A solar panel surface shelter identification and detection method based on an HSV color space model

Abstract

The application discloses a solar power panel surface shelter identification and detection method based on an HSV color space model, which comprises the following steps: step 1: for the solar power panel photovoltaic power station inspection requirement, an unmanned aerial vehicle is used to collect images of the solar power panel surface, and under the premise of meeting the unmanned aerial vehicle obstacle avoidance function and the image resolution, the distance between the unmanned aerial vehicle and the solar power panel is maintained at 2-3 meters when the unmanned aerial vehicle detects the solar power panel; for the complex environment of the solar power station, a four-rotor unmanned aerial vehicle is used to collect images, the solar power panel image collection speed and the image quality are improved, and based on a digital image processing technology, an improved three-component value method based on the HSV color space model is provided, green vegetation surface shelter defects in the solar power panel images under the complex working environment are segmented and extracted, and the efficiency of the defect features and the background in the segmented images is improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of solar power generation panels, and particularly relates to a solar power generation panel surface shelter identification and detection method based on an HSV color space model. BACKGROUND

[0002] With the continuous development of society and technology, the energy demand of countries is also increasing. The contradiction between ecological environment and energy acquisition is also increasing, therefore, the use of new energy and the development of related technologies are more and more valued by society. As a new energy industry developed by the country in recent years, the solar photovoltaic industry not only brings a large amount of energy to the country, but also promotes the transformation of the country's energy structure to clean and low carbon. With the increase of solar power stations, the operation and maintenance pressure of solar panels is increasing. The traditional photovoltaic power station operation and maintenance mainly relies on manual inspection, which has high cost, low efficiency and great safety hazards.

[0003] At present, commonly used detection methods include manual visual detection method, electroluminescence method and image processing. Among them, the manual visual detection method is that the detection personnel of the photovoltaic power station determines whether there is a defect on the surface of the solar cell panel, such as whether there is green vegetation growth to block the solar power generation panel, by manually checking one by one. However, for a slightly larger cell panel group, this detection method will consume a lot of labor and financial resources and there will be missed detection caused by eye fatigue, resulting in inaccurate detection results, which is not suitable for large photovoltaic power station detection. The method based on the electroluminescence method is to form a cell potential difference by applying an external voltage to the two ends of the cell to be detected, so that the electrons collide with each other to cause energy level transition to produce radiation light, thereby obtaining a detection image. The image processing method is to combine computer vision and image processing technology, and to determine the green plants on the surface of the solar power generation panel in the image by processing and recognizing the image by the computer.

[0004] The artificial visual detection method relies on human eyes to make judgments, is easily affected by fatigue and emotions, and is low in efficiency; the electroluminescence detection method uses a near-infrared camera to acquire images, and can detect internal defects of the battery piece, but in the process of exciting the battery piece to emit light, the battery piece needs to be powered, which is easy to cause secondary damage to the solar cell piece, and the detection method needs higher power consumption; the photoluminescence detection method and the phase-locked thermal map method also use a near-infrared camera to acquire images, have the advantage of non-contact compared with electroluminescence, but still have certain influence on the inside of the battery piece, and have insufficient robustness. Through the above analysis, it can be found that any detection method has its own advantages and application scenarios, and also has certain deficiencies for certain application requirements. The existing solar power panel surface green plant image segmentation algorithm has a long running time and cannot meet the real-time requirements of the system, and there are few defect recognition algorithms, especially for solar power panel surface green vegetation shielding defect image recognition and segmentation algorithms in solar power stations. SUMMARY

[0005] To solve the problems raised in the background art, the present application provides a solar power panel surface shielding object recognition and detection method based on an HSV color space model.

[0006] To achieve the above object, the present application provides the following technical scheme: a solar power panel surface shielding object recognition and detection method based on an HSV color space model, comprising the following steps:

[0007] Step 1: For the solar power panel photovoltaic power station inspection requirements, a UAV is used to collect images of the solar power panel surface, and under the premise of meeting the UAV obstacle avoidance function and image resolution, the distance between the UAV and the solar power panel is maintained at 2-3 meters during the detection of the solar power panel by the UAV;

[0008] Step 2: The solar power panel images collected by the UAV device are preprocessed;

[0009] Step 3: The green plants in the solar power panel images after image preprocessing are subjected to image segmentation operation;

[0010] Step 4: The size of the green vegetation in the image and the area of the minimum circumscribed rectangle are calculated to analyze the characteristics of the leaf green vegetation in the image and the size and shape of the green vegetation on the solar power panel surface.

[0011] In a preferred embodiment of the present application, in step 1, the solar power panel image collection module uses a quadcopter UAV to collect images, and the UAV is subjected to system debugging and path planning before use.

[0012] In a preferred embodiment of the present application, in step 2, the solar panel image preprocessing module further comprises the following steps:

[0013] Step 2.1: Perform motion deblurring operation on the solar panel image, and use Wiener filtering method to remove image blurring caused by the unmanned aerial vehicle itself or external environment during image acquisition;

[0014] Step 2.2: For the salt and pepper noise and Gaussian noise generated by the unmanned aerial vehicle during image acquisition, an improved adaptive median filtering method is used for image denoising, which can improve the algorithm running speed, reduce the number of pixel points participating in median sorting, and reduce the running time of the algorithm;

[0015] Step 2.3: Use histogram equalization processing to enhance the denoised solar panel image;

[0016] Step 24: For the case that the camera plane is not parallel to the object plane during image acquisition, distortion occurs during the projection of the target object, so perspective correction operation is performed on the image f(x, y), and the perspective corrected image is f'(x', y'), and the transformation model is,

[0017]

[0018] The transformation formula after dividing by z' is:

[0019]

[0020] Wherein, When any four points in the image are known, any unknown number can be solved.

[0021] In a preferred embodiment of the present application, in step 3, the solar panel image segmentation module uses an improved three-component value method based on the HSV color space model to segment the green vegetation surface shielding defects in the solar panel image under complex working environment.

[0022] For the blue solar panel and green vegetation in the image in the present application, the HSV green and blue color components are selected as the calculation points, and the normalized value range is as follows:

[0023]

[0024]

[0025] When there are multiple solar panels and green vegetation defects in the collected image, the image feature description method is improved to obtain the following segmentation principles:

[0026]

[0027] wherein R, G, B are three channels in the color original image color space, and the expression of k is:

[0028]

[0029] wherein: B i,j is the gray value of the B component in the RGB image color space at point (i, j); G max is the gray value corresponding to the peak point of the G component histogram in the RGB image color space; B max is the gray value corresponding to the peak point of the B component histogram in the RGB image color space; G max and B max are the gray values corresponding to the peak values of the image screening histogram, and the segmentation results according to the above segmentation principle are as follows:

[0030]

[0031] In a preferred embodiment of the present application, the solar power panel image segmentation module in step 3 has the following specific segmentation steps:

[0032] Step 31: Convert f(x, y) after image preprocessing to the HSV color space, and mark the green region in the solar panel image according to the value range of the HSV green component;

[0033] Step 32: Set the H(i, j), S(i, j), and V(i, j) three channels of the pixel points in the marked green region to 0, and set the processed image as f'(x, y);

[0034] Step 33: Extract the H single-channel image in the HSV image of f to process, set the target value range in the H channel component as [0.43-0.689], so as to obtain a binary image, and perform morphological processing by using morphological operation to improve the shape of the feature outline region in the image, and the image result is f''(x, y).

[0035] In a preferred embodiment of the present application, the solar power panel image recognition module in step 4 is based on the image f''(x, y) after morphological operation, and the green vegetation shielding area and the minimum circumscribed rectangle area of the green vegetation are calculated, so as to know the distribution type and distribution area of the green vegetation in the image.

[0036] In a preferred embodiment of the present application, for the shielding region after segmentation, the pixel value of the black region is 0, the pixel value of the white region is 1, the shielding region in the image is A, and the green vegetation shielding area is S:

[0037]

[0038] After the green plant sheltering feature is segmented, the connected region is contained in a minimum circumscribed rectangle, wherein the four vertices are (x1, y1), (x2, y2), (x3, y3), (x4, y4), and the minimum circumscribed area of the sheltering region can be calculated:

[0039] S w == H x L = (y2-y1) x (x3-x1)

[0040] The distribution type of the sheltering part is determined by comparing the length-width ratio of the sheltering region, and the length-width ratio of the sheltering area is:

[0041]

[0042] The distribution type and distribution area of the green vegetation on the solar panel in the image are obtained through the above calculation.

[0043] The present application solves the defects in the background art, and has the following beneficial effects:

[0044] 1. The solar panel surface sheltering object recognition and detection method based on the HSV color space model, for the complex environment of the solar power station, adopts a quadcopter unmanned aerial vehicle to collect images, improves the solar panel image collection speed and image quality, and based on the digital image processing technology, proposes an improved three-component value method based on the HSV color space model, extracts the green vegetation surface sheltering defects in the solar panel image under the complex working environment, and improves the efficiency of the defect features and background in the segmented image. Select 5 different solar panels for image collection, and statistically analyze the recognition speed and recognition accuracy. Compared with the existing traditional manual visual detection method, the effective effect is shown in the following table:

[0045] Recognition speed improvement Recognition accuracy improvement Solar panel a 43％ 61％ Solar panel b 41％ 58％ Solar panel c 40％ 60％ Solar panel d 43％ 59％ Solar panel e 44％ 60％ BRIEF DESCRIPTION OF DRAWINGS

[0046] The accompanying drawings are included to provide a further understanding of the present application, and constitute a part of the specification, and are used to explain the present application together with embodiments of the present application, and do not constitute a limitation of the present application. In the drawings:

[0047] Figure 1 is a flowchart of the present application; DETAILED DESCRIPTION

[0048] The present application will now be further described in detail in conjunction with the drawings and embodiments, and these drawings are all simplified schematic diagrams, and only schematically show the basic structure of the present application, and therefore only show the structures related to the present application.

[0049] Example 1

[0050] A solar panel surface shelter identification detection method based on an HSV color space model, comprising the following steps:

[0051] Step 1: According to the solar panel photovoltaic power station inspection requirements, the unmanned aerial vehicle is used to collect the solar panel surface image, and under the premise of meeting the unmanned aerial vehicle obstacle avoidance function and image resolution, the distance between the unmanned aerial vehicle and the solar panel is maintained at 2-3 meters during the detection of the solar panel by the unmanned aerial vehicle;

[0052] Step 2: The solar panel image collected by the unmanned aerial vehicle device is preprocessed;

[0053] Step 3: The image segmentation operation is performed on the green plants in the solar panel image after image preprocessing;

[0054] Step 4: The size of the green vegetation in the image and the area of the minimum circumscribed rectangle are calculated to analyze the characteristics of the leaf green vegetation in the image and the size and shape of the green vegetation on the surface of the solar panel.

[0055] In step 1, the solar panel image acquisition module uses a quadcopter unmanned aerial vehicle to collect images, and the unmanned aerial vehicle is debugged and path planned before use.

[0056] In step 2, the solar panel image preprocessing module further comprises the following steps:

[0057] Step 2.1: The solar panel image is subjected to a motion deblurring operation, and a Wiener filtering method is used to remove the image blurring caused by the unmanned aerial vehicle itself or external environment during image collection;

[0058] Step 2.2: For the salt and pepper noise and Gaussian noise generated by the unmanned aerial vehicle during image collection, an improved adaptive median filtering method is used for image denoising, which can improve the algorithm running speed, reduce the number of pixel points participating in median sorting, and reduce the running time of the algorithm;

[0059] Step 2.3: The histogram equalization processing is used to enhance the solar panel image after denoising;

[0060] Step 24: For the case that the camera plane is not parallel to the object plane during image collection, distortion occurs during the projection of the target object, so the perspective correction operation is performed on the image f(x,y), and the perspective corrected image is f'(x',y'), and the transformation model is,

[0061]

[0062] The transformation formula after dividing by z' is:

[0063]

[0064] wherein, When any four points in the image are known, any unknown can be solved.

[0065] In step 3, the solar panel image segmentation module adopts an improved three-component value method based on the HSV color space model to segment the green vegetation surface shielding defects in the solar panel image under complex working environment.

[0066] For the blue solar panel and green vegetation in the image, the HSV green and blue two color components are selected as the calculation points, and the value range after normalization is as follows:

[0067] Component H S V Green [0.19～0.50] [0.71～1] [0.18～1] Blue [0.56～0.71] [0.36～1] [0.36～1]

[0068] When there are multiple solar panels and green vegetation defects in the collected image, the image feature description method is improved to obtain the following segmentation principles:

[0069]

[0070] In the formula, R, G and B are three channels in the color space of the original color image, and the expression of k is:

[0071]

[0072] In the formula: B i,j is the gray value of the B component in the RGB image color space at point (i,j); G max is the gray value corresponding to the peak point of the G component histogram in the RGB image color space; B max is the gray value corresponding to the peak point of the B component histogram in the RGB image color space; G max and B max are the gray values corresponding to the peak values of the image screening histogram, and according to the above segmentation principles, the segmentation results are as follows:

[0073]

[0074] The specific segmentation steps of the solar panel image segmentation module in step 3 are as follows:

[0075] Step 31: Convert f(x,y) after image preprocessing to HSV color space, and mark the green area in the solar panel image according to the value range of the HSV green component;

[0076] Step 32: Set the H(i,j), S(i,j) and V(i,j) three channels of the pixel points in the marked green area to 0, and set the processed image as f'(x,y);

[0077] Step 33: Extract the H single channel image in the HSV image of f for processing, set the target value range in the H channel component as [0.43-0.689], so as to obtain the binary image, and adopt morphological operation for morphological processing, improve the shape of the feature outline area in the image, and the image result is f"(x, y).

[0078] The solar panel image recognition module of step 4 calculates the green vegetation shielding area and the minimum circumscribed rectangle area based on the image f"(x, y) processed by the morphological technique, so as to know the distribution type and distribution area of the green vegetation in the image.

[0079] For the shielding area after segmentation, the pixel value of the black area is 0, the pixel value of the white area is 1, the shielding area in the image is A, and the green vegetation shielding area is S:

[0080]

[0081] After the segmentation of the green plant shielding feature, the connected region is completely contained in a minimum circumscribed rectangle, and the four vertices are (x1, y1), (x2, y2), (x3, y3), and (x4, y4). The minimum circumscribed area of the shielding area can be calculated:

[0082] S w == H x L = (y2-y1) x (x3-x1)

[0083] The distribution type of the shielding part is determined by comparing the length-width ratio of the shielding area, and the length-width ratio of the shielding area is:

[0084]

[0085] Through the above calculation, the distribution type and distribution area of the green vegetation on the solar panel in the image are obtained.

[0086] The solar panel surface shielding detection method based on the HSV color space model uses a quadrotor unmanned aerial vehicle for image acquisition in view of the complex environment of the solar power station, improves the solar panel image acquisition speed and image quality, and based on the digital image processing technology, proposes an improved method for obtaining three components of the HSV color space model, extracts the green vegetation surface shielding defects in the solar panel image in the complex working environment, and improves the efficiency of the defect features and background in the segmented image. Select 5 different solar panels for image acquisition, and statistically analyze the recognition speed and recognition accuracy. Compared with the existing traditional manual visual detection method, the effective effect is shown in the following table:

[0087] Recognition speed improvement Recognition accuracy improvement Solar panel a 43％ 61％ Solar panel b 41％ 58％ Solar panel c 40％ 60％ Solar panel d 43％ 59％ Solar panel e 44％ 60％

[0088] Embodiment Two

[0089] The solar panel surface green vegetation detection method based on image processing of the embodiment comprises a road solar panel image acquisition module 1, a solar panel image preprocessing module 2, a solar panel image segmentation module 3, and a solar panel image recognition module 4. The solar panel image acquisition module 1 acquires images of the solar panel, the solar panel image preprocessing module 2 performs preprocessing operations on the images, the solar panel image segmentation module 3 segments and extracts green plant features in the solar panel, and the solar panel image recognition module 4 identifies and classifies the types of the green plant features in the solar panel.

[0090] The solar panel image acquisition module 1 acquires images of the solar panel using a quadcopter unmanned aerial vehicle, and the original image is denoted as g1(x,y).

[0091] The solar panel image preprocessing module 2 performs preprocessing on the acquired solar panel images, and the specific steps are as follows:

[0092] The Wiener filtering method is used to perform motion deblurring on the solar panel image g1(x,y), and the estimated value g2(x,y) of the defect original image g1(x,y) is found, so that the mean square error of the two is minimized:

[0093] e 2 (x)=|f2(x,y)-f3(x,y) 2

[0094] When e 2 (x) is minimized in the frequency domain, g2(x,y) can be calculated.

[0095] Noise analysis is performed on the image g2(x,y) that has been subjected to motion deblurring processing. Due to the harsh and complex working environment of the solar power station, a large amount of Gaussian noise and salt and pepper noise will be generated in the collected images. An improved adaptive median filtering method is used for image denoising to obtain the image g3(x,y).

[0096] Image enhancement is performed on the denoised solar panel image, and histogram equalization processing is used to achieve the effect of image enhancement to obtain the image g4(x,y).

[0097] Perspective correction is performed on the image g4(x,y) after the image enhancement operation. Distortion occurs during the projection of the target object, so perspective correction is performed on the image. The perspective-corrected image is g5(x',y'), and the transformation model is:

[0098]

[0099] The transformation formula after dividing by z' is:

[0100]

[0101] wherein, When the four points (x1, y1), (x2, y2), (x3, y3), (x4, y4) in the image are known, the remaining unknowns can be solved.

[0102] An improved three-component value method based on the HSV color space model is used to segment the green vegetation surface shielding defects in the solar panel image under complex working environment. In view of the characteristics of the blue solar panel and green vegetation in the image in the application, the HSV green and blue two color components are selected as the calculation points, and the value range after normalization is as follows:

[0103] Table 1: Value range of each HSV component

[0104] Component H S V Green [0.19～0.50] [0.71～1] [0.18～1] Blue [0.56～0.71] [0.36～1] [0.36～1]

[0105] Convert g5(x', y') which has been pre-processed into the HSV color space, and mark the green region in the solar panel image according to the value range of the HSV green component in Table 1;

[0106] Set the H(i, j), S(i, j), V(i, j) three channels of the pixel points in the marked green region to 0, and the processed image is g6(x, y);

[0107] Extract the H single-channel image in the HSV image of g6(x, y) for processing, and set the target value range in the H channel component to [0.43-0.689], so as to obtain a binary image, and perform morphological processing by using morphological operation, so as to improve the shape of the feature outline region in the image, and the image result is g7(x, y);

[0108] For the shielding region after segmentation, the pixel value of the black region is 0, the pixel value of the white region is 1, the shielding region in the image is A, and the green vegetation shielding area is S:

[0109]

[0110] The connected region of the green plant shielding feature after segmentation is completely contained in a minimum circumscribed rectangle, wherein the four vertices are (x1, y1), (x2, y2), (x3, y3), (x4, y4), and the minimum circumscribed area of the shielding region can be calculated:

[0111] S w == H x L = (y2-y1) x (x3-x1)

[0112] The distribution type of the occlusion part is determined by comparing the length-width ratio of the occlusion area, and the length-width ratio of the occlusion area is:

[0113]

[0114] The distribution type and the distribution area of the green vegetation on the solar panel in the image are obtained through the above calculation.

[0115] In the description of the present application, it should be understood that the terms "first", "second" are only for the purpose of description, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include one or more of the features. In the description of the present application, the meaning of "a plurality of" is two or more, unless otherwise specifically limited.

[0116] In the description of the present application, the description of the terms "one embodiment", "some embodiments", "embodiments", "example", "specific example" or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In the present application, the illustrative description of the above terms does not necessarily refer to the same embodiment or example.

[0117] Although the embodiments of the present application have been shown and described above, it should be understood that the above embodiments are exemplary and cannot be understood as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the above embodiments within the scope of the present application.

Claims

1. A method for detecting solar panel surface shading object recognition based on HSV color space model, characterized in that, It comprises the following steps: Step 1: for the solar panel photovoltaic power station inspection requirements, the unmanned aerial vehicle is used to collect the image of the solar panel surface, and under the premise of meeting the unmanned aerial vehicle obstacle avoidance function and the image resolution, the distance between the unmanned aerial vehicle and the solar panel is maintained at 2-3 meters during the detection of the solar panel by the unmanned aerial vehicle; Step 2: the solar panel image collected by the unmanned aerial vehicle device is preprocessed; Step 3: the image segmentation operation is performed on the green plants in the solar panel image after image preprocessing; Step 4: the size and shape of the green plants on the solar panel surface are analyzed by calculating the area of the green plants in the image and the minimum circumscribed rectangle area; In step 3, the solar panel image segmentation module uses an improved three-component value method based on the HSV color space model to segment the green plant surface defects in the solar panel image under complex working environment; When there are multiple solar panels and green plant defects in the collected image, the image feature description method is improved to obtain the following segmentation principles: ； Where R, G, B are three channels in the color space of the original color image, and the expression of k is: ； In the formula: is the gray value of the B component in the RGB image color space at the point ; is the gray value corresponding to the peak point of the G component histogram in the RGB image color space; is the gray value corresponding to the peak point of the B component histogram in the RGB image color space; and is the gray value corresponding to the peak of the image screening histogram, and the segmentation result according to the above segmentation principle is as follows: 。 2.The method of claim 1, wherein the method is characterized by: In step 1, the solar panel image acquisition module uses a quadcopter unmanned aerial vehicle to collect images, and the unmanned aerial vehicle is debugged and path planned before use.

3. The method for identifying and detecting obstructions on the surface of a solar panel based on the HSV color space model according to claim 1, characterized in that: In step 2, the solar panel image preprocessing module further comprises the following steps: Step 2.1: the solar panel image is subjected to motion deblurring operation, and the Wiener filter method is used to remove the image blurring caused by the unmanned aerial vehicle itself or external environment during image collection; Step 2.2: for the salt and pepper noise and Gaussian noise generated by the unmanned aerial vehicle during image collection, an improved adaptive median filter method is used for image denoising, which can improve the algorithm running speed, reduce the number of pixel points participating in median sorting, and reduce the running time of the algorithm; Step 2.3: histogram equalization processing is used to enhance the solar panel image after denoising; Step 24: In the case that the camera plane is not parallel to the object plane during image acquisition, distortion occurs in the target object projection process. Therefore, perspective correction operation is performed on the image , and the perspective-corrected image is , and the transformation model is ； Divide by 2 The transformation formula after that is: ； wherein , When any 4 points in the image are known, any unknown can be solved.

4. The method for detecting the shading object on the surface of the solar panel based on the HSV color space model according to claim 3, characterized in that: The solar panel image segmentation module in step 3 has the following specific segmentation steps: Step 31: The image pre-processed in step 30 is converted into HSV color space and the green region in the solar panel image is marked according to the value range of the HSV green component. converted into HSV color space and the green region in the solar panel image is marked according to the value range of the HSV green component. Step 32: set the three channel classifications of the marked green region pixel points to 0, and set the processed image as , , three channel classifications to 0, and set the processed image as ; Step 33: Extract the H single channel image in the HSV image of f for processing, set the target value range in the H channel component as [0.43~0.689], so as to obtain a binary image, and perform morphological operation for morphological processing, improve the shape of the feature outline region in the image, and the image result is .

5. The method for detecting the shading object on the surface of the solar panel based on the HSV color space model according to claim 1, characterized in that: The solar power panel image recognition module described in step 4 is based on the image after morphological technique operation The green vegetation shielding area and the minimum circumscribed rectangle area of the green vegetation are calculated, so as to know the distribution type and distribution area of the green vegetation in the image.

6. The method for detecting the shading object on the surface of the solar panel based on the HSV color space model according to claim 5, characterized in that: For the segmentation of the blocked area, the pixel value of the black area is 0, the pixel value of the white area is 1, the blocked area in the image is A, and the green vegetation blocked area is : ； The connected region of the green plant blocking feature after segmentation is contained in a minimum circumscribed rectangle, wherein four vertices are: , , , , the minimum circumscribed area of the blocking region can be calculated: ； The distribution type of the occluded part is determined by comparing the length-width ratio of the occluded area, and the length-width ratio of the occluded area is: ； The distribution type and distribution area of the green plants on the solar panel in the image are obtained through the above calculation.

Images