Photovoltaic panel crack detection device and method

Abstract

The invention provides a photovoltaic panel crack detection device and method.The device comprises an elastic expansion moving chassis main body structure and a crack detection module fixed to an elastic expansion moving chassis, the elastic expansion moving chassis main body structure is at least fixedly connected with one wet-type sweeping structure, and the crack detection module comprises an information processing system; the information processing system is electrically connected with an infrared spectrum camera module and a visible light spectrum camera module, and the infrared spectrum camera module and the visible light spectrum camera module send collected photovoltaic panel image information to the information processing system. The information processing system identifies the photovoltaic panel image information based on a multi-layer fusion infrared tiny target detection algorithm to obtain a crack identification result; interference of noise such as water spots can be eliminated, and the overall comprehensive performance of the cleaning robot is improved.

Description

Technical Field

[0001] The invention relates to the technical field of photovoltaic assembly cleaning and maintenance, and in particular to a photovoltaic panel crack detection device and method. Background Art

[0002] In recent years, with the continuous deepening of the research on solar energy by scholars from various countries, the photoelectric conversion efficiency has been significantly improved, the application of photovoltaic power generation technology has become more and more extensive, and the potential of solar energy to become one of the key energy sources to replace various other energy sources has gradually been tapped. By the end of 2022, the cumulative installed capacity of photovoltaic power stations worldwide will be about 1,101GW, of which China's cumulative installed capacity ranks first. This shows that photovoltaic power generation technology is widely used worldwide and has become one of the main clean energy sources in today's era. Studies have shown that if 4.05 grams of dust accumulates on the surface of photovoltaic panels per square meter, its photoelectric conversion efficiency may be reduced by up to 40%. In a humid environment, dust with acidic or alkaline chemical properties will react chemically with the glass or metal materials on the surface of the component, corrode the structure and function of the component, and reduce the mechanical strength and electrical performance of the component. The surface of the corroded component will become rough, causing light to form diffuse reflection on its surface, affecting the light transmittance and photoelectric conversion rate of the component. At the same time, photovoltaic components usually use tempered glass, which has a certain self-explosion rate. If there are quality defects in the component, such as stones, impurities, bubbles, etc., especially impurities in the glass, these weak points are prone to self-explosion in harsh environments. Photovoltaic cells are not preheated at low temperatures, and then suddenly expand when exposed to high temperatures, which may cause hidden cracks, such as excessive welding temperature, unreasonable setting of parameters such as lamination temperature, etc. In addition, due to the difficulty in managing other management personnel of photovoltaic power generation, weak quality awareness, and irregular training of new employees and qualification and training assessment of operators, hidden cracks may occur in the components during the production process. The current photovoltaic cleaning device mainly pursues the cleaning efficiency and cleanliness of photovoltaic panels, improves cleaning efficiency and reduces maintenance costs through modular design, and adopts an integrated double roller brush design to improve the cleaning effect. The market demand for photovoltaic cleaning robots is large and increasing year by year. Summary of the invention

[0003] The present invention solves the problem that the existing photovoltaic cleaning devices have single functions and the photovoltaic panel components have acquired cracks, and proposes a photovoltaic panel crack detection device and method, which can effectively solve the problem of the single structure and function of the photovoltaic cleaning device. At the same time, through the image processing of the micro-crack detection algorithm based on multi-spectral fusion, the interference of noise such as water stains can be eliminated, thereby improving the overall comprehensive performance of the cleaning robot.

[0004] In order to achieve the above objectives, the following technical solutions are proposed: A photovoltaic panel crack detection device comprises an elastically extended mobile chassis main structure and a crack detection module fixed on the elastically extended mobile chassis, wherein the elastically extended mobile chassis main structure is fixedly connected with at least one wet cleaning structure, and the crack detection module comprises an information processing system, wherein the information processing system is electrically connected with an infrared spectrum camera module and a visible light spectrum camera module, and the infrared spectrum camera module and the visible light spectrum camera module send collected photovoltaic panel image information to the information processing system, and the information processing system identifies the photovoltaic panel image information based on a multi-layer fusion infrared small target detection algorithm to obtain a crack identification result.

[0005] The present invention has high infrared small target detection and recognition performance, and the cleaning structure and the crack detection system of the multi-spectral fusion small target detection algorithm have a certain degree of independence. When the photovoltaic cleaning device is working, each part has a certain degree of coordination and integrity.

[0006] The multi-spectral fusion small target detection algorithm of the present invention utilizes interactive feature extraction to effectively fuse feature information of different dimensions, thereby improving the detection performance of irregularly shaped small cracks. The present invention effectively improves the power generation of the power station through automated cleaning and intelligent operation and maintenance, and achieves zero hidden cracks and damage during cleaning through a new design, thus ensuring the safety and efficiency of photovoltaic modules.

[0007] Preferably, the wet cleaning structures are provided with two and are respectively connected to both sides of the elastically extended mobile chassis main structure through cleaning mechanism fixings, and each wet cleaning structure is provided with an independent driving system to form a dual-drive wet cleaning structure.

[0008] The wet cleaning structure of the present invention is provided with two structures respectively connected to the two sides of the main structure of the elastically extended mobile chassis, and each wet cleaning structure is provided with an independent driving system, so that the two wet cleaning structures constitute a dual-drive wet cleaning structure, and each wet cleaning structure is connected to the main structure of the elastically extended mobile chassis through a cleaning mechanism fixing part.

[0009] Preferably, the wet cleaning structure includes a roller brush bracket and a roller brush shell, the roller brush bracket is rotatably connected to a cleaning roller brush, one end of the cleaning roller brush is connected to a driven synchronous pulley, a roller brush drive motor is fixed to the outside of the roller brush bracket, the output end of the roller brush drive motor is connected to an active synchronous pulley, the driven synchronous pulley and the active synchronous pulley are connected through a synchronous belt, an atomizing nozzle is provided on the roller brush shell, a water storage bucket and a micro liquid pump are provided on the main structure of the elastically expanded mobile chassis, and the output end of the micro liquid pump is connected to the atomizing nozzle through a water pipe.

[0010] The wet cleaning structure includes a roller brush bracket, a roller brush housing connected to the roller brush bracket, a roller brush driving motor connected to the roller brush bracket, an active synchronous pulley connected to the roller brush driving motor, a synchronous belt, a driven synchronous pulley, a cleaning roller brush connected to the driven synchronous pulley, an atomizing nozzle installed on the roller brush housing, a micro liquid pump and a water storage bucket installed on the vehicle body housing, and a roller brush motor housing fixedly connected to the roller brush bracket; the cleaning roller brush is connected to the roller brush bracket; the torque of the roller brush driving motor is transmitted to the cleaning roller brush through the active synchronous pulley, the synchronous belt, and the driven synchronous pulley; Preferably, the main structure of the elastically expandable mobile chassis includes a mobile chassis and a body shell, the body shell is fixed on the mobile chassis to form a cavity for storing and fixing various components, and the mobile chassis is a crawler-type mobile device.

[0011] A mobile chassis and a vehicle body shell fixedly connected to the mobile chassis, a driving motor fixedly connected to the mobile chassis, a driving shaft, a driving wheel connected to the driving shaft, a driving wheel shell, a driven wheel, a driven wheel shell, a driven shaft, a track and a mobile battery pack, wherein the torque of the driving motor is transmitted to the driving wheel via the driving shaft, and the torque of the driving motor is transmitted to the driven wheel via the driving wheel, the track and the driven shaft; the driving wheel shell is fixedly connected to the mobile chassis and connected to the driving shaft, the driven wheel shell is fixedly connected to the mobile chassis and connected to the driven shaft, and the mobile battery pack is fixedly connected to the vehicle body shell; the driving wheel is a circular track wheel connected to the driving motor via the driving shaft, and the driven wheel is a circular track wheel connected to the driven shaft.

[0012] Preferably, the crawler-type mobile device includes a pair of drive motors fixed on a mobile chassis, the output end of the drive motor is connected to a drive shaft, the drive shaft is fixed with a drive wheel, the mobile chassis is rotatably connected to a driven shaft, the driven shaft is fixed with a driven wheel, and the drive wheel and the driven wheel are connected by a crawler track.

[0013] Preferably, a driving wheel housing is provided on the outer side of the driving wheel, a driven wheel housing is provided on the outer side of the driven wheel, and the driving wheel housing and the driven wheel housing are fixed on the mobile chassis.

[0014] Preferably, the information processing system comprises: An information preprocessing unit obtains the initial coordinates of the section of the photovoltaic panel to be detected; An image preprocessing unit, which preprocesses the photovoltaic panel image information; The image detection unit detects the pre-processed photovoltaic panel image information based on the multi-layer fusion infrared small target detection algorithm to obtain the detection result; A data storage unit for storing raw data, pre-processed data and test results; Communication unit, communicates with external control systems.

[0015] A photovoltaic panel crack detection method, using the above photovoltaic panel crack detection device, comprises the following steps: S1, setting the photovoltaic cleaning path and the initial coordinates of the photovoltaic panel section to be detected; S2, using a multi-branch method of parallel infrared and visible spectrum image sequences to extract features and capture the multi-scale features of cracks; S3, performing target positioning and adaptive fusion of detection data; S4, based on the multi-layer fusion infrared small target detection algorithm, data processing and noise interference prevention are performed to obtain the crack recognition results.

[0016] Preferably, the S3 specifically comprises the following steps: Use the attention mechanism to locate the cracks and perform adaptive feature enhancement; Adaptive channel selection and fusion using dimension-aware selective integration module; The spatial features are captured by several depth-wise separable convolutional layers to finely model the differences between background and objects.

[0017] Preferably, the S4 specifically comprises the following steps: Using a multi-layer fusion infrared small target detection algorithm, the acquired real-time temperature features are compared with the image infrared threshold, and the image information of the hidden crack area and the corresponding coordinate information are uploaded to the data storage unit.

[0018] The beneficial effects of the present invention are: The present invention has high infrared small target detection and recognition performance, and the cleaning structure and the crack detection system of the multi-spectral fusion small target detection algorithm have a certain degree of independence. When the photovoltaic cleaning device is working, each part has a certain degree of coordination and integrity.

[0019] The multi-spectral fusion small target detection algorithm of the present invention utilizes interactive feature extraction to effectively fuse feature information of different dimensions, thereby improving the detection performance of irregularly shaped small cracks. The present invention effectively improves the power generation of the power station through automated cleaning and intelligent operation and maintenance, and achieves zero hidden cracks and damage during cleaning through a new design, thus ensuring the safety and efficiency of photovoltaic modules. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a schematic structural diagram of the photovoltaic cleaning device of the present invention.

[0021] Figure 2 It is a schematic diagram of the wet cleaning structure of the present invention.

[0022] Figure 3 It is a schematic diagram of the structure of the mobile chassis of the present invention.

[0023] Figure 4 It is a schematic diagram of the main structure of the elastically expandable mobile chassis of the present invention.

[0024] Among them: 1. Mobile chassis; 2. Vehicle body shell; 3. Driving motor; 4. Driving shaft; 5. Driving wheel; 6. Driving wheel shell; 7. Driven wheel; 8. Driven wheel shell; 9. Driven shaft; 10. Track; 11. Roller brush bracket; 12. Roller brush shell; 13. Roller brush driving motor; 14. Active synchronous pulley; 15. Synchronous belt; 16. Driven synchronous pulley; 17. Cleaning roller brush; 18. Atomizing nozzle; 19. Micro liquid pump; 20. Cleaning mechanism fixing parts; 21. Infrared spectrum camera module; 22. Visible light spectrum camera module; 23. Information processing system; 24. Camera module bracket; 25. Mobile battery pack; 26. Roller brush motor shell. DETAILED DESCRIPTION

[0025] Embodiment 1: This embodiment provides a photovoltaic panel crack detection device, referring to Figure 1 , including an elastically extended mobile chassis main structure, a crack detection module and at least one wet cleaning structure, wherein the crack detection module is fixed on the elastically extended mobile chassis main structure, and the wet cleaning structure is fixedly connected to the outside of the elastically extended mobile chassis main structure, wherein the crack detection module includes an infrared spectrum camera module 21, a visible light spectrum camera module 22 and an information processing system 23, wherein the infrared spectrum camera module 21 and the visible light spectrum camera module 22 are electrically connected to the information processing system 23, and the collected photovoltaic panel image information is sent to the information processing system 23, wherein the information processing system 23 is provided with a multi-layer fusion infrared small target detection algorithm, and the photovoltaic panel image information is identified through the multi-layer fusion infrared small target detection algorithm, and finally the crack identification result is obtained.

[0026] The information processing system 23 of the present invention includes an information preprocessing unit for obtaining the initial coordinates of the photovoltaic panel section to be detected, an image preprocessing unit for preprocessing the photovoltaic panel image information, an image detection unit for detecting the preprocessed photovoltaic panel image information based on a multi-layer fusion infrared small target detection algorithm to obtain a detection result, a data storage unit for storing original data, preprocessed data and detection results, and a communication unit for communicating with an external control system.

[0027] The information processing system 23 based on multi-layer fusion infrared micro-target detection pre-processes the collected images, and uses a multi-spectral fusion micro-target detection algorithm to eliminate noise interference such as residual water stains during cleaning, and identify areas where hidden cracks in photovoltaic panels may exist. The information processing system 23 based on multi-layer fusion infrared micro-target detection is connected to the external control system through a communication unit, and can upload the identified crack target data to the target system.

[0028] This embodiment also provides a photovoltaic panel crack detection method, which uses the photovoltaic panel crack detection device mentioned above and includes the following steps: Step 1: Information preprocessing: manually setting the photovoltaic cleaning path of the photovoltaic cleaning device and the initial coordinates of the photovoltaic panel section to be detected; Step 2: Detection data processing based on multi-layer fusion infrared small target detection algorithm: adopt the multi-branch method of parallel infrared and visible spectrum image sequences to extract features and capture the multi-scale features of cracks; Step 3: Target positioning and adaptive fusion: Use the attention mechanism to locate the target and perform adaptive feature enhancement; use the dimension-aware selective integration module to perform adaptive channel selection and fusion; use multiple depth-separable convolutional layers to capture the spatial features of different perception fields and finely model the difference between objects and backgrounds; Step 4: Data processing and noise prevention: Based on the multi-layer fusion infrared small target detection algorithm, the real-time temperature characteristics are obtained through the infrared spectrum camera module. The image information and coordinate information of the hidden crack area are uploaded to the background end through image infrared threshold recognition for subsequent photovoltaic panel repair and maintenance work.

[0029] The present invention has high infrared small target detection and recognition performance, and the cleaning structure and the crack detection system of the multi-spectral fusion small target detection algorithm have a certain degree of independence. When the photovoltaic cleaning device is working, each part has a certain degree of coordination and integrity.

[0030] The multi-spectral fusion small target detection algorithm uses interactive feature extraction to effectively fuse feature information of different dimensions, improving the detection performance of irregular-shaped small cracks. Through automated cleaning and intelligent operation and maintenance, the power generation of the power station has been effectively improved, and the new design has achieved zero hidden cracks and damage during cleaning, ensuring the safety and efficiency of photovoltaic modules.

[0031] Embodiment 2: This embodiment improves the main structure of the elastically extended mobile chassis on the basis of embodiment 1, and proposes a photovoltaic panel crack detection device. Figure 1, including an elastically extended mobile chassis main structure, a crack detection module and at least one wet cleaning structure, wherein the crack detection module is fixed on the elastically extended mobile chassis main structure, and the wet cleaning structure is fixedly connected to the outside of the elastically extended mobile chassis main structure, wherein the crack detection module includes an infrared spectrum camera module 21, a visible light spectrum camera module 22 and an information processing system 23, wherein the infrared spectrum camera module 21 and the visible light spectrum camera module 22 are electrically connected to the information processing system 23, and the collected photovoltaic panel image information is sent to the information processing system 23, wherein the information processing system 23 is provided with a multi-layer fusion infrared small target detection algorithm, and the photovoltaic panel image information is identified through the multi-layer fusion infrared small target detection algorithm, and finally the crack identification result is obtained.

[0032] The information processing system 23 of the present invention includes an information preprocessing unit for obtaining the initial coordinates of the photovoltaic panel section to be detected, an image preprocessing unit for preprocessing the photovoltaic panel image information, an image detection unit for detecting the preprocessed photovoltaic panel image information based on a multi-layer fusion infrared small target detection algorithm to obtain a detection result, a data storage unit for storing original data, preprocessed data and detection results, and a communication unit for communicating with an external control system.

[0033] The information processing system 23 based on multi-layer fusion infrared micro-target detection pre-processes the collected images, and uses a multi-spectral fusion micro-target detection algorithm to eliminate noise interference such as residual water stains during cleaning, and identify areas where hidden cracks in photovoltaic panels may exist. The information processing system 23 based on multi-layer fusion infrared micro-target detection is connected to the external control system through a communication unit, and can upload the identified crack target data to the target system.

[0034] refer to Figure 1 and Figure 3 A mobile chassis 1 and a vehicle body shell 2 fixedly connected to the mobile chassis 1, a driving motor 3 fixedly connected to the mobile chassis 1, a driving shaft 4, a driving wheel 5 connected to the driving shaft 4, a driving wheel shell 6, a driven wheel 7, a driven wheel shell 8, a driven shaft 9, a track 10 and a mobile battery pack 25, wherein the torque of the driving motor 3 is transmitted to the driving wheel 5 via the driving shaft 4, and the torque of the driving motor 3 is transmitted to the driven wheel 7 via the driving wheel 5, the track 10 and the driven shaft 9; the driving wheel shell 6 is fixedly connected to the mobile chassis 1 and connected to the driving shaft 4, the driven wheel shell 8 is fixedly connected to the mobile chassis 1 and connected to the driven shaft 9, and the mobile battery pack 25 is fixedly connected to the vehicle body shell 2; the driving wheel 5 is a circular track wheel connected to the driving motor via the driving shaft 4, and the driven wheel 7 is a circular track wheel connected to the driven shaft 9.

[0035] refer to Figure 4A camera module bracket 24 is fixedly installed on the surface of one end of the body shell 2. The camera bracket is fixedly connected to the infrared spectrum camera module 21 and the visible light spectrum camera module 22. A through hole is left at the bottom of the body shell 2. The infrared spectrum camera module 21 and the visible light spectrum camera module 22 can collect images of the photovoltaic panels at the bottom of the cleaning robot through the hole, and transmit the images to the multi-layer fusion infrared small target detection information processing system 23 via cables.

[0036] refer to Figure 4 A mobile battery pack 25 is installed on one side of the driving motor 3 at the bottom of the body shell 2, and a multi-layer fusion infrared small target detection information processing system 23 is installed on one side of the mobile battery pack 25 at the bottom of the body shell 2. The mobile battery pack 25 is used to power the electrical components in the robot. A charging port is provided on the mobile battery pack 25. The multi-layer fusion infrared small target detection information processing system 23 is used to perform program processing and electrical component control to realize the cleaning robot's autonomous walking, autonomous cleaning, and hidden crack detection functions.

[0037] The driving motors 3 are symmetrically fixedly installed at the front and rear ends of one side of the bottom end of the body shell 2. The output ends of the two driving motors 3 respectively penetrate the mobile chassis 1 and the front and rear ends of the body shell 2 and are fixedly connected to the driving shaft 4. The driving shaft 4 is fixedly installed with a driving wheel 5. The driving motor 3 rotates to drive the driving shaft 4, and then drives the driving wheel 5 to rotate. One side of the driving wheel 5 is located on the front and rear end surfaces of the body shell 2. A driven shaft 9 fixedly connected to the driven wheel shell 8 is installed. A driven wheel 7 is installed on the driven shaft 9. Tracks 10 are installed on the surfaces of the driving wheel 5 and the driven wheel 7. The driving wheel 5 drives the driven wheel 7 to rotate through the track 10, thereby realizing the walking function of the cleaning robot.

[0038] This embodiment also provides a photovoltaic panel crack detection method, which uses the photovoltaic panel crack detection device mentioned above and includes the following steps: Step 1: Information preprocessing: manually setting the photovoltaic cleaning path of the photovoltaic cleaning device and the initial coordinates of the photovoltaic panel section to be detected; Step 2: Detection data processing based on multi-layer fusion infrared small target detection algorithm: adopt the multi-branch method of parallel infrared and visible spectrum image sequences to extract features and capture the multi-scale features of cracks; Step 3: Target positioning and adaptive fusion: Use the attention mechanism to locate the target and perform adaptive feature enhancement; use the dimension-aware selective integration module to perform adaptive channel selection and fusion; use multiple depth-separable convolutional layers to capture the spatial features of different perception fields and finely model the difference between objects and backgrounds; Step 4: Data processing and noise prevention: Based on the multi-layer fusion infrared small target detection algorithm, the real-time temperature characteristics are obtained through the infrared spectrum camera module. The image information and coordinate information of the hidden crack area are uploaded to the background end through image infrared threshold recognition for subsequent photovoltaic panel repair and maintenance work.

[0039] The present invention has high infrared small target detection and recognition performance, and the cleaning structure and the crack detection system of the multi-spectral fusion small target detection algorithm have a certain degree of independence. When the photovoltaic cleaning device is working, each part has a certain degree of coordination and integrity.

[0040] The multi-spectral fusion small target detection algorithm uses interactive feature extraction to effectively fuse feature information of different dimensions, improving the detection performance of irregular-shaped small cracks. Through automated cleaning and intelligent operation and maintenance, the power generation of the power station has been effectively improved, and the new design has achieved zero hidden cracks and damage during cleaning, ensuring the safety and efficiency of photovoltaic modules.

[0041] Embodiment 3: This embodiment improves the structure of the wet cleaning structure on the basis of embodiment 2, and proposes a photovoltaic panel crack detection device. Figure 1 , including an elastically extended mobile chassis main structure, a crack detection module and at least one wet cleaning structure, wherein the crack detection module is fixed on the elastically extended mobile chassis main structure, and the wet cleaning structure is fixedly connected to the outside of the elastically extended mobile chassis main structure, wherein the crack detection module includes an infrared spectrum camera module 21, a visible light spectrum camera module 22 and an information processing system 23, wherein the infrared spectrum camera module 21 and the visible light spectrum camera module 22 are electrically connected to the information processing system 23, and the collected photovoltaic panel image information is sent to the information processing system 23, wherein the information processing system 23 is provided with a multi-layer fusion infrared small target detection algorithm, and the photovoltaic panel image information is identified through the multi-layer fusion infrared small target detection algorithm, and finally the crack identification result is obtained.

[0042] The information processing system 23 of the present invention includes an information preprocessing unit for obtaining the initial coordinates of the photovoltaic panel section to be detected, an image preprocessing unit for preprocessing the photovoltaic panel image information, an image detection unit for detecting the preprocessed photovoltaic panel image information based on a multi-layer fusion infrared small target detection algorithm to obtain a detection result, a data storage unit for storing original data, preprocessed data and detection results, and a communication unit for communicating with an external control system.

[0043] The information processing system 23 based on multi-layer fusion infrared micro-target detection pre-processes the collected images, and uses a multi-spectral fusion micro-target detection algorithm to eliminate noise interference such as residual water stains during cleaning, and identify areas where hidden cracks in photovoltaic panels may exist. The information processing system 23 based on multi-layer fusion infrared micro-target detection is connected to the external control system through a communication unit, and can upload the identified crack target data to the target system.

[0044] refer to Figure 1 and Figure 3 A mobile chassis 1 and a vehicle body shell 2 fixedly connected to the mobile chassis 1, a driving motor 3 fixedly connected to the mobile chassis 1, a driving shaft 4, a driving wheel 5 connected to the driving shaft 4, a driving wheel shell 6, a driven wheel 7, a driven wheel shell 8, a driven shaft 9, a track 10 and a mobile battery pack 25, wherein the torque of the driving motor 3 is transmitted to the driving wheel 5 via the driving shaft 4, and the torque of the driving motor 3 is transmitted to the driven wheel 7 via the driving wheel 5, the track 10 and the driven shaft 9; the driving wheel shell 6 is fixedly connected to the mobile chassis 1 and connected to the driving shaft 4, the driven wheel shell 8 is fixedly connected to the mobile chassis 1 and connected to the driven shaft 9, and the mobile battery pack 25 is fixedly connected to the vehicle body shell 2; the driving wheel 5 is a circular track wheel connected to the driving motor via the driving shaft 4, and the driven wheel 7 is a circular track wheel connected to the driven shaft 9.

[0045] refer to Figure 4 A camera module bracket 24 is fixedly installed on the surface of one end of the body shell 2. The camera bracket is fixedly connected to the infrared spectrum camera module 21 and the visible light spectrum camera module 22. A through hole is left at the bottom of the body shell 2. The infrared spectrum camera module 21 and the visible light spectrum camera module 22 can collect images of the photovoltaic panels at the bottom of the cleaning robot through the hole, and transmit the images to the multi-layer fusion infrared small target detection information processing system 23 via cables.

[0046] refer to Figure 4 A mobile battery pack 25 is installed on one side of the driving motor 3 at the bottom of the body shell 2, and a multi-layer fusion infrared small target detection information processing system 23 is installed on one side of the mobile battery pack 25 at the bottom of the body shell 2. The mobile battery pack 25 is used to power the electrical components in the robot. A charging port is provided on the mobile battery pack 25. The multi-layer fusion infrared small target detection information processing system 23 is used to perform program processing and electrical component control to realize the cleaning robot's autonomous walking, autonomous cleaning, and hidden crack detection functions.

[0047] The driving motors 3 are symmetrically fixedly installed at the front and rear ends of one side of the bottom end of the body shell 2. The output ends of the two driving motors 3 respectively penetrate the mobile chassis 1 and the front and rear ends of the body shell 2 and are fixedly connected to the driving shaft 4. The driving shaft 4 is fixedly installed with a driving wheel 5. The driving motor 3 rotates to drive the driving shaft 4, and then drives the driving wheel 5 to rotate. One side of the driving wheel 5 is located on the front and rear end surfaces of the body shell 2. A driven shaft 9 fixedly connected to the driven wheel shell 8 is installed. A driven wheel 7 is installed on the driven shaft 9. Tracks 10 are installed on the surfaces of the driving wheel 5 and the driven wheel 7. The driving wheel 5 drives the driven wheel 7 to rotate through the track 10, thereby realizing the walking function of the cleaning robot.

[0048] refer to Figure 1 and Figure 2 The wet cleaning structure of the present invention is provided with two structures respectively connected to the two sides of the main structure of the elastically extended mobile chassis, and each wet cleaning structure is provided with an independent driving system, so that the two wet cleaning structures constitute a dual-drive wet cleaning structure, and each wet cleaning structure is connected to the main structure of the elastically extended mobile chassis through a cleaning mechanism fixing part 20.

[0049] The wet cleaning structure includes a roller brush bracket 11, a roller brush housing 12 connected to the roller brush bracket 11, a roller brush driving motor 13 connected to the roller brush bracket 11, an active synchronous pulley 14 connected to the roller brush driving motor 13, a synchronous belt 15, a driven synchronous pulley 16, a cleaning roller brush 17 connected to the driven synchronous pulley 16, an atomizing nozzle 18 installed on the roller brush housing 12, a micro liquid pump 19 and a water storage tank installed on the vehicle body housing 2, and a roller brush motor housing 26 fixedly connected to the roller brush bracket; the cleaning roller brush 17 is connected to the roller brush bracket 11; the torque of the roller brush driving motor 13 is transmitted to the cleaning roller brush 17 through the active synchronous pulley 14, the synchronous belt 15, and the driven synchronous pulley 16; A cleaning mechanism fixing part 20 is fixedly installed on the surface of one end of the body shell 2, and the cleaning mechanism fixing part 20 is fixedly connected to the roller brush shell 12, the roller brush shell 12 is fixedly connected to the roller brush bracket 11, the roller brush bracket 11 is connected to the roller brush driving motor 13 and the cleaning roller brush 17, the roller brush driving motor 13 is fixedly connected to the active synchronous pulley 14, and the driven synchronous pulley 16 is fixedly connected to the cleaning roller brush 17. The active synchronous pulley 14 and the driven synchronous pulley 16 are equipped with a synchronous belt 15. When the roller brush driving motor 13 rotates, the cleaning roller brush 17 can be driven to rotate to sweep away dust on the surface of the photovoltaic panel, thereby realizing the cleaning function of the cleaning robot.

[0050] This embodiment also provides a photovoltaic panel crack detection method, which uses the photovoltaic panel crack detection device mentioned above and includes the following steps: Step 1: Information preprocessing: manually setting the photovoltaic cleaning path of the photovoltaic cleaning device and the initial coordinates of the photovoltaic panel section to be detected; Step 2: Detection data processing based on multi-layer fusion infrared small target detection algorithm: adopt the multi-branch method of parallel infrared and visible spectrum image sequences to extract features and capture the multi-scale features of cracks; Step 3: Target positioning and adaptive fusion: Use the attention mechanism to locate the target and perform adaptive feature enhancement; use the dimension-aware selective integration module to perform adaptive channel selection and fusion; use multiple depth-separable convolutional layers to capture the spatial features of different perception fields and finely model the difference between objects and backgrounds; Step 4: Data processing and noise prevention: Based on the multi-layer fusion infrared small target detection algorithm, the real-time temperature characteristics are obtained through the infrared spectrum camera module. The image information and coordinate information of the hidden crack area are uploaded to the background end through image infrared threshold recognition for subsequent photovoltaic panel repair and maintenance work.

[0051] The present invention has high infrared small target detection and recognition performance, and the cleaning structure and the crack detection system of the multi-spectral fusion small target detection algorithm have a certain degree of independence. When the photovoltaic cleaning device is working, each part has a certain degree of coordination and integrity.

[0052] The multi-spectral fusion small target detection algorithm uses interactive feature extraction to effectively fuse feature information of different dimensions, improving the detection performance of irregular-shaped small cracks. Through automated cleaning and intelligent operation and maintenance, the power generation of the power station has been effectively improved, and the new design has achieved zero hidden cracks and damage during cleaning, ensuring the safety and efficiency of photovoltaic modules.

Claims

1. A photovoltaic panel crack detection device, characterized in that: The invention comprises an elastically expandable mobile chassis main structure and a crack detection module fixed on the elastically expandable mobile chassis, wherein the elastically expandable mobile chassis main structure is fixedly connected to at least one wet cleaning structure, and the crack detection module comprises an information processing system (23), wherein the information processing system (23) is electrically connected to an infrared spectrum camera module (21) and a visible light spectrum camera module (22), wherein the infrared spectrum camera module (21) and the visible light spectrum camera module (22) send collected photovoltaic panel image information to the information processing system (23), and the information processing system (23) identifies the photovoltaic panel image information based on a multi-layer fusion infrared small target detection algorithm to obtain a crack identification result.

2. A photovoltaic panel crack detection device according to claim 1, characterized in that: The wet cleaning structures are provided with two and are respectively connected to both sides of the elastically extended mobile chassis main structure through cleaning mechanism fixing parts (20), and each wet cleaning structure is provided with an independent driving system to form a dual-drive wet cleaning structure.

3. A photovoltaic panel crack detection device according to claim 1, characterized in that: The wet cleaning structure comprises a roller brush bracket (11) and a roller brush housing (12); the roller brush bracket (11) is rotatably connected to a cleaning roller brush (17); one end of the cleaning roller brush (17) is connected to a driven synchronous pulley (16); a roller brush driving motor (13) is fixed to the outside of the roller brush bracket (11); the output end of the roller brush driving motor (13) is connected to a driving synchronous pulley (14); the driven synchronous pulley (16) and the driving synchronous pulley (14) are connected via a synchronous belt (15); an atomizing nozzle (18) is provided on the roller brush housing (12); a water storage bucket and a micro liquid pump (19) are provided on the main structure of the elastically expanded mobile chassis; the output end of the micro liquid pump (19) is connected to the atomizing nozzle (18) via a water pipe.

4. A photovoltaic panel crack detection device according to claim 1, characterized in that: The main structure of the elastically expandable mobile chassis comprises a mobile chassis (1) and a vehicle body shell (2); the vehicle body shell (2) is fixed on the mobile chassis (1) to form a cavity for storing and fixing various components; the mobile chassis (1) is a crawler-type mobile device.

5. A photovoltaic panel crack detection device according to claim 4, characterized in that: The crawler-type mobile device comprises a pair of drive motors (3) fixed on a mobile chassis (1), the output end of the drive motor (3) is connected to a drive shaft (4), the drive shaft (4) is fixed with a drive wheel (5), the mobile chassis (1) is rotatably connected with a driven shaft (9), the driven shaft (9) is fixed with a driven wheel (7), and the drive wheel (5) and the driven wheel (7) are connected via a crawler (10).

6. A photovoltaic panel crack detection device according to claim 5, characterized in that: A driving wheel housing (6) is provided on the outside of the driving wheel (5), and a driven wheel housing (8) is provided on the outside of the driven wheel (7). The driving wheel housing (6) and the driven wheel housing (8) are fixed on the mobile chassis (1).

7. A photovoltaic panel crack detection device according to any one of claims 1 to 6, characterized in that: The information processing system (23) comprises: An information preprocessing unit obtains the initial coordinates of the section of the photovoltaic panel to be detected; An image preprocessing unit, which preprocesses the photovoltaic panel image information; The image detection unit detects the pre-processed photovoltaic panel image information based on the multi-layer fusion infrared small target detection algorithm to obtain the detection result; A data storage unit for storing raw data, pre-processed data and test results; Communication unit, communicates with external control systems.

8. A photovoltaic panel crack detection method, using a photovoltaic panel crack detection device as claimed in claim 7, characterized in that: The following steps are involved: S1, setting the photovoltaic cleaning path and the initial coordinates of the photovoltaic panel section to be detected; S2, using a multi-branch method that runs in parallel with infrared and visible spectrum sequences to extract features and capture multi-scale features of cracks; S3, target positioning and adaptive fusion of detection data; S4, based on the multi-layer fusion infrared small target detection algorithm, data processing and noise interference prevention are performed to obtain the crack recognition results.

9. A photovoltaic panel crack detection method according to claim 8, characterized in that: The S3 specifically includes the following steps: using an attention mechanism to locate the crack target and perform adaptive feature enhancement; Adaptive channel selection and fusion using dimension-aware selective integration module; The spatial features are captured by several depth-wise separable convolutional layers to finely model the differences between background and objects.

10. A photovoltaic panel crack detection method according to claim 8, characterized in that: The S4 specifically includes the following steps: using a multi-layer fusion infrared small target detection algorithm, the acquired real-time temperature features are compared with the image infrared threshold, and the image information of the area judged to be a hidden crack area and the corresponding coordinate information are uploaded to the data storage unit.

Images