Infrared weak and small target detection method, system and device based on dual-band color ratio residual fusion and medium

Abstract

An infrared weak and small target detection method, system and device based on dual-band color ratio residual fusion and a medium belong to the field of infrared application, and the method comprises the following steps: selecting dual bands in an infrared spectrum, obtaining an infrared scene original image containing a weak and small target and corresponding to the dual bands, then calculating spatial filtering to obtain a prediction image, and carrying out prediction on the prediction image; the method comprises the following steps of: obtaining a pixel color ratio segmentation threshold value through a constant segmentation rate algorithm, carrying out pixel-level image fusion on the residual image to obtain a residual fusion image, approaching a noise-containing pixel value in the residual fusion image by using a quadric surface, determining parameters of a quadric surface equation, and obtaining a noise-containing pixel value in the residual fusion image by using the quadric surface. A target volume characteristic value is calculated through parameters, a threshold value is set according to the characteristic value, and the infrared weak and small target position is screened out from the finally obtained infrared image; the system, the equipment and the medium are used for implementing the method. The method has good effects of clutter suppression, image noise reduction and weak and small target detection capability improvement.

Description

Technical Field

[0001] The present invention relates to the field of infrared technology applications, and specifically relates to an infrared small and weak target detection method, system, device and medium based on dual-band color ratio residual fusion. Background Technique

[0002] Passive infrared detection systems are concealable and can provide a considerable detection range. However, modern low-observable targets generally adopt technologies such as infrared stealth, making the detectable signals extremely weak. In particular, infrared stealth technology greatly reduces the infrared signature signals of targets through advanced infrared stealth measures such as using turbofan engines, changing the shape and direction of the exhaust system, installing infrared baffle plates, improving fuel composition, adding special combustion agents, and using heat-absorbing and heat-insulating materials.

[0003] Existing infrared detection systems mainly operate within a single band of the infrared window. Using the average radiation energy within the band can only complete the detection of conventional targets, and the detection range is severely limited. Since when the detection range increases, the received radiation energy value decreases, and the interference of background clutter will become very obvious. Therefore, the detection and recognition of weak infrared signals in complex backgrounds face huge challenges.

[0004] Existing infrared stealth technologies focus on suppressing the infrared radiation intensity of targets within specific infrared bands and cannot guarantee the suppression effect within the entire infrared spectrum range, which provides a key technical breakthrough for the infrared detection and extraction of low-observable targets.

[0005] Chinese patent application with the publication number CN201711303933.8 discloses an infrared target detection method based on an improved Tri-edge operator. However, since this method only selects a single infrared band, the detection ability of infrared small and weak targets is insufficient, and the detection range is severely limited. Summary of the Invention

[0006] In order to overcome the above-mentioned disadvantages of the prior art, the purpose of the present invention is to provide an infrared small and weak target detection method, system, device and medium based on dual-band color ratio residual fusion. When detecting infrared small and weak targets, the present invention selects two independent spectral bands within the mid-wave infrared band, obtains the original infrared scene images corresponding to the dual bands containing small and weak targets through an infrared imaging system, performs spatial filtering calculations for non-linear background prediction to obtain the residual images corresponding to the dual bands, completes the fusion of the dual-band images according to the pixel color ratio calculation results of the dual-band images, and finally realizes the detection and extraction of infrared small and weak targets through denoising.

[0007] To achieve the above purpose, the technical solution adopted by the present invention is:

[0008] An infrared dim and small target detection method based on dual-band color ratio residual fusion, comprising the following steps:

[0009] Step 1: Select band 1 and band 2 respectively within the infrared spectral bands of 3-4μm and 4-5μm, and obtain the original infrared scene images containing dim and small targets corresponding to band 1 and band 2 respectively through infrared imaging;

[0010] Step 2: Perform spatial filtering calculation based on non-linear background prediction on the original images obtained in Step 1 to obtain the predicted images of band 1 and band 2;

[0011] Step 3: Subtract the predicted images obtained in Step 2 from the original images obtained in Step 1 to obtain the residual images of band 1 and band 2;

[0012] Step 4: Calculate the pixel color ratio of the images of band 1 and band 2, obtain the pixel color ratio segmentation threshold E through the constant segmentation rate algorithm th , and then perform pixel-level image fusion on the residual images of band 1 and band 2 obtained in Step 3 based on the pixel color ratio segmentation threshold E th to obtain a residual fusion image;

[0013] Step 5: Approximate the pixel values with noise in the residual fusion image obtained in Step 4 using a quadratic surface, determine the parameters of the quadratic surface equation, and then calculate the target volume eigenvalue through the parameters of the quadratic surface equation;

[0014] Step 6: Set a threshold according to the target volume eigenvalue calculated in Step 5, and screen out the positions of infrared dim and small targets in the finally obtained infrared image.

[0015] Both band 1 and band 2 selected in Step 1 are located in the mid-wave infrared spectral band, independent of each other and without spectral overlap.

[0016] The specific content of Step 2 includes:

[0017] Perform spatial filtering calculation based on non-linear background prediction on the original images obtained in Step 1 according to formula (1) to obtain the predicted images of band 1 and band 2;

[0018]

[0019] In the formula:

[0020] X - The input image with size M×N;

[0021] Y - The predicted image with size M×N;

[0022] W j - The weight matrix of the j-th level, j=(m - 1)×M + n;

[0023] S j —— Set of local background selection points;

[0024] m, n, l, k —— Indexes.

[0025] Step 3 specifically includes:

[0026] Subtract the predicted image obtained in Step 2 from the original image obtained in Step 1 according to formula (2) to obtain the residual images of Band 1 and Band 2;

[0027]

[0028] In the formula:

[0029] E1 —— Residual image of size M×N corresponding to Band 1;

[0030] E2 —— Residual image of size M×N corresponding to Band 2;

[0031] X1 —— Input image of size M×N corresponding to Band 1;

[0032] Y1 —— Predicted image of size M×N corresponding to Band 1;

[0033] X2 —— Input image of size M×N corresponding to Band 2;

[0034] Y2 —— Predicted image of size M×N corresponding to Band 2.

[0035] Step 4 specifically includes:

[0036] Step 4.1: Calculate the pixel color ratio of the images of Band 1 and Band 2 according to formula (3);

[0037]

[0038] In the formula:

[0039] X1 —— Input image of size M×N corresponding to Band 1;

[0040] X2 —— Input image of size M×N corresponding to Band 2;

[0041] —— Local average value of size L×L of image X1 at pixel (u, v);

[0042] —— Local average value of size L×L of image X2 at pixel (u, v);

[0043] I R (u, v) —— Pixel color ratio of the images of Band 1 and Band 2;

[0044] Step 4.2: On the pixel color ratio I obtained in Step 4.1 R , use the constant segmentation rate algorithm to obtain the pixel color ratio segmentation threshold E th , and the pixel color ratio segmentation threshold E th is unique for each group of infrared images of Band 1 and Band 2;

[0045] Step 4.3: Based on the pixel color ratio segmentation threshold E obtained in Step 4.2 th , perform pixel-level image fusion on the residual images of Band 1 and Band 2 according to formula (4);

[0046]

[0047] In the formula:

[0048] E th —— Pixel color ratio segmentation threshold;

[0049] I F —— Residual fusion image of Band 1 and Band 2;

[0050] E1 —— Residual image of size M×N corresponding to Band 1;

[0051] E2 —— Residual image of size M×N corresponding to Band 2.

[0052] The specific steps of Step 5 include:

[0053] Step 5.1: In the residual fusion image I F (m,n), approximate the pixel value with noise using a quadratic surface, and determine the parameters of the quadratic surface equation according to the minimization index function of formula (5);

[0054]

[0055] In the formula:

[0056] A —— Coefficient matrix;

[0057] x —— Quadratic surface equation parameter vector;

[0058] a, b, c, d, e, f —— Quadratic surface equation parameters;

[0059] x′, y′ —— Current pixel coordinates;

[0060] k —— Neighborhood half-width of the current pixel;

[0061] Step 5.2: Calculate the target volume eigenvalue according to formula (6) and the quadratic surface equation parameters determined in Step 5.1;

[0062]

[0063] In the formula:

[0064] V——Target volume eigenvalue.

[0065] An infrared dim small target detection system based on dual-band color ratio residual fusion, comprising:

[0066] Data import module: Select band 1 and band 2 respectively within the infrared spectrum bands of 3-4μm and 4-5μm, and obtain the original infrared scene images containing dim small targets corresponding to band 1 and band 2 respectively through infrared imaging;

[0067] Image processing module: Perform spatial filtering calculation based on non-linear background prediction on the original images to obtain the predicted images of band 1 and band 2, then subtract the original images from the predicted images to obtain the residual images of band 1 and band 2, then calculate the pixel color ratio of the images of band 1 and band 2, obtain the pixel color ratio segmentation threshold through the constant segmentation rate algorithm, and finally perform pixel-level image fusion on the residual images of band 1 and band 2 based on the pixel color ratio segmentation threshold to obtain the residual fusion image;

[0068] Detection, tracking and recognition module: Approximate the pixel values with noise in the residual fusion image by a quadratic surface, and determine the parameters of the quadratic surface equation, then calculate the target volume eigenvalue through the parameters of the quadratic surface equation, set a threshold according to the target volume eigenvalue, and screen out the positions of infrared dim small targets in the finally obtained infrared image.

[0069] An infrared dim small target detection device based on dual-band color ratio residual fusion, comprising:

[0070] Memory: Used to store the computer program for implementing the infrared dim small target detection method based on dual-band color ratio residual fusion as described above;

[0071] Processor: Used to implement the infrared dim small target detection method based on dual-band color ratio residual fusion when executing the computer program.

[0072] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the infrared dim small target detection method based on dual-band color ratio residual fusion as described above are implemented.

[0073] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0074] 1. In the present invention, spatial filtering calculation based on non-linear background prediction is performed on the dual-band original infrared scene image to obtain the dual-band predicted image, completing the first-step noise reduction processing and target enhancement, and being able to effectively suppress background clutter.

[0075] 2. In the present invention, pixel-level image fusion is performed on the dual-band residual image, and the pixel color ratio segmentation threshold is calculated according to the constant segmentation rate algorithm, completing the second-step noise reduction processing and target enhancement, and having a good image noise reduction effect.

[0076] 3. In the present invention, volume detection is performed on the residual fusion image, that is, the pixel values containing noise are approximated by a quadratic surface, completing the third-step noise reduction processing and target enhancement, and improving the detection ability of weak and small targets.

[0077] In summary, the present invention has good effects of clutter suppression, image noise reduction, and improving the detection ability of weak and small targets. Description of the Drawings

[0078] Figure 1 It is a flowchart of the dual-band color ratio residual fusion processing of the present invention.

[0079] Figure 2 In the figure: (a) is the original infrared scene image of band 1, (b) is the original infrared scene image of band 2, (c) is the segment residual map of band 1, (d) is the segment residual map of band 2, (e) is the target and clutter color ratio discrimination result map, and (f) is the detection and extraction result of the infrared weak and small target. Detailed Embodiment

[0080] The following will describe the present invention in detail with reference to the drawings.

[0081] A method for detecting infrared weak and small targets based on dual-band color ratio residual fusion in the present invention, on the basis of the original dual-band infrared scene image, through three core calculation methods including spatial filtering calculation based on non-linear background prediction, pixel-level image fusion of dual-band residual images, and volume detection of residual fusion images, effectively filters the background clutter in the infrared scene and greatly enhances the detection and extraction of infrared weak and small targets.

[0082] According to the spectral radiation caused by the rotational and vibrational energy level transitions of H2O and CO2 molecules in the engine plume, it is determined that there are two-color radiation peaks in the infrared mid-wave spectral band, and the infrared bands 1 and 2 corresponding to the two-color radiation peaks are used as the infrared detection characteristic spectra of low-observable targets. As Figure 1 shown, the steps of the present invention are as follows:

[0083] Step 1: Select Band 1 and Band 2 within the infrared spectrum bands of 3 - 4μm and 4 - 5μm respectively, and obtain the original infrared scene images containing weak targets corresponding to Band 1 and Band 2 respectively through infrared imaging;

[0084] Step 2: Conduct spatial filtering calculation based on non - linear background prediction for the original images obtained in Step 1 to obtain the predicted images of Band 1 and Band 2;

[0085] Step 3: Subtract the predicted images obtained in Step 2 from the original images obtained in Step 1 to obtain the residual images of Band 1 and Band 2;

[0086] Step 4: Calculate the pixel color ratio of the images of Band 1 and Band 2, and obtain the pixel color ratio segmentation threshold E through the constant segmentation rate algorithm th , and then based on the pixel color ratio segmentation threshold E th perform pixel - level image fusion on the residual images of Band 1 and Band 2 obtained in Step 3 to obtain the residual fusion image;

[0087] Step 5: Approximate the pixel values with noise in the residual fusion image obtained in Step 4 using a quadratic surface, determine the parameters of the quadratic surface equation, and then calculate the target volume eigenvalue through the parameters of the quadratic surface equation;

[0088] Step 6: Set a threshold according to the target volume eigenvalue calculated in Step 5, and screen out the positions of the infrared weak targets in the finally obtained infrared image.

[0089] Both Band 1 and Band 2 selected in Step 1 are located in the mid - wave infrared spectral band, independent of each other and without spectral overlap.

[0090] The specific content of Step 2 includes:

[0091] Conduct spatial filtering calculation based on non - linear background prediction for the original images obtained in Step 1 according to formula (1) to obtain the predicted images of Band 1 and Band 2;

[0092]

[0093] In the formula:

[0094] X - The input image with size M×N;

[0095] Y - The predicted image with size M×N;

[0096] W j - The weight matrix of the j - th level, j=(m - 1)×M + n;

[0097] S j - The set of local background selection points.

[0098] Step 3 specifically includes:

[0099] Subtract the original image obtained in Step 1 from the predicted image obtained in Step 2 according to Formula (2) to obtain the residual images of Band 1 and Band 2;

[0100]

[0101] In the formula:

[0102] E1 - The residual image corresponding to Band 1 with a size of M×N;

[0103] E2 - The residual image corresponding to Band 2 with a size of M×N;

[0104] X1 - The input image corresponding to Band 1 with a size of M×N;

[0105] Y1 - The predicted image corresponding to Band 1 with a size of M×N;

[0106] X2 - The input image corresponding to Band 2 with a size of M×N;

[0107] Y2 - The predicted image corresponding to Band 2 with a size of M×N.

[0108] Step 4 specifically includes:

[0109] Step 4.1: Calculate the pixel color ratio of the images of Band 1 and Band 2 according to Formula (3);

[0110]

[0111] In the formula:

[0112] - The local average value of size L×L of image X1 at pixel (u, v);

[0113] - The local average value of size L×L of image X2 at pixel (u, v);

[0114] I R (u, v) - The pixel color ratio of the images of Band 1 and Band 2;

[0115] Step 4.2: On the pixel color ratio I R obtained in Step 4.1, use the constant segmentation rate algorithm to obtain the pixel color ratio segmentation threshold E th , and the pixel color ratio segmentation threshold E th is unique for each group of infrared images of Band 1 and Band 2;

[0116] Step 4.3: Based on the pixel color ratio segmentation threshold E obtained in Step 4.2th , perform pixel-level image fusion on the residual images of band 1 and band 2 according to formula (4);

[0117]

[0118] In the formula:

[0119] E th —— Pixel color ratio segmentation threshold;

[0120] I F —— Residual fusion image of band 1 and band 2.

[0121] The specific steps of step 5 include:

[0122] Step 5.1: In the residual fusion image I F (m,n), approximate the pixel value containing noise with a quadratic surface, and determine the parameters of the quadratic surface equation according to the minimization index function of formula (5);

[0123]

[0124] In the formula:

[0125] A - Coefficient matrix;

[0126] x - Parameter vector of the quadratic surface equation;

[0127] a, b, c, d, e, f - Parameters of the quadratic surface equation;

[0128] x′, y′ - Current pixel coordinates;

[0129] k - Neighborhood half-width of the current pixel;

[0130] Step 5.2: Calculate the target volume eigenvalue according to formula (6) and the parameters of the quadratic surface equation determined in step 5.1;

[0131]

[0132] In the formula:

[0133] V - Target volume eigenvalue.

[0134] An infrared dim small target detection system based on dual-band color ratio residual fusion, comprising:

[0135] Data import module: Select band 1 and band 2 respectively within the infrared spectrum of 3 - 4μm and 4 - 5μm, and obtain the original infrared scene images containing dim small targets corresponding to band 1 and band 2 respectively through infrared imaging, so as to implement step 1 of an infrared dim small target detection method based on dual-band color ratio residual fusion;

[0136] Image processing module: It performs spatial filtering calculation on the original image based on non-linear background prediction to obtain the predicted images of Band 1 and Band 2. Then, it subtracts the original image from the predicted images to obtain the residual images of Band 1 and Band 2. Next, it calculates the pixel color ratio of the images of Band 1 and Band 2, and obtains the pixel color ratio segmentation threshold through the constant segmentation rate algorithm. Finally, it performs pixel-level image fusion on the residual images of Band 1 and Band 2 based on the pixel color ratio segmentation threshold to obtain the residual fusion image, which is used to implement Steps 2 to 4 of an infrared dim small target detection method based on dual-band color ratio residual fusion.

[0137] Detection, tracking and recognition module: It approximates the pixel values with noise in the residual fusion image using a quadratic surface and determines the parameters of the quadratic surface equation. Then, it calculates the target volume eigenvalue through the parameters of the quadratic surface equation, sets a threshold according to the target volume eigenvalue, and screens out the positions of infrared dim small targets in the finally obtained infrared image, which is used to implement Steps 5 to 6 of an infrared dim small target detection method based on dual-band color ratio residual fusion.

[0138] An infrared dim small target detection device based on dual-band color ratio residual fusion includes:

[0139] Memory: It is used to store the computer program for implementing an infrared dim small target detection method based on dual-band color ratio residual fusion.

[0140] Processor: It is used to implement an infrared dim small target detection method based on dual-band color ratio residual fusion when executing the computer program.

[0141] The present invention also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the steps of an infrared dim small target detection method based on dual-band color ratio residual fusion.

[0142] To verify the effectiveness of the present invention, Figure 2 Figures (a) and (b) in it respectively show the dual-band original infrared scene, and the corresponding dual-band residual maps (c) and (d); the pixel color ratio segmentation threshold is obtained as 0.03 through the constant segmentation rate algorithm, and the Figure 2 target and clutter color ratio discrimination result map shown in Figure (e) in it is obtained; by fusing the dual-band residual maps, the detection and extraction result map of infrared dim small targets shown in Figure (f) in Figure 2 it is finally obtained. The results show that compared with the existing methods, the method of the present invention can effectively improve the infrared dim small target detection probability by 20% - 40%, and greatly improve the detection and extraction ability of the optoelectronic system for infrared dim small targets.

[0143] The method for detecting weak and small infrared targets by using dual-band color ratio residual fusion can perform optoelectronic detection and extraction of advanced infrared stealth targets. The present invention can obtain reasonable and reliable results for detecting and extracting weak and small infrared targets, and can be applied to the semi-physical simulation calculation of actual optoelectronic engineering.

Claims

1. An infrared dim and small target detection method based on dual-band color ratio residual fusion, characterized in that, It includes the following steps: Step 1: Select Band 1 and Band 2 respectively within the infrared spectrum bands of 3 - 4μm and 4 - 5μm, and obtain the original infrared scene images containing weak targets corresponding to Band 1 and Band 2 respectively through infrared imaging; Step 2: Perform spatial filtering calculation based on non - linear background prediction on the original images obtained in Step 1 to obtain the predicted images of Band 1 and Band 2; Step 3: Subtract the predicted images obtained in Step 2 from the original images obtained in Step 1 to obtain the residual images of Band 1 and Band 2; Step 4: Calculate the pixel color ratio of the images of Band 1 and Band 2, and obtain the pixel color ratio segmentation threshold E through the constant segmentation rate algorithm th , and then based on the pixel color ratio segmentation threshold E th perform pixel-level image fusion on the residual images of Band 1 and Band 2 obtained in Step 3 to obtain a residual fusion image; Step 5: Approximate the pixel values with noise in the residual fusion image obtained in Step 4 using a quadratic surface, determine the parameters of the quadratic surface equation, and then calculate the target volume eigenvalue through the parameters of the quadratic surface equation; Step 6: Set a threshold according to the target volume eigenvalue calculated in Step 5, and screen out the positions of infrared weak targets in the finally obtained infrared image.

2. The infrared dim small target detection method based on dual-band color ratio residual fusion according to claim 1, characterized in that Both Band 1 and Band 2 selected in Step 1 are located in the mid - wave infrared spectral band, independent of each other and without spectral overlap.

3. A method for detecting a dim and small infrared target based on dual-band color ratio residual fusion according to claim 1, wherein, The specific content of Step 2 includes: Perform spatial filtering calculation based on non - linear background prediction on the original images obtained in Step 1 according to formula (1) to obtain the predicted images of Band 1 and Band 2; In the formula: X - The input image with size M×N; Y - The predicted image with size M×N; W j —— The weight matrix of the j-th level, where j = (m - 1) × M + n; S j —— Set of local background selection points; m, n, l, k - Indexes.

4. A method for detecting small and dim infrared targets based on dual-band color ratio residual fusion according to claim 1, characterized in that, The specific content of Step 3 includes: Subtract the predicted images obtained in Step 2 from the original images obtained in Step 1 according to formula (2) to obtain the residual images of Band 1 and Band 2; In the formula: E1 - The residual image with size M×N corresponding to Band 1; E2 - The residual image with size M×N corresponding to Band 2; X1 - The input image with size M×N corresponding to Band 1; Y1 - The predicted image with size M×N corresponding to Band 1; X2 - The input image with size M×N corresponding to Band 2; Y2 - The predicted image with size M×N corresponding to Band 2.

5. A method for detecting dim and small infrared targets based on dual-band color ratio residual fusion according to claim 1, characterized in that, The specific content of Step 4 includes: Step 4.1: Calculate the pixel color ratio of the images of Band 1 and Band 2 according to formula (3); In the formula: X1 - The input image with size M×N corresponding to Band 1; X2 - The input image with size M×N corresponding to Band 2; —— Local average value of size L×L of image X1 at pixel (u, v); —— the local average value of size L×L of image X2 at pixel (u, v); I R (u, v) — Pixel color ratio of images in Band 1 and Band 2 Step 4.2: On the pixel color ratio I obtained in Step 4.1 R use the constant segmentation rate algorithm to obtain the pixel color ratio segmentation threshold E th , and the pixel color ratio segmentation threshold E th is unique for each group of infrared images of Band 1 and Band 2; Step 4.3: Calculate the pixel color ratio segmentation threshold E based on Step 4.2 th , and perform pixel-level image fusion on the residual images of Band 1 and Band 2 according to Formula (4); In the formula: E th —— Pixel color ratio segmentation threshold; I F —— Residual fusion image of band 1 and band 2; E1 - The residual image with size M×N corresponding to Band 1; E2 - The residual image with size M×N corresponding to Band 2.

6. The infrared dim small target detection method based on dual-band color ratio residual fusion according to claim 1, wherein The specific content of Step 5 includes: Step 5.1: In the residual fusion image I F (m,n), approximate the pixel value with noise by a quadratic surface, and determine the parameters of the quadratic surface equation according to the minimization index function of formula (5); In the formula: A - Coefficient matrix; x - Quadratic surface equation parameter vector; a, b, c, d, e, f - Quadratic surface equation parameters; x′, y′ - Current pixel coordinates; k - Neighborhood half - width of the current pixel; Step 5.2: Calculate the target volume eigenvalue according to formula (6) and the quadratic surface equation parameters determined in Step 5.1; In the formula: V - Target volume eigenvalue.

7. An infrared dim and small target detection system based on dual-band color ratio residual fusion according to the method described in any one of claims 1 to 6, characterized in that, It includes: Data import module: Select Band 1 and Band 2 respectively within the infrared spectrum bands of 3 - 4μm and 4 - 5μm, and obtain the original infrared scene images containing weak targets corresponding to Band 1 and Band 2 respectively through infrared imaging; Image processing module: Perform spatial filtering calculation based on non-linear background prediction on the original image to obtain the predicted images of Band 1 and Band 2. Then, subtract the original image from the predicted images to obtain the residual images of Band 1 and Band 2. Next, calculate the pixel color ratio of the images of Band 1 and Band 2, and find the pixel color ratio segmentation threshold through the constant segmentation rate algorithm. Finally, perform pixel-level image fusion on the residual images of Band 1 and Band 2 based on the pixel color ratio segmentation threshold to obtain the residual fusion image; Detection, tracking and recognition module: Approximate the pixel values with noise in the residual fusion image using a quadratic surface, and determine the parameters of the quadratic surface equation. Then, calculate the target volume eigenvalue through the parameters of the quadratic surface equation, set a threshold according to the target volume eigenvalue, and screen out the positions of infrared dim small targets in the finally obtained infrared image.

8. An infrared dim and small target detection device based on dual-band color ratio residual fusion, characterized in that, Including: Memory: Used to store a computer program for implementing an infrared dim small target detection method based on dual-band color ratio residual fusion as described in any one of claims 1 to 6; Processor: Used to implement an infrared dim small target detection method based on dual-band color ratio residual fusion as described in any one of claims 1 to 6 when executing the computer program.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, and when the computer program is executed by the processor, the steps of an infrared dim small target detection method based on dual-band color ratio residual fusion as described in any one of claims 1 to 6 are implemented.

Images