensing (2024).

[0106] TWTVR comes from Zhao, Enzhong, et al. "Infrared Maritime Target Detection Based on Temporal Weight and Total Variation Regularization Under Strong WaveInterferences." IEEE Transactions on Geoscience and Remote Sensing (2024).

[0107] Figure 5 The image shows the original infrared image as an example frame and the results of thermal infrared weak target detection using the methods of this invention, GSWLCM, NRAM, STT-TRNR, SSTTC, and TWTVR. Figure 5 The target detection results images show that STT-TRNR and TWTVR have weak background suppression capabilities, while GSWLCM and NRAM exhibit residual noise and clutter in the detection results images. For SSTTC, a "long-tail effect" appears in the target detection results. Conversely, the infrared weak target detection method based on frequency domain data spherization proposed in this invention can effectively suppress background and noise components, significantly enhancing target saliency. Furthermore, according to the index results based on the 3D-ROC evaluation system shown in Table 1, the method proposed in this invention achieves the best and second-best performance in target detection capability, background suppression capability, and overall capability. In comparison, although STT-TRNR has the best target detection capability, it largely sacrifices background suppression capability, thus its overall performance is weaker, with a lower AUC. OAD =1.8614. Based on the above qualitative and quantitative analyses, the infrared weak target detection method based on frequency domain data spheroidization proposed in this invention possesses superior target detection capability, background suppression capability, and overall effectiveness.

[0108] Table 1. Quantitative indicators of the detection results of thermal infrared image example sequences using GSWLCM, NRAM, STT-TRNR, SSTTC, TWTVR, and the method of this invention.

[0109] method <![CDATA[AUC (D,F) ]]> <![CDATA[AUC ADP ]]> <![CDATA[AUC BDP ]]> <![CDATA[AUC JAD ]]> <![CDATA[AUC JBS ]]> <![CDATA[AUC ADBS ]]> <![CDATA[AUC SBPR ]]> <![CDATA[AUC OAD ]]> This invention 1.0000 0.9960 1.0000 1.9960 2.0000 0.9960 0.9960 1.9960 GSWLCM 1.0000 0.9950 1.0000 1.9950 2.0000 0.9949 0.9950 1.9949 NRAM 1.0000 0.9960 0.9997 1.9960 1.9997 0.9957 0.9963 1.9957 STT-TRNR 1.0000 1.0000 0.8614 2.0000 1.8614 0.8614 1.1609 1.8614 SSTTC 1.0000 0.9957 1.0000 1.9957 2.0000 0.9957 0.9957 1.9957 TWTVR 1.0000 0.9000 0.6777 1.8999 1.6777 0.5777 1.3280 1.5777

[0110] The accompanying drawings illustrating the embodiments of the present invention will make the objectives, technical solutions, and advantages of the present invention clearer. It should be noted that the specific embodiments described herein are merely illustrative and not intended to limit the scope of the invention. All equivalent substitutions and improvements made within the framework of the methods and principles provided by the present invention should be included within the protection scope of the present invention.

Claims

1. A method for detecting weak infrared targets based on frequency domain data spherization, characterized in that, Includes the following steps: Step 1): For the original thermal infrared image sequence, take the current infrared image frame D to be detected. t Centered on D t The adjacent preceding and following image frames are constructed into a three-dimensional spatiotemporal tensor. Step 2): Convert the three-dimensional spacetime tensor Transforming from the spatial domain to the frequency domain yields the three-dimensional spatiotemporal tensor in the frequency domain. Step 3): Design a high-pass filter and filter... The low-frequency information is extracted, while the high-frequency target information is preserved to obtain the high-pass filter tensor. Step 4): Construct the frequency domain spherical space Ω FDS and the high-pass filter tensor Mapping to the frequency domain spheroidized space yields the frequency domain spheroidized tensor. Step 5): Sphere the frequency domain tensor Transform to the spatial domain to obtain the spherical tensor. Step 5) specifically refers to: For the frequency domain spheroidized tensor obtained in step 4) The spherical tensor is obtained by transforming back to the spatial domain using the inverse Fourier transform. For the frequency domain spheric tensor Frontal slice have Where x represents the abscissa in the spatial domain, y represents the ordinate in the spatial domain, and (u,v) represents the frequency domain coordinates. For frequency domain spherization tensor Frontal slice The signal value with coordinates (u, v) in the middle. Let represent the spherical image corresponding to the tk-th frame in the original infrared image sequence, and let -l ≤ k ≤ l. The spherical tensor is thus obtained. Step 6): Design an information filter and apply it to the spherical tensor. The center front slice Obtain the target detection result matrix t t Further, the target detection result matrix of each frame of infrared image in the original thermal infrared image sequence is obtained, thereby obtaining the target detection sequence T, realizing the detection of weak infrared targets; Step 6) specifically refers to: For the spheric tensor obtained in step 5) The center front slice An information filter is constructed to filter out low-response pixel values, thereby enhancing high-frequency target information and obtaining the infrared image frame D to be detected. t The corresponding target detection result T t : in, Indicates will An operator that projects elements whose index value belongs to set ξ onto themselves and projects all other elements to zero. Indicates obtaining The operator for the index set ξ corresponding to the largest element in the middle κ, κ = αn1n2, where α is an adjustable positive constant; For each infrared image frame to be detected in the original thermal infrared image sequence, steps 1) to 6) are repeated to finally obtain the target detection result sequence T corresponding to the original thermal infrared image sequence, which serves as the infrared weak target detection result of the original infrared thermal infrared image sequence, thus realizing infrared weak target detection based on frequency domain data spherization.

2. The infrared weak target detection method based on frequency domain data spherization according to claim 1, characterized in that, Step 1) specifically refers to: For the original thermal infrared image sequence, take the current infrared image frame D to be detected. t Centered on the image D of the preceding and following frames, the images adjacent to it are... t-l ,…,D t-2 D t-1 D t+1 D t+2 ,…,D t+l Following the temporal order of the image frames, 2l+1 consecutive images are stacked sequentially to construct a three-dimensional spacetime tensor. Its size is n1×n2×n3, where n1 represents The width, n2 represents The height, n3 represents The thickness, and n3 = 2l + 1; where, D t-1 Indicates that it is located at D t The first frame image before, D t-2 Indicates that it is located at D t The second frame image before, D t-l Indicates that it is located at D t The l-th frame image before, D t+1 Indicates that it is located at D t The first frame image after D t+2 Indicates that it is located at D t The second frame image after D t+l Indicates that it is located at D t The image of the lth frame after that.

3. The infrared weak target detection method based on frequency domain data spherization according to claim 1, characterized in that, Step 2) specifically refers to: Three-dimensional spacetime tensor Each frame of image D t-k Transforming from the spatial domain to the frequency domain yields the frequency domain representation F. t-k : Where tk represents the tk-th frame in the original thermal infrared image sequence, and -l ≤ k ≤ l, x represents the horizontal coordinate in the spatial domain, y represents the vertical coordinate in the spatial domain, (u, v) represents the frequency domain coordinates, and F t-k (u,v) represents the signal value at coordinates (u,v) in the frequency domain image; Thus, the three-dimensional spacetime tensor is realized. Transforming from the spatial domain to the frequency domain yields the three-dimensional spatiotemporal tensor in the frequency domain.

4. The infrared weak target detection method based on frequency domain data spherization according to claim 1, characterized in that, Step 3) specifically refers to: Design a high-pass filter H(u,v), defined as: Where d0 is the cutoff frequency, and d(u,v) is the distance from the frequency point to the frequency center, we have: Using a high-pass filter H(u,v) to filter the three-dimensional spatiotemporal tensor in the frequency domain Low-frequency information is preserved, while high-frequency target information is retained; for the three-dimensional spatiotemporal tensor in the frequency domain... Each frontal slice F t-k ,have This yields the high-pass filter tensor.

5. The infrared weak target detection method based on frequency domain data spherization according to claim 1, characterized in that, Step 4) specifically refers to: For high-pass filter tensor The centered high-pass filter tensor is obtained by performing a centered process. Where m(·) represents the mean calculation operator; Center the high-pass filter tensor Expand into a two-dimensional matrix get Where reshape(·) represents the data reconstruction operator, and [n1n2,n3] represents... The dimensions are: width n1n2, height n3; Based on two-dimensional matrix Construct the corresponding spheroidized space in the frequency domain; first calculate The covariance matrix ∑: in, Representation matrix Transpose of; Perform eigenvalue decomposition on the covariance matrix ∑: ∑=UΛU T (7) Where U represents the eigenvector matrix and Λ represents the diagonal matrix; Using U and Λ, establish the frequency domain spherical space Ω FDS The projection matrix S corresponding to this frequency domain spherical space is Two-dimensional matrix Mapped to the frequency domain spherical space Ω FDS In the process, the frequency domain sphericization matrix is ​​obtained. Among them, Ω FFS (·) denotes the frequency domain spherification operator; furthermore, the frequency domain spherification matrix F spjered Reconstructed into a frequency domain spherical tensor of size n1×n2×n3 6. An infrared weak target detection device based on frequency domain data spherization implementing the method of claim 1, characterized in that, include: The 3D spatiotemporal tensor construction module is used to construct a 3D spatiotemporal tensor from the infrared image frame to be detected and its adjacent infrared image frames. The spatial domain to frequency domain conversion module is used to convert the three-dimensional spatiotemporal tensor from the spatial domain to the frequency domain, so as to obtain the three-dimensional spatiotemporal tensor in the frequency domain. The high-pass filter module is used to design high-pass filters and filter three-dimensional spatiotemporal tensors in the frequency domain; The frequency domain data spheroidization module is used to construct the frequency domain spheroidization space and perform frequency domain data spheroidization on the high-pass filter tensor. The information filtering module is used to design information filters and apply them to the spherical tensor. The target detection result is obtained from the central front slice; The target detection result output module is used to output the infrared weak target detection result map of each frame of the infrared image sequence.

Images