Two-dimensional polarization distance underwater imaging method inspired by bionic vision

Abstract

The invention discloses a two-dimensional polarization distance underwater imaging method under bionic vision inspiration, and relates to the technical field of underwater imaging, and the method comprises the following steps: S1, obtaining a polarization image; s2, extracting a low-frequency part; s3, calculating back scattering light AOP; s4, calculating backscattered light; s5, calculating the polarization sensitivity of the background; s6, calculating the polarization sensitivity of the target; s7, calculating the activity value of the background interneurons; s8, calculating the activity value of the target interneurons; and S9, obtaining a final image. According to the two-dimensional polarization distance underwater imaging method inspired by bionic vision, background interference can be effectively suppressed, and tiny polarization difference between a target and a background can be captured, so that target information can be clearly extracted in a complex water area environment, and compared with a traditional polarization imaging method, the method has the advantage that the imaging efficiency is improved. The method has obvious advantages in the aspects of image details, definition, stability and the like, and shows higher adaptability and reliability.

Description

Technical Field

[0001] The invention relates to the technical field of underwater imaging, and in particular to a two-dimensional polarization distance underwater imaging method inspired by bionic vision. Background Art

[0002] Many organisms possess polarization vision, which enables them to detect and use tiny polarization information in the environment for navigation, prey detection and communication, and have evolved body patterns to perceive polarization. These biological systems exhibit extraordinary abilities to discern tiny polarization differences even in challenging underwater conditions. For example, some coral reef fish, such as damselfish, black-tailed mudfish and three-spotted tropical fish, have a 60° periodic variation in their sensitivity to linearly polarized light at different angles, enabling them to have excellent vision even in visually complex tropical marine environments.

[0003] Current underwater polarization imaging algorithms mostly focus on using polarization information to enhance image details. However, the actual difficulty lies in effectively acquiring and utilizing underwater polarization information and reducing the estimation error of background scattered light. At the same time, existing algorithms have difficulties in effectively acquiring and utilizing underwater polarization information. The estimation error of background scattered light often affects the accuracy of imaging. A single polarization parameter can easily lead to inaccuracies and omissions when identifying targets and backgrounds. This reliance on a single parameter weakens the ability to extract image details and limits the accuracy of target detection. Secondly, although many studies have improved the performance of polarization image processing with the help of biological learning feedback and bionic models, these processes are often complex and challenging to implement. Summary of the invention

[0004] The purpose of the present invention is to provide a two-dimensional polarization distance underwater imaging method inspired by bionic vision to solve the problem of how to use biological polarization vision to effectively obtain and utilize polarization information and reduce background scattered light estimation errors raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solution: a two-dimensional polarization distance underwater imaging method inspired by bionic vision, comprising the following steps:

[0006] S1: Acquire 0°, 60°, and 120° polarization images and calculate the Stokes vector, AOP, and DOP;

[0007] S2: Perform frequency domain transformation on the Stokes vector and extract the low-frequency part as the polarization information of the backscattered light;

[0008] S3: Calculate the backscattered light AOP using the background Stocks vector;

[0009] S4: Calculate the maximum value, minimum value and DOP of the backscattered light;

[0010] S5: Calculate the polarization sensitivity of the background;

[0011] S6: Calculate the polarization sensitivity of the target;

[0012] S7: Calculate the background interneuron activity value;

[0013] S8: Calculate the activity value of the target interneuron;

[0014] S9: Calculate the two-dimensional polarization distance to obtain the final image.

[0015] Preferably, in S1, the Stokes vector can be used to describe the polarization state of light, and a polarization camera is used to obtain polarization images at 0°, 60°, and 120°. The Stokes vector can be calculated as:

[0016]

[0017] The degree of polarization (DOP) d and angle of polarization (AOP) a of the image can be expressed as:

[0018]

[0019]

[0020] Thus, polarization images of 0°, 60°, and 120° can be obtained and the Stokes vector, AOP, and DOP can be calculated.

[0021] Preferably, in S2, when polarized light irradiates a partially polarized light target, the target information is always retained in S1 and S2. When calculating the polarization information of the backscattered light, we transform the Stokes vector into the frequency domain and extract the low-frequency part. This process can be expressed as:

[0022]

[0023] Among them, S0, S1, S2 are Stokes vectors calculated from the original polarization image, which contain the target and background polarization information, FT and IFT are Fourier transform and inverse Fourier transform. represents an image containing only the polarization information of the backscattered light, H(u,v) is a low-pass filter, and (u,v) represents the coordinates in the frequency domain.

[0024] Preferably, in S3, after obtaining the Stokes vector of the background, the AOP of the backscattered light can be calculated:

[0025]

[0026] The AOP of backscattered light can be calculated according to the public information.

[0027] Preferably, in S4, using a B Calculate the maximum and minimum backscattered light and calculate the DOP of the backscattered light:

[0028]

[0029]

[0030]

[0031] In the calculation of two-dimensional polarization distance, after obtaining the background polarization state parameters, the parameter calculation of the target part can be expressed by using the DOP and AOP calculated by the Stocks vector of the entire image, and the relevant parameters of the target and the background are used as the photoreceptor cell R θ (a, d) Inputs used to calculate the activity value of the interneuron.

[0032] Preferably, in S5, when the photoreceptors in the biological retina receive light from the environment and the target, the incident light can be considered to be linearly polarized light. According to the relevant research on biological vision theory and electrophysiology, the polarization sensitivity of the photoreceptor cell R can be expressed by the cosine function of the orientation angle:

[0033]

[0034] Where φ is the measured polarization direction relative to the reference direction, i.e., the E vector axis, and φ max is the value of φ that makes R(φ) reach its maximum value, S p is the polarization sensitivity of each photoreceptor. For most high-precision polarization imaging systems, S p =10.

[0035] Preferably, in S6, when the incident light is partially polarized light, the proportion of polarized light in the incident light must be considered, d is the degree of polarization of the incident light, and R(φ,d) is expressed as:

[0036]

[0037] Used to get the proportion of incident light.

[0038] Preferably, in S7, in the organism, φ max The physical meaning of is the orientation of the microvilli of the photoreceptor cells, which can be further expressed by the polarization angle a in the model. Therefore, the polarization sensitivity of polarized light at different angles can be expressed as:

[0039]

[0040] Where, k indicates that the input polarization information is target light (O) or background light (B), a O ,d O AOP and DOP calculated for the entire image, a B ,d B AOP and DOP calculated for the background.

[0041] Preferably, in S8, when performing a target-based visual task, polarized light from the target and the background is detected by photoreceptor cells located at different positions of the eye, each photoreceptor cell is composed of cell pairs with different orientations, and the sensitive directions of polarized light corresponding to different photoreceptor cells are also different. After obtaining the polarization signal, the photoreceptor cell R transmits the information to the intermediate neuron for processing, and the change of its receptor potential is approximately the natural logarithm of the receptor activity. The activity values ​​of the intermediate neurons of the target and the background are calculated respectively. The three-channel model requires two intermediate neurons for the target and the background respectively, and their activity values ​​can be expressed as:

[0042]

[0043]

[0044] in, is the activity value of the intermediate neuron of the target polarization signal of 60° and 120°, The target polarization signal interneuron activity value is 0° and 120°, is the activity value of the intermediate neuron of the 60° and 120° background polarization signals, are the neuron activity values ​​for the 0° and 120° background polarization signals.

[0045] Preferably, in S9, the simultaneously acquired target polarization signal intermediate neuron activity value and background polarization signal interneuron activity value Convert to mathematical space and map to coordinate values, that is, convert the target and background images into a coordinate matrix in which each pixel position is a coordinate value, and calculate the distance between the corresponding coordinates, which is the two-dimensional polarization distance between the target and the background:

[0046]

[0047] According to the public, the distance between the coordinates can be calculated as the two-dimensional polarization distance between the target and the background.

[0048] Compared with the prior art, the beneficial effects of the present invention are: the two-dimensional polarization distance underwater imaging method inspired by bionic vision can effectively suppress background interference, capture tiny polarization differences between the target and the background, and thus clearly extract target information in complex water environments. In addition, the present invention performs excellently in processing multi-target scenes with different materials and polarization characteristics, and can achieve good recognition of both simple-shaped and complex-detailed targets, thereby enhancing the amount of information in the image and maintaining high-quality imaging effects. Compared with traditional polarization imaging methods, the present invention has obvious advantages in image details, clarity and stability, and exhibits stronger adaptability and reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a technical flow chart of the present invention;

[0050] Figure 2 A schematic diagram of the eye structure and polarization anatomical structure of the mantis shrimp of the present invention;

[0051] Figure 3 This is a block diagram of a specific implementation case of the present invention;

[0052] Figure 4 This is the 0° polarization muddy water image;

[0053] Figure 5 This is a 60° polarized muddy water image;

[0054] Figure 6 It is a 120° polarized muddy water image;

[0055] Figure 7 is a two-dimensional polarization distance image. DETAILED DESCRIPTION

[0056] 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.

[0057] See also Figure 1-Figure 7 The present invention provides a technical solution: a two-dimensional polarization distance underwater imaging method inspired by bionic vision, comprising the following steps:

[0058] S1: Acquire 0°, 60°, and 120° polarization images and calculate the Stokes vector, AOP, and DOP;

[0059] S2: Perform frequency domain transformation on the Stokes vector and extract the low-frequency part as the polarization information of the backscattered light;

[0060] S3: Calculate the backscattered light AOP using the background Stocks vector;

[0061] S4: Calculate the maximum value, minimum value and DOP of the backscattered light;

[0062] S5: Calculate the polarization sensitivity of the background;

[0063] S6: Calculate the polarization sensitivity of the target;

[0064] S7: Calculate the background interneuron activity value;

[0065] S8: Calculate the activity value of the target interneuron;

[0066] S9: Calculate the two-dimensional polarization distance to obtain the final image.

[0067] Example

[0068] Taking the crustacean mantis shrimp as an example, its polarization vision is a complex and sophisticated visual mechanism involving the unique structure and function of its compound eyes. Its eye structure and polarization anatomy are as follows Figure 2 As shown in the figure, the mantis shrimp's eyes are divided into three areas: the middle zone, the dorsal hemisphere and the ventral hemisphere. There are three clearly visible pseudo pupils in the eyes. Mantis shrimp has compound eyes composed of hundreds of visual units, each of which is composed of a cornea covering the lens. There are eight photoreceptor cells behind the lens. The microtubule bundles in the photoreceptor cells have an anatomically complex array of polarization receptors, which makes these cells highly sensitive to the polarization state of light. In the entire eyeball, the direction of the VH's horizontal stripes is offset by 45° compared to the direction of the DH's horizontal stripes, resulting in four unique microvilli directions. There are four orientations between these four orientations. The comparison exists at the neural level, which enables the reconstruction of some linearly polarized light. The incident light undergoes different degrees of polarization, and the rhodopsin molecules in the photoreceptor cells interact with the polarization direction of the light, resulting in different optical responses and the generation of neural signals. By integrating signals from multiple ommatidia, mantis shrimps can form comprehensive visual information about the polarization distribution of light. This visual mechanism enables them to effectively forage, navigate and avoid enemies in underwater environments, and may play a role in intraspecific communication. The evolutionary adaptability of polarized vision gives mantis shrimps a strong survival advantage in the changing underwater light environment.

[0069] Understanding the basic anatomy of polarized light vision provides the possibility to establish a model of the sensitivity of animals to polarized light signals. In the simple visual system of mantis shrimp and other organisms, there are two orthogonal polarization channels, which are connected in an opposing manner (similar to the dichromatic vision model) and produce a single horizontal neural activity. In this system, polarization contrast can be measured by comparing the neural activity between two different areas in the field of view (one observing the target and the other observing the background). This measurement is called "polarization distance". Polarization distance is based on a hypothetical model of biological vision and simulates the mechanism of biological processing of polarization information. Polarization distance can be used to quantify and analyze the difference in polarization information between the target and the background.

[0070] In biological visual systems, the processing of color and polarization shows significant similarities. This similarity is not only reflected in their perceptual mechanisms and neural structures, but also in the three-factor stimulation model and information processing strategy they rely on. By comparing these two visual forms, we can deeply understand their commonalities. The following will explore the similarities between color and polarization vision from multiple angles.

[0071] (1) Similarity of neural mechanisms: In the visual systems of many organisms, the neural networks of color vision and polarization vision are similar. For example, in the compound eye structure of mantis shrimp, the DH and VH regions are highly sensitive to changes in polarized light, and the MB region is highly sensitive to changes in color. The neuronal structures for color and polarization information processing are similar, indicating that the two systems rely on similar anatomical and physiological characteristics.

[0072] (2) Three-factor stimulation model: Biological processes for color and polarization are similar, because both polarization vision and color vision have three-factor characteristics. Color vision is involved in three spectral types of receptors. Any light can be composed and described by three linearly independent parameters, namely brightness, hue and saturation. For all known biological photoreceptors, polarized light is actually a three-factor stimulus. The polarization state can also be described by intensity, angle, and polarization degree. Their responses are also linearly independent. Unpolarized light is similar to white light, and linearly polarized light is similar to chromatic aberration. The angle (φ) and polarization degree (d) are similar to hue and saturation, respectively. Bees and ants have been proven to have both three-factor color vision and polarization vision, and they can work independently to achieve a variety of biological functions.

[0073] (3) Information processing strategy: In the comparison of related experiments, it was found that the polarization response of the receptors in the polarization vision system and the spectral response of the receptors in the color vision system have basically the same change trend, that is, the spectral sensitivity as a function of hue and saturation is similar to the polarization sensitivity as a function of polarization angle (φ) and polarization degree (d). Therefore, the two systems are likely to use the same information processing strategy.

[0074] Based on the commonalities in the perception and information processing mechanisms of color and polarization among the above-mentioned organisms, we can refer to the relevant methods of color space to construct an analysis method suitable for polarization vision. For color recognition tasks, color features are often described by constructing a color space, and "color distance" is calculated based on different color features to measure color differences. The core principle is to convert colors into mathematical space and quantify color differences by calculating the spatial distance between color points. Therefore, inspired by color vision and combined with the theoretical basis of biological polarization vision, a two-dimensional polarization distance model is constructed to quantify and highlight the significant differences between polarized targets and their backgrounds.

[0075] In summary, inspired by polarization vision and color vision, combined with polarization parameters, and mapping the polarization signal activity values ​​of the target and the background to the mathematical space, a two-dimensional polarization distance model is proposed to evaluate the polarization difference. Through frequency domain analysis and polarization state reconstruction, the global distribution of the polarization angle (AOP) and degree of polarization (DOP) of the background scattered light is obtained to reduce the estimation error of the background scattered light. At the same time, based on the polarization vision system model of aquatic organisms, the biological perception mechanism is simulated to solve the problem of obtaining and utilizing polarization information underwater. In addition, combined with color vision inspiration and traditional polarization parameters, a two-dimensional polarization distance model is proposed to highlight the polarization information difference between objects and backgrounds in turbid water bodies, and frequency domain analysis and polarization state reconstruction are introduced in the calculation of background polarization states to obtain The global distribution of the polarization angle (AOP) and degree of polarization (DOP) of the background scattered light reduces the estimation error, so that the two-dimensional polarization distance underwater imaging method can have excellent background noise suppression ability when used, so that it can more clearly extract the polarization information of the target in complex water environments, and focus on mining the tiny polarization differences between the target and the background. This makes it possible to better distinguish the characteristics of the target object even in high turbidity waters, and at the same time perform well in processing multi-target scenes with different materials and different polarization characteristics. In the experiment, whether it is a target with a simple shape or a target with complex details, it can better reflect its recognition ability and imaging effect. The content not described in detail in this description belongs to the prior art known to professional and technical personnel in this field.

[0076] Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A two-dimensional polarization distance underwater imaging method inspired by bionic vision, characterized in that: The following steps are involved: S1: Acquire 0°, 60°, and 120° polarization images and calculate the Stokes vector, AOP, and DOP; S2: Perform frequency domain transformation on the Stokes vector and extract the low-frequency part as the polarization information of the backscattered light; S3: Calculate the backscattered light AOP using the background Stocks vector; S4: Calculate the maximum value, minimum value and DOP of the backscattered light; S5: Calculate the polarization sensitivity of the background; S6: Calculate the polarization sensitivity of the target; S7: Calculate the background interneuron activity value; S8: Calculate the activity value of the target interneuron; S9: Calculate the two-dimensional polarization distance to obtain the final image.

2. The method for two-dimensional polarization distance underwater imaging inspired by bionic vision according to claim 1, characterized in that: In S1, the Stokes vector can be used to describe the polarization state of light. A polarization camera is used to obtain polarization images at 0°, 60°, and 120°. The Stokes vector can be calculated as: The degree of polarization (DOP) d and angle of polarization (AOP) a of the image can be expressed as: Thus, polarization images of 0°, 60°, and 120° can be obtained and the Stokes vector, AOP, and DOP can be calculated.

3. The method for two-dimensional polarization distance underwater imaging inspired by bionic vision according to claim 1, characterized in that: In S2, when polarized light illuminates a partially polarized target, the target information is always retained in S1 and S2. When calculating the polarization information of the backscattered light, we transform the Stokes vector into the frequency domain and extract the low-frequency part. This process can be expressed as: Among them, S0, S1, S2 are Stokes vectors calculated from the original polarization image, which contain the target and background polarization information, FT and IFT are Fourier transform and inverse Fourier transform. represents an image containing only the polarization information of the backscattered light, H(u,v) is a low-pass filter, and (u,v) represents the coordinates in the frequency domain.

4. The method for two-dimensional polarization distance underwater imaging inspired by bionic vision according to claim 1, characterized in that: In S3, after obtaining the Stokes vector of the background, the AOP of the backscattered light can be calculated: The AOP of backscattered light can be calculated according to the public information.

5. The method for two-dimensional polarization distance underwater imaging inspired by bionic vision according to claim 1, characterized in that: In S4, using a B Calculate the maximum and minimum backscattered light and calculate the DOP of the backscattered light: In the calculation of two-dimensional polarization distance, after obtaining the background polarization state parameters, the parameter calculation of the target part can be expressed by using the DOP and AOP calculated by the Stocks vector of the entire image, and the relevant parameters of the target and the background are used as the photoreceptor cell R θ (a, d) Inputs used to calculate the activity value of the interneuron.

6. The method for two-dimensional polarization distance underwater imaging inspired by bionic vision according to claim 1, characterized in that: In S5, when the photoreceptors in the biological retina receive light from the environment and the target, the incident light can be considered to be linearly polarized light. According to the relevant research on biological vision theory and electrophysiology, the polarization sensitivity of the photoreceptor cell R can be expressed by the cosine function of the orientation angle: Where φ is the measured polarization direction relative to the reference direction, i.e., the E vector axis, and φ max is the value of φ that makes R(φ) reach its maximum value, S p is the polarization sensitivity of each photoreceptor. For most high-precision polarization imaging systems, S p =10.

7. The method for two-dimensional polarization distance underwater imaging inspired by bionic vision according to claim 1, characterized in that: In S6, when the incident light is partially polarized light, the proportion of polarized light in the incident light must be considered, d is the polarization degree of the incident light, and R(φ,d) is expressed as: Used to get the proportion of incident light.

8. The method for two-dimensional polarization distance underwater imaging inspired by bionic vision according to claim 1, characterized in that: In the S7, in the organism, φ max The physical meaning of is the orientation of the microvilli of the photoreceptor cells, which can be further represented by the polarization angle a in the model. Therefore, the polarization sensitivity of polarized light at different angles can be expressed as: Where, k indicates that the input polarization information is target light (O) or background light (B), a O ,d O AOP and DOP calculated for the entire image, a B ,d B AOP and DOP calculated for the background.

9. The method for two-dimensional polarization distance underwater imaging inspired by bionic vision according to claim 1, characterized in that: In S8, when performing a target-based visual task, polarized light from the target and background is detected by photoreceptor cells located at different positions of the eye. Each photoreceptor cell is composed of cell pairs with different orientations. Different photoreceptor cells are sensitive to different polarized light. After obtaining the polarization signal, the photoreceptor cell R transmits the information to the intermediate neuron for processing. The change in its receptor potential is approximately the natural logarithm of the receptor activity. The activity values ​​of the intermediate neurons of the target and background are calculated respectively. The three-channel model requires two intermediate neurons for the target and background respectively, and their activity values ​​can be expressed as: in, The activity values ​​of the intermediate neurons of the target polarization signals of 60° and 120° are: The target polarization signal interneuron activity value is 0° and 120°, is the activity value of the neuron between the background polarization signals of 60° and 120°. The interneuron activity values ​​are 0° and 120° background polarization signals.

10. The method for two-dimensional polarization distance underwater imaging inspired by bionic vision according to claim 1, characterized in that: In S9, the target polarization signal intermediate neuron activity value obtained simultaneously and background polarization signal interneuron activity value Convert to mathematical space and map to coordinate values, that is, convert the target and background images into a coordinate matrix in which each pixel position is a coordinate value, and calculate the distance between the corresponding coordinates, which is the two-dimensional polarization distance between the target and the background: According to the public, the distance between the coordinates can be calculated as the two-dimensional polarization distance between the target and the background.

Images