Hot spot rapid detection method based on light noise removing algorithm and significant feature extraction

Abstract

The invention relates to a hot spot rapid detection method based on a light noise removing algorithm and significant feature extraction, and the method comprises the steps: obtaining an image of a photovoltaic module in a photovoltaic power station, and constructing a data set D1; smoothing the image by using a median filtering algorithm; converting an RGB color mode in the smoothed data set D1 image into an HSL mode, and removing the light influence in the data set D1 image by using illumination irrelevant features to obtain a light-free photovoltaic module image; extracting a saliency seed region from the light-free photovoltaic module image by using a saliency characteristic extraction algorithm; and according to hot spot features, screening and segmenting hot spot regions from the significant seed region, and issuing a hot spot detection result. According to the invention, the light influence of strong sunlight irradiation in the photovoltaic module image can be eliminated, the accuracy of photovoltaic module detection is improved, manual detection is replaced, and the hot spot detection efficiency can be improved; the existing hot spot detection equipment does not need to be upgraded, and the hot spot can be accurately and quickly detected based on the image.

Description

Technical Field

[0001] The invention belongs to the field of photovoltaic power generation detection, and in particular relates to a hot spot rapid detection method based on a bright light noise removal algorithm and significant feature extraction. Background Art

[0002] Photovoltaic power generation has developed rapidly due to its low cost and the convenience of laying photovoltaic power generation equipment. Photovoltaic power stations are usually built in mountains, deserts, or on the surface of water, on the top of buildings and other areas. This geographical environment brings great difficulties to the safety inspection of equipment in photovoltaic power stations, especially the detection of hot spots in photovoltaic modules. Manual detection methods are difficult to implement and inefficient. Using fixed cameras or drones in photovoltaic power plants to capture images of the working status of photovoltaic modules, and then using manual detection methods to detect and identify hot spots, this method has a large lag and false detection rate and cannot meet the real-time requirements of hot spots. The existing photovoltaic hot spot detection method based on artificial intelligence algorithms, due to the sun-facing nature of photovoltaic power stations, the imaging of photovoltaic modules is easily affected by the strong sunlight, forming bright light in the photovoltaic module image, which seriously affects the accuracy of photovoltaic hot spot detection. Summary of the invention

[0003] The purpose of the present invention is to address the above-mentioned problems and propose a method for rapid hot spot detection based on a bright light noise removal algorithm and significant feature extraction. The bright light influence of photovoltaic module images is removed by using bright light irrelevant features, and then a significant feature extraction algorithm is used to extract significant seed areas from photovoltaic module images without bright light, and then the hot spot areas are screened and segmented to replace manual detection, thereby improving the accuracy and real-time performance of photovoltaic hot spot detection.

[0004] To achieve the above object, the technical solution of the present invention is: The method for rapid detection of hot spots based on bright light noise removal algorithm and significant feature extraction includes the following steps: S1: Acquire images of photovoltaic modules in photovoltaic power plants and construct dataset D1; S2: Use the median filter algorithm to smooth the image of data set D1; S3: Convert the RGB color mode in the smoothed image of the dataset D1 to the HSL mode, and use the illumination-independent feature to remove the bright light effect in the image of the dataset D1, so as to obtain a photovoltaic module image without bright light noise; S4: Use the significant feature extraction algorithm to extract significant seed areas from the dark photovoltaic module image; based on the hot spot characteristics, filter and segment the hot spot areas from the significant seed areas, and publish the hot spot detection results.

[0005] Furthermore, step S2 includes the following sub-steps: S201: Determine an image sliding window; S202: after sorting the red, green and blue components of the pixels in the sliding window according to the brightness, calculate the median of the red, green and blue components of the pixels in the sliding window; S203: Calculate the new pixel value X according to the median of the red, green and blue components obtained in step S202 new , replace the pixel value at the center of the sliding window with the new pixel value X new ; S204: Repeat steps S202 and S203 until the image smoothing process is completed.

[0006] Preferably, the sliding window is an 8*8 rectangular window.

[0007] Furthermore, in step S202, the median of the red, green and blue components of the sliding window pixel is calculated using the following formula: ; In the formula, Indicates the brightness value of the mth color of the pixel; m=1,2,3 represents red, green or blue. Represents the color value of the 32nd pixel in the sequence sorted by brightness. Represents the color value of the 33rd pixel in a sequence sorted by brightness.

[0008] Furthermore, step S3 includes the following sub-steps: S301: Convert the original image RGB color mode to HSL color mode. The conversion formula is: ; ; ; In the formula, Represents the maximum value of the brightness of the red, green, and blue channels, Represents the minimum brightness of the red, green and blue channels. R, G and B represent the brightness values ​​of the red, green and blue channels respectively. H, L and S represent the hue, brightness and saturation of the pixel respectively. S302: Generate a brightness-independent luminance map of the image using the brightness-independent feature, and the calculation formula is: ; In the formula, represents the illumination-independent brightness characteristic value of pixel p, b is the illumination parameter, represents the illumination transformation function under illumination condition E, represents the illumination parameter value of pixel p, B represents the maximum value of illumination parameter b, k is a statistical constant, Indicates the brightness range of pixel p.

[0009] Preferably, step S4 includes the following sub-steps: S401: using a significant feature extraction algorithm, calculating the significance value of each pixel in the region of the photovoltaic module image without bright light, taking a pixel region in a specified range of significance values, and obtaining a significant seed region; S402: According to the hot spot characteristics, a region growing algorithm is used to screen and segment the hot spot region from the significant seed region, and the hot spot detection result is released.

[0010] Preferably, in step S401, the significance value of each pixel of the photovoltaic module image without bright light in the region is calculated, and the significance value of the current image pixel is calculated using a histogram and a Euclidean distance function, and the calculation formula is: ; In the formula, S ( ) is the significance value function, represents the kth photovoltaic module image, m represents the pixel value of the current pixel, and n represents the brightness in the brightness histogram of the photovoltaic module image, ranging from [0,255]; Indicates the saliency value of the pixel in the image The frequency of occurrence in represents the Euclidean distance function.

[0011] Compared with the prior art, the beneficial effects of the present invention include: (1) The present invention converts the original image into the HSL color mode, removes the influence of light in the original image according to the illumination-independent features, and obtains the hot spot area of ​​the photovoltaic module using the significant feature extraction algorithm, thereby eliminating the influence of bright light in the photovoltaic module image caused by strong sunlight, and improving the accuracy of photovoltaic module detection.

[0012] (2) The present invention can efficiently and accurately complete the rapid detection of hot spots of photovoltaic modules, has real-time and robustness, and is highly practical in actual photovoltaic module hot spot detection work.

[0013] (3) The present invention uses a significant feature extraction algorithm to obtain a significant seed area of ​​a suspected hot spot based on eliminating light noise, and then uses a region growing algorithm to screen and segment the hot spot area from the significant seed area, thereby achieving accurate detection of photovoltaic hot spots. The cost is low and there is no need to upgrade the existing hot spot detection equipment. The present invention can achieve accurate and rapid detection of photovoltaic hot spots based on photovoltaic module images, thereby achieving real-time online detection of photovoltaic module hot spots. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0015] Figure 1 This is a flow chart of the method for rapid detection of hot spots in photovoltaic modules according to an embodiment of the method. DETAILED DESCRIPTION

[0016] like Figure 1 As shown, the hot spot rapid detection method based on the bright light noise removal algorithm and significant feature extraction includes the following steps: S1: Acquire images of photovoltaic modules in photovoltaic power plants and construct dataset D1; S2: Use the median filter algorithm to smooth the image of data set D1; S3: Convert the RGB color mode in the smoothed image of the dataset D1 to the HSL mode, and use the light-independent feature to remove the light effect in the image of the dataset D1, so as to obtain a photovoltaic module image without light noise; S4: Use the significant feature extraction algorithm to extract significant seed areas from the dark photovoltaic module image; based on the hot spot characteristics, filter and segment the hot spot areas from the significant seed areas, and publish the hot spot detection results.

[0017] Step S2 includes the following sub-steps: S201: Determine an image sliding window, where the sliding window is a rectangular window of 8*8.

[0018] S202: after sorting the red, green and blue components of the pixels in the sliding window according to the brightness, calculate the median of the red, green and blue components of the pixels in the sliding window; S203: Calculate the new pixel value X according to the median of the red, green and blue components obtained in step S202 new , replace the pixel value at the center of the sliding window with the new pixel value X new ; S204: Repeat steps S202 and S203 until the image smoothing process is completed.

[0019] In step S202, the median of the red, green and blue components of the pixel point in the sliding window is calculated using the following formula: ; In the formula, Indicates the brightness value of the mth color of the pixel; m=1,2,3 represents red, green or blue. Represents the color value of the 32nd pixel in the sequence sorted by brightness. Represents the color value of the 33rd pixel in a sequence sorted by brightness.

[0020] Step S3 includes the following sub-steps: S301: Convert the original image RGB color mode to HSL color mode. The conversion formula is: ; ; ; In the formula, Represents the maximum value of the brightness of the red, green, and blue channels, Represents the minimum brightness of the red, green and blue channels. R, G and B represent the brightness values ​​of the red, green and blue channels respectively. H, L and S represent the hue, brightness and saturation of the pixel respectively. S302: Generate a brightness-independent luminance map of the image using the brightness-independent feature, and the calculation formula is: ; In the formula, represents the illumination-independent brightness characteristic value of pixel p, b is the illumination parameter, represents the illumination transformation function under illumination condition E, represents the illumination parameter value of pixel p, B represents the maximum value of illumination parameter b, k is a statistical constant, Indicates the brightness range of pixel p.

[0021] Step S4 includes the following sub-steps: S401: using a significant feature extraction algorithm, calculating the significance value of each pixel in the region of the photovoltaic module image without bright light, taking a pixel region in a specified range of significance values, and obtaining a significant seed region; S402: According to the hot spot characteristics, a region growing algorithm is used to screen and segment the hot spot region from the significant seed region, and the hot spot detection result is released.

[0022] In step S401, the significance value of each pixel of the photovoltaic module image without bright light in the region is calculated, and the significance value of the current image pixel is calculated using a histogram and a Euclidean distance function. The calculation formula is: ; In the formula, S ( ) is the significance value function, represents the kth photovoltaic module image, m represents the pixel value of the current pixel, and n represents the brightness in the brightness histogram of the photovoltaic module image, ranging from [0,255]; Indicates the saliency value of the pixel in the image The frequency of occurrence in represents the Euclidean distance function.

[0023] The implementation effect shows that the present invention can eliminate the influence of bright light in the photovoltaic module image caused by strong sunlight, improve the accuracy of photovoltaic module detection, replace manual detection, and greatly improve the efficiency of photovoltaic hot spot detection; the present invention uses a significant feature extraction algorithm to obtain a significant seed area of ​​the suspected hot spot on the basis of eliminating light noise, and then uses a region growing algorithm to screen and segment the hot spot area from the significant seed area, further improving the accuracy of photovoltaic hot spot detection, and does not need to upgrade the existing hot spot detection equipment, can realize accurate and rapid detection of photovoltaic hot spots based on photovoltaic module images, and realize real-time online detection of photovoltaic module hot spots.

Claims

1. A fast hot spot detection method based on light noise removal algorithm and significant feature extraction, characterized in that: The following steps are involved: S1: Acquire images of photovoltaic modules in photovoltaic power plants and construct dataset D1; S2: Use the median filter algorithm to smooth the image of data set D1; S3: Convert the RGB color mode in the smoothed image of the dataset D1 to the HSL mode, and use the illumination-independent feature to remove the bright light effect in the image of the dataset D1, so as to obtain a photovoltaic module image without bright light noise; S4: Use a significant feature extraction algorithm to extract significant seed areas from the photovoltaic module image without bright light noise; according to the hot spot characteristics, filter and segment the hot spot area from the significant seed area, and publish the hot spot detection results.

2. The method for rapid hot spot detection based on bright light noise removal algorithm and significant feature extraction according to claim 1, characterized in that: The step S2 comprises the following sub-steps: S201: Determine an image sliding window; S202: after sorting the red, green and blue components of the pixels in the sliding window according to the brightness, calculate the median of the red, green and blue components of the pixels in the sliding window; S203: Calculate the new pixel value X according to the median of the red, green and blue components obtained in step S202 new , replace the pixel value at the center of the sliding window with the new pixel value X new ; S204: Repeat steps S202 and S203 until the image smoothing process is completed.

3. The method for rapid hot spot detection based on bright light noise removal algorithm and significant feature extraction according to claim 1 or 2, characterized in that: The sliding window is a rectangular window of 8*8.

4. The method for rapid hot spot detection based on bright light noise removal algorithm and significant feature extraction according to claim 3 is characterized in that: In step S202, the median of the red, green and blue components of the sliding window pixel is calculated using the following formula: ； In the formula, Indicates the brightness value of the mth color of the pixel; m=1,2,3 means red or green or blue, Represents the color value of the 32nd pixel in the sequence sorted by brightness. Represents the color value of the 33rd pixel in a sequence sorted by brightness.

5. The method for rapid detection of hot spots based on bright light noise removal algorithm and significant feature extraction according to claim 1, 2 or 4, characterized in that: Step S3 includes the following sub-steps: S301: Convert the original image RGB color mode to HSL color mode. The conversion formula is: ； ； ； In the formula, Represents the maximum value of the brightness of the red, green, and blue channels, Represents the minimum brightness of the red, green and blue channels, R, G and B represent the brightness values ​​of the red, green and blue channels respectively; H, L, and S represent the hue, brightness, and saturation of the pixel, respectively; S302: Generate an illumination-independent brightness map of the image using the illumination-independent feature, and the calculation formula is: ； In the formula, represents the illumination-independent brightness characteristic value of pixel p, b is the illumination parameter, represents the illumination transformation function under illumination condition E, represents the illumination parameter value of pixel p, B represents the maximum value of illumination parameter b, k is a statistical constant, Indicates the brightness range of pixel p.

6. The method for rapid hot spot detection based on bright light noise removal algorithm and significant feature extraction according to claim 5, characterized in that: The step S4 comprises the following sub-steps: S401: using a significant feature extraction algorithm, calculating the significance value of each pixel in the region of the photovoltaic module image without bright light, taking a pixel region in a specified range of significance values, and obtaining a significant seed region; S402: According to the hot spot characteristics, a region growing algorithm is used to screen and segment the hot spot region from the significant seed region, and the hot spot detection result is released.

7. The method for rapid hot spot detection based on bright light noise removal algorithm and significant feature extraction according to claim 6, characterized in that: In step S401, the significance value of each pixel of the photovoltaic module image without bright light in the region is calculated, and the significance value of the current image pixel is calculated using a histogram and a Euclidean distance function, and the calculation formula is: ; In the formula, S ( ) is the significance value function, represents the kth photovoltaic module image, m represents the pixel value of the current pixel, and n represents the brightness in the brightness histogram of the photovoltaic module image, ranging from [0,255]; Indicates the saliency value of the pixel in the image The frequency of occurrence in represents the Euclidean distance function.

Images