Polarization Imaging Method for Underwater Low-Visibility Metal Target Using Focus Dividing Plane

Featured Application

Based on the principle of polarization imaging, a split-focus plane polarization imaging system is developed, and an image clarity method is proposed. Compared with traditional light intensity imaging detection methods, it is easier to detect metal targets from sediment backgrounds in underwater low-visibility scenes.

Abstract

Aiming at the problems of brightness attenuation and contrast reduction in the target image caused by underwater low-visibility environments, a metal target detection method based on split-focus plane polarization imaging is proposed. Firstly, a hybrid enhancement method is proposed to clarify the degraded polarization image. In this study, the GrayWorld method is improved to compensate the attenuation difference of the total light intensity of the polarization image. Variational contrast and saturation enhancement algorithms are used to reduce the underwater scattering effect; secondly, a split-focus plane polarization imaging system is built to complete the control of camera parameters, polarization image acquisition and information processing. Under different underwater low-visibility conditions, polarization imaging of targets with different materials can be realized; finally, an image quality evaluation system is constructed to compare the light intensity and degree of polarization images that are collected by the focal plane polarization imaging system. The polarization characteristics of metal and nonmetal target plates are analyzed. The results show that under the condition of low visibility, the obtained polarization image contrast of the metal target is relatively high, and its EME, information entropy and average gradient are increased by 183.82%, 53.46% and 586.22% on average relative to the image of light intensity. In an underwater low-visibility scene, the method of focal plane polarization imaging proposed in this paper can reduce the difficulty of metal target detection.

1. Introduction

During the imaging of targets in underwater low-visibility environments, the images are affected by the absorption and scattering of light by water and lighting equipment. There are problems such as image brightness attenuation and contrast reduction. The problems in these images will bring some difficulties to underwater environment monitoring, military reconnaissance and other target detection activities [1,2]. Polarization imaging is a new optical imaging technology. The polarization characteristic distribution can be obtained based on the target light intensity information, including the degree of polarization, the polarization angle and other information [3]. According to the polarization information difference of objects with different shapes, materials and surface roughness, it can enhance the edge contour, texture details and other information of underwater targets [4]. It can effectively detect targets in complex underwater environments [5].

Besker used the method of rotating the polarizer. He conducted polarization imaging on targets with six different materials. It is proved that the polarization image contains enough brightness information and rich detail information that is easy for human eyes to recognize [6]. Bao Fucheng studied the influence of different materials on polarization imaging through an underwater polarization experiment. The conclusion was that the degree of polarization of the target with a smooth surface is higher than that of the target with a rough surface [7]. Han Jiefei designed a polarization imaging experimental system based on LED auxiliary lighting. For targets with different polarization characteristics, different processing methods are used to obtain high-resolution images [8]. Liu Zhe proposed a material classification method based on polarized light imaging. In the study, it was observed that there are different reflectivity ratios between metals and nonmetals within a certain observation angle range, and a Fresnel reflectivity ratio method was proposed to distinguish metal and nonmetal targets [9]. Xiong Zhihang studied the polarization degree information of metal aluminum plates, cloth, rubber, ceramics, sediment and paper through a linear polarization imaging experiment. It was concluded that the polarization degree of a metal target is obviously higher than that of other objects. It can be used to quickly identify metal chips [10]. In a haze environment, Xu Xian conducted polarization imaging on plastic objects and metal objects under low background illumination. It was concluded that polarization imaging can be used as a complementary approach to obtain target details [11].

In the analysis of underwater target imaging technology, underwater polarization imaging technology can be studied by detectors with polarizers [12,13]. Tominaga studied the spatial distribution of the degree of polarization of the reflected light on the target surface, and obtained that the metal and nonmetal material targets can be classified by using the degree of the polarization image [14]. Qiang Fu used multibands as factors; the benefit of polarization imaging in a sea fog environment was assessed objectively using contrast, information entropy, degree of polarization, and other evaluation indices [15]. Through theoretical analysis, Ren conducted polarization imaging experiments on water bodies with different turbidity and different materials. The mechanism of active linear polarization imaging was demonstrated [16]. Wang proposed an underwater polarization imaging method based on a Mueller imager. In this method, optical comodulation of the polarization state of the light source and the direction of the polarizer in front of the camera is established. It can not only filter the underwater scattered light, but also improve the image quality [17].

To improve the detection capability of underwater metal targets, in this paper, an underwater polarization imaging experimental platform is built, and a method of underwater low-visibility metal target polarization imaging with a split-focus plane is proposed. Polarization imaging experiments were carried out on targets made of different materials and metals in a low-visibility environment. The image quality evaluation system was constructed, and the experimental results were analyzed quantitatively. Through polarization imaging technology, it can effectively detect metal targets in low-visibility underwater environments. It has certain reference and application values for underwater target detection.

2. Hybrid Enhanced Polarization Image Clarity

2.1. Principle of Polarization Imaging

The Stokes vector method is a commonly used method to describe polarized light. The Stokes vector has four components $S={(I,Q,U,V)}^{\mathrm{T}}$. It contains full polarization information of light. Among them, $I$ is light intensity, $Q$ and $U$ are linear polarization information and $V$ is circular polarization information. Aiming at the problem of weak and difficult measurements of circular polarization information in underwater measurement environments, In this paper, polarization imaging detection is based on linear polarization information. The linear polarization component of the Stokes vector can be expressed as:

$$S=\left[\begin{array}{l}I\\ Q\\ U\end{array}\right]=\left[\begin{array}{c}1/2\left(I_0+I_{45}+I_{90}+I_{135}\right)\\ I_{45}-I_{135}\\ I_0-I_{90}\end{array}\right]$$

(1)

In Equation (1), $I_0$, $I_{45}$, $I_{90}$ and $I_{135}$ are polarization images with polarization directions of 0°, 45°, 90° and 135°, respectively. According to the definition, DOP represents the ratio of linearly polarized light intensity to total light intensity, which can be calculated by Equation (2).

$$DOLP=\frac{\sqrt{Q^2+U^2}}{I}$$

(2)

The classical water degradation image-sharpening method has the following problems: the prior assumption of the underwater imaging model is the premise of image sharpening; image clarity is highly dependent on the accuracy of prior conditions. Therefore, the classical method is difficult to adapt to complex underwater scenes and its low robustness. In this paper, a hybrid enhanced water-descent polarization image-sharpening method is proposed. Figure 1 shows the algorithm flow. The pixel value is adjusted by analyzing the feature information of underwater image. Traditional avoidance methods are limited by complex prior conditions and cumbersome model parameter estimation. The algorithm implementation process is as follows:

Firstly, according to the Stokes vector method, the total intensity image I of 0°, 45°, 90° and 135° polarization images is obtained using Equation (1).

Secondly, an improved GrayWorld method combined with color compensation is established. The image attenuation difference caused by underwater propagation is compensated. Finally, the white balance image W is used as the guide image. By constructing a variational contrast and saturation enhancement algorithm, the image contrast and saturation are improved. The turbid appearance caused by underwater scattering is eliminated. After iterative calculation, a clear image D is finally obtained.

2.2. Improved Grayworld Method for Color Compensation

By observing a large number of underwater scene images, the optical properties of water are analyzed. It can be concluded that the red channel of underwater image has the most serious attenuation, while the retention of the green channel is relatively good. The blue channel’s information is not affected. The red channel and green channel of the underwater image have opposite color information. Therefore, this paper compensates the green channel to the red channel to better restore the overall color of the image. In order to avoid the subsequent use of the GrayWorld method to produce red channel supersaturation, only the highly attenuated areas in the image are compensated.

After the above analysis, the compensated red channel image ($I_{RC}\left(x\right)$) can be expressed as:

$$I_{RC}\left(x\right)=I_R\left(x\right)+\beta \left(\overline{I_G}-\overline{I_R}\right)\left(1-I_R\left(x\right)\right)I_G\left(x\right)$$

(3)

In Equation (3), $I_R\left(x\right)$ and $I_G\left(x\right)$ are the red and green channels of the image, while $\overline{I_R}$ and $\overline{I_G}$ represent the average values of $I_R\left(x\right)$ and $I_G\left(x\right)$. β is the adjustment parameter. Experiments show that the effect is better when β is 1.

After compensating the attenuation of the red channel of the image, the GrayWorld method is used to restore the color of the underwater image, as shown in Equation (4).

$$\overline{Gray}=\frac{\overline{I_{RC}}+\overline{I_G}+\overline{I_B}}{3}$$

(4)

In Equation (4), $\overline{Gray}$ is the uniform gray value defined by the GrayWorld method. $\overline{I_{RC}}$, $\overline{I_G}$ and $\overline{I_B}$ are the average values of $I_{RC}$, $I_G$ and $I_B$. $I_B$ is the blue channel of the image.

The gain coefficients of the three color channels can be obtained by Equation (5).

$$\{\begin{array}{c}gai{n}_R=\frac{\overline{Gray}}{\overline{I_{RC}}}\\ gai{n}_G=\frac{\overline{Gray}}{\overline{I_G}}\\ gai{n}_B=\frac{\overline{Gray}}{\overline{I_B}}\end{array}$$

(5)

According to the von Kries diagonal model, the pixel values $W_R$, $W_G$, $W_B$ of the RGB three channels color of image W processed by GrayWorld method are obtained, as shown in Equation (6).

$$\{\begin{array}{c}{W}_R=gai{n}_R\times I_{RC}\\ W_G=gai{n}_G\times I_G\\ W_B=gai{n}_B\times I_B\end{array}$$

(6)

In the white balance processing of the water degradation image, compared with the traditional method, the improved GrayWorld method proposed in this paper greatly reduces the red artifacts caused by overcorrection of the areas with serious underwater color deviation. At the same time, this method can still effectively restore the original color of the target and improve the contrast in the harsh underwater environment.

2.3. Variational Contrast and Saturation Enhancement Algorithm

The improved GrayWorld method can play a corresponding role in the process of image color restoration, aiming at the problems of image turbidity and loss of edge details caused by underwater strong scattering environment. To better restore the visibility of water degradation images, a variational model algorithm with single data item and double regular item is proposed. It is used to enhance the contrast and saturation of water degradation image.

First, by comparing the difference between the clarified image D and the improved GrayWorld method image W, the color distortion of the output result image is prevented. Secondly, the contrast and saturation parameters in the image are enhanced by using the biregular term. The method of image contrast enhancement is as follows: by calculating the weighted difference between all pixels in the image and the whole image, the method of image saturation enhancement is to calculate the difference between the intensity of a single channel and other channels in the R, G, B channels of the image, respectively.

$$\begin{array}{ll}R\left(D_m\right)=& \frac{1}{2}{\displaystyle \sum _x{\left(D_m\left(x\right)-W_m\left(x\right)\right)}^2}\\ & -\frac{\mu }{2}{\displaystyle \sum _{x,y}u\left(x,y\right)S\left(D_m\left(x\right)-D_m\left(y\right)\right)}\\ & -\frac{\eta }{2}{\displaystyle \sum _{x,y}\left({\left(D_m\left(x\right)-D_{m+1}\left(x\right)\right)}^2+{\left(D_m\left(x\right)-D_{m+2}\left(x\right)\right)}^2\right)}\end{array}$$

(7)

In Equation (7), $m\in R_3$, represents an integer field with 3 as a cycle, which corresponds to the RGB three channels color of the image ($D_1=D_R$, $D_2=D_G$, $D_3=D_B$). D is the clear output image. W is the guide image obtained by improving the GrayWorld method. μ and η are regularization parameters. $u\left(x,y\right)$ is the Euclidean distance weight, which weights the image’s local-to-global clarity. The filtering result of the image edge pixel can be improved by introducing weight function $u\left(x,y\right)$. The halo around the image boundary caused by underwater scattering is avoided. It can be expressed by Equations (8) and (9):

$$u\left(x,y\right)=\frac{E\left(x\right)}{\Vert x-y\Vert },x\ne y$$

(8)

$$E\left(x\right)={\left({\displaystyle \sum _{y\in \mathsf{\Omega }}\frac{1}{\Vert x-y\Vert }}\right)}^{-1}$$

(9)

3. Experimental Platform for Underwater Polarization Imaging

3.1. Polarization Imaging of Split Focus Plane

The subfocal plane imaging system developed in this paper consists of two subsystems. They are the polarization detection subsystem and the image acquisition and processing subsystem. Figure 2 shows the schematic diagram of the imaging system.

According to the polarization imaging theory, the Stokes vector can only be obtained through four polarization images. At present, polarization imaging methods can be divided into time division, amplitude division, aperture division and focal plane division. The experimental platform built in this paper uses the color polarization camera TRIO50S-QC, produced by the Canadian LUCID company, and the 75 mm fixed focus industrial lens VM7528MP5, produced by Beijing Zhonglian Kechuang Trade Co., Ltd (Beijing, China). The real object of the split-focus plane polarization camera is shown in Figure 3.

Based on the camera SDK, the host computer software is developed using the QT framework based on C++ language and OpenCV library. The software implementation process is shown in Figure 4. First, the camera connection and parameter setting are completed, and the polarization image of the target is collected; the image stream is turned on to start the image stream transmission engine of the camera. The subsequently acquired polarization images will be sent to the output buffer queue in turn. Then, they are read by the user and sent to the input buffer queue.

The software of the split-focus plane polarization imaging system is shown in Figure 5. It is mainly composed of human–computer interaction module and real-time display module. The human–computer interaction part includes the functions of connecting equipment, image acquisition, camera parameters and image processing; the real-time display module can display 0°, 45°, 90°, 135° polarization images and I, DOP, AOP result images.

3.2. Experimental Platform and Method

The underwater polarization imaging experimental platform consists of five parts: polarization light source, polarization camera, target, experimental water tank and computer. Figure 6 is a schematic diagram of underwater polarization imaging composed of five parts.

In Figure 6, a polarized light source is generated by rotating the polarizer in front of the LED light source. Reflected light is generated when polarized light shines on the surface of underwater target. The reflected light is received by the polarization camera to complete the polarization detection. Then, the polarization image of underwater target can be obtained. The corresponding degree of the polarization image can be calculated using Formulas (1) and (2).

In this paper, three objective image quality evaluation indexes, namely image detail enhancement measure (EME), information entropy H and average gradient $\overline{G}$, are introduced to quantitatively analyze the experimental results of underwater polarization imaging. In general, when the image EME value is larger, the image shows more details; when the information entropy H value is larger, the image contains more information; when the average gradient $\overline{G}$ is larger, the image contains richer texture details and higher definition.

4. Experimental Results and Analysis

4.1. Pixel Response Test of System Polarization

The polarization pixel performance of the split-focus plane polarization imaging system is related to three physical quantities. For a given polarization pixel, its polarization response parameter is fixed. Therefore, the direction of polarized light can be changed by rotating the polarizer. The polarization image sensor can obtain different polarization pixel digital signals. This section verifies the effectiveness of the above response model through relevant experiments.

When the linearly polarized light $I_{in}$ obtained after polarizing the polarizer passes through the polarizer, its outgoing light $I_{out}$ meets Marius’ law, as shown in Equation (10).

$$I_{out}=I_{in}{\cos }^2\chi$$

(10)

In Equation (10), $\chi$ is the angle between the polarization direction of the incident polarized light $I_{in}$ and the optical axis of the polarizer. According to Equation (11), when the intensities of $I_{out}$ and $I_{in}$ are equal, it means that the polarization direction of the incident polarized light is consistent with that of the polarizer, $\chi =0°$; when the intensity of $I_{out}$ is zero, it indicates that the polarization directions of the incident polarized light and the polarizer are orthogonal to each other, $\chi =90°$.

Based on the above theoretical analysis, a polarization pixel response test platform is built, as shown in Figure 7.

As shown in Figure 7, the devices numbered 1–5 represent the Ethernet switch, polarization camera, linear polarizer, LED light source and upper computer, respectively. An LED uniform light is used as the input light source of the test platform. A calibrated linear polarizer is placed between the light source and the focal plane polarizing camera, which can be used as a polarizer to generate linearly polarized light. The micropolarizer array in the polarization sensor can be regarded as a polarizer. During the test, the polarization angle increment is set to 10° and the polarizer is rotated. Under the same exposure time and gain conditions, 19 groups of 0°, 45°, 90° and 135° polarization grayscale images are collected by the split-focus plane polarization camera at 0°, 10°, 20°. The least-squares method is used to fit the gray value of the pixel response of the polarization image. The relationship between the gray value and the rotation angle is shown in Figure 8.

According to the polarization pixel response fitting curve shown in Figure 8, it can be concluded that the gray value distribution of the theoretical analysis results is consistent with the actual measured values; the corresponding relationship between the response of polarization pixels and the polarization direction is obtained. The validity of the model can be further verified according to the results.

4.2. Polarization Imaging Experiment of Underwater Targets with Different Materials

The polarization characteristics of targets made of different materials are quite different. The degree of polarization image contains a lot of information that is difficult to detect in the light intensity image. In the underwater target imaging experiment with different materials, in order to explore the polarization characteristics of metal and non-metal targets, four different materials of targets are tested in this paper. The materials of the targets are, respectively, brass, red copper, ceramics and plastics, and the size of the targets is 100 mm × 20 mm × 2 mm. Polarization imaging was performed in clear water with a depth of 30 cm. The light intensity image and degree of polarization image obtained are shown in Figure 9. By observing the degree of polarization image, it can be seen that some colors of red copper and brass metal materials are white, while some colors of ceramic and plastic nonmetallic materials are black. According to the relationship between the gray value of the digital image and color, it can be concluded that the degree of polarization of metal materials is higher, while that of the nonmetallic materials is lower. It is proved that polarization imaging technology can be used to effectively distinguish metal and nonmetal targets.

In Figure 9, the targets made of ceramics, red copper, brass and plastics are arranged from top to bottom. A single target is extracted by clipping the polarization degree images of different materials. The quality of polarization degree image is evaluated, respectively. The data distribution ladder diagram of evaluation results is shown in Figure 10.

According to the results shown in Figure 10, it can be concluded that the information entropy values of all polarization degree images are high. The values of EME and average gradient of ceramic and plastic materials are far lower than those of red copper and brass. It is further verified that the polarization characteristics of metal and nonmetal materials are quite different.

4.3. Polarization Imaging Experiment of Metal Target in Low-Visibility Underwater Scene

The metal target detection experiment of the underwater low-visibility scene consists of the following parts. First, the bottom of the experimental water tank is covered with irregular sediment to simulate the real underwater background. The metal target is placed on the upper part of the sediment irregularly. Then, clean water is injected into the experimental water tank, and the water injection depth is about 30 cm. Above the water surface, a tripod is used to fix the polarization camera and the light source. Finally, by adjusting the focal length of the camera, the brightness of the light source and the rotation angle of the polarizer, the original polarization image of the metal target is collected by the upper computer software. The polarization information and image quality evaluation data are calculated by the processing module. This experiment uses different metal targets of silver, gold and black, as shown in Figure 11.

The experimental results are shown in Figure 12, including the light intensity I and DOP images of three metal targets. It can be seen that the targets in the background can be easily recognized without image enhancement processing. Compared with the light intensity image, the contrast enhancement of the polarization image is significant, the texture details are rich and the contour features are significant. This is due to the polarization characteristics of the reflected light on the material surface. Sediment and other objects are depolarized due to multiple scattering due to surface roughness, while metal materials have good polarization retention and a high degree of polarization. Therefore, the degree of polarization image is suitable for the detection of metal targets in the underwater complex sediment background.

The experimental image quality evaluation data are shown in

Table 1. Quality evaluation of experimental images in low-visibility underwater scenes. Information entropy average gradient.

	Metal 1			Metal 2			Metal 3
	EME	H	$\overline{\mathit{G}}$	EME	H	$\overline{\mathit{G}}$	EME	H	$\overline{\mathit{G}}$
Images I	4.3865	5.3077	1.0372	4.8567	5.1512	0.9687	8.4450	3.9828	1.4285
Images DOP	12.2876	7.1146	6.5648	15.8649	7.1613	7.3524	20.6643	7.4604	9.5242

. Among them, the values of EME, information entropy and average gradient of DOP images of three kinds of metal targets are significantly higher than those of light intensity image I, further verify the effectiveness of using polarization imaging to detect metal targets in underwater low-visibility scenes.

4.4. Underwater Imaging Experiment with Different Turbidity

In the imaging experiment under different water turbidities, metal badges are placed as the targets. By evenly diluting a certain amount of milk into clear water, the simulation of water with different turbidity is realized. In the experiment, six kinds of turbid solutions with milk concentrations of 0.3 mL/L, 0.6 mL/L, 0.9 mL/L, 1.2 mL/L, 1.5 mL/L and 1.8 mL/L are prepared. The light intensity image and the degree of polarization image are collected for the target at the depth of 30 cm. The results are shown in Figure 13. Through the subjective evaluation method, it can be concluded that with the increase in the concentration of milk in the turbid solution, the clarity of the light intensity image and the degree of polarization image gradually decrease, and the saturation of the light intensity image is affected by the concentration of the milk in the solution. With the enhancement of the water scattering, the saturation decreases seriously, but the brightness will gradually brighten as the scattering increases. However, the correlation between brightness and the scattering of the polarization image is small.

The change line of image quality evaluation data in different environments is shown in Figure 14. Obviously, with the increase in turbidity in the experimental solution, the EME, information entropy and average gradient in the light intensity image and the polarization image have decreased. It is verified that the scattering effect of medium particles is intensified with the increase in water turbidity. The increase in image noise leads to a reduction in target information and the degradation of image quality. In addition, the quality of the polarization image is generally higher than that of the light intensity image, and it can still maintain good detection ability in low-turbidity water.

5. Discussion

Based on the split-focus plane imaging system, underwater polarization imaging experiments are carried out for metal targets of different materials and low-visibility scenes. The results show that the degree of polarization of metal objects is higher than that of nonmetals. Polarization imaging can effectively distinguish underwater metal and nonmetal targets. Under the condition of low underwater visibility, the brightness and contrast of the light intensity image drop sharply, but the contrast of the polarization image is high due to the great difference in the polarization characteristics of different objects. Therefore, polarization imaging can effectively detect metal targets in harsh underwater environments. It has certain application prospects in underwater environment monitoring, military reconnaissance and other fields.