Photovoltaic panel hot spot detection optimization method based on double model dynamic architecture

Abstract

The present application relates to the technical field of hot spot detection, and in particular to a photovoltaic panel hot spot detection optimization method based on a double model dynamic architecture.The present application screens a target visible light resolution according to the hot spot confidence of all hot spot regions in a visible light image corresponding to hot spot matching regions in different super-resolution images, the number of hot spot matching regions, and the number of hot spot regions in each super-resolution image; uses the target visible light resolution as an initial infrared resolution; and judges whether the initial infrared resolution needs to be adjusted according to the gradient distribution, temperature distribution, and corresponding hot spot confidence of different positions in infrared images of the photovoltaic panel between the initial infrared resolution and the next super-resolution. If adjustment is needed, the next super-resolution is used as a new initial infrared resolution until adjustment is no longer needed, and a target infrared resolution is obtained; and the detection of photovoltaic panel hot spots is realized. The present application improves the efficiency of hot spot detection by obtaining an accurate detection resolution.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of hot spot detection, and particularly relates to a photovoltaic panel hot spot detection optimization method based on a double-model dynamic architecture. BACKGROUND

[0002] The hot spot phenomenon refers to the fact that a local area of a photovoltaic panel is shielded or has performance differences, so that the area is called a load instead of a power generation unit, and is reversely charged by other normally operating cell pieces, thereby generating high temperature. Long-term hot spots can directly burn the affected cell pieces, and ultimately cause the entire photovoltaic panel to be scrapped. Therefore, the hot spots in the photovoltaic panel need to be treated in a timely manner to ensure the power generation efficiency of the photovoltaic system and reduce economic losses.

[0003] In the prior art, a target detection model and infrared hot spot detection are used to directly up-sample photovoltaic panel hot spots for detection by means of a drone flying regularly. However, considering that up-sampling is prone to loss of details, and the resolution of an image directly affects the flight power consumption of the drone and the time consumption of the model during the flight of the drone, a larger resolution requires a model to have stronger computing power and requires greater power consumption, and a smaller resolution may have poor image quality and reduced detection accuracy, so an unsuitable image resolution has a poor effect on photovoltaic panel hot spot detection. SUMMARY

[0004] In order to solve the technical problem of poor effect of an unsuitable image resolution on photovoltaic panel hot spot detection, the purpose of the present application is to provide a photovoltaic panel hot spot detection optimization method based on a double-model dynamic architecture, and the technical solution adopted is as follows:

[0005] The present application provides a photovoltaic panel hot spot detection optimization method based on a double-model dynamic architecture, which comprises the following steps:

[0006] Obtaining a visible light image and an infrared image of a photovoltaic panel containing a hot spot area in dynamic operation, wherein the hot spot area contains a hot spot confidence level;

[0007] Obtaining a plurality of super-resolution images of the visible light image in the initial dynamic operation, and obtaining a hot spot matching area of each hot spot area in the visible light image in different super-resolution images according to the position distribution of the hot spot area between the visible light image and different super-resolution images;

[0008] According to the hot spot confidence level of all hot spot areas in the visible light image corresponding to the hot spot matching areas in different super-resolution images, the number of hot spot matching areas, and the number of hot spot areas in each super-resolution image, a target visible light resolution is selected;

[0009] The target visible light resolution is taken as an initial infrared resolution, and whether adjustment is needed for the initial infrared resolution is determined according to gradient distribution, temperature distribution and corresponding hot spot confidence of the photovoltaic panel in infrared images between the initial infrared resolution and a next super resolution; if adjustment is needed, the next super resolution is taken as a new initial infrared resolution until adjustment is not needed, and the target infrared resolution is obtained;

[0010] The hot spot of the photovoltaic panel is detected according to the target infrared resolution and the target visible light resolution.

[0011] Further, the method for obtaining the hot spot matching area comprises:

[0012] The center positions of the hot spot areas in different adjacent super resolution images are matched, the corresponding matched hot spot areas are taken as the hot spot matching areas, and the hot spot matching areas of each hot spot area in the visible light image in different super resolution images are obtained.

[0013] Further, the method for obtaining the target visible light resolution comprises:

[0014] According to the hot spot confidence of each hot spot area in the visible light image in the hot spot matching areas in different super resolution images and the number of the hot spot matching areas, the local information loss degree of each hot spot area in each super resolution image is obtained.

[0015] According to the local information loss degree of all hot spot areas in each super resolution image and the difference value of the number of the hot spot areas on each super resolution image and the visible light image, the overall information loss degree of each super resolution image is obtained.

[0016] The resolution corresponding to the super resolution image with the minimum overall information loss degree value is selected as the target visible light resolution.

[0017] Further, the method for obtaining the local information loss degree comprises:

[0018] For any hot spot area in the visible light image, the difference value of the hot spot confidence of each hot spot area in the hot spot matching areas between each super resolution image and a next super resolution image is obtained as a first confidence difference value.

[0019] The comparison difference value of the number of the hot spot matching areas of each hot spot area in different super resolution images and the order of each super resolution image is obtained, the sum of the positive integer 1 and the comparison difference value is calculated as a first sum value, and the ratio of the first sum value and the number of the hot spot matching areas is calculated as an information loss weight.

[0020] The product of the first confidence difference and the information loss weight is obtained, and normalized mapping is performed, as the local information loss degree of each hot spot region in each super-resolution image.

[0021] Further, the method for obtaining the overall information loss degree comprises:

[0022] The sum of the difference between the positive integer 1 and the number of hot spot regions is obtained, and the sum value is divided by the number of hot spot regions, as a loss coefficient;

[0023] The accumulation value of the local information loss degrees of all hot spot regions in each super-resolution image is obtained, as a loss accumulation value;

[0024] The product of the loss accumulation value and the loss coefficient is calculated, as the overall information loss degree of each super-resolution image.

[0025] Further, the method for determining whether the initial infrared resolution needs to be adjusted comprises:

[0026] According to the gradient distribution, temperature distribution and corresponding hot spot confidence of different positions in the infrared image under any resolution, the feature saliency under the corresponding resolution is obtained;

[0027] According to the feature saliency, gradient distribution and corresponding hot spot confidence of the infrared image between the initial infrared resolution and the next resolution, the reference value of the initial infrared resolution is obtained;

[0028] According to the reference value of the initial reference resolution, it is determined whether the initial infrared resolution needs to be adjusted.

[0029] Further, the method for obtaining the feature saliency comprises:

[0030] For the hot spot region or non-hot spot region in the infrared image under any resolution, the average of the gradient amplitudes of all pixel points in each region is obtained, as the sharpness of each region;

[0031] The difference between the temperature averages of all pixel points between the hot spot region and the non-hot spot region in the infrared image is obtained, as the temperature difference;

[0032] The ratio of the sharpness difference and the temperature difference between the hot spot region and the non-hot spot region is obtained, the product of the ratio result and the sharpness of the hot spot region is calculated, as the adjustment necessity under the corresponding resolution; the ratio of the hot spot confidence and the adjustment necessity of the corresponding infrared image is obtained, and normalized mapping is performed, as the feature saliency under the corresponding frequency.

[0033] Further, the method for obtaining the reference value comprises:

[0034] Obtaining the sharpness ratio of the non-defect area between the initial infrared resolution and the next super-resolution as an adjustment factor; obtaining the product of the adjustment factor and the feature saliency of the super-resolution, calculating the difference between the feature saliency of the initial infrared resolution and the product result as the saliency change amount between the corresponding resolutions;

[0035] Obtaining the saliency change amount between the corresponding resolutions by comparing the confidence of the infrared image on the initial infrared resolution with the saliency change amount, and performing normalized mapping as the reference value of the initial infrared resolution.

[0036] Further, the obtaining method for judging whether the initial infrared resolution needs to be adjusted according to the reference value of the initial reference resolution comprises:

[0037] If the reference value of the initial reference resolution is less than or equal to the preset reference threshold, it is judged that the corresponding initial reference resolution needs to be adjusted.

[0038] Further, the obtaining method for the target infrared resolution comprises:

[0039] After judging that adjustment is needed, the next super-resolution is taken as a new initial infrared resolution, the reference value of the super-resolution is obtained for judgment, and this process is repeated until adjustment is no longer needed, and the new initial infrared resolution is taken as the target infrared resolution.

[0040] The present application has the following beneficial effects:

[0041] According to the position distribution of the hot spot area between the visible light image and the different super-resolution images, the present application obtains the hot spot matching area of each hot spot area in the visible light image in the different super-resolution images, ensures that the same physical hot spot is analyzed under different resolutions, and according to the hot spot confidence of the hot spot area corresponding to the hot spot matching area in the different super-resolution images, the number of the hot spot matching areas and the number of the hot spot areas in each super-resolution image, the target visible light resolution is screened out, the resolution beneficial to the hot spot detection and analysis is screened out, which is helpful for more accurate analysis of the photovoltaic panel hot spot, the target visible light resolution is taken as the initial infrared resolution, whether the initial infrared resolution needs to be adjusted is judged according to the gradient distribution, temperature distribution and corresponding hot spot confidence of different positions in the infrared image between the initial infrared resolution and the next super-resolution, the balance point of precision and efficiency is efficiently found based on the infrared image-specific and hot spot-essentially-related features, if adjustment is needed, the next super-resolution is taken as a new initial infrared resolution, until adjustment is no longer needed, the target infrared resolution is obtained, and the photovoltaic panel hot spot detection is realized. The present application improves the efficiency of the hot spot detection by obtaining an accurate detection resolution. BRIEF DESCRIPTION OF DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, and the advantages thereof, the following will briefly introduce the drawings needed in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 A flow chart of a photovoltaic panel hot spot detection optimization method based on a double-model dynamic architecture provided by an embodiment of the present application;

[0044] Figure 2 A flow chart of a local information loss degree acquisition method provided by an embodiment of the present application;

[0045] Figure 3 A flow chart of a method for judging whether the initial infrared resolution needs to be adjusted provided by an embodiment of the present application. DETAILED DESCRIPTION

[0046] In order to further illustrate the technical means and effects adopted by the present application to achieve the predetermined purposes, the following will combine the drawings and the preferred embodiments to specifically describe the specific implementation, structure, features and effects of the photovoltaic panel hot spot detection optimization method based on a double-model dynamic architecture according to the present application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0047] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0048] The following will specifically describe the specific scheme of the photovoltaic panel hot spot detection optimization method based on a double-model dynamic architecture provided by the present application in combination with the drawings.

[0049] Please refer to Figure 1 which shows a flow chart of the photovoltaic panel hot spot detection optimization method based on a double-model dynamic architecture provided by an embodiment of the present application, and the specific method includes:

[0050] Step S1: acquiring the visible light image and the infrared image containing the hot spot area of the photovoltaic panel in dynamic operation, and the hot spot area contains the hot spot confidence.

[0051] In the embodiment of the present application, in order to reduce the energy consumption problem in the flight process of the unmanned aerial vehicle, improve the processing capacity for the reflection problem and the hot spot problem, a double model dynamic architecture is adopted, the double model is Binomial-PV1, the basic framework is YOLOX-Nano, and the model belongs to a photovoltaic feature enhancement layer PV-LEA, which solves the reflection mis-detection problem in the photovoltaic image; the Binomial-ThermoScannerV1 model, the basic framework is NanoDet, and the model belongs to a temperature gradient attention TGAM, which solves the low recognition rate of micro-hot spots in the detection process; the corresponding image obtained is input into the model to obtain the hot spot position and hot spot confidence in the image and other information; the visible light image and the infrared image containing the hot spot area of the photovoltaic panel in the dynamic operation are obtained, and the hot spot confidence is contained in the hot spot area.

[0052] The hot spot confidence can measure the possibility of the existence of hot spot defects in the image. The greater the confidence, the more the hot spot affects the operation of the photovoltaic panel. For the visible light image, input into the PV-LEA, for the infrared image, input into the TGAM, obtain the corresponding hot spot confidence, mark the hot spot confidence of each hot spot area, and obtain the hot spot confidence. The specific means are well known to those skilled in the art and will not be described here.

[0053] Step S2: obtaining a plurality of super-resolution images of the visible light image in the initial dynamic operation, and obtaining the hot spot matching area of each hot spot area in the visible light image in different super-resolution images according to the position distribution of the hot spot area between the visible light image and different super-resolution images.

[0054] The selection of the image resolution affects the performance of the double model. High resolution causes information redundancy and increases computing power consumption. Low resolution may have poor image quality and reduced detection accuracy. Therefore, the resolution of the image needs to be adjusted to better identify the photovoltaic hot spot while improving the identification efficiency. A plurality of super-resolution images of the visible light image in the initial dynamic operation are obtained.

[0055] It should be noted that, in an embodiment of the present application, a super-resolution algorithm is used to obtain a plurality of super-resolution images of the visible light image, that is, the scaling factor of super-resolution is used to process the visible light image in the initial dynamic operation to obtain a plurality of super-resolution images. The scaling factor can be set according to specific conditions, and the specific means are well known to those skilled in the art and will not be described here.

[0056] The center position of the hot spot region can determine the accurate position of the hot spot region. By analyzing the distribution of the center positions of the hot spot regions, the performance of the same hot spot at different resolutions can be associated, and the hot spot matching region of each hot spot region in the visible light image in different super-resolution images is obtained according to the position distribution of the hot spot regions between the visible light image and different super-resolution images.

[0057] Preferably, in an embodiment of the present application, the method for obtaining the hot spot matching region comprises:

[0058] The center positions of the hot spot regions in different adjacent super-resolution images are matched, and the corresponding matching hot spot region is taken as the hot spot matching region to obtain the hot spot matching region of each hot spot region in the visible light image in different super-resolution images.

[0059] It should be noted that the center positions of the hot spot regions in different super-resolution images are matched by using the SIFT algorithm. If the center positions of the hot spot regions in the previous super-resolution image and the next super-resolution image are matched, the corresponding hot spot region is taken as the hot spot matching region, and thus the hot spot matching region of each hot spot region in the visible light image in different super-resolution images can be obtained. The specific means is a technology familiar to those skilled in the art, and will not be described here.

[0060] Step S3: According to the hot spot confidence of the hot spot regions in the visible light image corresponding to the hot spot matching regions in different super-resolution images, the number of hot spot matching regions, and the number of hot spot regions in each super-resolution image, the overall information loss degree of each super-resolution image is obtained, and the target visible light resolution is screened out.

[0061] The hot spot confidence can evaluate the possibility of the existence of hot spots in the hot spot region. If an image at a certain super-resolution is selected, and the difference between the hot spot confidence of the image and the next super-resolution image is small, it indicates that the change in resolution has no obvious change on the hot spot identification of the photovoltaic panel, and the information loss degree is smaller. The more the number of hot spot matching regions is, the more the hot spot defects can be identified at different resolutions, and the current image resolution is closer to a larger image resolution, the more the information of the image itself is, and the smaller the information loss degree is. The smaller the number of hot spot regions is, the less the image has hot spots, and the smaller the information loss degree is. According to the hot spot confidence of the hot spot regions in the visible light image corresponding to the hot spot matching regions in different super-resolution images, the number of hot spot matching regions, and the number of hot spot regions in each super-resolution image, the target visible light resolution is screened out.

[0062] Preferably, in an embodiment of the present application, the method for obtaining the target visible light resolution comprises:

[0063] According to the hot spot confidence of each hot spot region in the visible light image corresponding to the hot spot matching region in different super-resolution images, and the number of hot spot matching regions, the local information loss degree of each hot spot region in each super-resolution image is obtained.

[0064] Preferably, in an embodiment of the present application, the local information loss degree is obtained by referring to Figure 2 which shows a flow chart of a method for obtaining a local information loss degree, comprising:

[0065] Step S201: For any hot spot region in the visible light image, the difference of the hot spot confidence of each hot spot region corresponding to the hot spot matching region between each super-resolution image and the next super-resolution image is obtained as a first confidence difference.

[0066] The hot spot confidence can evaluate the hot spot probability of the hot spot region, and the closer the hot spot confidence between each super-resolution image and the next super-resolution image, the more similar the information features.

[0067] It should be noted that in the embodiments of the present application, the difference represents the absolute value of the difference.

[0068] Step S202: Obtain the number of hot spot matching regions of each hot spot region in different super-resolution images and the comparison difference of the order of each super-resolution image; calculate the sum of the positive integer 1 and the comparison difference as a first sum value.

[0069] The ratio of the first sum value and the number of hot spot matching regions is calculated as an information loss weight.

[0070] It should be noted that the larger the image resolution, the more information it has, and the closer it is to the position of the subsequent super-resolution image, the greater the impact on information; in order to analyze the impact of the relative position of the image on information, the sum of the positive integer 1 and the comparison difference is calculated as the first sum value to avoid the comparison difference being 0; the larger the comparison difference, the smaller the order of the super-resolution image relative to the number of hot spot matching regions, the less information the super-resolution image has, and the greater the information loss weight.

[0071] Step S203: Obtain the product of the first confidence difference and the information loss weight, and perform normalized mapping as the local information loss degree of each hot spot region in each super-resolution image.

[0072] It should be noted that in some embodiments of the present application, normalization can be performed by linear normalization or a normalization function, and the specific means is a technology known to those skilled in the art, which will not be described here.

[0073] According to the local information loss degree of all hot spot regions in each super-resolution image and the difference between the number of hot spot regions on the visible light image and each super-resolution image, the overall information loss degree of each super-resolution image is obtained.

[0074] Preferably, in one embodiment of the present application, the method for obtaining the overall information loss degree comprises:

[0075] The sum of the difference between the positive integer 1 and the number of hot spot regions is obtained, and the ratio of the sum value to the number of hot spot regions is calculated as a loss coefficient; the accumulation value of the local information loss degree of all hot spot regions in each super-resolution image is obtained as a loss accumulation value;

[0076] The product between the loss accumulation value and the loss coefficient is calculated as the overall information loss degree of each super-resolution image.

[0077] It should be noted that the greater the local information loss degree, the greater the influence on the overall information loss degree; the smaller the number of hot spot regions at the increased resolution than the number of hot spot regions on the visible light image at the initial time, the greater the difference between the number of hot spot regions, the more hot spot region details are lost, and the greater the overall information loss degree at this resolution; therefore, the correlation between the local information loss degree, the difference between the number of hot spot regions and the overall information loss degree is constructed based on the above basic mathematical operations, that is, the greater the local information loss degree and the greater the difference between the number of hot spot regions, the greater the overall information loss degree.

[0078] The overall information loss degree reflects the accuracy of image analysis at the resolution, the smaller the overall information loss degree, the more details are displayed, the more accurate the judgment of the defect position, and the target visible light resolution is selected, that is, the super-resolution image with the smallest overall information loss degree value is selected, and the corresponding resolution of the corresponding super-resolution image is taken as the target visible light resolution.

[0079] Step S4: Taking the target visible light resolution as the initial infrared resolution, according to the gradient distribution, temperature distribution and corresponding hot spot confidence of the photovoltaic panel at different positions in the infrared image between the initial infrared resolution and the next super-resolution, it is judged whether the initial infrared resolution needs to be adjusted; if adjustment is needed, the next super-resolution is taken as the new initial infrared resolution, until no adjustment is needed, and the target infrared resolution is obtained.

[0080] In order to improve the accuracy of the photovoltaic panel hot spot identification, the infrared image needs to be analyzed to provide temperature information for more accurate judgment. The gradient distribution reflects the texture complexity and edge strength of the image. Therefore, the greater the gradient of the hot spot region relative to the non-hot spot region, the greater the texture complexity and edge strength, and the greater the sharpness, which can better represent the obvious hot spot characteristics. Due to the existence of hot spots, there is a significant temperature difference between the hot spot region and the non-hot spot region of the photovoltaic panel. The greater the temperature difference, the greater the probability of showing hot spots. Therefore, the target visible light resolution is used as the initial infrared resolution, and the gradient distribution, temperature distribution and corresponding hot spot confidence of the photovoltaic panel at different positions in the infrared image between the initial infrared resolution and the next super-resolution are used to determine whether the initial infrared resolution needs to be adjusted.

[0081] Preferably, in one embodiment of the present application, whether the initial infrared resolution needs to be adjusted is determined by referring to Figure 3 , which shows a flow chart of a method for determining whether the initial infrared resolution needs to be adjusted, comprising:

[0082] Step S301: According to the gradient distribution, temperature distribution and corresponding hot spot confidence of different positions in the infrared image at any resolution, the feature saliency at the corresponding resolution is obtained.

[0083] Preferably, in one embodiment of the present application, the feature saliency acquisition method comprises:

[0084] For the hot spot region or non-hot spot region in the infrared image at any resolution, the average value of the gradient amplitude of all pixel points in each region is obtained as the sharpness of each region. The difference between the average temperature of all pixel points in the hot spot region and the non-hot spot region in the infrared image is obtained as the temperature difference.

[0085] The ratio of the sharpness difference and the temperature difference between the hot spot region and the non-hot spot region is obtained. The product of the ratio result and the sharpness of the hot spot region is calculated as the adjustment necessity at the corresponding resolution.

[0086] The ratio of the hot spot confidence and the adjustment necessity of the corresponding infrared image is obtained and normalized to map as the feature saliency at the corresponding frequency.

[0087] It should be noted that the gradient amplitude can be obtained according to the existing Sobel operator or Prewitt operator. The specific means are well known to those skilled in the art, and will not be repeated here.

[0088] Based on this, the greater the sharpness of the image, the smaller the sharpness difference between the next super resolution, the greater and more consistent the gradient performance, the clearer the image under the resolution, and the less the necessity of adjustment; the temperature in the infrared image represents the heat state of the object surface, the greater the temperature difference, the more obvious the thermal feature difference between the hot spot area and the non-hot spot area, and the less the necessity of adjustment; the higher the hot spot confidence, the greater the possibility of evaluating the hot spot, the greater the feature saliency, and the more accurate the identification of the hot spot defect of the photovoltaic panel under the corresponding resolution.

[0089] Step S302: Obtain the reference value of the initial infrared resolution according to the feature saliency, the gradient distribution of the infrared image, and the corresponding hot spot confidence between the initial infrared resolution and the next super resolution.

[0090] Preferably, as the resolution of the image increases, the gradient becomes greater, making the contour more obvious and the details richer, that is, the sharpness is greater, the reliability of hot spot detection is increased, and the reference value is greater; if there is a significant difference in the saliency change between the increased resolutions, it indicates that there is a significant change in the identification of the hot spot after the resolution adjustment, the infrared image resolution is low, and the increase processing is more needed, and the reference value is smaller; the greater the hot spot confidence, the better the infrared image under the resolution can represent the feature of the hot spot area, and the greater the reference value; in an embodiment of the present application, the method for obtaining the reference value comprises:

[0091] obtaining the sharpness ratio of the non-defect area between the initial infrared resolution and the next super resolution as an adjustment factor;

[0092] obtaining the product of the adjustment factor and the feature saliency of the super resolution, calculating the difference between the feature saliency of the initial infrared resolution and the product result as the saliency change amount between the corresponding resolutions;

[0093] obtaining the confidence ratio of the infrared image under the initial infrared resolution and the saliency change amount between the corresponding resolutions, and performing normalized mapping as the reference value of the initial infrared resolution.

[0094] Step S303: Determine whether the initial infrared resolution needs to be adjusted according to the reference value of the initial reference resolution.

[0095] Preferably, in an embodiment of the present application, if the reference value of the initial reference resolution is less than or equal to the preset reference threshold, it is determined that the corresponding initial reference resolution needs to be adjusted.

[0096] It should be noted that in an embodiment of the present application, the size of the preset value threshold is 0.8; in other embodiments of the present application, the size of the preset value threshold can be set according to specific circumstances, which is not limited and described here.

[0097] If adjustment is needed, the next super-resolution is taken as the new initial infrared resolution until adjustment is not needed, and the target infrared resolution is obtained.

[0098] It should be noted that, in an embodiment of the present application, the method for obtaining the target infrared resolution comprises: judging whether adjustment is needed, taking the next super-resolution as the new initial infrared resolution, judging the reference value of the super-resolution until adjustment is not needed, and taking the new initial infrared resolution as the target infrared resolution.

[0099] Step S5: inputting the infrared image under the target infrared resolution and the visible light image under the target visible light resolution into the corresponding model to realize detection of the hot spot of the photovoltaic panel.

[0100] Therefore, the target infrared resolution is optimal for identifying the hot spot on the infrared image, and is helpful for accurately analyzing the authenticity of the hot spot; the target visible light resolution can effectively obtain an image with the minimum information loss, and is helpful for obtaining a comprehensive hot spot region for analysis; therefore, after the appropriate target infrared resolution and target visible light resolution are obtained, the visible light image under the target visible light resolution and the infrared image under the target infrared resolution are registered, accurate hot spot positioning is performed, the demand for computing power of the double model is reduced, and the efficiency and accuracy of detection are improved.

[0101] In summary, according to the position distribution of the hot spot region between the visible light image and the different super-resolution images, the present application obtains the hot spot matching region of each hot spot region in the visible light image in the different super-resolution images; according to the hot spot confidence of the hot spot matching region corresponding to all hot spot regions in the visible light image in the different super-resolution images, the number of hot spot matching regions, and the number of hot spot regions in each super-resolution image, the target visible light resolution is screened out; the target visible light resolution is taken as the initial infrared resolution, whether adjustment of the initial infrared resolution is needed is judged according to the gradient distribution, temperature distribution and corresponding hot spot confidence of different positions in the infrared image of the photovoltaic panel between the initial infrared resolution and the next super-resolution; if adjustment is needed, the next super-resolution is taken as the new initial infrared resolution until adjustment is not needed, and the target infrared resolution is obtained; and detection of the hot spot of the photovoltaic panel is realized. The present application improves the efficiency of hot spot detection by obtaining accurate detection resolution.

[0102] It should be noted that the above-mentioned embodiments of the present application are only in the order of description, and do not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or may be advantageous.

[0103] The various embodiments described in this specification are presented by way of example, and each embodiment is not inherently more important than any other embodiment.

Claims

1. An optimization method for photovoltaic panel hot spot detection based on a dual-model dynamic architecture, characterized in that, The method includes: Acquire visible light and infrared images of a photovoltaic panel containing a hot spot region during dynamic operation, wherein the hot spot region contains hot spot confidence; The method for obtaining the hot spot confidence includes: inputting visible light images into a Binomial-PV1 model containing PV-LEA, and inputting infrared images into a Binomial-ThermoScannerV1 model containing TGAM, and marking the hot spot confidence of each hot spot region; Multiple super-resolution images of the visible light image during initial dynamic operation are obtained. Based on the positional distribution of hot spot regions between the visible light image and different super-resolution images, the hot spot matching region of each hot spot region in the visible light image is obtained in different super-resolution images. The target visible light resolution is selected based on the hot spot confidence, the number of hot spot matching regions, and the number of hot spot regions in each super-resolution image corresponding to all hot spot regions in the visible light image. Using the target visible light resolution as the initial infrared resolution, the gradient distribution, temperature distribution, and corresponding hot spot confidence of the photovoltaic panel at different locations in the infrared image between the initial infrared resolution and the next super-resolution determine whether the initial infrared resolution needs to be adjusted. If adjustment is required, the next super-resolution is used as the new initial infrared resolution until no adjustment is needed, thus obtaining the target infrared resolution. Based on the infrared image at the target infrared resolution and the visible light image at the target visible light resolution, the corresponding models are input to realize the detection of hot spots on photovoltaic panels; The method for obtaining the hot spot matching region includes: The SIFT algorithm is used to match the center positions of hot spot regions in different adjacent super-resolution images. The corresponding matched hot spot regions are used as hot spot matching regions to obtain the hot spot matching regions of each hot spot region in the visible light image in different super-resolution images. The determination of whether the initial infrared resolution needs to be adjusted includes: Based on the gradient distribution, temperature distribution, and corresponding hot spot confidence at different locations in an infrared image at any resolution, the feature saliency at the corresponding resolution is obtained. The reference value of the initial infrared resolution is obtained based on the saliency of features between the initial infrared resolution and the next resolution, the gradient distribution of the infrared image, and the corresponding hot spot confidence. Based on the reference value of the initial reference resolution, determine whether the initial infrared resolution needs to be adjusted; The method for obtaining the saliency of the features includes: For hot spot or non-hot spot regions in an infrared image at any resolution, the mean value of the gradient magnitude of all pixels in each region is obtained as the sharpness of each region. The difference in the average temperature of all pixels between hot spot and non-hot spot regions in an infrared image is obtained as the temperature difference. The ratio of the sharpness difference between the hot spot region and the non-hot spot region to the temperature difference is obtained. The product of the ratio result and the sharpness of the hot spot region is calculated as the adjustment necessity at the corresponding resolution. The ratio of the hot spot confidence and the adjustment necessity of the corresponding infrared image is obtained and normalized and mapped as the feature saliency at the corresponding frequency. The methods for obtaining the reference value include: The sharpness ratio of the non-defect region between the initial infrared resolution and the next super-resolution is obtained as an adjustment factor; Obtain the product of the adjustment factor and the feature significance of the super-resolution, calculate the difference between the feature significance of the initial infrared resolution and the product result, and use it as the amount of significance change between corresponding resolutions; The confidence level of the infrared image at the initial infrared resolution is obtained as a percentage of the significant change between the corresponding resolutions, and then normalized and mapped to serve as a reference value for the initial infrared resolution. The method for determining whether the initial infrared resolution needs adjustment based on its reference value includes: If the reference value of the initial reference resolution is less than or equal to the preset reference threshold, it is determined that the corresponding initial reference resolution needs to be adjusted.

2. The photovoltaic panel hot spot detection optimization method based on a dual-model dynamic architecture according to claim 1, characterized in that, The method for obtaining the target visible light resolution includes: Based on the hot spot confidence of each hot spot region in the visible light image corresponding to the hot spot matching region in different super-resolution images, and the number of hot spot matching regions, the local information loss degree of each hot spot region in each super-resolution image is obtained. The overall information loss of each super-resolution image is obtained based on the local information loss of all hot spot regions in each super-resolution image and the difference in the number of hot spot regions in each super-resolution image and the visible light image. The super-resolution image with the smallest overall information loss is selected, and the resolution of the corresponding super-resolution image is taken as the target visible light resolution.

3. The photovoltaic panel hot spot detection optimization method based on a dual-model dynamic architecture according to claim 2, characterized in that, The method for obtaining the local information loss includes: For any hot spot region in a visible light image, the hot spot confidence difference of each hot spot region in the hot spot matching region between each super-resolution image and the next super-resolution image is obtained as the first confidence difference. Obtain the number of hotspot matching regions for each hotspot region in different super-resolution images and the comparison difference between the order of each super-resolution image; calculate the sum of the positive integer 1 and the comparison difference as the first sum; calculate the ratio of the first sum to the number of hotspot matching regions as the information loss weight. The product of the first confidence difference and the information loss weight is obtained and normalized to form the local information loss of each hotspot region in each super-resolution image.

4. The photovoltaic panel hot spot detection optimization method based on a dual-model dynamic architecture according to claim 3, characterized in that, The method for obtaining the overall information loss includes: The sum of the differences between the positive integer 1 and the number of hot spot regions is obtained, and the sum is divided by the number of hot spot regions as the loss coefficient. The cumulative value of the local information loss of all hot spot regions in each super-resolution image is obtained as the cumulative loss value; The product of the accumulated loss and the loss coefficients is calculated as the overall information loss for each super-resolution image.

5. The photovoltaic panel hot spot detection optimization method based on a dual-model dynamic architecture according to claim 1, characterized in that, The method for obtaining the target infrared resolution includes: Once it is determined that adjustment is needed, the next super-resolution is used as the new initial infrared resolution. The reference value of the super-resolution is used for further judgment until no adjustment is needed, at which point the new initial infrared resolution is used as the target infrared resolution.

Images