An automated detection method and system based on the microcrack characteristics of photovoltaic modules

Abstract

This invention discloses an automated detection method and system based on the microcrack characteristics of photovoltaic modules, belonging to the field of automated detection technology. It includes a drone equipped with an infrared thermal imager, a polarization camera, and a vibration sensor to simultaneously collect the thermal distribution, surface texture, and micro-vibration signals of the photovoltaic module, constructing a multi-dimensional microcrack feature dataset. This invention achieves local data processing and microcrack probability value output by deploying a lightweight convolutional network on the drone, reducing cloud dependence and latency. The microcrack probability value is calculated by weighting infrared thermal anomaly factors, polarization texture anomaly factors, and edge sharpness factors, effectively suppressing environmental noise and improving detection accuracy. Multi-angle re-inspection tasks are inserted into high-probability microcrack areas, low-risk areas are scanned quickly, and medium-probability areas are determined based on power and priority. When there are many re-inspection areas, the globally optimal path is recalculated. When a high-probability microcrack is detected, the current path is immediately interrupted, and a re-inspection task is inserted to avoid delays.

Description

Technical Field

[0001] This invention belongs to the field of automated testing technology, specifically referring to an automated testing method and system based on the microcrack characteristics of photovoltaic modules. Background Technology

[0002] As the global energy structure transitions towards a low-carbon model, photovoltaic power generation, as one of the core technologies of clean energy, relies on the long-term stable operation of its components for the efficiency and economy of power supply.

[0003] However, existing automated detection methods for microcrack characteristics of photovoltaic modules still have certain shortcomings. Existing methods rely on cloud processing, which suffers from high latency and large bandwidth consumption. UAV inspection uses fixed altitude and focal length, making it difficult to achieve both high precision and wide coverage. The inspection path is a preset route and cannot be dynamically adjusted according to real-time detection results. The detection only outputs the location of microcracks without classification, resulting in inaccurate maintenance strategies. The detection relies on a single scan, leading to a high false positive rate. Therefore, an automated detection method and system based on the microcrack characteristics of photovoltaic modules is proposed. Summary of the Invention

[0004] The purpose of this invention is to provide an automated detection method and system based on the microcrack characteristics of photovoltaic modules, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an automated detection method based on the microcrack characteristics of photovoltaic modules, comprising the following steps:

[0006] S1. The UAV is equipped with an infrared thermal imager, a polarization camera and a vibration sensor to simultaneously collect the thermal distribution, surface texture and micro-vibration signals of photovoltaic modules and construct a multi-dimensional hidden crack feature dataset.

[0007] S2. Run a lightweight convolutional network through edge computing nodes to analyze thermal anomaly regions in infrared images and microcrack textures in polarization images in real time, and output the probability value of hidden cracks.

[0008] S3. Automatically adjust the drone's flight altitude and camera focal length based on the probability value of hidden cracks;

[0009] S4. Dynamically update the inspection path based on the detection results, conduct multi-angle re-inspection of high-probability hidden crack areas, and adopt a rapid scanning mode for low-risk areas.

[0010] S5. By fusing vibration sensor data and image analysis results, electrical and mechanical microcracks are distinguished and a classification diagnosis report is generated.

[0011] S6. Perform local high-precision rescanning on the areas marked as microcracks to verify the detection results and correct misjudged samples;

[0012] S7. Upload the verified hidden crack samples and environmental parameters to the cloud and feed them back to the edge computing nodes to update the convolutional network.

[0013] In step S1, the UAV is equipped with an infrared thermal imager, a polarization camera, and a vibration sensor to simultaneously collect the thermal distribution, surface texture, and micro-vibration signals of the photovoltaic module, constructing a multi-dimensional microcrack feature dataset. A timing-type GPS module is integrated on the UAV, outputting a second pulse signal as a global time reference. The PPS signal is distributed to the infrared thermal imager, polarization camera, and vibration sensor via a time synchronization controller, triggering simultaneous data collection by all three and automatically resetting the time counter. Combined with an inertial navigation board, the UAV's real-time position, attitude, and velocity are recorded. The installation positions, angles, and interrelationships of the infrared thermal imager, polarization camera, and vibration sensor are pre-calibrated to establish a unified three-dimensional coordinate system. The UAV's position and attitude data are acquired in real-time through the POS system. Combined with sensor installation parameters, the three-dimensional spatial coordinates of the data collected by each sensor are calculated. When the UAV hovers above the photovoltaic module, after the GPS triggers the PPS signal, the infrared thermal imager, polarization camera, and vibration sensor simultaneously begin collecting data. The collected thermal distribution, polarization texture, and vibration signal data are aligned according to timestamps to construct a multi-dimensional microcrack feature dataset.

[0014] S2 involves running a lightweight convolutional network through edge computing nodes to analyze thermal anomaly regions in infrared images and polarization microcrack textures in polarization images in real time, outputting a hidden crack probability value. A lightweight edge computing unit is deployed on the UAV to perform preliminary processing on the multidimensional hidden crack feature dataset, calculating the hidden crack probability value based on infrared thermal anomalies and polarization microcrack textures. Let the infrared thermal anomaly factor be T. df The polarization microcrack texture anomaly factor is P. ay The edge sharpness factor is S. ed T df Represented as T lcl T represents the average temperature of the target area. bd P represents the average temperature of the background area on the component surface. ay Represented as θ∈{0°,45°,90°,135°} represents the four polarization directions of the polarization camera, I θ I represents the pixel intensity under the polarization direction θ. θ,avg S represents the average pixel intensity of the background region under the same polarization direction. ed Represented as This represents the gradient magnitude of the i-th pixel. This represents the sum of gradient magnitudes of all pixels within the target area, and k represents the gradient amplification factor, which is adjusted according to the contrast of the photovoltaic module surface texture. The value of k is k∈[1.2,2.0].

[0015] S2, based on the infrared thermal anomaly factor T df Polarization microcrack texture anomaly factor P ay and edge sharpness factor S ed The probability value of microcracks is calculated as follows:

[0016]

[0017] In the formula, P ck T represents the probability value of a hidden crack, exp() represents the natural exponential function, α, β, γ represent weight parameters, α takes values ​​of α∈[0.1,0.5], β takes values ​​of β∈[0.2,0.8], γ takes values ​​of γ∈[0.1,0.5], and T df P represents the infrared anomaly factor. ay S represents the polarization microcrack texture anomaly factor. ed This represents the edge sharpness factor.

[0018] In step S3, the probability value of the hidden crack is obtained, and the area is divided into low-probability, medium-probability, and high-probability regions based on the probability value. When the region is low-probability, the UAV gradually increases its flight altitude to expand the coverage of a single shot. When a high-probability hidden crack region is detected, the UAV gradually decreases its flight altitude. In the high-probability hidden crack region, the camera switches to telephoto mode to focus on the details of the target area. In the low-probability region, the camera switches to wide-angle mode. The flight altitude and focal length are corrected in real time through GPS, inertial measurement unit, and visual feedback. The hidden crack probability is calculated once for each frame of image, and the UAV status is dynamically adjusted according to the result.

[0019] Specifically, in step S4, the inspection path is dynamically updated based on the detection results. High-probability hidden crack areas are re-inspected from multiple angles, while low-risk areas are scanned using a fast scanning mode. In high-probability hidden crack areas, after the UAV completes the initial scan, it automatically triggers a multi-angle re-inspection task, generating a new flight path around the target area. In low-probability areas, the UAV skips detailed re-inspection and directly enters the fast scanning mode, quickly covering the area along a preset route. The medium-probability areas are determined based on the remaining battery power and task priority. When a high-probability hidden crack area is detected, the current path is immediately interrupted, and a re-inspection task is inserted. A dynamic threshold for the number of re-inspection areas is preset. When the number of re-inspection areas is large, the globally optimal path is recalculated, and global path optimization is performed asynchronously while the UAV is hovering.

[0020] Specifically, in step S5, vibration sensor data and image analysis results are fused to distinguish between electrical and mechanical microcracks, generating a classification diagnosis report. Multi-angle image data and vibration sensor data are acquired, and vibration features and image texture features of electrical and mechanical microcracks are extracted respectively. The features of the vibration data are matched with the texture features of the image to analyze whether a coupling mode exists. There is a strong correlation between the vibration features of electrical microcracks and thermal anomalies, and a strong correlation between the vibration features of mechanical microcracks and crack morphology. The coupling strength between vibration features and image features is quantified using the Pearson correlation coefficient to distinguish between the two types of microcracks. Classification is performed based on the analysis and matching, generating a classification diagnosis report, and marking the microcrack areas.

[0021] The timing-synchronized GPS module outputs a second pulse signal to trigger synchronous data acquisition by the infrared thermal imager, polarization camera, and vibration sensor, achieving a timestamp accuracy of 1ms. The real-time attitude of the UAV is acquired through the POS system, and the three-dimensional spatial coordinates are calculated based on sensor installation parameters to ensure data alignment in the spatial dimension. Let x be the peak point of the micro-vibration in the vibration signal. k ,y k Let a and b be the fitting model parameters, and t be the threshold. The formula for determining the intrapoint distance between vibration data and image data is as follows:

[0022]

[0023] In the formula, x k ,y k This represents the peak point of micro-vibration in the vibration signal, where a and b are the fitting model parameters, and t is the threshold value when d ck If ≤ t, then it is determined to be an interior point;

[0024] According to d ck The effective distance for filtering, aligning the vibration signal with the time axis of the image sequence, eliminating nonlinear misalignment, and achieving dynamic time warping are as follows:

[0025] D(n,m)=d(T(n),R(m)+min{D(n-1,m),D(n-1,m-1),D(n,m-1)}),

[0026] In the formula, D(n,m) represents the dynamically regularized path, d(T(n),R(m)) represents the Euclidean distance, calculates the difference between the nth frame of the vibration signal T(n) and the nth frame of the image sequence R(m), and min{} represents finding the minimum value to ensure the optimal path; let the vibration feature vector be V and the image feature vector be I, calculate the projection ratio similarity between the vibration feature vector V and the image feature vector I, and implement the formula as follows:

[0027]

[0028] In the formula, Sim(V,I) represents the projection similarity between the dynamic feature vector V and the image feature vector I, and n represents the feature dimension.

[0029] In step S6, a local high-precision rescan is performed on the region marked as a hidden crack to verify the detection results and correct misjudged samples; the marked hidden crack region is obtained, and a local high-precision rescan is performed if the hidden crack probability value is within a preset threshold; when the matching degree between the classification result and the sensor data is lower than the preset threshold, a manual review process is triggered to verify and correct the misjudged results.

[0030] In step S7, the verified hidden crack samples and environmental parameters are uploaded to the cloud. The verified hidden crack samples are then uploaded to the remote end via the HTTP protocol. The local model updates of multiple drones are aggregated through horizontal federated learning. The cloud serves as the receiver of the model gradients of each edge node. After the drones are updated, they are tested using known hidden crack samples to compare the accuracy of the new and old models.

[0031] One example is an automated detection system based on the microcrack characteristics of photovoltaic modules, characterized by comprising a data acquisition module, an edge computing and real-time analysis module, a flight state adjustment module, a dynamic path planning module, a classification and diagnosis module, a result verification and correction module, and a cloud-based collaborative optimization module.

[0032] Compared with the prior art, the beneficial effects of the present invention are:

[0033] 1. This invention achieves local data processing and output of hidden crack probability values ​​by deploying a lightweight convolutional network on a drone, reducing cloud dependence and latency. The hidden crack probability value is calculated by weighting infrared thermal anomaly factor, polarization texture anomaly factor and edge sharpness factor, which effectively suppresses environmental noise and improves detection accuracy.

[0034] 2. This invention balances coverage and resolution by dynamically adjusting flight altitude and camera focal length based on the probability value of microcracks. It triggers an adjustment for each frame of image through GPS, IMU, and visual feedback to ensure dynamic response, expand the coverage area in low-risk areas, focus on details in high-risk areas, and reduce redundant scanning.

[0035] 3. This invention inserts multi-angle re-inspection tasks into high-probability hidden crack areas, quickly scans low-risk areas, and makes decisions on medium-probability areas based on power and priority. When there are many re-inspection areas, the global optimal path is recalculated to ensure efficiency. When a high-probability hidden crack is detected, the current path is immediately interrupted and a re-inspection task is inserted to avoid delays.

[0036] 4. This invention verifies the detection results by performing local high-precision rescanning on the hidden crack area. When the classification results significantly contradict the sensor data, a manual review process is triggered to correct misjudged samples and update the local database, thereby improving the accuracy of subsequent detections. Attached Figure Description

[0037] Figure 1 The present invention provides an automated detection method for photovoltaic module microcrack characteristics. Figure 1 ;

[0038] Figure 2 The present invention provides an automated detection method for photovoltaic module microcrack characteristics. Figure 2 ;

[0039] Figure 3 The present invention provides an automated detection method for photovoltaic module microcrack characteristics. Figure 3 ;

[0040] Figure 4 The present invention provides an automated detection method for photovoltaic module microcrack characteristics. Figure 4 ;

[0041] Figure 5 This is a schematic diagram of the structure of an automated detection system based on the microcrack characteristics of photovoltaic modules according to the present invention. Detailed Implementation

[0042] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0043] Example

[0044] Please see Figures 1-5 As shown, the present invention provides a technical solution comprising the following steps:

[0045] S1. The UAV is equipped with an infrared thermal imager, a polarization camera and a vibration sensor to simultaneously collect the thermal distribution, surface texture and micro-vibration signals of photovoltaic modules and construct a multi-dimensional hidden crack feature dataset.

[0046] S2. Run a lightweight convolutional network through edge computing nodes to analyze thermal anomaly regions in infrared images and microcrack textures in polarization images in real time, and output the probability value of hidden cracks.

[0047] S3. Automatically adjust the drone's flight altitude and camera focal length based on the probability value of hidden cracks;

[0048] S4. Dynamically update the inspection path based on the detection results, conduct multi-angle re-inspection of high-probability hidden crack areas, and adopt a rapid scanning mode for low-risk areas.

[0049] S5. By fusing vibration sensor data and image analysis results, electrical and mechanical microcracks are distinguished and a classification diagnosis report is generated.

[0050] S6. Perform local high-precision rescanning on the areas marked as microcracks to verify the detection results and correct misjudged samples;

[0051] S7. Upload the verified hidden crack samples and environmental parameters to the cloud and feed them back to the edge computing nodes to update the convolutional network.

[0052] In step S1, the UAV is equipped with an infrared thermal imager, a polarization camera, and a vibration sensor to simultaneously collect the thermal distribution, surface texture, and micro-vibration signals of the photovoltaic module, constructing a multi-dimensional microcrack feature dataset. A timing-type GPS module is integrated on the UAV, outputting a second pulse signal as a global time reference. The PPS signal is distributed to the infrared thermal imager, polarization camera, and vibration sensor via a time synchronization controller, triggering simultaneous data collection by all three and automatically resetting the time counter. Combined with an inertial navigation board, the UAV's real-time position, attitude, and velocity are recorded. The installation positions, angles, and interrelationships of the infrared thermal imager, polarization camera, and vibration sensor are pre-calibrated to establish a unified three-dimensional coordinate system. The UAV's position and attitude data are acquired in real-time through the POS system. Combined with sensor installation parameters, the three-dimensional spatial coordinates of the data collected by each sensor are calculated. When the UAV hovers above the photovoltaic module, after the GPS triggers the PPS signal, the infrared thermal imager, polarization camera, and vibration sensor simultaneously begin collecting data. The collected thermal distribution, polarization texture, and vibration signal data are aligned according to timestamps to construct a multi-dimensional microcrack feature dataset.

[0053] S2 involves running a lightweight convolutional network through edge computing nodes to analyze thermal anomaly regions in infrared images and polarization microcrack textures in polarization images in real time, outputting a hidden crack probability value. A lightweight edge computing unit is deployed on the UAV to perform preliminary processing on the multidimensional hidden crack feature dataset, calculating the hidden crack probability value based on infrared thermal anomalies and polarization microcrack textures. Let the infrared thermal anomaly factor be T. df The polarization microcrack texture anomaly factor is P. ay The edge sharpness factor is S. ed T df Represented as T lcl T represents the average temperature of the target area. bd P represents the average temperature of the background area on the component surface. ay Represented as θ∈{0°,45°,90°,135°} represents the four polarization directions of the polarization camera, I θ I represents the pixel intensity under the polarization direction θ. θ,avg S represents the average pixel intensity of the background region under the same polarization direction. ed Represented as This represents the gradient magnitude of the i-th pixel. This represents the sum of gradient magnitudes of all pixels within the target area, and k represents the gradient amplification factor, which is adjusted according to the contrast of the photovoltaic module surface texture. The value of k is k∈[1.2,2.0].

[0054] S2, based on the infrared thermal anomaly factor T df Polarization microcrack texture anomaly factor P ay and edge sharpness factor S ed The probability value of microcracks is calculated as follows:

[0055]

[0056] In the formula, P ck T represents the probability value of a hidden crack, exp() represents the natural exponential function, α, β, γ represent weight parameters, α takes values ​​of α∈[0.1,0.5], β takes values ​​of β∈[0.2,0.8], γ takes values ​​of γ∈[0.1,0.5], and T df P represents the infrared anomaly factor. ay S represents the polarization microcrack texture anomaly factor. ed This represents the edge sharpness factor.

[0057] In step S3, the probability value of the hidden crack is obtained, and the area is divided into low-probability, medium-probability, and high-probability regions based on the probability value. When the region is low-probability, the UAV gradually increases its flight altitude to expand the coverage of a single shot. When a high-probability hidden crack region is detected, the UAV gradually decreases its flight altitude. In the high-probability hidden crack region, the camera switches to telephoto mode to focus on the details of the target area. In the low-probability region, the camera switches to wide-angle mode. The flight altitude and focal length are corrected in real time through GPS, inertial measurement unit, and visual feedback. The hidden crack probability is calculated once for each frame of image, and the UAV status is dynamically adjusted according to the result.

[0058] Specifically, in step S4, the inspection path is dynamically updated based on the detection results. High-probability hidden crack areas are re-inspected from multiple angles, while low-risk areas are scanned using a fast scanning mode. In high-probability hidden crack areas, after the UAV completes the initial scan, it automatically triggers a multi-angle re-inspection task, generating a new flight path around the target area. In low-probability areas, the UAV skips detailed re-inspection and directly enters the fast scanning mode, quickly covering the area along a preset route. The medium-probability areas are determined based on the remaining battery power and task priority. When a high-probability hidden crack area is detected, the current path is immediately interrupted, and a re-inspection task is inserted. A dynamic threshold for the number of re-inspection areas is preset. When the number of re-inspection areas is large, the globally optimal path is recalculated, and global path optimization is performed asynchronously while the UAV is hovering.

[0059] Specifically, in step S5, vibration sensor data and image analysis results are fused to distinguish between electrical and mechanical microcracks, generating a classification diagnosis report. Multi-angle image data and vibration sensor data are acquired, and vibration features and image texture features of electrical and mechanical microcracks are extracted respectively. The features of the vibration data are matched with the texture features of the image to analyze whether a coupling mode exists. There is a strong correlation between the vibration features of electrical microcracks and thermal anomalies, and a strong correlation between the vibration features of mechanical microcracks and crack morphology. The coupling strength between vibration features and image features is quantified using the Pearson correlation coefficient to distinguish between the two types of microcracks. Classification is performed based on the analysis and matching, generating a classification diagnosis report, and marking the microcrack areas.

[0060] The timing-synchronized GPS module outputs a second pulse signal to trigger synchronous data acquisition by the infrared thermal imager, polarization camera, and vibration sensor, achieving a timestamp accuracy of 1ms. The real-time attitude of the UAV is acquired through the POS system, and the three-dimensional spatial coordinates are calculated based on sensor installation parameters to ensure data alignment in the spatial dimension. Let x be the peak point of the micro-vibration in the vibration signal. k ,y k Let a and b be the fitting model parameters, and t be the threshold. The formula for determining the intrapoint distance between vibration data and image data is as follows:

[0061]

[0062] In the formula, x k ,y k This represents the peak point of micro-vibration in the vibration signal, where a and b are the fitting model parameters, and t is the threshold value when d ck If ≤ t, then it is determined to be an interior point;

[0063] According to d ck The effective distance for filtering, aligning the vibration signal with the time axis of the image sequence, eliminating nonlinear misalignment, and achieving dynamic time warping are as follows:

[0064] D(n,m)=d(T(n),R(m)+min{D(n-1,m),D(n-1,m-1),D(n,m-1)}),

[0065] In the formula, D(n,m) represents the dynamically regularized path, d(T(n),R(m)) represents the Euclidean distance, calculates the difference between the nth frame of the vibration signal T(n) and the nth frame of the image sequence R(m), and min{} represents finding the minimum value to ensure the optimal path; let the vibration feature vector be V and the image feature vector be I, calculate the projection ratio similarity between the vibration feature vector V and the image feature vector I, and implement the formula as follows:

[0066]

[0067] In the formula, Sim(V,I) represents the projection similarity between the dynamic feature vector V and the image feature vector I, and n represents the feature dimension.

[0068] In step S6, a local high-precision rescan is performed on the region marked as a hidden crack to verify the detection results and correct misjudged samples; the marked hidden crack region is obtained, and a local high-precision rescan is performed if the hidden crack probability value is within a preset threshold; when the matching degree between the classification result and the sensor data is lower than the preset threshold, a manual review process is triggered to verify and correct the misjudged results.

[0069] In step S7, the verified hidden crack samples and environmental parameters are uploaded to the cloud. The verified hidden crack samples are then uploaded to the remote end via the HTTP protocol. The local model updates of multiple drones are aggregated through horizontal federated learning. The cloud serves as the receiver of the model gradients of each edge node. After the drones are updated, they are tested using known hidden crack samples to compare the accuracy of the new and old models.

[0070] Among them, an automated detection system based on the microcrack characteristics of photovoltaic modules is characterized by including a data acquisition module, an edge computing and real-time analysis module, a flight state adjustment module, a dynamic path planning module, a classification and diagnosis module, a result verification and correction module, and a cloud-based collaborative optimization module.

[0071] The cloud-based collaborative optimization module uploads the hidden crack samples and environmental parameters to the cloud via HTTPS protocol. The cloud uses the FedAvg algorithm to aggregate the model gradients of multiple drones and update the global convolutional network parameters. The cloud then distributes the updated model difference packets to each drone to reduce bandwidth usage. The drones use known hidden crack samples to test the new and old models and calculate the accuracy improvement. If the accuracy improvement is greater than 2%, the new model is enabled; otherwise, the original model is maintained.

[0072] Working principle: The drone is equipped with an infrared thermal imager, a polarization camera and a vibration sensor. The timing GPS module outputs a second pulse signal to trigger the synchronous acquisition of data by the three devices. Combined with the inertial navigation board and sensor calibration parameters, the acquired thermal, optical and vibration data are mapped to a unified three-dimensional coordinate system to construct a multi-dimensional hidden crack feature dataset containing spatial location information.

[0073] By running a lightweight convolutional network model on the edge computing nodes of the drone, multidimensional data is processed in real time. Anomalies in thermal distribution are identified by comparing the average temperature difference between the target and background regions. Differences in pixel intensity between the polarized image and the background are analyzed to capture microcrack texture features. The clarity of crack edges is quantified by combining the sum of image gradient magnitudes and the magnification factor. A weighted calculation of the probability value of hidden cracks is performed by integrating thermal anomaly factors, texture anomaly factors, and edge sharpness factors. Adaptive altitude and focal length adjustments are made based on this probability value. In low-probability areas, the flight altitude is increased, and the camera switches to wide-angle mode for rapid scanning of large areas. In high-probability areas, the flight altitude is decreased, and the camera switches to telephoto mode to focus on details. Through GPS, inertial measurement unit, and visual feedback, a hidden crack probability calculation is triggered for each frame of image, and the drone's state is dynamically adjusted accordingly. The inspection path is dynamically updated. When the detection result is a high-probability hidden crack area, a multi-angle re-inspection task is inserted to generate a local spiral mesh path. When the detection result is a low-probability area, detailed re-inspection is skipped, and coverage is quickly achieved along the preset route. When the detection result is a medium-probability area, whether to re-inspect is determined based on the remaining power and task priority. When there are many re-inspection areas, the global optimal path is recalculated to balance coverage and energy consumption. Through the coupled analysis of vibration data and image features, a classification diagnosis report is generated, and hidden crack areas are marked. Local high-precision scanning is performed on the marked hidden crack areas. When the classification result is significantly different from the sensor data, a manual review process is triggered to correct misjudged samples and update the local database. The verified hidden crack samples and environmental parameters are uploaded to the cloud via HTTP protocol. After the UAV is updated, the accuracy of the new and old models is tested using known hidden crack samples.

[0074] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their likenesses.

[0075] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. An automated detection method based on the microcrack characteristics of photovoltaic modules, characterized in that, Includes the following steps: S1. The UAV is equipped with an infrared thermal imager, a polarization camera and a vibration sensor to simultaneously collect the thermal distribution, surface texture and micro-vibration signals of photovoltaic modules and construct a multi-dimensional hidden crack feature dataset. S2. Run a lightweight convolutional network through edge computing nodes to analyze thermal anomaly regions in infrared images and microcrack textures in polarization images in real time, and output the probability value of hidden cracks. S3. Automatically adjust the drone's flight altitude and camera focal length based on the probability value of hidden cracks; S4. Dynamically update the inspection path based on the detection results, conduct multi-angle re-inspection of high-probability hidden crack areas, and adopt a rapid scanning mode for low-risk areas. S5. By fusing vibration sensor data and image analysis results, electrical and mechanical microcracks are distinguished and a classification diagnosis report is generated. S6. Perform local high-precision rescanning on the areas marked as microcracks to verify the detection results and correct misjudged samples; S7. Upload the verified hidden crack samples and environmental parameters to the cloud and feed them back to the edge computing nodes to update the convolutional network; In S1, the UAV is equipped with an infrared thermal imager, a polarization camera, and a vibration sensor to simultaneously collect the thermal distribution, surface texture, and micro-vibration signals of the photovoltaic module, and construct a multi-dimensional microcrack feature dataset. A timing-type GPS module is integrated on the UAV to output a second pulse signal as a global time reference. The PPS signal is distributed to the infrared thermal imager, polarization camera, and vibration sensor through a time synchronization controller, triggering the three to collect data simultaneously and automatically resetting the time counter. Combined with the inertial navigation board, the real-time position, attitude, and speed of the UAV are recorded. The installation positions, angles, and interrelationships of the infrared thermal imager, polarization camera, and vibration sensor are pre-calibrated to establish a unified three-dimensional coordinate system. The position and attitude data of the UAV are acquired in real time through the POS system. Combined with the sensor installation parameters, the three-dimensional spatial coordinates of the data collected by each sensor are calculated. When the UAV hovers above the photovoltaic module, the infrared thermal imager, polarization camera, and vibration sensor start collecting data simultaneously after the GPS triggers the PPS signal. The collected heat distribution, polarization texture, and vibration signal data are aligned according to the timestamp to construct a multi-dimensional hidden crack feature dataset. S2 involves running a lightweight convolutional network through edge computing nodes to analyze thermal anomaly regions in infrared images and polarization microcrack textures in polarization images in real time, outputting a hidden crack probability value. A lightweight edge computing unit is deployed on the UAV to perform preliminary processing on the multidimensional hidden crack feature dataset, calculating the hidden crack probability value based on infrared thermal anomalies and polarization microcrack textures. Let the infrared thermal anomaly factor be... The polarization microcrack texture anomaly factor is Edge sharpness factor , Represented as , This represents the average temperature of the target area. This indicates the average temperature of the background area on the component surface. Represented as , These represent the four polarization directions of a polarization camera. Indicates polarization direction Pixel intensity below This represents the average pixel intensity of the background region under the same polarization direction. Represented as , This represents the gradient magnitude of the i-th pixel. This represents the sum of gradient magnitudes of all pixels within the target area, where k represents the gradient amplification factor, adjusted according to the contrast of the photovoltaic module's surface texture. k takes values ​​of... ; S2, based on the infrared thermal anomaly factor Polarization microcrack texture anomaly factor and edge sharpness factor The probability value of microcracks is calculated as follows: ， In the formula, This represents the probability value of microcracks. Represents the natural exponential function Represents the weight parameters. Values , Value , Values , Indicates infrared anomalous factor, Indicator of polarization microcrack texture anomaly factor Indicates the edge sharpness factor; In step S5, vibration sensor data and image analysis results are fused to distinguish between electrical and mechanical microcracks, generating a classification diagnosis report. Multi-angle image data and vibration sensor data are acquired, and vibration features and image texture features of electrical and mechanical microcracks are extracted respectively. The features of vibration data are matched with the texture features of images to analyze whether there is a coupling mode. There is a strong correlation between the vibration features of electrical microcracks and thermal anomalies, and a strong correlation between the vibration features of mechanical microcracks and crack morphology. The coupling strength between vibration features and image features is quantified by the Pearson correlation coefficient to distinguish between the two types of microcracks. Classification is performed based on the analysis and matching, a classification diagnosis report is generated, and the microcrack areas are marked.

2. The automated detection method based on the microcrack characteristics of photovoltaic modules according to claim 1, characterized in that: In step S3, the probability value of the hidden crack is obtained, and the area is divided into low-probability, medium-probability, and high-probability regions based on the probability value. When the area is low-probability, the UAV gradually increases its flight altitude to expand the coverage of a single shot. When a high-probability hidden crack area is detected, the UAV gradually decreases its flight altitude. In the high-probability hidden crack area, the camera switches to telephoto mode to focus on the details of the target area. In the low-probability area, the camera switches to wide-angle mode. The flight altitude and focal length are corrected in real time through GPS, inertial measurement unit, and visual feedback. The hidden crack probability is calculated once for each frame of image, and the UAV state is dynamically adjusted according to the result.

3. The automated detection method based on the microcrack characteristics of photovoltaic modules according to claim 1, characterized in that: S4 dynamically updates the inspection path based on the detection results, performs multi-angle re-inspection on high-probability hidden crack areas, and adopts a fast scanning mode for low-risk areas. In high-probability hidden crack areas, after the UAV completes the initial scan, it automatically triggers a multi-angle re-inspection task and generates a new flight path around the target area. In low-probability areas, the UAV skips detailed re-inspection and directly enters the fast scanning mode to quickly cover along the preset route. The medium-probability areas are determined based on the remaining battery power and task priority. When a high-probability hidden crack area is detected, the current path is immediately interrupted and a re-inspection task is inserted. A dynamic threshold for the number of re-inspection areas is preset. When the number of re-inspection areas is large, the global optimal path is recalculated. When the UAV hovers, global path optimization is performed asynchronously.

4. The automated detection method based on the microcrack characteristics of photovoltaic modules according to claim 1, characterized in that: In step S6, a local high-precision rescan is performed on the region marked as a hidden crack to verify the detection results and correct misjudged samples; the marked hidden crack region is obtained, and a local high-precision rescan is performed if the hidden crack probability value is within a preset threshold; when the matching degree between the classification result and the sensor data is lower than the preset threshold, a manual review process is triggered to verify and correct the misjudged results.

5. The automated detection method based on the microcrack characteristics of photovoltaic modules according to claim 1, characterized in that: In step S7, the verified hidden crack samples and environmental parameters are uploaded to the cloud. The verified hidden crack samples are obtained and uploaded to the remote end via the HTTP protocol. The local model updates of multiple drones are aggregated through horizontal federated learning. The cloud serves as the receiver of the model gradients of each edge node. After the drone is updated, it is tested using known hidden crack samples to compare the accuracy of the new and old models.

6. An automated detection system based on the microcrack characteristics of photovoltaic modules, implemented according to the method of claim 1, characterized in that: It includes a data acquisition module, an edge computing and real-time analysis module, a flight status adjustment module, a dynamic path planning module, a classification and diagnosis module, a result verification and correction module, and a cloud-based collaborative optimization module.

Images