Rattlesnake bionic PCNN image fusion method based on snake optimization algorithm

Abstract

The invention discloses a rattlesnake bionic PCNN image fusion method based on a snake optimization algorithm, comprising: preprocessing a visible light source image and an infrared source image respectively through an ON central receptive field model, extracting common information and unique information of the infrared source image and the visible light source image, and using the obtained infrared unique information and visible light unique information as a new infrared source image and a new visible light source image respectively; enhancing the preprocessed image, the new infrared source image and the new visible light source image based on the six dual-mode cell mechanisms of the rattlesnake to obtain enhanced image feature information; fusing the enhanced image feature information with a PCNN model, and adaptively adjusting the parameters of the PCNN model through a snake optimization algorithm to obtain a fused image. The invention can obtain more feature information and improve the image fusion effect.

Description

Technical Field

[0001] The invention relates to the technical field of bionic infrared light and visible light image fusion, and more specifically to a rattlesnake bionic PCNN image fusion method based on a snake optimization algorithm. Background Art

[0002] Image fusion refers to the synthesis of multiple images obtained by different sensors through a certain algorithm to generate a new fused image with complementary feature information. The generated fused image has the feature information of multiple source images at the same time, so that the scene content of the image has a more comprehensive supplement, which is more conducive to human eye recognition and machine detection. Therefore, image fusion technology is an important application in computer vision. And according to the different types of sensors or different lens focus areas, the research fields of image fusion are also different, such as multi-focus image fusion, multi-modal image fusion, etc. Among them, multi-focus image fusion is aimed at images with different focus areas when shooting in the same scene, and images with different focus areas are synthesized into a clear image with full scene content. Multi-modal image fusion is aimed at images obtained by different types of sensor devices, such as medical images of magnetic resonance imaging (MRI), infrared images and visible light images. The fusion method of infrared image and visible light image for multi-modal image fusion terminal is divided into bionic method and non-bionic method. The present invention belongs to the field of bionic infrared light and visible light image fusion.

[0003] Regarding the image fusion technology in the field of bionics, some scholars have used rattlesnakes as research objects to simulate their visual imaging mechanisms. For example, AMW Axman et al. proposed a classic model for the fusion of visible light and infrared images based on the working principle of rattlesnake visual imaging and the established visual receptive field model. The classic fusion structure is composed of ON / OFF central receptive fields. The bionic image fusion models based on rattlesnakes proposed so far all process the source image through the established rattlesnake model and then output three images to produce a pseudo-color fused image. However, the fusion effect and visual effect of some of the generated pseudo-color fused images are not very ideal.

[0004] Therefore, the purpose of this patent is to propose a PCNN image fusion model based on rattlesnake bionics, which is first processed by the bionic rattlesnake model, and then the processed image is input into the multi-channel PCNN to output the final fused image, so that the fused image has more feature information of the source image. Summary of the invention

[0005] In view of this, the present invention provides a rattlesnake bionic PCNN image fusion method based on snake optimization algorithm, abandoning the current situation that the processed image is mapped to RGB channels to generate a pseudo-color fused image in the prior art, and proposes a method of first processing the source image through an established ON / OFF central receptive field and a rattlesnake bionic model composed of six dual-mode cells, and then fusing the processed images through a multi-channel PCNN, thereby obtaining more feature information and improving the image fusion effect.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] The rattlesnake bionic PCNN image fusion method based on the snake optimization algorithm includes:

[0008] The visible light source image and the infrared source image are preprocessed respectively by the ON central receptive field model, and the common information and unique information of the input infrared source image and the visible light source image are extracted, and the obtained infrared unique information and visible light unique information are used as the new infrared source image and the new visible light source image respectively;

[0009] Based on the six dual-mode cell mechanisms of rattlesnake, the pre-processed images, new infrared source images and new visible light source images are enhanced to obtain enhanced image feature information;

[0010] The PCNN model fuses the enhanced image feature information, and the PCNN model is adaptively adjusted through the snake optimization algorithm to obtain a fused image.

[0011] Preferably, the ON center receptive field model is:

[0012]

[0013] The OFF center receptive field model is:

[0014]

[0015] Among them, A is a constant, D is the cell basal activity, E and F are polarization constants, and X k (i, j) and They are images after the ON center receptive field model and the OFF center receptive field model, and the receptive field center area C k (i, j) and the surrounding area S of the receptive field k (i, j) obeys Gaussian distribution, respectively:

[0016]

[0017]

[0018] Among them, σc represents the central region space constant, σ s represents the space constant of the surrounding area, I k (i, j) represents the input image, i, j represent the pixel position; the Gaussian kernel of the central area is set to m×n, and the Gaussian kernel of the surrounding area is set to p×q;

[0019] Extract the common information of the input infrared source image and visible light source image:

[0020] I ir (i, j)∩I v (i, j) = min{I ir (i, j), I v (i, j)}

[0021] Among them, I ir (i, j) represents the infrared source image, I v (i, j) represents the visible light source image, I ir (i, j)∩I v (i, j), ∩ taking the intersection means taking the common feature information in the two images;

[0022] Extract unique information from the input infrared and visible light source images:

[0023] I IR (i, j) = I ir (i, j)-I ir (i, j)∩I v (i, j)

[0024] I V (i, j) = I v (i, j)-I ir (i, j)∩I v (i, j)

[0025] Among them I IR (i, j) represents the unique feature information of the infrared source image, I V (i, j) represents the unique feature information of the visible light source image.

[0026] Preferably, the six dual-mode cell mechanisms of rattlesnake include the visible light enhanced infrared cell mathematical model, the visible light inhibited infrared cell mathematical model, the infrared enhanced visible light cell mathematical model, the infrared inhibited visible light cell mathematical model, the and cell mathematical model and the or cell mathematical model.

[0027] Preferably, the infrared enhanced visible light cell mathematical model:

[0028] I IR+V (i, j) = I IR(i,j)exp[I V (i, j)]

[0029] Among them, I IR+V (i, j) represents the image obtained by enhancing visible light by infrared, and exp represents the exponential function;

[0030] Mathematical model of infrared suppression of visible light cells:

[0031] I IR-V (i, j) = I IR (i, j)log[I V (i, j)+1]

[0032] Among them, I IR-V (i, j) represents the image obtained by suppressing visible light by infrared;

[0033] Mathematical model of visible light enhanced infrared cells:

[0034] I V+IR (i, j) = I V (i, j)exp[I IR (i, j)]

[0035] Among them, I V+IR (i, j) represents the image obtained by enhancing infrared with visible light;

[0036] Mathematical model of visible light inhibiting infrared cells:

[0037] I V-IR (i, j) = I V (i, j)log[I IR (i, j)+1]

[0038] Among them, I V-IR (i, j) represents the image obtained by suppressing infrared by visible light;

[0039] With Cell Mathematical Models:

[0040] When I V (i, j)<I IR (i, j), it is:

[0041] I AND (i, j) = mI V (i,j)+nI IR (i, j)

[0042] When I V (i, j)>I IR When (i, j), it is:

[0043] I AND (i,j)=nIV (i,j)+mI IR (i, j)

[0044] Where m>0.5, n<0.5, I AND (i, j) represents the image obtained by weighting the infrared source image and the visible light source image;

[0045] Or a cell mathematical model:

[0046] When I V (i, j)<I IR When (i, j), it is:

[0047] I OR (i,j)=nI V (i,j)+mI IR (i, j)

[0048] When I V (i, j)>I IR When (i, j), it is:

[0049] I OR (i, j) = mI V (i,j)+nI IR (i, j)

[0050] Where m>0.5, n<0.5, I OR (i, j) represents the image obtained by weighted OR of the infrared source image and the visible light source image.

[0051] Preferably, the PCNN model has three basic parts: receiving domain, modulation domain and pulse generator;

[0052] The purpose of the receiving domain is to receive two types of input:

[0053]

[0054]

[0055] Among them, F ij and L ij Represents feedback input and link input, S ij represents external stimulus, M ij kl and W ij kl Represents the connection weight matrix, which represents the connection weight matrix of the gain strength of the feedback input and the connection input between neurons i and j, that is, the contribution of each neuron in the neighborhood to the central neuron, which can adjust the influence of each neuron in the neighborhood of the central neuron. Generally, M ij kl =W ij kl , α L and α FRespectively represent the time decay constants of the F channel and the L channel, which determine the decay speed of the F and L channels, usually α L >α F , V F and V L is the intrinsic potential constant, is the amplitude coefficient of the feed input and the connection input, represents the amplitude adjustment constant of the connection domain, and can scale the energy transmitted to the central neuron by the firing neurons in the neighborhood. The subscript ij represents the position of the central pixel of the PCNN, and the subscript kl represents the position of the adjacent pixel corresponding to the central pixel; n represents the number of iterations, and Y kl Represents the output of adjacent pixels;

[0056] In the modulation domain:

[0057] U ij (n) = F ij (n)[1+βL ij (n)]

[0058] Among them, U ij represents the internal state of the neuron, β represents the connection coefficient of the modulation domain;

[0059] The function of the pulse generator g is to generate pulse output, which consists of a threshold regulator, a comparator and a pulse generator:

[0060]

[0061]

[0062] Among them, T ij (n) represents the dynamic threshold, α T represents the time decay constant, V T Expressed as a coefficient, usually a constant, when the internal state of the neuron U ij Greater than the threshold T ij , that is, the neuron satisfies the condition U ij (n)>T ij (n), the neuron generates a pulse, also called a firing.

[0063] Preferably, the snake optimization algorithm comprises:

[0064] S1: Initialize the population. The initialization formula is:

[0065] X i =X min +r×(X max -X min )

[0066] Among them, X max and X minare the upper and lower bounds of the function to be optimized. The population is divided into female and male parts in a 1:1 ratio. r is a random number between 0 and 1. X i represents the position of the i-th snake;

[0067] S2: Define temperature Temp and food quantity Q:

[0068]

[0069]

[0070] Where t represents the current iteration, T represents the maximum number of iterations, and c1 = 0.5, which is the global optimal fitness;

[0071] S3: Determine whether the food quantity Q is less than 0.25. If so, execute S4 for global search; otherwise, execute S5 for local search.

[0072] S4: The snake selects a random position to search for food and updates the position of the snake relative to the food. The random search formula is:

[0073]

[0074] in, Indicates the male position, represents the randomly selected male position, c2 = 0.05, and rand represents a random number between 0 and 1;

[0075]

[0076] in, is the randomly selected male position The fitness value of yes The fitness value of

[0077] The global search formula of female snakes is basically the same as that of males;

[0078]

[0079] in, Indicates the female position, represents the randomly selected female position, c2 = 0.05, and rand represents a random number between 0 and 1;

[0080]

[0081] in, is a randomly selected female position The fitness value of yes The fitness value of

[0082] S5: Determine whether the temperature Temp is greater than 0.6. If it is greater, the snake will move toward the location of the food, that is, closer to the global optimal fitness:

[0083] X i，j (t+1)=X food ±c3×Temp×rand×(X food -X i，j (t))

[0084] Among them, X food is the global optimal position, c3=2,,X i，j (t) is the location of all male and female individuals in the snake group;

[0085] If the temperature Temp is less than 0.6, the snake will fight or mate. Fighting mode:

[0086]

[0087]

[0088] in, is the ith male position, FM is the male fighting value, It is the best position for females; is the best individual position in the female snake group The fitness value, f i is the fitness value of individual snake;

[0089] The global search formula of female snakes is basically the same as that of males;

[0090]

[0091]

[0092] in, is the ith female position, FF is the female fighting value, It is the optimal position for males; Best position for males in the group The fitness value, f i is the fitness value of individual snake;

[0093] Mating Mode:

[0094]

[0095]

[0096] Among them, M m Indicates male mating ability. and Represent the best individual fitness in the female and male populations, is the position of the ith female. If there are snake eggs hatching, the worst individual is selected to replace them;

[0097]

[0098]

[0099] Among them, M f Indicates the female's ability to mate. For the ith male position, if there are snake eggs hatching, the worst individual is selected for replacement;

[0100] S6: Determine whether the current iteration t is less than the maximum number of iterations T. If so, continue iterating and execute S2; otherwise, the process ends.

[0101] It can be seen from the above technical solutions that, compared with the prior art, the present invention discloses a rattlesnake bionic PCNN image fusion method based on a snake optimization algorithm, which firstly performs image preprocessing: infrared images and visible light images are processed through the ON / OFF central receptive field, and the common information and unique information of the infrared source image and the visible light source image are extracted, thereby improving the quality of the fused image; secondly, the six dual-mode cell fusion mechanisms of the rattlesnake are utilized to better simulate the visual perception mechanism of the rattlesnake; and the PCNN model is used for fusion again to improve the detail feature information of the final fused image, and the entire model is adaptively adjusted through the snake optimization algorithm, without the need to manually set parameters, so that the obtained fused image has improved visual effects, and the detail feature information is more obvious and the target is more prominent. BRIEF DESCRIPTION OF THE DRAWINGS

[0102] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0103] Figure 1 The accompanying drawing is a schematic flow chart of the rattlesnake bionic PCNN image fusion method based on the snake optimization algorithm provided by the present invention.

[0104] Figure 2 The attached figure is a schematic diagram of the structure of the ON central receptive field model.

[0105] Figure 3 The attached figure is a structural schematic diagram of the OFF central receptive field model.

[0106] Figure 4The attached figure is a schematic diagram of the PCNN model structure.

[0107] Figure 5 The attached figure is a flow chart of the snake optimization algorithm.

[0108] Figure 6 The accompanying drawing is a framework diagram of the rattlesnake bionic PCNN image fusion method based on the snake optimization algorithm provided by the present invention. DETAILED DESCRIPTION

[0109] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions 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 are within the scope of protection of the present invention.

[0110] The embodiment of the present invention discloses a rattlesnake bionic PCNN image fusion method based on a snake optimization algorithm, such as Figure 1 As shown, including:

[0111] The visible light source image and the infrared source image are preprocessed respectively by the ON central receptive field model, and the common information and unique information of the input infrared source image and the visible light source image are extracted, and the obtained infrared unique information and visible light unique information are used as the new infrared source image and the new visible light source image respectively;

[0112] Based on the six dual-mode cell mechanisms of rattlesnakes, the pre-processed images, new infrared source images and new visible light source images are enhanced to obtain enhanced image feature information; the six dual-mode cell mechanisms of rattlesnakes are to further process the obtained image feature information based on the mathematical model of cells by simulating the visual imaging mechanism of rattlesnakes;

[0113] The three images (visible light enhanced image, infrared visible enhanced image, infrared enhanced image) are fused through a multi-channel PCNN model to obtain the final fused image. Among them, the pulse coupled neural network PCNN is a neural network model based on the neuron signal of the cat visual cortex. It is an artificial neural network that uses the synchronous pulse group of neurons to process images without training;

[0114] The snake optimization algorithm is used to adaptively set the parameters of the PCNN model, so that manual parameter setting is no longer required. The snake optimization algorithm is a swarm intelligence optimization algorithm based on snake behavior simulation. It simulates the foraging, fighting, mating, and egg-laying behaviors of snakes in nature, and can solve the optimal solution of the objective function according to requirements, so that the model can achieve better fusion effect.

[0115] In this embodiment, image preprocessing includes:

[0116] First, the visible light source image is preprocessed by the ON center receptive field model, and the infrared source image is preprocessed by the OFF center receptive field model:

[0117] like Figure 2 As shown, the ON central receptive field model is:

[0118]

[0119] like Figure 3 As shown, the OFF center receptive field model is:

[0120]

[0121] Among them, A is a constant, D is the cell basal activity, E and F are polarization constants, and X k (i, j) and It is the image after the ON / OFF receptive field. The central area of ​​the receptive field C k (i, j) and the surrounding area S of the receptive field k (i, j) obeys Gaussian distribution, respectively:

[0122]

[0123]

[0124] σ c represents the central region space constant, σ s represents the space constant of the surrounding area, I k (i, j) represents the input image, i, j represent the pixel position; the Gaussian kernel of the central area is set to m×n, and the Gaussian kernel of the surrounding area is set to p×q;

[0125] Then, the common information of the input infrared source image and the visible light source image is extracted. The common information is the common feature information in both the infrared source image and the visible light source image:

[0126] I ir (i, j)∩I v (i, j) = min{I ir (i, j), I v (i, j)}

[0127] Among them I ir (i, j) represents the infrared source image, I v (i, j) represents the visible light source image, I ir (i, j)∩I v (i, j), ∩ taking the intersection means taking the common feature information in the two images;

[0128] Extract the unique information of the input infrared source image and visible light source image. The unique information is the unique feature information of the infrared source image and the visible light source image:

[0129] I IR (i, j) = I ir (i, j)-I ir (i, j)∩I v (i, j)

[0130] I V (i, j) = I v (i, j)-I ir (i, j)∩I v (i, j)

[0131] Among them I IR (i, j) represents the unique feature information of the infrared source image, I V (i, j) represents the unique feature information of the visible light source image;

[0132] In this embodiment, the six dual-mode cell mechanisms of rattlesnakes include a visible light enhanced infrared cell mathematical model, a visible light inhibited infrared cell mathematical model, an infrared enhanced visible light cell mathematical model, an infrared inhibited visible light cell mathematical model, an AND cell mathematical model, and an OR cell mathematical model:

[0133] Infrared enhanced visible light cell mathematical model:

[0134] I IR+V (i, j) = I IR (i, j)exp[I V (i, j)]

[0135] Among them, I IR+V (i, j) represents the image obtained by enhancing visible light by infrared;

[0136] Mathematical model of infrared suppression of visible light cells:

[0137] I IR-V (i, j) = I IR (i, j)log[I V (i, j)+1]

[0138] Among them, I IR-V (i, j) represents the image obtained by suppressing visible light by infrared;

[0139] Mathematical model of visible light enhanced infrared cells:

[0140] I V+IR (i, j) = I V(i, j)exp[I IR (i, j)]

[0141] Among them, I V+IR (i, j) represents the image obtained by enhancing infrared with visible light;

[0142] Mathematical model of visible light inhibiting infrared cells:

[0143] I V-IR (i, j) = I V (i, j)log[I IR (i, j)+1]

[0144] Among them, I V-IR (i, j) represents the image obtained by suppressing infrared by visible light;

[0145] With Cell Mathematical Models:

[0146] When I V (i, j)<I IR When (i, j), it is:

[0147] I AND (i, j) = mI V (i,j)+nI IR (i, j)

[0148] When I V (i, j)>I IR When (i, j), it is:

[0149] I AND (i,j)=nI V (i,j)+mI IR (i, j)

[0150] Where m>0.5, n<0.5, I AND (i, j) represents the image obtained by weighting the infrared image and the visible light image;

[0151] Or a cell mathematical model:

[0152] When I V (i, j)<I IR When (i, j), it is:

[0153] I OR (i,j)=nI V (i,j)+mI IR (i,j)

[0154] When I V (i,j)>I IR (i,j), it is:

[0155] I OR (i,j)=mI V (i,j)+nI IR (i,j)

[0156] Where m>0.5, n<0.5, I OR (i, j) represents the image obtained by weighted OR of the infrared source image and the visible light source image.

[0157] like Figure 6 As shown, the new visible light source image and the new infrared source image are processed by six rattlesnake dual-mode cell mathematical models to obtain or output signal I OR (i, j), and the output signal I AND (i, j), infrared enhanced visible light output signal I IR+V (i, j), infrared suppression visible light output signal I IR-V (i, j), visible light enhanced infrared output signal I V+IR (i, j) and visible light suppresses infrared output signal I V-IR (i, j) Output signals of six rattlesnake dual-mode cell models.

[0158] Subtract the preprocessed visible light source image from the output signal I AND (i, j), and enhanced by the ON central receptive field model, the output signal I OR (i, j), and the output signal I AND (i, j) is fed into the OFF central receptive field model, and finally the visible light enhanced image is generated through the ON central receptive field;

[0159] Infrared enhances visible light output signal I IR+V (i, j), infrared suppression visible light output signal I IR-V (i, j), visible light enhanced infrared output signal I V+IR (i, j) and visible light suppresses infrared output signal I V-IR (i, j) are fed into the ON central receptive field model respectively to generate two enhanced information, and finally the infrared visible enhanced image is obtained by combining with the cell;

[0160] Subtract the preprocessed infrared source image from the output signal I AND (i, j), and enhanced by the ON central receptive field, and then the output signal I OR (i, j), and the output signal I AND (i, j) is fed into the OFF center receptive field model, and finally the two generate an infrared enhanced image through the ON center receptive field.

[0161] In this embodiment, if Figure 4As shown, the PCNN model has three basic parts: receiving domain, modulation domain and pulse generator.

[0162] The purpose of the receiving domain is to receive two types of input:

[0163]

[0164]

[0165] Among them, F ij and L ij Represents feedback input and link input, S ij represents external stimulus, M ijkl and W ijkl Represents the connection weight matrix, which represents the connection weight matrix of the gain strength of the feedback input and the connection input between neurons i and j, that is, the contribution of each neuron in the neighborhood to the central neuron, which can regulate the influence of each neuron in the neighborhood of the central neuron. L and α F Indicates the time decay constant of the L channel and the F channel, which determines the decay speed of the F and L channels. Usually α L >α F . V F and V L is the intrinsic potential constant, is the amplitude coefficient of the feed input and the connection input, and represents the amplitude adjustment constant of the connection domain, which can scale the energy transmitted to the central neuron by the firing neurons in the neighborhood. The subscript ij represents the position of the central pixel of the PCNN. The subscript kl represents the position of the adjacent pixel corresponding to the central pixel, n represents the number of iterations, and Y kl Represents the output of neighboring pixels.

[0166] In the modulation domain:

[0167] U ij (n) = F ij (n)[1+βL ij (n)]

[0168] Among them, U ij represents the internal state of the neuron, and β represents the connection coefficient of the modulation domain.

[0169] The function of the pulse generator g is to generate pulse output, which consists of a threshold regulator, a comparator and a pulse generator:

[0170]

[0171]

[0172] Among them, T ij (n) represents the dynamic threshold, αT represents the time decay constant, V T Expressed as a coefficient, usually a constant. When the internal state of the neuron U ij Greater than the threshold T ij , that is, the neuron satisfies the condition U ij (n)>T ij (n), the neuron will generate a pulse, also called a firing. The three processed images are input into PCNN, and finally the images are fused according to the firing map obtained by PCNN to obtain the final fusion result.

[0173] In this embodiment, the snake optimization algorithm is a simple and effective optimization algorithm proposed by mathematically simulating and modeling the foraging and reproduction behaviors and patterns of snakes. Figure 5 As shown, including:

[0174] First, a uniformly distributed random population is generated so that the optimization algorithm can be started. The initialization formula is:

[0175] X i =X min +r×(X max -X min )

[0176] Among them, X max and X min are the upper and lower bounds of the function to be optimized, and then the population is divided into female and male parts in a 1:1 ratio. r is a random number between 0 and 1. X i Represents the position of the i-th snake.

[0177] The behavior of the snake is mainly affected by the temperature Temp and the amount of food Q. The temperature Temp and the amount of food Q are defined as:

[0178]

[0179]

[0180] Among them, t represents the current iteration, T represents the maximum number of iterations, c1=0.5, and food is the global optimal fitness.

[0181] The algorithm is divided into two stages, global search and local search:

[0182] (1) Global search (no food)

[0183] When Q < 0.25, the snake will choose a random position to search for food and update the position of the snake relative to the food. The random search formula is:

[0184]

[0185] in, Indicates the male position, represents the randomly selected male position, c2=0.05, and rand represents a random number between 0 and 1.

[0186]

[0187] in, is the randomly selected male position The fitness value of yes The fitness value of the female snake group; the global search formula of the female snake group is basically the same as that of the male, as follows:

[0188]

[0189] in, Indicates the female position, represents the randomly selected female position, c2 = 0.05, and rand represents a random number between 0 and 1;

[0190]

[0191] in, is a randomly selected female position The fitness value of yes The fitness value of

[0192] (2) Local search (with food, Q>0.25)

[0193] If there is food and the current temperature is hot, that is, Q>0.25, Temp>0.6, the snake will move towards the location of the food, that is, close to the global optimal fitness:

[0194] X i,j (t+1)=X food ±c3×Temp×rand×(X food -X i,j (t))

[0195] Among them, X food is the global optimal position, c3=2, X i,j (t) is the position of all individuals in the snake group.

[0196] If there is food and the current temperature is low (temperature < 0.6), the snake will fight or mate. Fighting mode:

[0197]

[0198]

[0199] in, is the ith male position, FM is the male fighting value, It is the best position for females; is the best individual position in the female snake group The fitness value, f i is the fitness value of the individual snake.

[0200] The global search formula of female snakes is basically the same as that of males, as follows:

[0201]

[0202]

[0203] in, is the ith female position, FF is the female fighting value, It is the optimal position for males; Best position for males in the group The fitness value, f i is the fitness value of individual snake;

[0204] Mating Mode:

[0205]

[0206]

[0207] Among them, M m Indicates male mating ability. and Represent the best individual fitness in the female and male populations, respectively. is the position of the ith female. If there are snake eggs hatching, the worst individual is selected to replace them;

[0208] The formula for the mating pattern of female snakes is basically the same as that of males, as follows:

[0209]

[0210]

[0211] Among them, M f Indicates the female's ability to mate. For the i-th male position, if there are snake eggs hatching, choose the worst individual to replace it.

[0212] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0213] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A rattlesnake bionic PCNN image fusion method based on snake optimization algorithm, characterized in that: include: The visible light source image and the infrared source image are preprocessed respectively by the ON central receptive field model, and the common information and unique information of the input infrared source image and the visible light source image are extracted, and the obtained infrared unique information and visible light unique information are used as the new infrared source image and the new visible light source image respectively; Based on the six dual-mode cell mechanisms of rattlesnake, the pre-processed images, new infrared source images and new visible light source images are enhanced to obtain enhanced image feature information; The PCNN model fuses the enhanced image feature information, and the PCNN model is adaptively adjusted through the snake optimization algorithm to obtain a fused image; The snake optimization algorithm includes: S1: Initialize the population. The initialization formula is: X i =X min +r×(X max -X min ) Among them, X max and X min are the upper and lower bounds of the function to be optimized. The population is divided into female and male parts in a 1:1 ratio. r is a random number between 0 and 1. X i represents the position of the i-th snake; S2: Define temperature Temp and food quantity Q: Where t represents the current iteration, T represents the maximum number of iterations, and c1 = 0.5, which is the global optimal fitness; S3: Determine whether the food quantity Q is less than 0.

25. If so, execute S4 for global search; otherwise, execute S5 for local search. S4: The snake selects a random position to search for food and updates the position of the snake relative to the food. The random search formula is: in, Indicates the male position, represents the randomly selected male position, c2 = 0.05, and rand represents a random number between 0 and 1; in, is the randomly selected male position The fitness value of yes The fitness value of in, Indicates the female position, represents the randomly selected female position, c2 = 0.05, and rand represents a random number between 0 and 1; in, is a randomly selected female position The fitness value of yes The fitness value of S5: Determine whether the temperature Temp is greater than 0.

6. If it is greater, the snake will move toward the location of the food, that is, closer to the global optimal fitness: X i,j (t+1)=X food ±c3×Temp×rand×(X food -X i,j (t)) Among them, X food is the global optimal position, c3=2, X i,j (t) is the position of all individuals in the snake group; If the temperature Temp < 0.6, the snake will fight or mate. Fighting mode: in, is the ith male position, FM is the male fighting value, It is the best position for females; is the best individual position in the female snake group The fitness value, f i is the fitness value of individual snake; in, is the ith female position, FF is the female fighting value, It is the optimal position for males; Best position for males in the group The fitness value, f i is the fitness value of individual snake; Mating Mode: Among them, M m Indicates male mating ability. and Represent the best individual fitness in the female and male populations, is the position of the ith female. If there are snake eggs hatching, the worst individual is selected to replace them; Among them, M f Indicates the female's ability to mate. For the ith male position, if there are snake eggs hatching, the worst individual is selected for replacement; S6: Determine whether the current iteration t is less than the maximum number of iterations T. If so, continue iterating and execute S2; otherwise, the process ends.

2. The rattlesnake bionic PCNN image fusion method based on the snake optimization algorithm according to claim 1 is characterized in that: The ON central receptive field model is: The OFF center receptive field model is: Among them, A is a constant, D is the cell basal activity, E and F are polarization constants, and X k (i,j) and They are images after the ON center receptive field model and the OFF center receptive field model, and the receptive field center area C k (i, j) and the surrounding area S of the receptive field k (i,j) obeys Gaussian distribution, which are: Among them, σ c represents the central region space constant, σ s represents the space constant of the surrounding area, I k (i, j) represents the input image, i, j represents the pixel position; the Gaussian kernel of the central area is set to m×n, and the Gaussian kernel of the surrounding area is set to p×q; Extract the common information of the input infrared source image and visible light source image: IN ir (i,j)∩I v (i,j)=min{I ir (i,j),I v (i,j)} Among them, I ir (i,j) represents the infrared source image, I v (i,j) represents the visible light source image, I ir (i,j)∩I v (i, j), ∩ taking the intersection means taking the common feature information in the two images; Extract unique information from the input infrared and visible light source images: I IR (i,j)=I ir (i,j)-I ir (i,j)∩I v (i,j) I V (i,j)=I v (i,j)-I ir (i,j)∩I v (i,j) Among them I IR (i, j) represents the unique feature information of the infrared source image, I V (i, j) represents the unique feature information of the visible light source image.

3. The rattlesnake bionic PCNN image fusion method based on snake optimization algorithm according to claim 1 is characterized in that: The six dual-mode cell mechanisms of rattlesnake include the visible light enhanced infrared cell mathematical model, the visible light inhibited infrared cell mathematical model, the infrared enhanced visible light cell mathematical model, the infrared inhibited visible light cell mathematical model, the and cell mathematical model, and the or cell mathematical model.

4. The rattlesnake bionic PCNN image fusion method based on snake optimization algorithm according to claim 3 is characterized in that: Infrared enhanced visible light cell mathematical model: I IR+V (i,j)=I IR (i,j)exp[I V (i,j)] Among them, I IR+V (i, j) represents the image obtained by enhancing visible light by infrared, and exp represents the exponential function; Mathematical model of infrared suppression of visible light cells: I IR-V (i,j)=I IR (i,j)log[I V (i,j)+1] Among them, I IR-V (i, j) represents the image obtained by suppressing visible light through infrared; Mathematical model of visible light enhanced infrared cells: I V+IR (i,j)=I V (i,j)exp[I IR (i,j)] Among them, I V+IR (i, j) represents the image obtained by enhancing infrared with visible light; Mathematical model of visible light inhibiting infrared cells: I V-IR (i,j)=I V (i,j)log[I IR (i,j)+1] Among them, I V-IR (i, j) represents the image obtained by suppressing infrared by visible light; With Cell Mathematical Models: When I V (i,j) IR (i,j), it is:​ I AND (i,j)=mI V (i,j)+nI IR (i,j) When I V (i, j)>I IR (i, j), it is as follows: I AND (i,j)=nI V (i,j)+mI IR (i,j) Where m>0.5, n<0.5, I AND (i, j) represents the image obtained by weighting the infrared source image and the visible light source image; Or a cell mathematical model: When I V (i,j) IR (i,j), it is:​ I OR (i,j)=nI V (i,j)+mI IR (i,j) When I V (i, j) > I IR (i, j), it is as follows: I OR (i,j)=mI V (i,j)+nI IR (i,j) Where m>0.5, n<0.5, I OR (i, j) represents the image obtained by weighted OR of the infrared source image and the visible light source image.

5. The rattlesnake bionic PCNN image fusion method based on snake optimization algorithm according to claim 1 is characterized in that: The PCNN model has three basic parts: receiving domain, modulation domain and pulse generator; The purpose of the receiving domain is to receive two types of input: Among them, F ij and L ij Represents feedback input and link input, S ij represents external stimulus, M ijkl and W ijkl represents the connection weight matrix, which represents the connection weight matrix of the gain strength of the feedback input and the connection input between neurons i and j, that is, the contribution of each neuron in the neighborhood to the central neuron, which can regulate the influence of each neuron in the neighborhood of the central neuron, M ijkl =W ijkl , α F and α L They represent the time decay constants of the F channel and the L channel respectively, and determine the decay speed of the F and L channels. L >α F , V F and V L is the intrinsic potential constant, is the amplitude coefficient of the feed input and the connection input, represents the amplitude adjustment constant of the connection domain, and scales the energy transmitted to the central neuron by the firing neurons in the neighborhood. The subscript ij represents the position of the central pixel of the PCNN, and the subscript kl represents the position of the adjacent pixel corresponding to the central pixel; n represents the number of iterations, and Y kl Represents the output of adjacent pixels; In the modulation domain: YOU ij (n)=F ij (n)[1+βL ij (a)] Among them, U ij represents the internal state of the neuron, β represents the connection coefficient of the modulation domain; The function of the pulse generator is to generate pulse output, which consists of a threshold regulator, a comparator and a pulse generator: Among them, T ij (n) represents the dynamic threshold, α T represents the time decay constant, V T Expressed as a coefficient, is a constant, when the internal state of the neuron U ij Greater than the threshold T ij , that is, the neuron satisfies the condition U ij (n)>T ij (n), the neuron generates a pulse, also called a firing.

Images