Dark weak target energy enhancement method

Abstract

The invention discloses a faint target energy enhancement method, and relates to the technical field of image processing. The method is suitable for a multi-faint target image sequence obtained by an infrared detector, and comprises the following steps: S101, inputting an infrared faint target image group; s102, background component suppression calculation of the target image group; s103, carrying out multi-direction shift accumulation search on the dark weak target; s104, filtering and multiplying the maximum value of an accumulated result; and S105, searching accumulation graph spatial domain local energy enhancement, and realizing signal energy enhancement of the dark and weak target. Aiming at an infrared detector system to obtain a multi-faint target image sequence, faint target signal energy is enhanced through a multi-frame accumulation principle, increase of target gray level signals is enhanced by adopting a multi-directional shift search accumulation and spatial domain local energy gain strategy, the image signal-to-noise ratio of faint targets is improved, and the target detection accuracy is improved. The signal energy of a plurality of dim targets in different motion states in a field of view can be enhanced, and the dim target detection probability of an infrared detection system is effectively improved.

Description

Technical Field

[0001] The invention belongs to the field of image processing, and in particular to a method for enhancing the energy of a faint target. The method is particularly suitable for enhancing the energy of a plurality of infrared faint target images in an empty background acquired by an infrared detection system, and finally realizing accurate extraction of faint targets. Background Art

[0002] Compared with ground-based detection, radar detection and visible light detection, infrared detection systems are widely used in space environment monitoring and observation due to their many advantages. Infrared detection detectors are highly sensitive, and the background stray light of the target will inevitably enter the detector, resulting in uneven brightness in the spatial and temporal domains of the infrared image, thereby submerging the target in the strong clutter background, reducing the image contrast and signal-to-noise ratio of the target, and causing optical imaging or detection failure. In recent years, research on stray light suppression has mainly been in the design of optomechanical systems, but some stray light will still appear in actual imaging, resulting in uneven image brightness, which needs to be suppressed through image processing methods, and weak targets need to be enhanced to improve the accuracy of target detection.

[0003] The existing weak target enhancement methods are mainly single-frame algorithms and multi-frame algorithms. The single-frame weak target enhancement algorithms mainly include traditional image filtering algorithms and target enhancement algorithms based on human visual characteristics (HVS). Traditional algorithms include maximum median filtering, TopHat, median filtering algorithms, etc. This type of algorithm distinguishes the target from the background by extracting the grayscale feature information of the target area. The algorithm is simple and easy to implement, but it will weaken the target energy and has limited ability to suppress strong noise in the image; the target enhancement algorithm based on human visual characteristics (HVS) uses the local contrast characteristics of the target to design the contrast enhancement factor, which can enhance the grayscale signal of the target area, but has limited ability to enhance targets with extremely low signal-to-noise ratios and limited ability to suppress high-brightness clutter in the image background.

[0004] The multi-frame dim target enhancement algorithm is mainly based on the theory of multi-frame accumulation. The temporal and spatial distribution of noise in infrared dim target images is random and unrelated, while the correlation between target frames is strong. The multi-frame accumulation algorithm uses this feature to suppress random noise, thereby improving the target signal-to-noise ratio. The premise for this type of algorithm to take effect is to make the target accumulate energy in continuous multi-frame images along its motion path. However, the existing multi-frame accumulation target enhancement algorithm extracts the target motion trajectory and accumulates and enhances the target signal along the target motion trajectory. The performance of this type of method mainly depends on the accuracy of trajectory extraction. However, the low signal-to-noise, uncertain quantitative characteristics and rapid motion of the target in the actual image increase the difficulty of trajectory extraction, limiting the target enhancement ability of this type of method. Summary of the invention

[0005] The purpose of the present invention is to provide a method for enhancing the energy of dim and weak targets to solve the technical problems existing in the above-mentioned prior art. The present invention enhances the signal energy of dim and weak targets through the principle of multi-frame accumulation, adopts multi-directional shift search accumulation and spatial local energy gain strategy to enhance the increase of target grayscale signal, improves the image signal-to-noise ratio of dim and weak targets, and can realize the enhancement of the signal energy of multiple dim and weak targets with different motion states in the field of view, effectively improving the detection probability of dim and weak targets of infrared detection systems. The present invention is of great significance in the field of dim and weak multi-target enhancement and detection of infrared detection systems.

[0006] In order to achieve the above-mentioned object of the invention, the technical solution of the present invention is:

[0007] A method for enhancing the energy of a dim target, characterized in that it comprises the following steps:

[0008] S101, inputting an infrared dim target image group;

[0009] S102 target image group background component suppression calculation;

[0010] S103 Multi-directional shift accumulation search for dim targets;

[0011] S104 searches for the maximum value of the accumulated result, filters it, and multiplies it;

[0012] S105 searches for local energy enhancement in the spatial domain of the accumulated image to achieve energy enhancement of weak target signals.

[0013] The present invention mainly comprises the following steps:

[0014] (1) Input of target enhanced image sequence. According to the reference frame image F to be processed n , select m frames of image F from the target image sequence n-m ~F n-1 Composition of image groups n-m ~F n}.

[0015] (2) For the image group {F n-m ~F n}, the background component is suppressed to obtain the target background suppressed image group {M n-m ~M n}, background component suppression can be achieved by using related image processing algorithms such as mean filtering and median filtering.

[0016] (3) Estimate the possible movement angle and direction of the target, and calculate the velocity vector V in the X-axis and Y-axis directions according to the set search angle step and speed step x (k) and V y (k), which is calculated as follows:

[0017]

[0018] k1=1,…,n,k2=1,…,m,k=1,…,n*m

[0019] Where V x (k) is the speed of the target in the X-axis direction, V y (k) is the speed of the target in the Y-axis direction, To search for velocity components, The angle component to search for. The calculation formula is as follows:

[0020]

[0021] where v max is the maximum search speed, and n is the search speed resolution.

[0022] The calculation formula is as follows:

[0023]

[0024] angle s =360° / m

[0025] Where m is the search angle resolution.

[0026] (4) For the local mean filter image group {M n-m ~M n}, using the velocity vector V calculated in the previous step x (k) and V y (k), search for the pixel gray value of the pixel coordinate (x, y) on each local mean filter image after shifting on each possible motion vector and accumulate them to obtain the multi-directional shift accumulation image sequence MFA (x, y, k), the formula is as follows:

[0027]

[0028] Among them, M f (x f ,k,y f ,k) is the image M f Medium (x f ,k,y f , the pixel gray value at the coordinates of k, (x f ,k,y f ,k) is the coordinate of the target at coordinate (x, y) in the nfth frame image after shifting along the kth search direction, (x f ,k,y f , k) is calculated as follows:

[0029] x f,k =x+f*V x (k)

[0030] y f，k =y+f*V y (k)

[0031] (5) Perform maximum projection calculation on the shifted and accumulated three-dimensional image MFA(x, y, k) to obtain the multi-directional shifted and accumulated grayscale maximum image MFA max , the formula is as follows:

[0032] MFA max =max(MFA(x, y, k))

[0033] (6) MFA max The image is negatively suppressed and each pixel is multiplied to further enhance the target energy to obtain the multi-directional shift accumulation result map MFA out , the formula is as follows:

[0034] MFA out (x, y) = MFA′ max (x, y)×MFA′ max (x, y)×MFA′ max (x, y)

[0035]

[0036] (7) For the multi-directional shift accumulation result graph MFA out Take the pixel (x, y) in , extract the local neighborhood pixels around the pixel, and construct the central block S0 and the neighborhood background block S. The regions of S and S0 are represented as follows.

[0037]

[0038] R S ={(p, q)|max(1p-x|, |qy|≤s)}, s=4, 7, 10, 13

[0039] Where (i, j) and (p, q) are image MFA out In the pixel coordinates, (x, y) is the center pixel coordinate of the image block, s and l represent the radii of S and S0 respectively, and the value of l is determined by the size of the target to be detected.

[0040] (8) The mean and standard deviation of the grayscale values ​​of pixels in block S are calculated to characterize the background and noise components of the neighborhood background block. The formula is as follows:

[0041]

[0042] MFA out (p, q) is the grayscale value of the pixel at the coordinate (p, q) in block S, (x, y) is the coordinate of the center pixel of block S, MFA m (x, y) and MFA s (x, y) represent the mean and standard deviation of the grayscale values ​​of the pixel in the neighborhood block S of the pixel (x, y) respectively.

[0043] (9) Using the neighborhood background component value MFA calculated in the previous step m (x, y) suppresses the background component of the target area and obtains the target component map MFA t , the formula is as follows:

[0044] MFA t (p, q) = MFA out (p, q)-MFA m (x,y),(p,q)∈S0

[0045] (10) Calculate the energy accumulation value E of the pixel grayscale in block S0 t (x, y), amplify the signal of the target component and quickly enhance the target energy. The calculation formula is as follows:

[0046]

[0047] (11) Calculation of signal-to-noise ratio gain factor. Calculate the ratio of the energy accumulation value to the local background noise intensity as the signal-to-noise ratio gain factor M c , the formula is as follows:

[0048] M c (x, y) = E t (x, y) / MFA s (x, y)

[0049] (12) Calculate the energy concentration characteristics of the target area and calculate its inverse proportional factor as the energy concentration gain factor M E , which is used to suppress isolated strong noise pixels, and its formula is as follows:

[0050] M E =(1-T env ) 2

[0051] Where T env is the target energy concentration, and its formula is as follows:

[0052]

[0053] MFA t_max R after background suppression s The maximum grayscale value of the pixel.

[0054] (13) With the help of the signal-to-noise ratio gain factor M c and energy concentration gain factor M E Calculate the local energy gain value G of the pixel at coordinate (x, y) in the image L (x, y), the formula is as follows:

[0055] G L (x, y) = MFA t (x, y)×M c (x, y)×M E (x, y)

[0056] (14) Calculate the adaptive segmentation threshold Th of the local energy gain map for G L Perform binary segmentation to determine the target location. The threshold Th is defined as follows:

[0057] Th=μ+k×σ

[0058] Where k is the division coefficient. L When the value of an element in is greater than Th, it is set to 1, otherwise it is set to 0. The point set to 1 is judged as a candidate target. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 A flowchart of the implementation steps of a method for enhancing the energy of a dim target according to the present invention;

[0060] Figure 2 A flowchart of a method for enhancing energy of a dim target according to the present invention;

[0061] Figure 3 A schematic diagram of the enhancement effect of a single dim target infrared simulation image provided by an example of the present invention;

[0062] Figure 4 A schematic diagram of the enhancement effect of multiple infrared simulation images of dim targets provided by an example of the present invention; DETAILED DESCRIPTION

[0063] The technical solutions in the embodiments of the present invention will be described in detail below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work belong to the scope of protection of the present invention, and the changes, modifications, additions or replacements made within the scope of the technical solution of the present invention also belong to the scope of protection of the present invention.

[0064] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0065] like Figure 1 As shown, the multi-frame accumulation infrared dim moving target enhancement method provided by the embodiment of the present invention includes the following steps:

[0066] S101, inputting an infrared dim target image group;

[0067] S102, background component suppression calculation of target image group;

[0068] S103, multi-directional shift accumulation search for dim targets;

[0069] S104, searching for the maximum value of the accumulated result, filtering and multiplying;

[0070] S105, searching for local energy enhancement in the spatial domain of the accumulated graph to achieve energy enhancement of the weak target signal;

[0071] As a preferred embodiment, Figure 2 As shown, a method for enhancing energy of a dim target provided by an embodiment of the present invention specifically includes the following steps:

[0072] (1) Input of target enhanced image sequence. According to the reference frame image F to be processed n , select the neighborhood image F from the target image sequence n-m ~F n-1 Composition of image groups n-m ~F n |, parameter m=15.

[0073] (2) Based on step (1), the image background component is suppressed and the neighborhood block R is calculated for each pixel in the image. s The mean grayscale value of the internal pixels is used to obtain the mean filtered image group {M n-m ~M n}. Take the image frame F n For example, its mean filter graph M n The calculation of is as follows:

[0074]

[0075] Among them, F n (p, q) is the image F n The grayscale value of the pixel with coordinates (p, q) and (x, y) is the grayscale value of the block R s The center pixel coordinates of the block R s The radius of the , parameter s = 3.

[0076] (3) Estimate the possible movement angle and direction of the target, and calculate the velocity vector V in the X-axis and Y-axis directions according to the set search angle step and speed step x (k) and V y (k), which is calculated as follows:

[0077]

[0078] k1=1,…,n, k2=1,…,m, k=1,…,n*m

[0079] Where V x (k) is the speed of the target in the X-axis direction, V y (k) is the speed of the target in the Y-axis direction, To search for velocity components, The angle component to search for. The calculation formula is as follows:

[0080]

[0081] where v max is the maximum search speed, n is the search speed resolution, and parameter v max =4.25, n=17.

[0082] The calculation formula is as follows:

[0083]

[0084] angle s =360° / m

[0085] Wherein m is the search angle resolution, parameter m=8.

[0086] (4) Based on step (2), the local mean filter image group {M n-m ~M n}, using the velocity vector V calculated in step (3) x (k) and V y (k), calculate the multi-directional shift accumulation image sequence MFA (x, y, k), the formula is as follows:

[0087]

[0088] Among them, M f (x f,k ,y f,k ) is the image M f Medium (x f,k ,y f，k ) is the pixel grayscale value at the coordinates, (x f,k ,y f,k) is the coordinate data of the target at coordinate (x, y) in the nfth frame image after it is shifted along the kth search direction, (x f,k ,y f，k ) is calculated as follows:

[0089] x f,k =x+f*V x (k)

[0090] y f，k =y+f*V y (k)

[0091] (5) Based on step (4), the maximum projection calculation is performed on the three-dimensional image MFA (x, y, k) to obtain the multi-directional shifted cumulative grayscale maximum image MFA max , the formula is as follows:

[0092] MFA max =max(MFA(x, y, k))

[0093] (6) Based on step (5), use MFA max Image calculation multi-directional shift accumulation result graph MFA out , the formula is as follows:

[0094] MFA out (x, y) = MFA′ max (x, y)×MFA′ max (x, y)×MFA′ max (x, y)

[0095]

[0096] (7) Based on step (6), extract the multi-directional shift accumulation result graph MFA out The local neighborhood pixels around the pixel (x, y) in , construct the central block S0 and the neighborhood background block S. The regions of S and S0 are represented as follows:

[0097]

[0098] R s ={(p, q)|max(|px|, |qy|≤s)}, s=4, 7, 10, 13

[0099] Where (i, j) and (p, q) are image MFA out The pixel coordinates in (x, y) are the center pixel coordinates of the image block, s and l represent the radii of S and S0 respectively, and the parameters l=1, s=3×l+1.

[0100] (8) Based on step (7), the mean value and standard deviation of the grayscale of pixels in block S are calculated to characterize the background and noise components of the neighborhood background block. The formula is as follows:

[0101]

[0102] MFA out (p, q) is the grayscale value of the pixel at the coordinate (p, q) in block S, (x, y) is the coordinate of the center pixel of block S, MFA m (x, y) and MFA s (x, y) represent the mean and standard deviation of the grayscale values ​​of the pixel in the neighborhood block S of the pixel (x, y) respectively.

[0103] (9) Based on step (8), use the background component MFA of the neighborhood background block m (x, y) calculates the neighborhood background component value MFA m (x, y) suppresses the background component of the target area and obtains the target component map MFA t , the formula is as follows:

[0104] MFA t (p, q) = MFA out (p, q)-MFA m (x,y),(p,q)∈S0

[0105] (10) Based on step (9), use the target component graph MFA t Calculate the energy accumulation value E of the pixel grayscale in block S0 t (x, y), amplify the signal of the target component and quickly enhance the target energy. The calculation formula is as follows:

[0106]

[0107] (11) Based on step (9), the ratio of the energy accumulation value to the local background noise intensity is calculated as the signal-to-noise ratio gain factor M c , the formula is as follows:

[0108] M c (x, y) = E t (x, y) / MFA s (x, y)

[0109] (12) Based on step (9), the energy concentration characteristics of the target area are calculated, and its inverse proportional factor is calculated as the energy concentration gain factor M E , which is used to suppress isolated strong noise pixels, and its formula is as follows:

[0110] M E =(1-Tenv ) 2

[0111] Where T env is the target energy concentration, and its formula is as follows:

[0112]

[0113] MFA t_max R after background suppression s The maximum grayscale value of the pixel.

[0114] (13) Based on steps (11) and (12), the signal-to-noise ratio gain factor M c and energy concentration gain factor M E Calculate the local energy gain value G of the pixel at coordinate (x, y) in the image L (x, y), the formula is as follows:

[0115] G L (x, y) = MFA t (x, y)×M c (x, y) × ME (x, y)

[0116] (14) Calculate the adaptive segmentation threshold Th of the local energy gain map for G L Perform binary segmentation to determine the target location. The threshold Th is defined as follows:

[0117] Th=μ+k×σ

[0118] Where k is the division coefficient, and the recommended value is 20 to 30. L When the value of an element in is greater than Th, it is set to 1, otherwise it is set to 0. The point set to 1 is judged as a candidate target.

[0119] Figure 3 is a schematic diagram of the enhancement effect of a single dim target infrared simulation image provided by an embodiment of the present invention, Figure 3 (a) is the reference frame image F to be processed input in step (1) n , Figure 3 (b) is the result after step (6). Figure 3 (c) is the result diagram after processing in step (13).

[0120] Figure 4 Schematic diagram of the enhancement effect of multiple infrared simulation images of dim targets provided by an example of the present invention. Figure 4 (a) is the reference frame image F to be processed input in step (1) n , Figure 4 (b) is the result after step (6). Figure 4(c) is the result diagram after processing in step (13).

[0121] Evidence of the effects of the embodiments. The embodiments of the present invention have achieved some positive effects during the development or use process, and indeed have great advantages over the prior art. The following content is described in conjunction with the data, charts, etc. of the test process.

[0122] For space targets, due to the long detection distance, the target size is smaller than the minimum spatial resolution unit of the sensor, and the imaging appears as a spot, which is generally regarded as a point target. In addition, affected by the space-based optoelectronic imaging system, point target imaging will produce a diffusion phenomenon, which is represented by the point spread function. The deep space background is a 4K cold background, which is relatively pure, and the cosmic background radiation can be ignored. The noise in the infrared image that has a greater impact on the target is random noise, which can generally be regarded as approximately obeying a Gaussian distribution. In addition, in actual detection scenarios, there are usually multiple targets with different target characteristics and different motion conditions appearing simultaneously in the field of view. Based on the above analysis, the test image used in the present invention is mainly a sequence image obtained by adding multiple infrared simulation targets to a long-wave infrared background image.

[0123] Simulation environment: Matlab2023a;

[0124] Test input: infrared target image sequence, size 512×512, background is empty background, target is a weak point target, target size is 3×3, target movement speed is 0~4 pixels / frame.

[0125] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for enhancing the energy of a dim target, characterized in that: It includes the following steps: S101, inputting an infrared dim target image group; S102 target image group background component suppression calculation; S103 Multi-directional shift accumulation search for dim targets; S104 searches for the maximum value of the accumulated result, filters it, and multiplies it; S105 searches for local energy enhancement in the spatial domain of the accumulated image to achieve energy enhancement of weak target signals.

2. The method for enhancing the energy of a dim target according to claim 1, characterized in that: The specific steps of this method are as follows: (1) Input of target enhanced image sequence; Based on the reference frame image F to be processed n , select m frames of image F from the target image sequence n-m ~F n-1 Composing an image group n-m ~F n }; (2) For the image group {F n-m ~F n }, the background component is suppressed to obtain the target background suppressed image group {M n-m ~M n }, background component suppression can be achieved by using relevant image processing algorithms such as mean filtering and median filtering; (3) Estimate the possible movement angle and direction of the target, and calculate the velocity vector V in the X-axis and Y-axis directions according to the set search angle step and speed step x (k) and V y (k), which is calculated as follows: k1=1,…,n, k2=1,…,m, k=1,…,n*m Where V x (k) is the speed of the target in the X-axis direction, V y (k) is the speed of the target in the Y-axis direction, To search for velocity components, is the search angle component; The calculation formula is as follows: where v max is the maximum value of the search speed, and n is the search speed resolution; The calculation formula is as follows: angle s =360° / m Where m is the search angle resolution; (4) For the local mean filter image group {M n-m ~M n }, using the velocity vector V calculated in the previous step x (k) and V y (k), search for the pixel gray value of the pixel coordinate (x, y) on each local mean filter image after shifting on each possible motion vector and accumulate them to obtain the multi-directional shift accumulation image sequence MFA (x, y, k), the formula is as follows: Among them, M f (x f，k ,y f，k ) is the image M f Medium (x f，k ,y f，k ) is the pixel gray value at the coordinates, (x f，k ,y f，k ) is the coordinate of the target at (x, y) in the nfth frame image after it is shifted along the kth search direction, (x f，k ,y f，k ) is calculated as follows: x f，k =x+f*V x (k) y f，k =y+f*V y (k) (5) Perform maximum projection calculation on the shifted and accumulated three-dimensional image MFA(x, y, k) to obtain the multi-directional shifted and accumulated grayscale maximum image MFA max , the formula is as follows: MFA max =max(MFA(x,y,k)) (6) MFA max The image is negatively suppressed and each pixel is multiplied to further enhance the target energy to obtain the multi-directional shift accumulation result map MFA out , the formula is as follows: MFA out (x,y)=MFA′ max (x,y)×MFA′ max (x,y)×MFA′ max (x,y) (7) For the multi-directional shift accumulation result graph MFA out Take the pixel (x, y) in the image, extract the local neighborhood pixels around the pixel, and construct the central block S0 and the neighborhood background block S; the regions of S and S0 are represented as follows: R S ={(p,q)|max(|p-x|,|q-y|≤s)},s=4,7,10,13 Where (i, j) and (p, q) are image MFA out The pixel coordinates in (x, y) are the center pixel coordinates of the image block, s and l represent the radii of S and S0 respectively, and the value of l is determined by the size of the target to be detected; (8) The mean and standard deviation of the grayscale values ​​of pixels in block S are calculated to characterize the background and noise components of the neighborhood background block. The formula is as follows: MFA out (p, q) is the grayscale value of the pixel at the coordinate (p, q) in block S, (x, y) is the coordinate of the center pixel of block S, MFA m (x, y) and MFA s (x, y) represent the mean and standard deviation of the grayscale values ​​of the pixel in the neighborhood S of the pixel (x, y) respectively; (9) Using the neighborhood background component value MFA calculated in the previous step m (x, y) suppresses the background component of the target area and obtains the target component map MFA t , the formula is as follows: MFA t (p,q)=MFA out (p,q)-MFA m (x,y),(p,q)∈S0 (10) Calculate the energy accumulation value E of the pixel grayscale in block S0 t (x, y), amplify the signal of the target component and quickly enhance the target energy. The calculation formula is as follows: (11) Calculation of signal-to-noise ratio gain factor: Calculate the ratio of the energy accumulation value to the local background noise intensity as the signal-to-noise ratio gain factor M c , the formula is as follows: M c (x,y)=E t (x,y) / MFA s (x,y) (12) Calculate the energy concentration characteristics of the target area and calculate its inverse proportional factor as the energy concentration gain factor M E , which is used to suppress isolated strong noise pixels, and its formula is as follows: M E =(1-T env ) 2 Where T env is the target energy concentration, and its formula is as follows: MFA t_max R after background suppression s The maximum grayscale value of the middle pixel; (13) With the help of the signal-to-noise ratio gain factor M c and energy concentration gain factor M E Calculate the local energy gain value G of the pixel at coordinate (x, y) in the image L (x, y), the formula is as follows: G L (x,y)=MFA t (x,y)×M c (x,y)×M E (x,y) (14) Calculate the adaptive segmentation threshold Th of the local energy gain map for G L Perform binary segmentation processing to determine the target position; the threshold Th is defined as follows: Th=μ+k×σ Where k is the division coefficient; when G L When the value of an element in is greater than Th, it is set to 1, otherwise it is set to 0. The point set to 1 is judged as a candidate target.

Images