Photovoltaic Module Residual Ice and Snow Volume Identification Algorithm Based on Autonomous Cruise UAV

Abstract

The residual ice and snow volume recognition algorithm for photovoltaic modules based on autonomous cruise drones disclosed by the present invention: Step 1, use the drones arranged in the photovoltaic power station to collect the winter photovoltaic module ice and snow covered image set S1; Step 2, judge whether any ice and snow covered image I1 in the collected image set S1 is distorted: if it is distorted, perform preprocessing on I1 in sequence after distortion correction to obtain image I2; Step 3, perform image segmentation on the image I2 obtained in Step 2 by using the adaptive maximum inter-class variance method to obtain image I3; Step 4, perform morphological correction on the photovoltaic module icing area segmentation image I3 obtained in Step 3 to obtain the hole filling map I4 and the icing area correction map I5; Step 5, image feature extraction and residual ice and snow volume calculation. This method monitors the ice and snow on the components of photovoltaic power stations in remote areas and realizes the long-distance perception of the ice and snow state of photovoltaic modules.

Description

Technical Field

[0001] The present invention belongs to the technical field of online monitoring of icing on photovoltaic modules, and particularly relates to an algorithm for identifying the residual ice and snow volume on photovoltaic modules based on an autonomous cruising unmanned aerial vehicle (UAV). Background Art

[0002] Relevant data shows that the maximum monthly power generation loss of photovoltaic modules caused by icing and snow covering problems can exceed 80%, and the annual power generation loss can reach 30%. This causes serious economic losses and seriously affects the normal operation of photovoltaic power stations in cold regions. Currently, many distributed photovoltaic power stations are located on slopes, hills, etc. It is very difficult for workers to inspect and clean the site after snow. Summary of the Invention

[0003] The object of the present invention is to provide an algorithm for identifying the residual ice and snow volume on photovoltaic modules based on an autonomous cruising UAV. This method monitors the ice and snow on the components of photovoltaic power stations in remote areas and realizes the remote perception of the ice and snow state of photovoltaic modules.

[0004] The technical solution adopted by the present invention is that the algorithm for identifying the residual ice and snow volume on photovoltaic modules based on an autonomous cruising UAV is specifically implemented according to the following steps:

[0005] Step 1: Use the UAV arranged in the photovoltaic power station to collect the image set S1 of photovoltaic modules covered with ice and snow in winter;

[0006] Step 2: Determine whether any ice and snow covered image I1 in the collected image set S1 is distorted: if it is distorted, perform distortion correction on I1 and then perform preprocessing in sequence to obtain image I2;

[0007] Step 3: Use the adaptive maximum inter-class variance method to perform image segmentation on the image I2 obtained in Step 2 to obtain image I3;

[0008] Step 4: Perform morphological correction on the segmented image I3 of the ice-covered area of the photovoltaic module. Since the ice-covered curve of the ice-covered image of the photovoltaic panel after segmentation is not smooth and there are some holes in the ice-covered area, perform morphological correction on I3 to obtain the hole-filled image I4 and the corrected image I5 of the ice-covered area;

[0009] Step 5: Image feature extraction and residual ice and snow volume calculation. Select two feature quantities, color feature and shape feature, which are related to the ice-covered area of the photovoltaic module; select the ice-covered area, the shape factor of the ice-covered area, and the ice and snow coverage rate of the photovoltaic panel as parameters to calculate the residual ice and snow volume of the photovoltaic module.

[0010] The features of the present invention also lie in that

[0011] The process of drone inspection of photovoltaic modules in step 1 is as follows: take off the drone outside the safety range of the photovoltaic array, manually or automatically control the drone to fly above the photovoltaic modules, fly along the photovoltaic module array and collect image information of the photovoltaic modules. After the images of all checkpoints are collected and the image set S1 is obtained, the drone returns to the starting point.

[0012] The specific implementation of step 2 is:

[0013] 2.1) Correct the distortion of the captured image

[0014] The image I1 is distorted using the affine transformation method. In two-dimensional space, the formula for transforming a point (x, y) to a point (x', y') through affine transformation is as follows:

[0015] (1)

[0016] Where (T x ,T y ) represents the translation amount, and parameter A ij , where i = 0, 1, j = 0, 1, reflects the image rotation and scaling changes; the parameter T x , T y , A ij By calculating, we can get the coordinate transformation relationship of the transformed image.

[0017] 2.2) Grayscale the color image

[0018] The brightness values of the three-channel R, G, and B components in image I1 are multiplied by different weights according to their respective importance, and the weighted average of each channel is calculated as follows:

[0019] (2)

[0020] Where R(x, y), G(x, y), and B(x, y) are the pixel values of image R, image G, and image B at (x, y);

[0021] 2.3) Perform histogram normalization

[0022] The histogram normalization method is used to enhance the image I1. First, the grayscale of the image I1 is equalized:

[0023] (3)

[0024] in Represents the grayscale value of image I1 after equalization, r k Is the grayscale of image I1 at grayscale level k, rk Apply the corresponding rule T, denoted as T( r k ), represents the probability density function of the input image, n j represents the number of pixel points at the k gray level, n is the total number of pixel points, and L represents the total number of gray levels;

[0025] Then perform histogram equalization on the target image as well:

[0026] ; (4)

[0027] where represents the gray value after equalization of the target image; z k represents the gray level of the image to be specified at the k gray level. Apply the corresponding rule G to z k denoted as G( r k ), represents the probability density function that the desired output image is to have. In this way, and will have the same uniform density, that is For the inverse transformation function of Equation (4), substitute , so the gray level z of the target image can be obtained based on the gray value of the image after equalization of image I1 k :

[0028] (5)

[0029] and represent the objective function, and substituting the result can obtain the preprocessed image I2.

[0030] The specific implementation method of step 4 is as follows:

[0031] 4.1) Establish a planar disk-shaped structural element. Different sizes of the structural element will result in different filtering effects. Perform an opening operation on image I3 to remove isolated small dots, burrs, and small bridges, and finally remove the white fence in the background while keeping the remaining ice and snow forms unchanged;

[0032] 4.2) Process the holes in the icing area

[0033] Delete all connected components less than P from the image obtained in the previous step to obtain a preliminary picture of the icing area hole filling and generate the hole filling map I4; where P is the threshold for judging whether there are bright noise points in the image background;

[0034] 4.3) Remove the frame of the photovoltaic panel, input parameters by means of human-computer interaction or directly, retain the picture within the area where the parameters are located, and finally obtain the corrected icing area map I5 to prepare for the next calculation of the residual ice and snow volume.

[0035] The specific implementation manner of step 5 is as follows:

[0036] Method 1: When used for calculating the icing area of continuous photovoltaic modules: The statistical pixel count method is used to calculate the residual ice and snow volume. The specific method is as follows:

[0037] Statistically count the number of pixels of the foreground object in the icing area of the corrected icing area map I5 and the total number of pixels of the photovoltaic module, and calculate the percentage of the number of foreground image pixels in the total number of pixels of the entire image. Multiply the percentage number by the actual area number of the corrected icing area map I5, and the product is the icing area S. The calculation formula is as follows:

[0038] (11)

[0039] In the formula: S is the icing area, is the number of pixels in the ice and snow covered area, is the total number of pixels of the photovoltaic module, is the actual area of the corrected icing area map I5;

[0040] Among them, in formula (11):

[0041] Total number of pixels of the photovoltaic module : The corrected icing area map I5 is obtained by the aforementioned method. Use the method based on the number of pixels to traverse the entire image, find the number of pixels in the entire area, and record it as the total number of pixels ;

[0042] Number of pixels in the ice and snow covered area : Traverse the target area with connected domains segmented, and statistically obtain the number of pixels in the ice and snow covered area ;

[0043] Finally, substitute and into formula (11) to obtain the ice and snow covered area S;

[0044] Method 2: When used for calculating the ice and snow covered area of local photovoltaic modules, the circumscribed ellipse method is used to calculate the residual ice and snow volume. The specific method is as follows:

[0045] a) Determine the required ice and snow area according to image I5;

[0046] b) Take the direction of the photovoltaic module frame as the x-axis of the ellipse, and its perpendicular direction as the y-axis. Respectively, take the midpoints of the maximum pixel projections of the residual ice and snow along the x- and y-axis directions as the center of the ellipse, and the maximum pixel projection length as the two characteristic lengths of the ellipse;

[0047] c) Draw an ellipse with the ellipse midpoint and length parameters in step b;

[0048] d) If the corresponding area of the residual ice and snow pixels outside the ellipse is 10 - 20% of the ellipse area, it is considered to meet the requirements. Otherwise, go to step e;

[0049] e) Adjust the major axis, minor axis, and angle respectively with a certain step size, and return to step c.

[0050] The beneficial effects of the present invention are as follows:

[0051] 1) The method of the present invention aims at the problem that it is difficult to manually detect ice and snow in winter in field photovoltaic power stations. It uses drones and surveillance cameras to take pictures and identify the icing images on photovoltaic modules, and combines image processing algorithms to monitor the ice and snow on the components of remote photovoltaic power stations, realizing the long-distance perception of the ice and snow state of photovoltaic modules.

[0052] 2) The method of the present invention proposes a complete image recognition process, mainly including corner point positioning of photovoltaic modules, region segmentation, and calculation of the remaining ice and snow area, which can accurately identify the area covered with ice and snow on photovoltaic modules and the severity of the remaining ice and snow volume, meeting the requirements of power generation enterprises for identifying the ice and snow degree on the surface of photovoltaic modules, and is of great significance for power generation enterprises to carry out power generation efficiency prediction and component operation and maintenance under winter ice and snow conditions for photovoltaic power stations.

[0053] 3) The drone inspection of photovoltaic power stations has the advantages of less terrain restriction, high inspection efficiency, good inspection effect, rapid deployment, low inspection cost, simple operation, etc. It can effectively supplement manual inspection in terms of inspection scope, content, and frequency, and has been popularized and applied in many domestic photovoltaic power stations. Therefore, the method of the present invention takes pictures of the ice and snow on winter components through an autonomous cruising drone based on the layout of the photovoltaic power station, calculates the remaining ice and snow volume on the components during the ice and snow ablation process through image processing, realizes the long-distance perception of the ice and snow state of photovoltaic modules, and provides a reference basis for power production prediction of winter photovoltaic power stations. Brief Description of the Drawings

[0054] Figure 1 is the flowchart of the algorithm for identifying the remaining ice and snow volume of photovoltaic modules based on the autonomous cruising drone of the present invention;

[0055] Figure 2 is the flowchart of the recognition effect of the algorithm for identifying the remaining ice and snow volume of photovoltaic modules based on the autonomous cruising drone of the present invention;

[0056] Figure 3Schematic diagram of the circumscribed ellipse of the icing area in the embodiment of the present invention;

[0057] Figure 4 Original image collected in Embodiment 1 of the present invention;

[0058] Figure 5 Corrected image of the icing area in Embodiment 1 of the present invention;

[0059] Figure 6 Original image collected in Embodiment 2 of the present invention;

[0060] Figure 7 Corrected image of the icing area in Embodiment 2 of the present invention;

[0061] Figure 8 Original image collected in Embodiment 3 of the present invention;

[0062] Figure 9 Corrected image of the icing area in Embodiment 3 of the present invention;

[0063] Figure 10 Original image collected in Embodiment 4 of the present invention;

[0064] Figure 11 Corrected image of the icing area in Embodiment 4 of the present invention;

[0065] Figure 12 Original image collected in Embodiment 5 of the present invention;

[0066] Figure 13 Corrected image of the icing area in Embodiment 5 of the present invention;

[0067] Figure 14 Original image collected in Embodiment 6 of the present invention;

[0068] Figure 15 Corrected image of the icing area in Embodiment 6 of the present invention. Detailed implementation manners

[0069] The present invention will be described in detail below in conjunction with the accompanying drawings and specific implementation manners.

[0070] The present invention provides a recognition algorithm for the residual ice and snow volume of photovoltaic modules based on an autonomous cruising unmanned aerial vehicle (UAV), as Figure 1 shown, and is specifically implemented according to the following steps:

[0071] Step 1: Use the autonomous cruising UAV arranged in the photovoltaic power station to collect the winter icing and snow-covered image set S1 of the photovoltaic modules. The process of the UAV inspecting the photovoltaic modules is as follows: The UAV takes off outside the safety range of the photovoltaic array, controls the flight of the UAV to reach above the photovoltaic modules manually or automatically, and flies along the photovoltaic module array to collect the image information of the photovoltaic modules. After the image information of all inspection points is collected to obtain the image set S1, the UAV returns to the starting point.

[0072] Step 2. Determine whether any ice / snow-covered image I1 in the acquired image set S1 is distorted: If it is distorted, perform preprocessing on I1 in sequence after distortion correction, including image grayscale conversion, image enhancement, and image filtering to remove interference noise, enhance the image contrast, make the ice / snow area more prominent, and obtain image I2. The specific process is as follows:

[0073] 2.1) Perform distortion correction on the acquired image

[0074] Not every picture of the photovoltaic module collected by the drone is taken directly above the module, and some pictures will be distorted. To reduce the calculation error of the residual ice / snow volume, it is necessary to perform distortion correction on image I1 using the affine transformation method. In the two-dimensional space, the calculation formula for the point (x, y) to be transformed to the point (x', y') through affine transformation is as follows:

[0075] (1)

[0076] where (T x , T y ) represents the translation amount, and the parameter A ij , where i = 0, 1, j = 0, 1, then reflects the image rotation and scaling changes; by calculating the parameters T x , T y , A ij , the coordinate transformation relationship of the transformed image can be obtained.

[0077] 2.2) Convert the color image to grayscale

[0078] The present invention uses an algorithm based on weighted average to perform grayscale conversion on it, which can reduce the amount of image data and improve the operation efficiency.

[0079] Multiply the brightness values of the three channels R, G, and B in image I1 by different weight values according to their respective importance, and perform weighted average on the weight values of each channel. The calculation formula is as follows:

[0080] (2)

[0081] where R(x, y), G(x, y), and B(x, y) are the pixel values of image R, image G, and image B at (x, y);

[0082] 2.3) Perform histogram specification processing

[0083] In addition, since the boundary between the ice / snow edge and the background board of image I1 is blurred, the histogram specification method is used to enhance image I1. First, perform grayscale equalization on image I1:

[0084] (3)

[0085] Wherein represents the gray value after equalization of image I1, r k represents the gray level of image I1 at the k gray level, and for r k the corresponding rule T is applied, denoted as T( r k ), represents the probability density function of the input image, n j represents the number of pixel points at the k gray level, n is the total number of pixel points, and L represents the total number of gray levels.

[0086] Then, the target image is also subjected to gray level equalization:

[0087] ; (4)

[0088] Wherein represents the gray value after equalization of the target image; z k represents the gray level of the image to be specified at the k gray level. For z k the corresponding rule G is applied, denoted as G( r k ), represents the probability density function that the desired output image is to have. In this way, and have the same uniform density, that is, For the inverse transformation function of formula (4), is substituted, so the gray level z of the target image can be obtained according to the gray value of the image after equalization of image I1 k :

[0089] (5)

[0090] and represent the objective function. Substituting the result can obtain the preprocessed image I2. This figure reduces the amount of image data, the snow and ice covered areas are all retained intact, and the contrast between the ice covered area and the background area is enhanced and the clarity of the image is increased.

[0091] Step 3: The image I2 obtained in Step 2 is segmented by the adaptive maximum between-class variance method (OTSU). This method calculates the maximum variance between the background and the target by calculating the gray levels of the image, thereby obtaining the segmentation threshold of the image, and then performing binarization processing according to this threshold, and finally segmenting the ice covered area segmentation image I3 of the photovoltaic module.

[0092] Step 4. Morphologically correct the ice-covered area segmentation image I3 obtained in Step 3. Since the ice-covered curve of the ice-covered photovoltaic panel image after segmentation is not smooth and there are some holes in the ice-covered area, morphologically correct I3 to obtain a hole-filled image I4 and an ice-covered area corrected image I5. The specific process is as follows:

[0093] 4.1) Remove the background fence of the photovoltaic panel to simplify the image background. Establish a planar disc-shaped structuring element. Different sizes of the structuring element will result in different filtering effects (in this embodiment, the radius of the structuring element size is specified as 7). Perform an opening operation on image I3 to remove isolated small dots, burrs, and small bridges, and finally remove the white fence in the background while keeping the remaining ice and snow morphology unchanged;

[0094] 4.2) Process the holes in the ice-covered area

[0095] Delete all connected components with fewer than P (P is the threshold for judging bright noise points in the image background) from the image obtained in the previous step to obtain a preliminary ice-covered area hole-filled picture and generate the hole-filled image I4;

[0096] 4.3) Remove the photovoltaic panel border, input parameters by using a human-computer interaction method or directly input parameters, and retain the picture within the area where the parameters are located. Finally, obtain the ice-covered area corrected image I5 to prepare for calculating the remaining ice and snow volume in the next step.

[0097] Use Steps 1 - 4 to identify the ice- and snow-covered photovoltaic module image. The effect during the identification process is as Figure 2 shown.

[0098] Step 5. Image feature extraction and calculation of the remaining ice and snow volume. Select two feature quantities related to the ice-covered area of the photovoltaic module, namely color feature and shape feature. Generally, when the proportion of the white area is higher, the ratio of the ice-covered area is also larger (the ice-covered area represents the remaining ice and snow volume). Therefore, select the ice-covered area, the ice-covered area shape factor, and the ice and snow coverage rate of the photovoltaic panel as the parameters of the present invention to calculate the remaining ice and snow volume of the photovoltaic module.

[0099] Method 1. When used for calculating the ice-covered area of continuous photovoltaic modules: Use the method of counting the number of pixel points to calculate the remaining ice and snow volume. The specific method is as follows:

[0100] Count the number of pixels of the foreground object in the ice-covered area of the ice-covered area corrected image I5 and the total number of pixels of the photovoltaic module, and calculate the percentage of the number of foreground image pixels in the total number of pixels of the entire image. Multiply the percentage number by the actual area number of the ice-covered area corrected image I5. The product is the ice-covered area S. The calculation formula is as follows:

[0101] (11)

[0102] Where: S is the area of the icing region, is the number of pixels in the icing and snow region, is the total number of pixels of the photovoltaic module, is the actual area of the corrected map I5 of the icing region;

[0103] Among them, in formula (11):

[0104] The total number of pixels of the photovoltaic module : The corrected map I5 of the icing region is obtained by the aforementioned method. The method based on the number of pixel points is used to traverse the whole map to find the number of pixels in all regions and record it as the total number of pixels ;

[0105] The number of pixels in the icing and snow region : The target regions with connected domains are traversed and segmented, and the number of pixels in the icing and snow region is statistically obtained ;

[0106] Finally, and are substituted into formula (11) to obtain the area S of the icing and snow region;

[0107] Method 2: When calculating the area of the icing and snow region of a local photovoltaic module, the circumscribed ellipse method is used to calculate the residual ice and snow volume. Due to the uncertainty of the ice and snow type, when there is a lack of recognition of the ice and snow region, the ice and snow coverage region is delimited by an ellipse, which can reduce the error. This method has the value and significance for reference. The specific method is as follows:

[0108] a) Determine the required ice and snow region according to image I5;

[0109] b) Take the direction of the photovoltaic module border as the x-axis of the ellipse, and its perpendicular direction as the y-axis. Respectively, take the midpoints of the maximum pixel projections of the residual ice and snow along the x and y axes as the center of the ellipse, and the maximum pixel projection length as the two characteristic lengths of the ellipse;

[0110] c) Draw an ellipse with the ellipse midpoint and length parameters in step b;

[0111] d) If the corresponding area of the residual ice and snow pixels outside the ellipse is 10 - 20% (manually calibrated) of the ellipse area, it is considered to meet the requirements. Otherwise, go to step e;

[0112] e) Adjust the major axis, minor axis, and angle respectively with a certain step size (e.g., step size 2% for steps 1 to 5), and return to step c.

[0113] Experimental results: The circumscribed ellipse diagram of the icing region is as Figure 3 shown. In the figure, a represents the major axis length and b represents the minor axis length. Two methods for calculating the area of the icing region of the photovoltaic module are proposed in step 5:

[0114] Method 1 calculates the ice-covered area by counting the number of pixels in the target area of the image. By traversing the entire picture, all pixels in the target area and the total number of pixels in the entire image are counted, and finally the ice-covered area is obtained. The calculation results are shown in Table 1.

[0115] Table 1

[0116]

[0117] Among them, the original image of Example 1 is as shown in Figure 4 ; the corrected image of the ice-covered area is as shown in Figure 5 ; the original image of Example 2 is as shown in Figure 6 ; the corrected image of the ice-covered area is as shown in Figure 7 ; the original image of Example 3 is as shown in Figure 8 ; the corrected image of the ice-covered area is as shown in Figure 9 ; the original image of Example 4 is as shown in Figure 10 ; the corrected image of the ice-covered area is as shown in Figure 11 ; the original image of Example 5 is as shown in Figure 12 ; the corrected image of the ice-covered area is as shown in Figure 13 ; the original image of Example 6 is as shown in Figure 14 ; the corrected image of the ice-covered area is as shown in Figure 15 . Obtaining Figure 5 、 7 、9, 11, 13, 15, the residual ice and snow volume can be calculated by Method 1. Example 6 is a picture of the continuous residual ice and snow area of the photovoltaic module, and there is no situation of local photovoltaic modules, so Method 1 is used to calculate the residual ice and snow volume.

[0118] Method 2 uses the circumscribed ellipse method to calculate the residual ice and snow volume. The implementation examples are Examples 1 to 5. The original image I1 is still as shown in Figure 6 、 8 、10, 12, 14. After being processed by the aforementioned recognition process, and then according to Method 2, the calculated results of the residual ice and snow volume of Examples 1 to 5 are 34.079%, 46.153%, 22.814%, 17.457%, 71.156% respectively.

[0119] Analysis of the error of the residual ice and snow volume calculation model:

[0120] Comparing the two methods in step 5 to evaluate the accuracy of the residual ice and snow volume algorithm during the ice and snow ablation process, the present invention uses two indicators, accuracy rate and absolute error, to reflect the accuracy of the residual ice and snow volume;

[0121] (12)

[0122] Among them is the area of the ice and snow region manually segmented; is the area of the ice and snow region obtained by the algorithm; is the proportion of the actual ice-covered area obtained after manual segmentation; is the proportion of the ice-covered area obtained by the algorithm; is the absolute error; is the relative accuracy rate. Then, the calculation results of the two methods for different remaining ice and snow amounts of local components and continuous components are respectively counted, as shown in Table 2.

[0123] Table 2

[0124]

[0125] As can be seen from Table 2, both methods can calculate the ice and snow covered areas of continuous and local photovoltaic modules, and the maximum error does not exceed 5%, and the relative accuracy rate is also above 90%. In summary, the method of the present invention can accurately identify the ice and snow covered areas on photovoltaic modules and the severity of the remaining ice and snow amounts by using image processing technology, monitor the ice and snow on photovoltaic modules in remote areas, realize the long-distance perception of the ice and snow state of photovoltaic modules, save time and effort, and can be widely used in large areas of photovoltaic power stations and hilly and mountainous areas where photovoltaic power generation is remotely installed, reduce the labor intensity, and is of great significance to photovoltaic inspection.

Claims

1. An algorithm for identifying the residual ice and snow volume of photovoltaic modules based on an autonomous cruise unmanned aerial vehicle, characterized in that, The implementation is specifically carried out according to the following steps: Step 1: Use the unmanned aerial vehicle (UAV) arranged in the photovoltaic power station to collect the winter snow and ice covered photovoltaic module image set S1; Step 2: Determine whether any snow and ice covered image I1 in the collected image set S1 is distorted: If it is distorted, perform distortion correction on I1 and then perform preprocessing in sequence to obtain image I2; Step 3: Use the adaptive maximum inter-class variance method to perform image segmentation on the image I2 obtained in Step 2 to obtain image I3; Step 4: Perform morphological correction on the photovoltaic module ice covered area segmentation image I3 obtained in Step 3 to obtain the hole filling image I4 and the ice covered area correction image I5; Step 5: Image feature extraction and calculation of the remaining snow and ice volume. Select two feature quantities, namely color feature and shape feature, which are related to the ice covered area of the photovoltaic module; Select the ice covered area, the ice covered area shape factor, and the snow and ice coverage rate of the photovoltaic panel as parameters to calculate the remaining snow and ice volume of the photovoltaic module; The specific implementation method of Step 5 is as follows: Method 1: When used for calculating the ice covered area of continuous photovoltaic modules: Use the method of counting pixel points to calculate the remaining snow and ice volume. The specific method is as follows: Count the number of pixels of the foreground target of the ice covered area in the ice covered area correction image I5 and the total number of pixels of the photovoltaic module, and calculate the percentage of the number of foreground image pixels in the total number of pixels of the entire image. Multiply the percentage number by the actual area number of the ice covered area correction image I5. The product is the ice covered area S. The calculation formula is as follows: （11） Where: S is the area of the icing region, is the number of pixels in the icing and snow region, is the total number of pixels of the photovoltaic module, is the actual area of the corrected map I5 of the icing region; Among them, in formula (11): Total number of pixels of the photovoltaic module : The icing area correction map I5 is obtained by the foregoing method. The whole map is traversed using the method based on the number of pixel points to find the number of pixels in the entire area, which is recorded as the total number of pixels ; Number of pixels in the snow and ice covered area : It is the target area with connected regions obtained by traversing and segmentation, and the number of pixels in the snow and ice covered area is statistically obtained ; Finally, bring and into Equation (11) to obtain the area S of the snow and ice covered area; Method 2: When used for calculating the snow and ice covered area of local photovoltaic modules, use the circumscribed ellipse method to calculate the remaining snow and ice volume. The specific method is as follows: a) Determine the required snow and ice area according to image I5; b) Take the direction of the photovoltaic module border as the x-axis of the ellipse, and its perpendicular direction as the y-axis. Respectively take the midpoints of the maximum pixel projections of the remaining snow and ice along the x and y axes as the center of the ellipse, and the maximum pixel projection length as the two characteristic lengths of the ellipse; c) Draw an ellipse with the ellipse midpoint and length parameters in Step b; d) If the corresponding area of the remaining snow and ice pixels outside the ellipse is 10 - 20% of the ellipse area, it is considered to meet the requirements; otherwise, go to Step e; e) Adjust the major axis, minor axis, and angle respectively with a certain step length, and return to Step c.

2. The photovoltaic module residual ice and snow volume recognition algorithm based on an autonomous cruise unmanned aerial vehicle according to claim 1, wherein The inspection process of the UAV for the photovoltaic module in Step 1 is as follows: Take off the UAV outside the safe range of the photovoltaic array, control the flight of the UAV to reach above the photovoltaic module manually or automatically, and fly along the photovoltaic module array to collect the image information of the photovoltaic module. After the images of all inspection points are collected to obtain the image set S1, the UAV returns to the starting point.

3. The photovoltaic module residual ice and snow volume identification algorithm based on an autonomous cruise unmanned aerial vehicle according to claim 1, wherein The specific implementation method of Step 2 is as follows: 2.1) Perform distortion correction on the collected image Use the affine transformation method to perform distortion correction on image I1. In the two-dimensional space, the calculation formula for the point (x, y) to be transformed to the point (x', y') through affine transformation is as follows: (1) where (T x , T y ) represents the translation amount, and the parameter A ij , where i = 0, 1, j = 0, 1, then reflects the image rotation and scaling changes; the parameters T x , T y , A ij are calculated to obtain the coordinate transformation relationship of the transformed image; 2.2) Grayscale the color image Multiply the brightness values of the three channels R, G, and B components in image I1 by different weight values according to their respective importance, and perform weighted average on the weight values of each channel. The calculation formula is as follows: （2） Among them, R(x, y), G(x, y), and B(x, y) are the pixel values of image R, image G, and image B at (x, y); 2.3) Perform histogram specification processing Use the histogram specification method to enhance image I1. First, perform gray-level equalization on image I1: （3） Among them represents the gray value after equalization of image I1, r k represents the gray level of image I1 at the k gray level. For r k applying the corresponding rule T, denoted as T( r k ), represents the probability density function of the input image, n j represents the number of pixel points at the k gray level, n is the total number of pixel points, and L represents the total number of gray levels; Then perform gray-level equalization on the target image as well: ； （4） where represents the gray value after equalization of the target image; z k represents the gray value of the image to be specified at the k gray level. Applying the corresponding rule G to z k is denoted as G( r k ), represents the probability density function that the desired output image has. Thus, and have the same uniform density, that is For the inverse transformation function of Equation (4), substituting Therefore, the gray level z of the target image is obtained based on the gray value of the image after equalization of image I1 k : （5） and represents the objective function, and substituting the result can obtain the preprocessed image I2.

4. The photovoltaic module residual ice and snow volume identification algorithm based on an autonomous cruise unmanned aerial vehicle according to claim 1, wherein The specific implementation manner of step 4 is as follows: 4.1) Establish a planar disk-shaped structural element. Different sizes of the structural element will result in different filtering effects. Perform an opening operation on image I3 to remove isolated small dots, burrs, and small bridges. Finally, remove the white fences in the background while keeping the remaining ice and snow forms unchanged; 4.2) Process the holes in the icing area Delete all connected components with less than P from the image obtained in the previous step to obtain a picture with the holes in the preliminary icing area filled, and generate a hole filling map I4; where P is the threshold for judging whether there are bright noise points in the image background; 4.3) Remove the photovoltaic panel border, input parameters using the human-computer interaction method or directly input parameters, and retain the picture within the area where the parameters are located. Finally, obtain a corrected icing area map I5 to prepare for the next step of calculating the residual ice and snow volume.

Images