Pipeline crack visual detection method based on multispectral feature level fusion

Abstract

The application discloses a pipeline crack visual detection method based on multispectral feature level fusion, and relates to the technical field of pipeline cracks. The application solves the problems of low detection accuracy, poor robustness and difficulty in distinguishing defect types in the prior art. The application optimizes the fusion of multispectral images, and performs weighted fusion on the gradient profile and internal features of the overlapping area. The key feature information of each waveband is retained, and the feature difference of the hole area is highlighted through the coefficient factor. The optimized result more intuitively and accurately reflects the actual state of the pipeline defects. Through the collaborative application of multispectral information and the step-by-step feature processing, the accuracy, completeness and robustness of pipeline crack and hole area detection are significantly improved. The application can adapt to the pipeline detection requirements in complex environments, provides reliable technical support for pipeline safety evaluation and maintenance decision-making, and has strong engineering application value.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of pipeline crack, in particular to a pipeline crack visual detection method based on multi-spectrum feature level fusion. BACKGROUND

[0002] As the core infrastructure in the fields of energy transportation, urban water supply and drainage, industrial fluid transportation, etc., the operation safety of pipeline is directly related to the development of national economy and social public safety; with the extension of service life of pipeline, cracks, holes and other defects are prone to occur on the surface and inside of pipeline due to the influence of factors such as corrosion, fatigue load, external force impact and environmental erosion. If such defects are not detected and repaired in time, it may cause medium leakage, explosion and other major accidents, resulting in serious economic loss and environmental harm.

[0003] At present, pipeline crack detection technology has developed various implementation paths, and traditional detection methods mainly include manual inspection, ultrasonic detection, magnetic powder detection, eddy current detection, etc.; manual inspection relies on the experience judgment of detection personnel, which is low in efficiency and strong in subjectivity, and is difficult to meet the detection needs of large-scale pipeline network, and there are great safety hazards in complex scenes such as underground pipeline and high-altitude pipeline; although ultrasonic detection and magnetic powder detection have certain detection accuracy, they need to be in direct contact with the surface of pipeline, and have high requirements for the cleanliness of pipeline surface, and the detection range is limited, which is difficult to realize rapid and full coverage detection; eddy current detection is only suitable for metal pipeline, and has poor detection effect on non-metal pipeline or defects under pipeline coating.

[0004] Existing visual detection technology is mainly based on single spectrum band (such as visible light) for detection, and through gray processing, edge detection, contour extraction and other processing of visible light image of pipeline surface, the recognition and marking of crack area are realized; however, the single spectrum visual detection method has significant technical bottlenecks: the visible light band can only capture the macro morphological features of pipeline surface, and has poor detection effect on fine cracks (width less than 0.1mm), shallow internal cracks or cracks disturbed by surface stains and light changes; at the same time, for the closed contour area of pipeline surface, it is difficult to effectively distinguish whether it is ordinary scratch or hole with structural damage, which is easy to cause misjudgment or omission problem, resulting in that the detection accuracy and robustness are difficult to meet the actual needs in complex scenes.

[0005] To improve the comprehensive performance of visual detection, some studies try to introduce multispectral imaging technology to make up for the shortcomings of single spectrum by integrating image information of different wavebands. However, existing multispectral detection methods mostly stay at the data level fusion, only simply superimposing or splicing different waveband images, without fully exploiting the complementary characteristics of each waveband. In the defect recognition link, there is a lack of collaborative analysis of multispectral gradient features and grayscale features, making it difficult to accurately extract key features of crack and hole regions. Meanwhile, the methods for judging the coincidence of defect regions in multispectral images and feature optimization are relatively rough, resulting in problems such as unobvious features and inaccurate defect positioning in the fused detection results.

[0006] Therefore, how to design a detection method that can fully exploit the complementary advantages of multispectral images, realize accurate feature extraction and fusion optimization, and solve the problems of low detection accuracy, poor robustness, and difficulty in distinguishing defect types in the prior art, has become a technical difficulty that needs to be broken through in the field of pipeline visual detection. SUMMARY

[0007] To overcome the shortcomings of the prior art, the present application provides a multispectral feature-level fusion pipeline crack visual detection method, which solves the problems of low detection accuracy, poor robustness, and difficulty in distinguishing defect types in the prior art.

[0008] To achieve the above purpose, the present application realizes the following technical scheme: a multispectral feature-level fusion pipeline crack visual detection method, comprising the following steps:

[0009] Step 1: Use a multispectral device to acquire pipeline images of a specified area of the pipeline, sequentially confirm the visible light images, near-infrared images and thermal infrared images associated with the specified area, and generate a multispectral image set associated with the specified area of the pipeline, in particular:

[0010] When using a multispectral device to take pictures of the specified area, take pictures according to the preset wavelength characteristics, and confirm the pipeline images associated with the corresponding wavelength characteristics;

[0011] Integrate different pipeline images associated with different wavelength characteristics to generate a multispectral image set corresponding to the specified area of the pipeline;

[0012] Step 2: According to the multispectral image set associated with the specified area of the pipeline, use the Sobel algorithm to confirm the gradient pixel points associated with different images in the multispectral image set, and from the confirmed gradient pixel points, confirm and mark the crack regions associated with the corresponding images, in particular:

[0013] Grayscale processing of the pipeline images associated with the multispectral image set to confirm the grayscale images associated with the corresponding pipeline images;

[0014] Sobel algorithm is used again to confirm the vertical gradient and the horizontal gradient associated with different gray points in the gray image. According to the vertical gradient and the horizontal gradient associated with the corresponding gray point, the comprehensive gradient associated with the corresponding gray point is confirmed, which And it is confirmed whether the comprehensive gradient of the corresponding gray point satisfies: comprehensive gradient≥Y1, where Y1 is a preset value. If it satisfies, the corresponding gray point is marked as a gradient pixel point. If the comprehensive gradient is less than Y1, no marking is performed;

[0015] The adjacent gradient pixel points in the gray image are connected and processed to confirm the gradient edge line associated with the corresponding gradient pixel point. If the gradient edge line is in a closed state, the image area included in the gradient edge line is marked as a crack area. If the gradient edge line is not in a closed state, no marking is performed;

[0016] Step three, based on the crack area confirmed in the corresponding image, the gray feature associated with the crack area is checked and compared with the gray feature of the peripheral area. Based on the checking result, it is confirmed whether the crack area is a hole area, and the hole area is marked in the corresponding image at the same time. The specific method is:

[0017] According to the confirmed crack area, the external area of the crack area is recorded as a peripheral area, and the gray values associated with different gray points in the peripheral area are processed by mean value. The mean value associated with several gray values is confirmed, and the confirmed mean value is recorded as the gray feature TZ1 associated with the peripheral area. At the same time, the gray values associated with several gray points in the crack area are processed by mean value, and the gray feature TZ2 in the crack area is confirmed;

[0018] If the gray feature TZ1 and the gray feature TZ2 satisfy: |TZ1-TZ2|≥Y2, where Y2 is a preset value, if it satisfies, the confirmed crack area is recorded as a hole area, if it does not satisfy, no marking is performed;

[0019] Step four, according to the hole area marked in the image, the overlapping area associated with the hole area in multiple images is confirmed, and the gradient profile and internal feature associated with the overlapping area are comprehensively optimized, and the optimized hole area is displayed. The specific method is:

[0020] The hole areas associated with different images in the multispectral image set are confirmed, and then the corresponding hole areas in multiple images are overlapped to confirm whether there is an overlapping area. If there is an overlapping area, the overlapping area existing in the hole area is marked. If there is no overlapping area, the corresponding image is directly displayed. The displayed image is marked with a hole area and a crack area at the same time;

[0021] According to the marked coincident region in the multispectral image set, the edge profile of the coincident region is marked, and the gray value associated with each profile point on the edge profile is confirmed simultaneously, and the gray value associated with this profile point on the visible light image is recorded as ZH1 k The gray value associated with this profile point on the near-infrared image is recorded as ZH2 k The gray value associated with this profile point on the thermal infrared image is recorded as ZH3 k Wherein k represents different profile points, and ZH k = ZH1 k × C1+ ZH2 k × C2+ ZH3 k × C3 is adopted, and the comprehensive gray value ZH k associated with the corresponding profile point is confirmed, wherein C1, C2 and C3 are all preset fixed coefficient factors, and C1+C2+C3=1, and the gray value associated with the edge profile of the coincident region is adjusted again according to the comprehensive gray value associated with the corresponding profile point, so that the adjusted and optimized edge profile is obtained.

[0022] The same point associated with the coincident region is processed again by using the same processing method, and the gray value associated with the same point on the visible light image is recorded as ZZ1 i The gray value associated with this same point on the near-infrared image is recorded as ZZ2 i The gray value associated with this same point on the thermal infrared image is recorded as ZZ3 i Wherein i represents different same points, and XD i = ZZ1 i × A1+ ZZ2 i × A2+ ZZ3 i × A3 is adopted, and the comprehensive gray value associated with the corresponding same point is confirmed, wherein A1, A2 and A3 are all preset fixed coefficient factors, and A1+A2+A3=0.8, and the gray value associated with the same point in the coincident region is adjusted again according to the comprehensive gray value associated with the same point, so that the adjusted and optimized coincident region is obtained.

[0023] Preferably, the wavelength characteristics associated with the visible light image are 400-760nm, the wavelength characteristics associated with the near-infrared image are 760-1100nm, and the wavelength characteristics associated with the thermal infrared image are 8-14um.

[0024] The present application provides a multispectral feature level fusion pipeline crack visual detection method.

[0025] The application realizes the preliminary positioning of the crack area by combining the accurate extraction of gradient pixel points by the Sobel algorithm and the closed contour judgment rule in the crack area identification link, and further improves the fine degree of defect identification by distinguishing the hole area and the ordinary scratch through the gray feature check, effectively avoiding the misjudgment of non-structural defects.

[0026] For the fusion optimization of multispectral images, the gradient profile and internal feature weighted fusion (ZH k and the comprehensive gray calculation of XD i ) are combined, which not only retains the key feature information of each band, but also highlights the feature difference of the hole area through the coefficient factor, so that the optimized result more intuitively and accurately reflects the real state of the pipeline defects.

[0027] Through the cooperative application of multispectral information and step-by-step feature processing, the accuracy, completeness and robustness of pipeline crack and hole area detection are significantly improved, which can adapt to the pipeline detection demand in complex environment, provide reliable technical support for pipeline safety evaluation and maintenance decision, and has strong engineering application value. BRIEF DESCRIPTION OF DRAWINGS

[0028] Figure 1 The figure is a schematic diagram of the method of the application. DETAILED DESCRIPTION

[0029] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the application.

[0030] Please refer to Figure 1 , the application provides a multispectral feature-level fusion pipeline crack visual detection method, including the following steps:

[0031] Step 1: Use a multispectral device to acquire pipeline images of a specified area of the pipeline, and sequentially confirm the visible light image, near-infrared image and thermal infrared image associated with the specified area to generate a multispectral image set associated with the specified area of the pipeline:

[0032] When the multispectral device is used to take pictures of the specified area, the pictures are taken according to the preset wavelength characteristics, and the pipeline images associated with the corresponding wavelength characteristics are confirmed, wherein the wavelength characteristics associated with the visible light image are 400-760 nm, the wavelength characteristics associated with the near-infrared image are 760-1100 nm, and the wavelength characteristics associated with the thermal infrared image are 8-14 μm.

[0033] The different pipeline images associated with different wavelength characteristics are integrated to generate a multi-spectral image set corresponding to the specified area of the pipeline;

[0034] Specifically, when the corresponding multi-spectral device acquires images, different images are associated with different wavelength characteristics, and the corresponding wavelength characteristics are preset by relevant personnel in advance. The corresponding associated wavelength values are set in advance and images are acquired. In the actual operation process, the multi-spectral device can adjust the corresponding wavelength characteristics to actually complete the corresponding image acquisition process. Since multi-spectral image acquisition is relatively common in the prior art, further description is not provided here.

[0035] Step two, according to the multi-spectral image set associated with the specified area of the pipeline, the Sobel algorithm is used to confirm the gradient pixel points associated with different images in the multi-spectral image set, and from the confirmed gradient pixel points, the crack area associated with the corresponding image is confirmed and marked. Specifically, different spectral images exist in the corresponding multi-spectral image set, and each different spectral image needs to be grayed to confirm the corresponding gray image and confirm the gray value associated with different pixel points in the gray image. From the confirmed gray characteristics, the gradient pixel points are marked and confirmed, and then the marking process of the crack area in the corresponding image is completed from the marking process.

[0036] Specifically, the crack area associated with the image is marked as follows:

[0037] The multi-spectral image set associated with the pipeline image is grayed to confirm the gray image associated with the corresponding pipeline image (different image points are associated with different RGB values. According to the set weight, the gray value associated with the corresponding image point can be confirmed according to the corresponding RGB value, and the gray value = 0.299×R+0.587×G+0.114×B).

[0038] Then the Sobel algorithm is used to confirm the vertical gradient and horizontal gradient associated with different gray points in the gray image (the gray points are sorted in the form of a nine-square grid. Each gray point in the nine-square grid has a corresponding gray value. According to the weight assigned and the different gray values associated with the surrounding gray points, the vertical gradient and horizontal gradient associated with the intermediate gray point can be effectively confirmed. The weight factor is between -2 and 2, including 0. Since the Sobel algorithm confirms the gradient of the gray point in the prior art, further description is not provided here. and confirm whether the comprehensive gradient of the corresponding gray scale point satisfies: comprehensive gradient≥Y1, wherein Y1 is a preset value, and a specific value of Y1 is determined by an operator according to experience, if the condition is satisfied, the corresponding gray scale point is marked as a gradient pixel point, otherwise, no marking is performed;

[0039] The adjacent gradient pixel points in the gray scale image are connected and processed to confirm the gradient edge line associated with the corresponding gradient pixel point, if the gradient edge line is in a closed state, the image area included in the gradient edge line is marked as a crack area, if the gradient edge line is not in a closed state, no marking is performed;

[0040] Specifically, in the corresponding image, there are different gradient pixel points, and the different gradient pixel points can be in an adjacent state, so that the gradient pixel points form a gradient contour line, if the gradient pixel points are not in an adjacent state, the gradient contour line cannot be formed, when the gradient contour line is in a closed state, the corresponding closed area is a crack area, which can be an area included by a scratch or a hole area, and needs to be processed and confirmed by subsequent related processing steps;

[0041] Step three, based on the crack area confirmed in the corresponding image, the gray scale features associated with the crack area are checked and compared with the gray scale features of the peripheral area, based on the checking result, whether the crack area is a hole area is confirmed, and the hole area is marked in the corresponding image at the same time, specifically, the corresponding crack area can be an internal area caused by a scratch, when a scratch on the outer wall of the pipeline is in a closed state, the internal area of the scratch does not belong to a hole area, but is an internal area caused by the corresponding scratch, if the internal area of the corresponding scratch is a hole area, the marking process of the corresponding hole area can be completed according to the pixel point processing process;

[0042] Specifically, the hole area is marked in the corresponding image at the same time in the following manner:

[0043] According to the confirmed crack area, the external area of the crack area is recorded as a peripheral area, the gray scale values associated with different gray scale points in the peripheral area are subjected to mean value processing, the mean values associated with a plurality of gray scale values are confirmed, and the confirmed mean value is recorded as the gray scale feature TZ1 associated with the peripheral area, and the gray scale values associated with a plurality of gray scale points in the crack area are subjected to mean value processing, and the gray scale feature TZ2 in the crack area is confirmed;

[0044] If the gray scale feature TZ1 and the gray scale feature TZ2 satisfy: |TZ1-TZ2|≥Y2, wherein Y2 is a preset value, and a specific value of Y2 is determined by an operator according to experience, and Y2 is generally 100, the numerical range of the gray scale feature is 0-255, if the condition is satisfied, the confirmed crack area is recorded as a hole area, if the condition is not satisfied, no marking is performed;

[0045] Specifically, when the gray scale feature difference associated with the corresponding crack area and the peripheral area is large, it means that the corresponding crack area is a hole area, and the gray scale value needs to be confirmed and marked for subsequent feature fusion of the hole area and better display of the hole area.

[0046] Step four, according to the marked hole area in the image, confirm the overlapping area associated with the hole area in multiple images, and comprehensively optimize the gradient profile and internal features associated with the overlapping area, and display the optimized hole area;

[0047] Among them, the specific way of comprehensively optimizing the hole area is:

[0048] Confirm the hole area associated with different images in the multispectral image set, and then overlap the corresponding hole area in multiple images to confirm whether there is an overlapping area. If there is an overlapping area, mark the overlapping area in the hole area. If there is no overlapping area, directly display the corresponding image. The displayed image is marked with a hole area and a crack area. Specifically, when overlapping, the center point of the corresponding image can be directly translated and overlapped according to the center point of the corresponding image. When the image center point is obtained, the center point in the corresponding image can be marked in advance by combining the two-dimensional coordinate system and the edge profile of the corresponding image. Each profile point on the edge profile is associated with a set of two-dimensional coordinates. After averaging a plurality of two-dimensional coordinates, the average coordinate can be confirmed, and the center point of the corresponding image can be calibrated according to the point where the average coordinate is located.

[0049] According to the marked overlapping area in the multispectral image set, mark the edge profile of the overlapping area, and simultaneously confirm the gray scale value associated with each profile point on the edge profile. The gray scale value associated with this profile point on the visible light image is denoted as ZH1 k The gray scale value associated with this profile point on the near-infrared image is denoted as ZH2 k The gray scale value associated with this profile point on the thermal infrared image is denoted as ZH3 k Where k represents different profile points, and the comprehensive gray scale ZH k associated with the corresponding profile point is confirmed by ZH k = ZH1 k × C1 + ZH2 k × C2 + ZH3 k × C3.Wherein C1, C2 and C3 are preset fixed coefficient factors, the specific values of which are determined by the operator according to experience, and C1+C2+C3=1, the gray value associated with the edge profile of the overlapping region is adjusted again according to the comprehensive gray associated with the corresponding profile point, and the adjusted and optimized edge profile is obtained.

[0050] The same point is processed by using the same processing method, and the gray value associated with the same point on the visible light image is recorded as ZZ1 i The gray value associated with the same point on the near-infrared image is recorded as ZZ2 i The gray value associated with the same point on the thermal infrared image is recorded as ZZ3 i Wherein i represents different same points, and XD i =ZZ1 i ×A1+ZZ2 i ×A2+ZZ3 i ×A3 confirm the comprehensive gray associated with the corresponding same point, wherein A1, A2 and A3 are preset fixed coefficient factors, and A1+A2+A3=0.8, in order to make the broken area more obvious, the value of the weight factor is limited here, so that the gray value of the corresponding broken area is smaller, highlighting its broken feature, and the gray value associated with the same point in the overlapping region is adjusted again according to the comprehensive gray associated with the same point, and the adjusted and optimized overlapping region is obtained.

[0051] Specifically, in the actual adjustment and optimization process, each pixel point is associated with different gray features, and there are same points between pixel points. In order to effectively highlight the characteristics of multispectral image, the different gray features associated with different points in the multispectral image are comprehensively confirmed, so that the optimized and processed overlapping region can be obtained, so as to achieve the comprehensive optimization effect.

[0052] Some data in the above formula are dimensionless numerical calculations, and the contents not described in detail in the specification all belong to the prior art known to those skilled in the art.

[0053] The above examples are only used to illustrate the technical method of the present application, not to limit it. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical method of the present application.

Claims

1. A visual inspection method for pipe cracks based on multispectral feature-level fusion, characterized in that, Includes the following steps: Step 1: Use multispectral equipment to acquire pipeline images of a specified area of ​​the pipeline, and sequentially confirm the visible light image, near-infrared image and thermal infrared image associated with the specified area to generate a multispectral image set associated with the specified area of ​​the pipeline. Step 2: Based on the multispectral image set associated with the specified area of ​​the pipeline, use the Sobel algorithm to confirm the gradient pixels associated with different images in the multispectral image set, and from the confirmed gradient pixels, confirm and mark the crack area associated with the corresponding image. Step 3: Based on the cracked area confirmed in the corresponding image, verify and compare the grayscale features associated with the cracked area with the grayscale features of the surrounding area. Based on the verification results, confirm whether the cracked area is a hole area, and simultaneously mark the hole area in the corresponding image. Step 4: Based on the marked hole areas in the image, identify the overlapping areas associated with the hole areas in multiple images, and perform comprehensive optimization processing on the gradient contours and internal features associated with the overlapping areas, and then display the optimized hole areas.

2. The multispectral feature-level fusion method for visual inspection of pipe cracks according to claim 1, characterized in that, In step one, the specific method for confirming the multispectral image set is as follows: When using multispectral equipment to photograph a designated area, the image is taken based on preset wavelength characteristics to confirm the pipeline image associated with the corresponding wavelength characteristics. By integrating and processing different pipeline images associated with different wavelength features, a multispectral image set for a specified region of the corresponding pipeline is generated.

3. The multispectral feature-level fusion method for visual inspection of pipe cracks according to claim 2, characterized in that, The wavelength characteristics associated with visible light images are 400-760nm, those associated with near-infrared images are 760-1100nm, and those associated with thermal infrared images are 8-14μm.

4. The pipe crack visual inspection method based on multispectral feature-level fusion according to claim 1, characterized in that, In step two, the specific method for marking the cracked area is as follows: The pipeline images associated with the multispectral image set are converted to grayscale to confirm the grayscale images associated with the corresponding pipeline images. The Sobel algorithm is then used to confirm the vertical and horizontal gradients associated with different gray-level points within the grayscale image. Based on the vertical and horizontal gradients associated with each gray-level point, the overall gradient associated with that gray-level point is determined. And confirm whether the comprehensive gradient of the corresponding gray point satisfies: comprehensive gradient ≥ Y1, where Y1 is a preset value. If it is satisfied, then the corresponding gray point is marked as a gradient pixel point. Connect adjacent gradient pixels within the grayscale image to identify the gradient edge associated with each pixel. If the gradient edge is closed, mark the image area included by the gradient edge as a crack region. If the gradient edge is not closed, do not mark it.

5. The multispectral feature-level fusion method for visual inspection of pipe cracks according to claim 4, characterized in that, If the combined gradient is less than Y1, no labeling is performed.

6. The multispectral feature-level fusion method for visual inspection of pipe cracks according to claim 1, characterized in that, In step three, the specific method for marking the hole area within the corresponding image is as follows: Based on the confirmed crack area, the outer area of ​​the crack area is recorded as the outer region. The gray values ​​associated with different gray points in the outer region are averaged to confirm the average value associated with several gray values. The confirmed average value is recorded as the gray feature TZ1 associated with the outer region. Simultaneously, the gray values ​​associated with several gray points in the crack area are averaged to confirm the gray feature TZ2 in the crack area. If grayscale features TZ1 and TZ2 satisfy: |TZ1-TZ2|≥Y2, where Y2 is a preset value, the confirmed crack area is recorded as a hole area if the condition is met; otherwise, no marking is performed.

7. The multispectral feature-level fusion method for visual inspection of pipe cracks according to claim 1, characterized in that, In step four, the specific method for comprehensively optimizing the hole area is as follows: The hole areas associated with different images in the multispectral image set are identified. Then, the corresponding hole areas in multiple images are overlapped to confirm whether there are overlapping areas. If there are overlapping areas, the overlapping areas in the hole area are marked. If there are no overlapping areas, the corresponding images are displayed directly. The displayed images are simultaneously marked with hole areas and crack areas. Based on the overlapping regions marked in the multispectral image set, the edge contours of the overlapping regions are marked, and the gray values ​​associated with each contour point on the edge contour are simultaneously confirmed. The gray value associated with this contour point in the visible light image is denoted as ZH1. k The gray value associated with this contour point on the near-infrared image is denoted as ZH2. k The gray value associated with this contour point on the thermal infrared image is denoted as ZH3. k Where k represents different contour points, using: ZH k =ZH1 k ×C1+ZH2 k ×C2+ZH3 k ×C3, confirm the overall grayscale ZH associated with the corresponding contour point. k C1, C2 and C3 are preset fixed coefficient factors, and C1+C2+C3=1. Based on the comprehensive gray value associated with the corresponding contour point, the gray value associated with the edge contour of the overlapping area is readjusted to obtain the adjusted and optimized edge contour.

8. The multispectral feature-level fusion method for visual inspection of pipe cracks according to claim 7, characterized in that, In step four, the specific methods for comprehensively optimizing the hole area also include: Then, the same processing method is applied to the corresponding points in the overlapping area, and the gray values ​​associated with the corresponding points in the visible light image are simultaneously recorded as ZZ1. i The gray value associated with this similarity in the near-infrared image is denoted as ZZ2. i The grayscale value associated with this similarity in the thermal infrared image is denoted as ZZ3. i Where i represents different similarities, using: XD i =ZZ1 i ×A1+ZZ2 i ×A2+ZZ3 i ×A3 confirms the comprehensive grayscale associated with the corresponding identical points, where A1, A2, and A3 are all preset fixed coefficient factors, and A1+A2+A3=0.

8. Based on the comprehensive grayscale associated with the identical points, the grayscale values ​​associated with the identical points in the overlapping area are readjusted to obtain the adjusted and optimized overlapping area.

Images