Method and system for identifying accumulated dirt of photovoltaic cleaning robot based on machine vision

Abstract

The invention discloses a photovoltaic cleaning robot accumulated dirt identification method and system based on machine vision, relates to the technical field of photovoltaic accumulated dirt identification, and solves the technical problem that the actual accumulated dirt degree and distribution condition of a photovoltaic panel cannot be accurately judged. According to the method, the photovoltaic panel image is divided into nine equal parts, the gray value of each divided area is calculated and compared with the gray value of the normal image to screen the abnormal area, so that the position and the range of dirt accumulation can be accurately determined, invalid cleaning operation on the area without dirt accumulation is avoided, the pertinence and the effectiveness of cleaning are improved, and the cleaning efficiency is improved. Analyzing by combining the dirt retention color and texture of the abnormal region, obtaining an HSV value, comparing the HSV value with a preset matching interval to obtain a color matching result, extracting texture features by using a gray-level co-occurrence matrix, matching the texture features with a preset texture matching interval to generate a pre-selected result, and further matching by combining the light reflectivity to obtain a pre-selected result; the dirt accumulation type can be more accurately judged by integrating multiple features, and proper cleaning strategies and tools can be selected.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic dirt accumulation identification, and specifically to a method and system for identifying dirt accumulation of a photovoltaic cleaning robot based on machine vision. Background Art

[0002] With the rapid development of the photovoltaic industry, the scale of photovoltaic power stations has been continuously expanding. During the long-term outdoor operation of photovoltaic panels, various dirt will inevitably accumulate on the surface, such as dust, bird droppings, oil stains, etc. These accumulated dirt will seriously affect the photoelectric conversion efficiency of the photovoltaic panels, thereby reducing the power generation of the photovoltaic power generation system.

[0003] According to the publication number CN118840617A, a method for identifying the stain posture for intelligent cleaning of a photovoltaic panel is provided, including the following steps: Step 1, deploy a mobile operation robot inside the photovoltaic power station, and the robot carries a camera to scan the photovoltaic panel to obtain a stain image; Step 2, accurately identify the area where the stain is located in the stain image obtained in Step 1 and classify the stain; Step 3, perform edge detection on the image of the area where the stain is located identified in Step 2 through an adaptive edge detection algorithm to obtain the edge of the stain; Step 4, use a tiltable minimum circumscribed rectangle to draw the outer bounding contour of the stain for the image obtained after edge detection, and the maximum length, maximum width, and deflection angle are the stain postures.

[0004] However, when some existing dirt accumulation identification methods are used, they cannot accurately judge the actual dirt accumulation degree and distribution of the photovoltaic panel, and it is difficult to effectively clean the photovoltaic panel at the appropriate time, thus affecting the overall power generation efficiency and economic benefits of the photovoltaic power generation system. Summary of the Invention

[0005] Aiming at the deficiencies of the existing technology, the present invention provides a method and system for identifying dirt accumulation of a photovoltaic cleaning robot based on machine vision, which solves the problem of being unable to accurately judge the actual dirt accumulation degree and distribution of the photovoltaic panel.

[0006] To achieve the above objectives, the present invention is realized through the following technical solutions: A method for identifying dirt accumulation of a photovoltaic cleaning robot based on machine vision, the method specifically includes the following steps:

[0007] Step 1, obtain the basic information of the photovoltaic panel, and the basic information includes the area and shape of the photovoltaic panel;

[0008] Step 2, perform regional segmentation on the photovoltaic panel according to the obtained basic information of the photovoltaic panel to obtain multiple groups of segmented regions, and at the same time calculate the gray values of the segmented regions, and screen out abnormal regions according to the gray values;

[0009] Step 3: Analyze the obtained abnormal regions, judge the relevance of the abnormal regions by analyzing the adjacent conditions of the abnormal regions, and generate a relevance result;

[0010] Step 4: Analyze the associated region results in the relevance result, analyze the pollution accumulation characteristics corresponding to the abnormal regions by combining historical data, and perform comprehensive identification by combining the light reflectivity of the abnormal regions to obtain identification information.

[0011] As a further solution of the present invention, the specific method for calculating the gray value of the segmented region in Step 2 is as follows:

[0012] Obtain the photovoltaic panel and denote it as the target object. At the same time, obtain the image information of the target object. Then, segment the obtained image to obtain a segmented image, and label it as i, where i = 1, 2, …, j, and j represents the number label of the segmented images. At the same time, calculate the gray value of the segmented image i and denote it as Hi, and the specific calculation method is as follows:

[0013] Obtain the segmented image i. At the same time, convert the segmented image i to obtain a gray-scale segmented image i. Then, obtain the number of pixel points corresponding to the segmented image i and denote it as Ci, and obtain the intensity values of the red, green, and blue channels corresponding to a single pixel point, which are denoted as R, G, and B respectively. Then, substitute the obtained intensity values into the formula Gray = 0.299R + 0.587G + 0.114B to calculate the gray value Gray corresponding to a single pixel point. By analogy, sum the gray values Gray of all pixel points corresponding to the gray-scale segmented image i and calculate the average value, and at the same time, use the calculated average value as the gray value of the gray-scale segmented image i.

[0014] As a further solution of the present invention, the specific method for screening the abnormal regions according to the gray value in Step 2 is as follows:

[0015] Calculate the gray values of all gray-scale segmented images i. Then, screen the gray-scale segmented images i according to the gray values, obtain the gray values corresponding to the normal images, and calculate the gray difference between the gray value of the gray-scale segmented image i and the gray value of the normal image. At the same time, compare the obtained gray difference with a preset value;

[0016] If the gray difference is greater than the preset value, classify the corresponding gray-scale segmented image i as an abnormal image, and denote the corresponding region as an abnormal region. If the gray difference is less than the preset value, classify the corresponding gray-scale segmented image i as a normal image, and denote the corresponding region as a normal region.

[0017] As a further solution of the present invention, the specific method for analyzing the abnormal regions in Step 3 is as follows:

[0018] Obtain all abnormal areas and label them as n, where n = 1, 2, …, m, and m represents the number label of abnormal areas. Then, classify them into associated areas and non-associated areas according to whether the abnormal areas are adjacent, generate the results of associated areas and non-associated areas, and analyze the generated results of associated areas and non-associated areas respectively.

[0019] As a further solution of the present invention, the specific method for analyzing the results of associated areas in step four is:

[0020] Obtain the associated areas and the corresponding abnormal areas denoted as n. Then, obtain the fouling characteristics corresponding to the abnormal area n, where the fouling characteristics include fouling color and fouling texture, and analyze them by combining the fouling color and fouling texture;

[0021] Obtain the HSV value corresponding to the abnormal area n. At the same time, match and analyze the HSV value with a preset matching interval to obtain a color matching result. Then, obtain the fouling texture corresponding to the abnormal area n, extract the texture features of the image through a gray-level co-occurrence matrix, and match the obtained texture features with a preset texture matching interval, and generate a preliminary selection result at the same time. Then, analyze the preliminary selection result.

[0022] As a further solution of the present invention, the specific method for analyzing the preliminary selection result in step four is:

[0023] Obtain all the preliminary selection results and denote them as o, where o = 1, 2, …, p, and p represents the number label of the preliminary selection results. Then, obtain the light reflectivity of the abnormal area n denoted as Fn. At the same time, match the obtained light reflectivity Fn with the light reflectivity interval Fo corresponding to the preliminary selection result o to obtain matching information, and perform secondary analysis on the matching information.

[0024] As a further solution of the present invention, the specific method for performing secondary analysis on the matching information in step four is:

[0025] Obtain the light reflectivity of the photovoltaic panel under normal conditions denoted as F 0 , and at the same time, obtain the area of the abnormal area denoted as Sn. Determine the relationship formula according to the relationship between the change of light reflectivity and the fouling area to obtain where θ is a preset proportional coefficient, and calculate the fouling area Sn corresponding to the abnormal area according to the relationship formula. By analogy, calculate the fouling areas of all abnormal areas n, sum up the fouling areas to obtain the overall fouling area of the photovoltaic panel denoted as Sz, and generate identification information at the same time.

[0026] The fouling recognition system of a photovoltaic cleaning robot based on machine vision includes:

[0027] An information acquisition unit is used to obtain the basic information of the photovoltaic panel and transmit the obtained basic information of the photovoltaic panel to the information analysis unit;

[0028] An information analysis unit is used to analyze the fouling condition of the photovoltaic panel according to the obtained basic information, divide the photovoltaic panel into multiple groups of segmented areas, calculate the grayscale values of the segmented areas, screen out abnormal areas according to the grayscale values, analyze the obtained abnormal areas, judge the relevance of the abnormal areas by analyzing the adjacent situation of the abnormal areas, and generate a relevance result. Analyze the associated area results in the relevance result, analyze the fouling characteristics corresponding to the abnormal areas by combining historical data, and perform comprehensive identification by combining the light reflectivity of the abnormal areas to obtain identification information, and at the same time transmit the obtained identification information to the information output unit;

[0029] An information output unit is used to display the obtained identification information to the corresponding operator through a display device.

[0030] The present invention provides a fouling identification method and system for a photovoltaic cleaning robot based on machine vision. Compared with the prior art, it has the following beneficial effects:

[0031] The present invention divides the photovoltaic panel image into nine equal parts and calculates the grayscale values of each segmented area, and screens out abnormal areas by comparing with the grayscale values of normal images, which can accurately determine the location and scope of fouling. Compared with the traditional extensive cleaning method, it avoids the ineffective cleaning operation of non-fouled areas, improves the pertinence and effectiveness of cleaning, analyzes the fouling color and texture of abnormal areas, obtains the HSV value and compares it with the preset matching interval to obtain a color matching result, extracts texture features using the gray-level co-occurrence matrix and matches them with the preset texture matching interval to generate a preliminary selection result, and further matches by combining the light reflectivity. Combining multiple aspects of features can more accurately judge the fouling type, which helps to select appropriate cleaning strategies and tools;

[0032] Calculate the fouling area based on the light reflectivity of the abnormal area, and then obtain the overall fouling area of the photovoltaic panel. According to the fouling area and characteristics, intelligent judgment can be made for subsequent cleaning, specifically manifested as whether cleaning is required and determining the cleaning priority, thus avoiding the blindness of fixed-cycle cleaning, reducing unnecessary cleaning times, and reducing the loss and energy consumption of cleaning equipment. Description of the Drawings

[0033] Figure 1 It is a flow chart of the method steps of the present invention;

[0034] Figure 2 It is a diagram of the distribution of abnormal areas of the present invention;

[0035] Figure 3 It is a block diagram of the system principle of the present invention. Specific Embodiments

[0036] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without making creative efforts shall fall within the protection scope of the present invention.

[0037] Embodiment 1. Please refer to Figure 1 、 Figure 2 and Figure 3 . This application provides a method for identifying dirt accumulation of a photovoltaic cleaning robot based on machine vision. The method specifically includes the following steps:

[0038] Step 1: Obtain the basic information of the photovoltaic panel, and the basic information includes the area and shape of the photovoltaic panel.

[0039] Step 2: Perform region segmentation on the photovoltaic panel according to the obtained basic information of the photovoltaic panel to obtain multiple groups of segmented regions. At the same time, calculate the gray values of the segmented regions, and screen out abnormal regions according to the gray values.

[0040] Obtain the photovoltaic panel and denote it as the target object. At the same time, obtain the image information of the target object, and the image information here is obtained through a high-definition camera. Then, segment the obtained image to obtain a segmented image. Here, the image of the target object is evenly divided into nine equal parts and labeled as i, where i = 1, 2,..., j, and j represents the number label of the segmented image. At the same time, calculate the gray value of the segmented image i and denote it as Hi, and the specific calculation method is as follows:

[0041] Obtain the segmented image i. At the same time, convert the segmented image i to obtain a gray-scale segmented image i. Then, obtain the number of pixel points corresponding to the segmented image i and denote it as Ci, and obtain the intensity values of the red, green, and blue channels corresponding to a single pixel point, denoted as R, G, and B respectively. Then, substitute the obtained intensity values into the formula Gray = 0.299R + 0.587G + 0.114B to calculate the gray value Gray corresponding to a single pixel point. By analogy, sum and calculate the mean value of the gray values Gray of all pixel points corresponding to the gray-scale segmented image i, and use the calculated mean value as the gray value of the gray-scale segmented image i;

[0042] For example, there is an RGB pixel with R = 100, G = 120, and B = 80. Then, according to the above formula, Gray = 0.299 x 100 + 0.587 x 120 + 0.114 x 80 = 29.9 + 70.44 + 9.12 = 109.46. This value is the grayscale value after conversion of the pixel. For a local image area in the shape of a rectangle or a simple shape, its average grayscale value can be calculated to describe the overall grayscale situation of the area. Let the grayscale value of the pixels in the area be ga, where a represents the pixel number label. The calculation formula for the average grayscale value is where b is the number of pixels.

[0043] Similarly, for the above analysis method, calculate the grayscale values of all grayscale segmentation images i, and then screen the grayscale segmentation images i according to the grayscale values. The specific screening method is as follows: obtain the grayscale value corresponding to the normal image, where the normal image here refers to an image without fouling, and calculate the grayscale difference between the grayscale value corresponding to the grayscale segmentation image i and the grayscale value of the normal image. At the same time, compare the obtained grayscale difference with a preset value. The specific value of the preset value is set by the operator and is calculated in combination with historical data. If the grayscale difference is greater than the preset value, classify the corresponding grayscale segmentation image i as an abnormal image, and record the corresponding area as an abnormal area. If the grayscale difference is less than the preset value, classify the corresponding grayscale segmentation image i as a normal image, and record the corresponding area as a normal area. For example, if the grayscale value of a certain segmentation image calculated is 130, and the grayscale value of this area of the corresponding normal image is 120, and the grayscale difference is 10, which is greater than the preset value of 5, then this area is determined to be an abnormal area.

[0044] Step 3: Analyze the obtained abnormal areas, judge the relevance of the abnormal areas by analyzing the adjacent situation of the abnormal areas, and generate a relevance result.

[0045] Obtain all abnormal areas and label them as n, where n = 1, 2,..., m, and m represents the number label of the abnormal areas. Then, classify them into associated areas and non-associated areas according to whether the abnormal areas are adjacent, and generate an associated area result and a non-associated area result. Here, in combination with the appendix Figure 2 It can be known that in Case 1 of the appendix Figure 2 it means that the abnormal areas are adjacent, so the adjacent abnormal areas are classified as associated areas. If it is Case 2 in the appendix Figure 2 it means that the abnormal areas are not adjacent, so the non-adjacent abnormal areas are classified as non-associated areas. At the same time, analyze the generated associated area result and non-associated area result respectively.

[0046] Step 4: Analyze the associated area results in the relevance results, analyze the fouling characteristics corresponding to the abnormal area by combining historical data, and comprehensively identify the identification information by combining the light reflectivity of the abnormal area.

[0047] Obtain the associated area and obtain the corresponding abnormal area denoted as n, and here n has the same label as that in Step 3 above. Then, obtain the fouling characteristics corresponding to the abnormal area n, where the fouling characteristics include fouling color and fouling texture, and analyze by combining the fouling color and fouling texture;

[0048] Obtain the HSV value corresponding to the abnormal area n. Here, the HSV value is usually expressed in degrees. The specific acquisition method is to convert the abnormal image corresponding to the abnormal area into an HSV image, and further obtain the corresponding hue (H), saturation (S), and value (V). At the same time, match the HSV value with the preset matching interval. Here, the preset matching interval is obtained by integrating historical data. At the same time, when matching, the three groups of values are matched separately to obtain the color matching result. Then, obtain the fouling texture corresponding to the abnormal area n, extract the texture features of the image through the gray-level co-occurrence matrix, and match the obtained texture features with the preset texture matching interval, and generate a preliminary selection result. Here, the preset texture matching interval is set by the operator and is integrated by the operator according to historical data.

[0049] Obtain all the preliminary selection results and denote them as o, and o = 1, 2, …, p, where p represents the number label of the preliminary selection results. Then, obtain the light reflectivity of the abnormal area n denoted as Fn. Here, the light reflectivity Fn is measured by an optical sensor. At the same time, match the obtained light reflectivity Fn with the light reflectivity interval Fo corresponding to the preliminary selection result o. Here, the light reflectivity interval is obtained by combining different preliminary selection results to obtain the matching information. For example, if the light reflectivity is 0.35, according to the REFLECTIVITY_RANGES dictionary, it is within the interval (0.3, 0.5) corresponding to the preliminary selection result 2, so "Identification information: The abnormal area matches the preliminary selection result 2, and the light reflectivity is 0.35" will be output.

[0050] Then, perform a secondary analysis on the obtained matching information, and analyze the corresponding fouling area based on the light reflectivity of the abnormal area. The specific calculation method is as follows:

[0051] Obtain the light reflectivity of the photovoltaic panel under normal conditions denoted as F 0 , and at the same time obtain the area of the abnormal area denoted as Sn. Determine the relationship formula according to the relationship between the change of light reflectivity and the fouling area to obtain Where θ is a preset proportionality coefficient, and the fouling area Sn corresponding to the abnormal area is calculated according to the relational expression. By analogy, the fouling areas of all abnormal areas n are calculated, and the sum of the fouling areas is obtained as the overall fouling area of the photovoltaic panel, denoted as Sz. At the same time, identification information is generated.

[0052] Embodiment 2. As the second embodiment of the present invention, it is implemented on the basis of Embodiment 1, and the differences from Embodiment 1 are as follows:

[0053] In this embodiment, the non-region results in the correlation results are analyzed, and the analysis method here is the same as that for the region results in Embodiment 1. At the same time, identification information is generated.

[0054] Embodiment 3. As the third embodiment of the present invention, the key lies in combining the implementation processes of Embodiment 1 and Embodiment 2.

[0055] Embodiment 4. Please refer to Figure 3 , this application provides a fouling identification system for a photovoltaic cleaning robot based on machine vision, including: an information acquisition unit, an information analysis unit, and an information output unit. And in combination with the attached Figure 3 It can be known that the above functional units are connected in a one-way electrical connection.

[0056] The information acquisition unit is used to obtain the basic information of the photovoltaic panel and transmit the obtained basic information of the photovoltaic panel to the information analysis unit at the same time;

[0057] The information analysis unit is used to analyze the fouling situation of the photovoltaic panel according to the obtained basic information, divide the photovoltaic panel into multiple groups of divided areas, calculate the gray values of the divided areas at the same time, and screen out abnormal areas according to the gray values. And the processing method here is the same as the processing process of Step 2 in Embodiment 1. The obtained abnormal areas are analyzed, the relevance of the abnormal areas is judged by analyzing the adjacent situation of the abnormal areas, and a relevance result is generated. And the processing method here is the same as the processing process of Step 3 in Embodiment 1. The associated area results in the relevance result are analyzed, the fouling characteristics corresponding to the abnormal areas are analyzed by combining historical data, and comprehensive identification is performed by combining the light reflectivity of the abnormal areas to obtain identification information. And the processing method here is the same as the processing process of Step 4 in Embodiment 1. At the same time, the obtained identification information is transmitted to the information output unit;

[0058] The information output unit is used to display the obtained identification information to the corresponding operator through a display device.

[0059] Some data in the above formula are all numerically calculated after removing their dimensions, and the content not described in detail in this specification belongs to the prior art well known to those skilled in the art.

[0060] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical method of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A photovoltaic cleaning robot dirt recognition method based on machine vision, characterized in that: The method specifically comprises the following steps: Step 1: acquiring basic information of the photovoltaic panel, and the basic information includes the area and shape of the photovoltaic panel; Step 2: According to the obtained basic information of the photovoltaic panel, the photovoltaic panel is segmented to obtain multiple groups of segmented regions, and the grayscale values ​​of the segmented regions are calculated, and the abnormal regions are screened according to the grayscale values; Step 3: Analyze the abnormal area obtained, determine the correlation of the abnormal area by analyzing the adjacent situation of the abnormal area, and generate a correlation result; Step 4: Analyze the associated area results in the correlation results, analyze the pollution characteristics corresponding to the abnormal area by combining historical data, and obtain identification information by comprehensive identification combined with the light reflectivity of the abnormal area.

2. The photovoltaic cleaning robot dirt recognition method based on machine vision according to claim 1 is characterized in that: The specific method for calculating the grayscale value of the segmented area in step 2 is: The photovoltaic panel is obtained and recorded as the target object, and the image information of the target object is obtained at the same time. Then, the obtained image is segmented to obtain a segmented image, which is labeled as i, and i=1, 2, ..., j, where j represents the number of segmented images. At the same time, the gray value of the segmented image i is calculated and recorded as Hi, and the specific calculation method is as follows: The segmented image i is obtained, and the segmented image i is converted to obtain the grayscale segmented image i. Then, the number of pixels corresponding to the segmented image i is obtained and recorded as Ci, and the intensity values ​​of the red, green and blue channels corresponding to a single pixel are obtained, recorded as R, G and B respectively. Then, the obtained intensity value is substituted into the formula Gray = 0.299R + 0.587G + 0.114B to calculate the grayscale value Gray corresponding to a single pixel. Similarly, the grayscale values ​​Gray of all pixels corresponding to the grayscale segmented image i are summed and the mean is calculated. At the same time, the calculated mean is used as the grayscale value of the grayscale segmented image i.

3. The method for identifying accumulated dirt in a photovoltaic cleaning robot based on machine vision according to claim 1, characterized in that: In step 2, the specific method of filtering out abnormal areas according to grayscale values ​​is as follows: Calculate the grayscale values ​​of all grayscale segmentation images i, then screen the grayscale segmentation images i according to the grayscale values, obtain the grayscale values ​​corresponding to the normal images, and calculate the grayscale values ​​corresponding to the grayscale segmentation images i and the grayscale values ​​of the normal images to obtain the grayscale difference, and compare the obtained grayscale difference with the preset value; If the grayscale difference is greater than the preset value, the corresponding grayscale segmentation image i is classified as an abnormal image, and the corresponding area is recorded as an abnormal area. If the grayscale difference is less than the preset value, the corresponding grayscale segmentation image i is classified as a normal image, and the corresponding area is recorded as a normal area.

4. The method for identifying dirt accumulation in a photovoltaic cleaning robot based on machine vision according to claim 1, characterized in that: The specific method of analyzing the abnormal area in step 3 is: All abnormal regions are obtained and labeled as n, where n=1, 2, ..., m, where m represents the number of abnormal regions. Then, the abnormal regions are classified into associated regions and unassociated regions according to whether they are adjacent. Associated region results and unassociated region results are generated, and the generated associated region results and unassociated region results are analyzed respectively.

5. The photovoltaic cleaning robot dirt recognition method based on machine vision according to claim 1 is characterized in that: The specific method of analyzing the associated area results in step 4 is: Obtain the associated area and obtain the corresponding abnormal area, which is recorded as n. Then, obtain the pollution feature corresponding to the abnormal area n, wherein the pollution feature includes the pollution color and the pollution texture, and analyze the pollution color and the pollution texture in combination. Get the HSV value corresponding to the abnormal area n, and match and analyze the HSV value with the preset matching interval to get the color matching result. Then get the dirt texture corresponding to the abnormal area n, extract the texture features of the image through the grayscale co-occurrence matrix, and match the obtained texture features with the preset texture matching interval. Generate the pre-selected results and then analyze the pre-selected results.

6. The method for identifying dirt accumulation in a photovoltaic cleaning robot based on machine vision according to claim 5, characterized in that: The specific method of analyzing the pre-selection results in step 4 is: Obtain all pre-selected results and record them as o, where o=1, 2, ..., p, and p represents the number label of the pre-selected results. Then obtain the light reflectivity of the abnormal area n and record it as Fn. At the same time, match the obtained light reflectivity Fn with the light reflectivity interval Fo corresponding to the pre-selected result o to obtain matching information, and perform secondary analysis on the matching information.

7. The method for identifying dirt accumulation in a photovoltaic cleaning robot based on machine vision according to claim 6, characterized in that: The specific method of performing secondary analysis on the matching information in step 4 is: The light reflectivity of the photovoltaic panel under normal conditions is recorded as F0, and the area of ​​the abnormal area is recorded as Sn. The relationship between the change in light reflectivity and the area of ​​pollution is determined to obtain Wherein θ is the preset proportional coefficient, and the pollution area Sn corresponding to the abnormal area is calculated according to the relationship. Similarly, the pollution areas of all abnormal areas n are calculated, and the pollution areas are summed to obtain the overall pollution area of ​​the photovoltaic panel, recorded as Sz, and identification information is generated at the same time.

8. A photovoltaic cleaning robot dirt accumulation identification system based on machine vision, used to execute the photovoltaic cleaning robot dirt accumulation identification method based on machine vision according to any one of claims 1 to 7, characterized in that: include: An information collection unit is used to obtain basic information of the photovoltaic panels and transmit the obtained basic information of the photovoltaic panels to the information analysis unit; An information analysis unit is used to analyze the pollution accumulation of the photovoltaic panel according to the acquired basic information, perform regional segmentation on the photovoltaic panel to obtain multiple groups of segmented regions, calculate the grayscale values ​​of the segmented regions, and screen out abnormal regions according to the grayscale values, analyze the obtained abnormal regions, determine the correlation of the abnormal regions by analyzing the adjacent conditions of the abnormal regions, and generate correlation results, analyze the correlation region results in the correlation results, analyze the pollution accumulation characteristics corresponding to the abnormal regions by combining historical data, and perform comprehensive identification in combination with the light reflectivity of the abnormal regions to obtain identification information, and transmit the obtained identification information to the information output unit; The information output unit is used to display the acquired identification information to the corresponding operator through a display device.

Images