Infrared polarization imaging polarization channel selection method and system

Abstract

The invention relates to an infrared polarization imaging polarization channel selection method and system, and the method comprises the following steps: collecting a multi-channel infrared polarization image set of a target scene through a polarization camera, and carrying out the noise reduction preprocessing of the multi-channel infrared polarization image set; dividing the multi-channel infrared polarization image set subjected to noise reduction preprocessing into image sets with different channel numbers; stokes vector images of the image sets with different channel numbers are solved, and polarization characteristic parameter images under different channel numbers are obtained according to the stokes vector images; and taking the polarization characteristic parameter image with the maximum channel number as a reference image, and evaluating the similarity between other channel number images and the reference image by adopting an image similarity index to obtain an optimal polarization channel meeting the imaging requirement of the target scene. Compared with the prior art, effective guidance is provided for a channel number selection strategy in the time-sharing polarization imaging system, the resource utilization efficiency of polarization imaging is improved, and the detection precision of power inspection is enhanced.

Description

Technical Field

[0001] The present invention relates to the field of polarization imaging technology, and in particular to a polarization channel selection method and system for infrared polarization imaging. Background Art

[0002] In the field of power equipment defect detection, infrared thermal detection technology can be used to monitor the heat distribution of power equipment such as transformers, high-voltage transmission lines, switchgear, bus ducts, generators, etc., so as to detect potential faults or problems early and prevent possible power outages. However, the existing infrared thermal imaging detection technology is easily interfered by external radiation in complex light source environments, making it difficult to achieve the expected diagnostic effect. Infrared thermal imaging detection of power equipment has strict requirements on the ambient light source conditions, which seriously hinders the detection efficiency.

[0003] Polarization imaging technology is a new type of photoelectric detection technology. On the basis of traditional imaging, it adds polarization feature information determined by target morphology, surface roughness, and material characteristics. The ability to retain these polarization information is much greater than that of intensity information, and it has higher material and contour recognition capabilities than other imaging methods in specific application scenarios. Since there are large differences in the polarization-related surface features and inherent properties of artificial objects and natural objects, image segmentation of artificial targets and natural backgrounds can be achieved based on the image statistics and shape characteristics of the collected polarization images. Therefore, polarization imaging technology is expected to make infrared thermal imaging detection technology adapt to scenarios with radiation interference in complex light source environments, greatly improve the operation and maintenance efficiency of transmission lines and power equipment, and provide technical support for the efficient, digital and intelligent transformation and upgrading of defect detection in related industries.

[0004] Polarization imaging systems are mainly divided into four categories: time-sharing type, aperture-sharing type, amplitude-sharing type, and focal plane-sharing type. Among them, the time-sharing polarization imaging system can achieve the acquisition of polarization images at any angle by rotating the tuning optical element. The Stokes vector is a mathematical tool used to describe the polarization state of light, which usually contains four components S0, S1, S2, and S3, each of which reflects the intensity distribution of light in different polarization directions. The solution of polarization images is usually based on these four components to construct polarization images. Among them, S3 is usually used to describe circular polarization information and can usually be ignored in power inspections. Therefore, polarization imaging image processing only requires three-channel or four-channel data for Stokes vector solution. However, existing polarization image processing often blindly adopts three-channel and four-channel solution methods, which leads to the loss of a large amount of polarization information, which easily causes the recognition accuracy of target features to decrease, and greatly wastes the advantages of the time-sharing polarization imaging system in acquiring images at any polarization angle, resulting in a waste of resources. Moreover, in the existing three-channel and four-channel polarization image solving process, fixed channels of 0°, 60°, 120° and 0°, 45°, 90°, 135° are often used, which are not combined with the actual scene, resulting in poor imaging results. Summary of the invention

[0005] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and to provide a polarization channel selection method and system for infrared polarization imaging, to provide effective guidance for the channel number selection strategy in the time-sharing polarization imaging system, to improve the resource utilization efficiency of polarization imaging, and to enhance the detection accuracy of power inspection.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] A method for selecting a polarization channel for infrared polarization imaging comprises the following steps:

[0008] Using a polarization camera to collect a multi-channel infrared polarization image set of a target scene, and performing noise reduction preprocessing on the multi-channel infrared polarization image set;

[0009] Dividing the multi-channel infrared polarization image set that has undergone noise reduction preprocessing into image sets with different numbers of channels;

[0010] Solving the Stokes vector image of the image set with different channel numbers, and obtaining polarization characteristic parameter images under different channel numbers according to the Stokes vector image;

[0011] The polarization characteristic parameter image with the largest number of channels is used as a reference image, and an image similarity index is used to evaluate the similarity between images with other channel numbers and the reference image to obtain the best polarization channel that meets the imaging requirements of the target scene.

[0012] Furthermore, the configuration of the polarization camera includes various forms including integrating a multi-channel polarization filter chip at the camera end and configuring a rotatable polarizer or a polarization beam splitter prism in the optical path.

[0013] Furthermore, the noise reduction preprocessing step includes:

[0014] Based on the relationship between noise variance and image gradient covariance matrix, low-rank sub-blocks in the image are selected iteratively, and the noise variance is estimated using the maximum likelihood method.

[0015] According to the noise variance, the BM3D algorithm is used to denoise the multi-channel polarization image set.

[0016] Furthermore, the process of solving the Stokes vector image includes:

[0017] An overdetermined set of equations is constructed according to the relationship between the intensity value and the polarization angle at the same pixel point of the infrared polarization image; the overdetermined set of equations is solved by a pseudo-inverse method to obtain an optimal solution of the Stokes vector, and the optimal solution of the Stokes vector is used as a Stokes vector image.

[0018] Furthermore, the Stokes vector image includes an S0 image, an S1 image and an S2 image.

[0019] Furthermore, the relationship between the intensity value and the polarization angle of the polarization image at the same pixel point is:

[0020]

[0021] Where (u, v) is the pixel coordinate, is the intensity value of the polarization image at the pixel point (u, v), θ j is the polarization angle of the polarization image at the pixel point (u, v), S0(u, v), S1(u, v) and S2(u, v) are the Stokes vectors;

[0022] The overdetermined system of equations is:

[0023]

[0024] In the formula, is the intensity value of the polarization image of channel n at the pixel point (u, v), θ n It is the polarization angle of the polarization image of channel n at the pixel point (u,v).

[0025] Furthermore, the polarization characteristic parameter image includes a polarization degree image and a polarization angle image.

[0026] Furthermore, the polarization degree image is:

[0027]

[0028] Where ρ(u,v) is the polarization image at the pixel point (u,v), S0(u,v), S1(u,v) and S2(u,v) are Stokes vectors;

[0029] The polarization angle image is:

[0030]

[0031] In the formula, is the polarization angle image at the pixel point (u, v).

[0032] Furthermore, the image similarity index includes multiple ones of image mean square error, peak signal-to-noise ratio and structural similarity index.

[0033] According to another aspect of the present invention, there is provided an infrared polarization imaging polarization channel selection system, comprising:

[0034] A polarization image acquisition module uses a polarization camera to acquire a multi-channel infrared polarization image set of a target scene, and performs noise reduction preprocessing on the multi-channel infrared polarization image set;

[0035] An image set decomposition module, used for dividing the multi-channel infrared polarization image set that has undergone noise reduction preprocessing into image sets with different numbers of channels;

[0036] A polarization characteristic parameter image acquisition module is used to solve the Stokes vector image of the image set with different channel numbers, and obtain polarization characteristic parameter images under different channel numbers according to the Stokes vector image;

[0037] The polarization channel selection module uses the polarization characteristic parameter image with the largest number of channels as a reference image, and uses an image similarity index to evaluate the similarity between images with other channel numbers and the reference image to obtain the best polarization channel that meets the imaging requirements of the target scene.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] 1. The present invention collects multi-channel infrared polarization images of the target scene, uses the intensity and polarization angle information of the image to construct an overdetermined set of equations that satisfies the Stokes formula, and obtains the optimal solution of the Stokes vector as the Stokes vector image, thereby obtaining the polarization characteristic parameter image of the target scene. Finally, the quality of the polarization characteristic parameter images under different channel numbers is evaluated, and the optimal polarization channel that meets the imaging requirements of the target scene is found, providing effective guidance for the channel number selection strategy in the time-sharing polarization imaging system.

[0040] 2. The present invention uses a variety of image similarity indicators including image mean square error, peak signal-to-noise ratio and structural similarity index to evaluate the quality of polarization feature parameter images under different channel numbers, find the optimal polarization channel that meets the imaging requirements of the target scene, improve the recognition accuracy of target features, improve the resource utilization efficiency of polarization imaging, and enhance the detection accuracy of power inspection. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 A schematic diagram of a flow chart of a polarization channel selection method for infrared polarization imaging proposed by the present invention;

[0042] Figure 2 The polarization images of overhead transmission lines with different numbers of polarization channels, where (2a) is the polarization image of 3 channels, (2b) is the polarization image of 4 channels, (2c) is the polarization image of 6 channels, (2d) is the polarization image of 9 channels, (2e) is the polarization image of 12 channels, and (2f) is the polarization image of 18 channels;

[0043] Figure 3 These are the reference images of polarization characteristic parameters of overhead transmission lines under 36 channels, where (3a) is the S0 image, (3b) is the S1 image, (3c) is the S2 image, and (3d) is the degree of polarization image. DETAILED DESCRIPTION

[0044] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0045] Abbreviations involved:

[0046] Mean Squared Error, MSE

[0047] Peak Signal to Noise Ratio: Peak Signal to Noise Ratio, PSNR

[0048] Structural Similarity: Structural Similarity, SSIM

[0049] Example 1

[0050] This embodiment uses overhead transmission lines as infrared polarization detection objects, and the imaging detection module used is composed of a long-wave infrared detection core and a polarizer. The core resolution is 640×512, the pixel size is 12μm, and the spectral response range is 8-14μm; the polarizer uses the WP25M-IRC silicon-based metal wire grid polarizer developed by Thorlabs in the United States; the rotation scanning drive module in front of the lens can be controlled to accurately control the rotation angle of the polarizer and stabilize the brake, and images with any polarization angle can be collected.

[0051] This embodiment provides a polarization channel selection method for infrared polarization imaging, such as Figure 1 As shown, the following steps are included:

[0052] S1. Use a polarization camera to collect a multi-channel infrared polarization image set of a target scene, and perform noise reduction preprocessing on the multi-channel infrared polarization image set.

[0053] A time-sharing polarization imaging camera was used to collect an infrared polarization image of the target scene at an interval of 5°, and a total of 36 images were collected at 0°, 5°, 10°, ..., 175°.

[0054] Based on the relationship between noise variance and image gradient covariance matrix, the 36-channel polarization image set is iteratively selected with low-rank sub-blocks in the image, and the noise variance is estimated using the maximum likelihood method. According to the noise variance, the 36-channel polarization image set is denoised using the BM3D algorithm.

[0055] S2. Divide the multi-channel infrared polarization image set that has been pre-processed with noise reduction into image sets with different numbers of channels.

[0056] According to the denoised image set obtained in step S1, it is divided into 3 channels (0°, 60°, 120°), 4 channels (0°, 45°, 90°, 135°), 6 channels (0°, 30°, 60°, 90°, 120°), 9 channels (0°, 20°, 40°, 60°, 80°, 100°, 120°, 140°, 160°), 12 channels (0°, 15 The image sets are 3D images with 18 channels (0°, 10°, 20°, 30°, 40°, 50°, 60°, 70°, 80°, 90°, 100°, 110°, 120°, 130°, 140°, 150°, 160°, 170°).

[0057] S3. Solve the Stokes vector image of the image set with different channel numbers, and obtain the polarization characteristic parameter image under different channel numbers according to the Stokes vector image.

[0058] For the image sets of 3 channels, 4 channels, 6 channels, 9 channels, 12 channels and 18 channels in step S2, their Stokes parameter images are calculated respectively, ignoring the circular polarization component, and the intensity value I and polarization angle θ of the collected polarization image at the pixel point (u, v) are calculated. j The relationship is:

[0059]

[0060] Where (u, v) is the pixel coordinate, is the intensity value of the polarization image at the pixel point (u, v), θ j is the polarization angle of the polarization image at the pixel point (u, v), S0(u, v), S1(u, v) and S2(u, v) are the Stokes vectors;

[0061] The intensity value I and polarization angle θ of the polarization image at the pixel point (u, v) j The relationship between the overdetermined equations is:

[0062]

[0063] In the formula, is the intensity value of the polarization image of channel n at the pixel point (u, v), θ n It is the polarization angle of the polarization image of channel n at the pixel point (u,v).

[0064] The optimal solution of the Stokes vector is obtained by the pseudo-inverse method:

[0065]

[0066] The optimal solution of the Stokes vector is used as the Stokes vector image, including the S0 image, the S1 image and the S2 image.

[0067] According to the Stokes vector, the polarization characteristic parameter image is obtained, including the polarization degree image and the polarization angle image.

[0068] The polarization degree image is:

[0069]

[0070] Where ρ(u,v) is the polarization image at the pixel point (u,v), S0(u,v), S1(u,v) and S2(u,v) are Stokes vectors;

[0071] The polarization angle image is:

[0072]

[0073] In the formula, is the polarization angle image at the pixel point (u, v).

[0074] Take the polarization degree images with different polarization channel numbers as an example, Figure 2 As shown, Figure 2 (2a) is the polarization image of 3 channels, Figure 2 (2b) is the polarization image of 4 channels. Figure 2 (2c) is the polarization image of 6 channels. Figure 2 (2d) is the polarization image of 9 channels, Figure 2 (2e) is the polarization image of 12 channels, Figure 2 (2f) is an 18-channel polarization image. As the number of polarization channels increases, the contrast between the high-voltage transmission line and the background is significantly improved, and the outline is clearer. Especially when the number of channels is large, such as Figure 2 As shown in (2f), the separation effect between the target and the background is more obvious, the ghosting phenomenon in the sky is significantly weakened, and the background is purer, thereby enhancing the prominence of the subject.

[0075] S4. The polarization characteristic parameter image with the largest number of channels is used as the reference image, and the image similarity index is used to evaluate the similarity between images with other channel numbers and the reference image to obtain the optimal polarization channel that meets the imaging requirements of the target scene.

[0076] The 36-channel polarization characteristic parameter image is used as the reference image, such as Figure 3 As shown, Figure 3 (3a) is the S0 image, Figure 3 (3b) is the S1 image, Figure 3 (3c) is the S2 image, Figure 3 (3d) is the polarization image. Image similarity indexes are used to evaluate the similarity between other 3-channel, 4-channel, 6-channel, 9-channel, 12-channel and 18-channel polarization feature parameter images and the reference image, such as image mean square error MSE, peak signal-to-noise ratio PSNR, structural similarity index SSIM, etc. The better the similarity index evaluation, the closer the two images being evaluated are. The MSE calculation formula is as follows:

[0077]

[0078] In the formula, I(x, y) and I′(x, y) represent the grayscale values ​​of the reference image and the comparison image, respectively, and m and n represent the image length and width. The smaller the MSE value, the more similar the two images are, and the higher the image quality of the channel number.

[0079] The PSNR calculation formula is as follows:

[0080]

[0081] In the formula, max represents the maximum grayscale value of the contrast image pixel. The larger the PSNR value, the smaller the introduced noise error, the less distortion, and the higher the image quality.

[0082] The SSIM calculation formula is as follows:

[0083]

[0084] In the formula, μ x and μ y Represent the pixel mean of the reference image and the comparison image respectively, and Respectively represent the variance of the two images, σ xy represents the covariance of the two images, C1 and C2 are small constants. The closer the SSIM value is to 1, the more similar the reconstructed image is to the reference image, and the higher the image quality. Taking the S0 image as an example, the final evaluation results are shown in Table 1.

[0085] Table 1 Comparison of evaluation index results of S0 images with different channel numbers

[0086]

[0087]

[0088] According to the obtained evaluation index values, the optimal number of channels is selected. As the number of channels increases, the image errors are significantly reduced, the signal-to-noise ratio is improved, and the structural similarity is enhanced. Especially when the number of channels increases from 4 to 9, the image quality is improved most significantly. When the number of channels continues to increase, the image quality still improves, but the improvement gradually decreases. In fact, when the number of channels is low, the SSIM value is extremely close to 1, which means that the image is extremely close to the real image in terms of brightness, contrast and structural information. It is difficult for subjective vision to distinguish the difference, resulting in channel redundancy and reduced work efficiency. It can be seen that in this scenario, selecting 9 channels for image processing can not only make more complete use of polarization image information and ensure image quality, but also reduce information redundancy and avoid unnecessary waste of resources.

[0089] Example 2

[0090] This embodiment provides a polarization channel selection system for infrared polarization imaging, including:

[0091] A polarization image acquisition module uses a polarization camera to acquire a multi-channel infrared polarization image set of a target scene, and performs noise reduction preprocessing on the multi-channel infrared polarization image set;

[0092] An image set decomposition module, used for dividing the multi-channel infrared polarization image set that has undergone noise reduction preprocessing into image sets with different numbers of channels;

[0093] A polarization characteristic parameter image acquisition module is used to solve the Stokes vector image of the image set with different channel numbers, and obtain polarization characteristic parameter images under different channel numbers according to the Stokes vector image;

[0094] The polarization channel selection module uses the polarization characteristic parameter image with the largest number of channels as a reference image, and uses an image similarity index to evaluate the similarity between images with other channel numbers and the reference image to obtain the best polarization channel that meets the imaging requirements of the target scene.

[0095] The rest is the same as in Example 1.

[0096] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A polarization channel selection method for infrared polarization imaging, characterized in that: The following steps are involved: Using a polarization camera to collect a multi-channel infrared polarization image set of a target scene, and performing noise reduction preprocessing on the multi-channel infrared polarization image set; Dividing the multi-channel infrared polarization image set that has undergone noise reduction preprocessing into image sets with different numbers of channels; Solving the Stokes vector image of the image set with different channel numbers, and obtaining polarization characteristic parameter images under different channel numbers according to the Stokes vector image; The polarization characteristic parameter image with the largest number of channels is used as a reference image, and an image similarity index is used to evaluate the similarity between images with other channel numbers and the reference image to obtain the best polarization channel that meets the imaging requirements of the target scene.

2. The method for selecting polarization channels for infrared polarization imaging according to claim 1, characterized in that: The polarization camera is configured in various forms including integrating a multi-channel polarization filter chip at the camera end and configuring a rotatable polarizer or a polarization beam splitter prism in the optical path.

3. The method for selecting polarization channels for infrared polarization imaging according to claim 1, characterized in that: The steps of the noise reduction preprocessing include: Based on the relationship between noise variance and image gradient covariance matrix, low-rank sub-blocks in the image are selected iteratively, and the noise variance is estimated using the maximum likelihood method. According to the noise variance, the BM3D algorithm is used to denoise the multi-channel polarization image set.

4. The method for selecting polarization channels for infrared polarization imaging according to claim 1, characterized in that: The process of solving the Stokes vector image includes: An overdetermined set of equations is constructed according to the relationship between the intensity value and the polarization angle at the same pixel point of the infrared polarization image; the overdetermined set of equations is solved by a pseudo-inverse method to obtain an optimal solution of the Stokes vector, and the optimal solution of the Stokes vector is used as a Stokes vector image.

5. The method for selecting polarization channels for infrared polarization imaging according to claim 4, characterized in that: The Stokes vector image includes an S0 image, an S1 image and an S2 image.

6. The method for selecting polarization channels for infrared polarization imaging according to claim 5, characterized in that: The relationship between the intensity value and the polarization angle of the polarization image at the same pixel point is: I θj (u,v)=0.5·S0(u,v)+0.5·S1(u,v)·cos(2θ j )+0.5·S2(u,v)·sin(2θ j ) Where (u, v) is the pixel coordinate, is the intensity value of the polarization image at the pixel point (u, v), θ j is the polarization angle of the polarization image at the pixel point (u, v), S0(u, v), S1(u, v) and S2(u, v) are the Stokes vectors; The overdetermined system of equations is: In the formula, is the intensity value of the polarization image of channel n at the pixel point (u, v), θ n It is the polarization angle of the polarization image of channel n at the pixel point (u,v).

7. The method for selecting polarization channels for infrared polarization imaging according to claim 1, characterized in that: The polarization characteristic parameter image includes a polarization degree image and a polarization angle image.

8. The method for selecting polarization channels for infrared polarization imaging according to claim 7, characterized in that: The polarization degree image is: Where ρ(u,v) is the polarization image at the pixel point (u,v), S0(u,v), S1(u,v) and S2(u,v) are Stokes vectors; The polarization angle image is: In the formula, is the polarization angle image at the pixel point (u, v).

9. The method for selecting polarization channels for infrared polarization imaging according to claim 1, characterized in that: The image similarity index includes multiple ones of image mean square error, peak signal-to-noise ratio and structural similarity index.

10. An infrared polarization imaging polarization channel selection system, characterized in that: include: A polarization image acquisition module uses a polarization camera to acquire a multi-channel infrared polarization image set of a target scene, and performs noise reduction preprocessing on the multi-channel infrared polarization image set; An image set decomposition module, used for dividing the multi-channel infrared polarization image set that has undergone noise reduction preprocessing into image sets with different numbers of channels; A polarization characteristic parameter image acquisition module is used to solve the Stokes vector image of the image set with different channel numbers, and obtain polarization characteristic parameter images under different channel numbers according to the Stokes vector image; The polarization channel selection module uses the polarization characteristic parameter image with the largest number of channels as a reference image, and uses an image similarity index to evaluate the similarity between images with other channel numbers and the reference image to obtain the best polarization channel that meets the imaging requirements of the target scene.

Images