A method and device for enhancing image of power drone inspection with adaptive light intensity adjustment

Abstract

The present invention provides a method and device for enhancing inspection images of electric power drones with adaptive light intensity adjustment, the method comprising: performing super-pixel segmentation on the collected inspection images to obtain N superpixel regions; perform intrinsic image decomposition on the superpixel regions to obtain the spectral reflection components of the superpixel regions; combine the spectral reflection components of each superpixel region to obtain the preliminary image enhancement result; use the bright channel feature extraction model to process the inspection image to obtain the bright saliency map; make a decision fusion between the preliminary image enhancement result and the bright saliency map to obtain the final restored image with enhanced brightness. The present invention can ensure that even under extreme lighting conditions (such as backlight and weak light), the electric drone can obtain high-quality inspection images, improve the safety and reliability of intelligent operation and inspection of power facilities, and can be widely used in fields such as electric drone inspection, with broad application prospects and market value.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a method and device for enhancing an inspection image of an electric power unmanned aerial vehicle with adaptive light intensity adjustment. Background Art

[0002] Power drones play an important role in power overhead line inspection tasks due to their ease of operation and efficient operation capabilities. Especially in difficult-to-reach places with complex terrain (such as mountainous areas), drones can effectively overcome natural obstacles and achieve comprehensive monitoring of transmission lines. However, under harsh lighting conditions, such as backlight or low-light environments, the quality of inspection images taken by drones may deteriorate, causing key features in the image to be blurred and difficult to identify.

[0003] At this time, the application of image enhancement technology is particularly important. This type of technology can significantly improve image quality and make the details in the image clearer and more discernible. By adopting image enhancement technology, power drones can provide accurate and reliable inspection information even under adverse lighting conditions, ensuring the safe and stable operation and effective maintenance of power facilities.

[0004] In summary, there is an urgent need for a method and device for enhancing the inspection image of an electric power drone with adaptive adjustment of light intensity to solve the problems existing in the prior art. Summary of the invention

[0005] The purpose of the present invention is to provide a method for enhancing the inspection image of an electric power drone with adaptive adjustment of light brightness, aiming to suppress the interference of backlight and weak light phenomena on the imaging of high-quality inspection images and meet the requirements for high-quality imaging of electric power inspection images. The specific technical scheme is as follows:

[0006] A method for enhancing an image of an electric power drone inspection with adaptive light intensity adjustment, comprising:

[0007] Collected inspection images Y Perform superpixel segmentation and obtain N Superpixel area;

[0008] Perform intrinsic image decomposition on the super-pixel region to obtain the spectral reflectance component of the super-pixel region; combine the spectral reflectance components of each super-pixel region to obtain the preliminary image enhancement result ;

[0009] The bright channel feature extraction model is used to extract the inspection image Y Processing to obtain a bright saliency map ;

[0010] The image is initially enhanced. and Brightness Saliency Map Perform decision fusion to obtain the final restored image with enhanced light intensity .

[0011] In the above technical solution, the inspection image is preferably Y The superpixel area is expressed as:

[0012] (1),

[0013] in, N is the number of superpixel regions, Indicates superpixel area.

[0014] In the above technical solution, the super pixel area is preferably calculated according to formula (3): Perform intrinsic image decomposition:

[0015] (3),

[0016] in, Indicates inspection image Y Middle i The first superpixel region n pixels, Indicates i The first superpixel region n The spectral reflectance component of each pixel, Indicates i The first superpixel region n The brightness component of each pixel.

[0017] In the above technical solution, the preferred method is to convert the formula (3) into and To solve:

[0018] (4),

[0019] in, Indicates solution and The objective function of In pixels c Centered at A local window of pixels, Parameters to control the size of the local window; Represents pixels c and pixels v Similarity of spectral angles between

[0020] According to formula (3) and formula (4), the super pixel area is obtained The spectral reflectance component of each pixel in the final super-pixel area is obtained The spectral reflectance component .

[0021] In the above technical solution, the spectral reflection components of each super-pixel area are combined to generate a preliminary image enhancement result. :

[0022] (6),

[0023] in, Indicates i The spectral reflectance component of the super-pixel area is N is the number of superpixel regions.

[0024] The above technical solution is preferably used to obtain a bright saliency map. The specific method is:

[0025] Through the bright channel feature extraction model, the inspection image Y Extract bright channel features , for bright channel characteristics Perform average fusion to obtain the atlas , for the atlas Perform Fourier transform to generate amplitude spectrum , for the amplitude spectrum Perform logarithmic calculations to generate logarithmic spectra , and based on the logarithmic spectrum Generate average spectrum , and finally according to the logarithmic spectrum and average spectrum Obtain a bright saliency map without dark components .

[0026] Preferably, in the above technical solution, the bright channel feature extraction model is:

[0027] (7),

[0028] Bright channel characteristics Perform average fusion to obtain the atlas for:

[0029] (8),

[0030] Pair Spectrum Perform Fourier transform to generate amplitude spectrum for:

[0031] (9),

[0032] Based on the logarithmic spectrum Generate average spectrum for:

[0033] (10),

[0034] According to the logarithmic spectrum and average spectrum Obtain a bright saliency map without dark components for:

[0035] (12),

[0036] in, is the red channel of the inspection image, is the green channel of the inspection image, is the blue channel of the inspection image, is the hue component of the inspection image in the HUE hue-saturation-intensity change, is the saturation component of the inspection image in the HUE hue-saturation-intensity change, is the intensity component of the inspection image in the HUE hue-saturation-intensity change, p Bright channel characteristics M The number of channels, Express The function to perform Fourier transform, For local k×k Mean filter feature extractor, k represents the parameters of the mean filter feature extractor, is the inverse Fourier transform operation.

[0037] In the above technical solution, the local k×k Mean filter feature extractor It is expressed as:

[0038] (11).

[0039] The preferred technical solution above is to use the image preliminary enhancement result according to formula (13): and Brightness Saliency Map Perform decision fusion to obtain the final restored image with enhanced light intensity :

[0040] (13).

[0041] The present invention also provides an electric drone inspection image enhancement device with adaptive light intensity adjustment, comprising a memory and a processor. When the processor runs the computer instructions stored in the memory, it executes the electric drone inspection image enhancement method with adaptive light intensity adjustment.

[0042] The application of the technical solution of the present invention has the following beneficial effects:

[0043] On the one hand, the present invention fully mines the spatial structure similarity information and intrinsic image spectral reflectance information of the image through superpixel segmentation and intrinsic image decomposition algorithm to obtain the preliminary image enhancement result; on the other hand, by mining the saliency information of the image brightness, a bright saliency map is generated; finally, the preliminary image enhancement result and the bright saliency map are decision-fused, and the image spatial structure similarity information, the intrinsic image spectral reflectance information, and the saliency information of the image brightness are jointly mined to further improve the image enhancement effect.

[0044] The present invention can ensure that even under extreme lighting conditions (such as backlight and weak light), the electric power drone can obtain high-quality inspection images, thereby improving the safety and reliability of intelligent operation and inspection of power facilities. It can be widely used in fields such as electric power drone inspection and has broad application prospects and market value.

[0045] In addition to the above-described purposes, features and advantages, the present invention has other purposes, features and advantages. The present invention will be further described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The drawings constituting a part of this application are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0047] Figure 1 It is a flow chart of the image enhancement method for power UAV inspection with adaptive light intensity adjustment of the present invention;

[0048] Figure 2a This is an imaging effect diagram of the first inspection scene when the enhancement method of the present invention is not used;

[0049] Figure 2b This is an imaging effect diagram when the enhancement method of the present invention is used in the first inspection scene;

[0050] Figure 3a This is an imaging effect diagram of the second inspection scene when the enhancement method of the present invention is not used;

[0051] Figure 3b This is an imaging effect diagram of the second inspection scene when the enhancement method of the present invention is used. DETAILED DESCRIPTION

[0052] In order to facilitate the understanding of the present invention, the present invention will be described more fully below, and preferred embodiments of the present invention are provided. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present invention more thorough and comprehensive.

[0053] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0054] Example:

[0055] See also Figure 1 This embodiment provides a method for enhancing an image of an electric power drone inspection with adaptive light intensity adjustment, comprising the following steps:

[0056] Step S1: Perform superpixel segmentation on the collected inspection image to obtain N Superpixel region;

[0057] Preferably, in step S1, a power drone is used to photograph the power equipment to obtain an original visible light image in jpg format. Y (i.e. inspection image Y ). Specifically, the superpixel region is a homogeneous region in the inspection image, and the pixels in the homogeneous region have similar spatial information. In this embodiment, the entropy rate superpixel segmentation technology (ERS) is used to obtain the superpixel region in the inspection image; first, the principal component analysis algorithm is used to analyze the inspection image. Y The first principal component is processed and the first principal component is retained. Then, the graph optimization problem is modeled for the first principal component, and a series of superpixel edges are generated using the ERS algorithm to obtain several homogeneous regions (i.e., superpixel regions).

[0058] For the entropy rate superpixel segmentation technology (ERS), please refer to the prior art "Entropy rate superpixelsegmentation", Liu M, Tuzel O, Ramalingam S, et al, Proc of Processing IEEE Confense Computer Vision Pattern Recognition, 2011, 2097–2104; for the principal component analysis algorithm, please refer to the prior art "Two-stage image denoising by principal component analysis withlocal pixel grouping", Zhang L, Dong W, Zhang D, et al, Pattern Recognition, 2010, 43(4):1531–1549; these two technologies will not be further described in this embodiment.

[0059] Specifically, the inspection image Y The superpixel area is expressed as:

[0060] (1),

[0061] in, N is the number of superpixel regions, Indicates superpixel area.

[0062] Step S2: Decompose the intrinsic image of the super-pixel region to obtain the spectral reflection component of the super-pixel region; combine the spectral reflection components of each super-pixel region to obtain a preliminary image enhancement result. ;

[0063] Specifically, inspection images Y The intrinsic image can be decomposed into spectral reflectance components R and brightness component S , the process is expressed as:

[0064] (2),

[0065] in: Indicates the inspection image pixels, Indicates the inspection image The spectral reflectance component of each pixel, Indicates the inspection image The brightness component of each pixel;

[0066] Furthermore, , , ; Respectively represent inspection images Y No. The elements of a pixel in the red, green and blue channels, Respectively represent the first The spectral reflectance components of each pixel in the red, green and blue channels are Respectively represent the first The brightness components of each pixel in the red, green and blue channels. r represents the red channel, g represents the green channel, b Represents the blue channel.

[0067] Furthermore, since the pixel value of the image brightness component is directly related to factors such as scene lighting and shadows, and the spectral reflectance component is directly related to the material of the object itself, removing the brightness component in the image and retaining only the spectral reflectance component in the image can better suppress the interference of extreme lighting in the image. The intrinsic image is decomposed into:

[0068] (3),

[0069] in, Indicates inspection image Y Middle i The first superpixel region n pixels, Indicates i The first superpixel region n The spectral reflectance component of each pixel, Indicates i The first superpixel region n The brightness component of each pixel.

[0070] Furthermore, in formula (3) and It can be obtained by solving the following optimization equation:

[0071] (4),

[0072] in, Indicates solution and The objective function of In pixels c Centered at A local window of pixels, Parameters to control the size of the local window; Represents pixels c and pixels v The similarity of the spectral angles.

[0073] According to formula (3) and formula (4), the super pixel area is obtained The spectral reflectance component of each pixel in , and finally the entire superpixel area is obtained The spectral reflectance component :

[0074] (5),

[0075] Where sup represents the superpixel area The number of pixels inside.

[0076] Combine the spectral reflectance components of each superpixel area to generate a preliminary image enhancement result :

[0077] (6),

[0078] in, Indicates i The spectral reflectance component of the super-pixel area is N is the number of superpixel regions.

[0079] Step S3: Process the inspection image using the bright channel feature extraction model to obtain a bright saliency map ;

[0080] Specifically, the bright channel feature extraction model constructed in this embodiment is:

[0081] (7),

[0082] in, is the bright channel feature, is the red channel of the inspection image, is the green channel of the inspection image, is the blue channel of the inspection image, is the hue component of the inspection image in the HUE hue-saturation-intensity change, is the saturation component of the inspection image in the HUE hue-saturation-intensity change, It is the intensity component of the inspection image in the HUE hue-saturation-intensity change.

[0083] By extracting the characteristic band with the maximum brightness of the inspection image, the bright channel feature can be obtained. , for bright channel characteristics Perform average fusion to reduce the dimension of features and obtain the atlas :

[0084] (8),

[0085] in, p Bright channel characteristics M The number of channels.

[0086] To further explore the graph The saliency map with bright features in the image suppresses the dark features in the image. Perform Fourier transform to generate amplitude spectrum :

[0087] (9),

[0088] in, Express Function that performs Fourier transform.

[0089] Furthermore, for the amplitude spectrum Perform logarithmic calculations to generate logarithmic spectra , and generate the average spectrum according to formula (10) :

[0090] (10),

[0091] in, For local k×k Mean filter feature extractor.

[0092] Preferably, the local k×k Mean filter feature extractor for:

[0093] (11),

[0094] in, k is the parameter of the mean filter feature extractor. k =3.

[0095] Furthermore, the average spectrum It can reflect the redundant dark area features in the image, and then restore the bright saliency map of the dark component through formula (12) :

[0096] (12),

[0097] in, is the inverse Fourier transform operation.

[0098] Step S4: Initially enhance the image and Brightness Saliency Map Perform decision fusion to obtain the final restored image with enhanced light intensity .

[0099] Specifically, according to formula (13), the image preliminary enhancement result and Brightness Saliency Map Decision fusion is performed to suppress the interference of backlight and weak light in the image on high-quality inspection image imaging:

[0100] (13).

[0101] See also Figure 2a , Figure 2b , Figure 3a and Figure 3b ,in Figure 2a and Figure 2b For a group of the same inspection scenes, Figure 2a and Figure 2b It can be seen that when the enhancement method of this embodiment is not used, the overall image is dark, and the features in the image are blurred and difficult to identify. After the enhancement method of this embodiment is used, the overall image is bright, and the features in the image are clearer and easier to identify. Figure 3a and Figure 3b For a group of the same inspection scenes, Figure 3a and Figure 3b It can be seen that when the enhancement method of this embodiment is not used, the image is dark as a whole, the features on the lower right side are blurred, and the smoke in the image is difficult to identify. After the enhancement method of this embodiment is used, the image is brighter, the features on the lower right side are clearer and easier to identify, and the smoke on the lower right side of the image can be clearly seen. It can be seen that the enhancement method of this embodiment can effectively enhance the brightness of the light in the image. Even under extreme lighting conditions (such as backlight and weak light), the electric drone can obtain high-quality inspection images, improve the safety and reliability of intelligent operation and inspection of power facilities, and can be widely used in fields such as electric drone inspection, with broad application prospects and market value.

[0102] This embodiment also provides an electric drone inspection image enhancement device with adaptive light brightness adjustment, comprising a memory and a processor. When the processor runs the computer instructions stored in the memory, it executes the electric drone inspection image enhancement method with adaptive light brightness adjustment to realize processing of the inspection image.

[0103] This embodiment also provides another electric drone inspection image enhancement device with adaptive light brightness adjustment, including an image acquisition unit and a data calculation and processing unit. The image acquisition unit is used to photograph electric equipment to obtain inspection images; the data calculation and processing unit processes the collected inspection images according to the electric drone inspection image enhancement method with adaptive light brightness adjustment.

[0104] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for enhancing the inspection image of an electric power drone with adaptive light intensity adjustment, characterized in that: include: Collected inspection images Y Perform superpixel segmentation and obtain N Superpixel region; Perform intrinsic image decomposition on the super-pixel region to obtain the spectral reflectance component of the super-pixel region; combine the spectral reflectance components of each super-pixel region to obtain the preliminary image enhancement result ; The bright channel feature extraction model is used to extract the inspection image Y Processing to obtain a bright saliency map ; The image is initially enhanced. and Brightness Saliency Map Perform decision fusion to obtain the final restored image with enhanced light intensity ; Among them, the bright saliency map is obtained The specific method is: the inspection image is extracted by the bright channel feature extraction model Y Extract bright channel features , for bright channel characteristics Perform average fusion to obtain the atlas , for the atlas Perform Fourier transform to generate amplitude spectrum , for the amplitude spectrum Perform logarithmic calculations to generate logarithmic spectra , and based on the logarithmic spectrum Generate average spectrum , and finally according to the logarithmic spectrum and average spectrum Obtain a bright saliency map without dark components ; Bright channel characteristics Perform average fusion to obtain the atlas for: （8）， in, p Bright channel characteristics M The number of channels.

2. The method for enhancing the inspection image of an electric power drone with adaptive light intensity adjustment according to claim 1 is characterized in that: The inspection image Y The superpixel area is expressed as: （1）， in, N is the number of superpixel regions, Indicates superpixel area.

3. The method for enhancing the inspection image of an electric power drone with adaptive light intensity adjustment according to claim 1 is characterized in that: According to formula (3), the superpixel area Perform intrinsic image decomposition: （3）， in, Indicates inspection image Y Middle i The first superpixel region n pixels, Indicates i The first superpixel region n The spectral reflectance component of each pixel, Indicates i The first superpixel region n The brightness component of each pixel.

4. The method for enhancing the inspection image of an electric power drone with adaptive light intensity adjustment according to claim 3 is characterized in that: According to formula (4), and To solve: （4）， in, Indicates solution and The objective function of In pixels c Centered at A local window of pixels, Parameters to control the size of the local window; Represents pixels c and pixels v Similarity of spectral angles between According to formula (3) and formula (4), the super pixel area is obtained The spectral reflectance component of each pixel in the final super-pixel area is obtained The spectral reflectance component .

5. The method for enhancing the inspection image of an electric power drone with adaptive light intensity adjustment according to claim 4 is characterized in that: Combine the spectral reflectance components of each superpixel area to generate a preliminary image enhancement result : （6）， in, Indicates i The spectral reflectance component of the super-pixel area is N is the number of superpixel regions.

6. The method for enhancing the inspection image of an electric power drone with adaptive light intensity adjustment according to claim 1 is characterized in that: The bright channel feature extraction model is: （7）， Pair Spectrum Perform Fourier transform to generate amplitude spectrum for: （9）， Based on the logarithmic spectrum Generate average spectrum for: （10）， According to the logarithmic spectrum and average spectrum Obtain a bright saliency map without dark components for: （12）， in, is the red channel of the inspection image, is the green channel of the inspection image, is the blue channel of the inspection image, is the hue component of the inspection image in the HUE hue-saturation-intensity change, is the saturation component of the inspection image in the HUE hue-saturation-intensity change, is the intensity component of the inspection image in the HUE hue-saturation-intensity change, Express The function to perform Fourier transform, For local k×k Mean filter feature extractor, k represents the parameters of the mean filter feature extractor, is the inverse Fourier transform operation.

7. The method for enhancing the inspection image of an electric power drone with adaptive light intensity adjustment according to claim 6 is characterized in that: The local k×k Mean filter feature extractor It is expressed as: （11）。 8. The method for enhancing the inspection image of an electric power drone with adaptive light intensity adjustment according to claim 1 is characterized in that: According to formula (13), the image preliminary enhancement result and Brightness Saliency Map Perform decision fusion to obtain the final restored image with enhanced light intensity : （13）。 9. An image enhancement device for power drone inspection with adaptive light intensity adjustment, characterized in that: It comprises a memory and a processor. When the processor runs the computer instructions stored in the memory, it executes the power drone inspection image enhancement method with adaptive light brightness adjustment as described in any one of claims 1 to 8.

Images