Photovoltaic panel counting method and device, counting equipment and storage medium

Abstract

The invention relates to a photovoltaic panel counting method and device, counting equipment and a storage medium, and belongs to the technical field of photovoltaic power station management monitoring, and the photovoltaic panel counting method comprises the steps: obtaining a real-time image of a photovoltaic power station, and carrying out the preprocessing of the real-time image of the photovoltaic power station, and obtaining a real-time image of a target photovoltaic power station; the real-time image of the target photovoltaic power station is input into a photovoltaic panel recognition model, a recognition result output by the photovoltaic panel recognition model is obtained, and the recognition result comprises a bounding box of each photovoltaic panel in the real-time image of the target photovoltaic power station and the confidence coefficient of the bounding box of each photovoltaic panel; and determining a photovoltaic panel counting result corresponding to the real-time image of the target photovoltaic power station based on the identification result. According to the invention, the counting efficiency and identification precision of the photovoltaic panel are effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of photovoltaic power station management and monitoring, and particularly relates to a photovoltaic panel counting method and device, a counting equipment and a storage medium. BACKGROUND

[0002] With the rapid development of the photovoltaic power generation industry, the scale of photovoltaic power stations is continuously expanding, and the demand for real-time monitoring and management of the number of photovoltaic panels is increasing. Traditional photovoltaic panel counting methods mainly include manual inspection and automated counting based on traditional image processing technology, but the traditional photovoltaic panel counting method has the following problems: Manual inspection is inefficient and lacks real-time performance: it relies on manual counting one by one, which is time-consuming and labor-intensive, and in large-scale power stations, it is easy to miss or repeat counting. The processing speed of existing systems is slow, and it cannot meet the demand for large-scale real-time monitoring. Traditional algorithms have great limitations: image processing methods based on edge detection or template matching are difficult to adapt to complex scenes such as changes in illumination, angle differences, and occlusions, and have insufficient recognition accuracy and robustness.

[0003] Therefore, how to improve the efficiency and recognition accuracy of photovoltaic panel counting has become a technical problem that needs to be solved. SUMMARY

[0004] Therefore, it is necessary to provide a photovoltaic panel counting method, device, counting equipment and storage medium to solve the problem of low efficiency and recognition accuracy of the existing photovoltaic panel counting method.

[0005] In order to solve the above problems, in a first aspect, the present application provides a photovoltaic panel counting method, comprising: obtaining a real-time image of a photovoltaic power station and preprocessing the real-time image of the photovoltaic power station to obtain a target real-time image of the photovoltaic power station; inputting the target real-time image of the photovoltaic power station into a photovoltaic panel recognition model to obtain a recognition result output by the photovoltaic panel recognition model, the recognition result including a bounding box of each photovoltaic panel in the target real-time image of the photovoltaic power station and a confidence score of each photovoltaic panel bounding box; determining a photovoltaic panel counting result corresponding to the target real-time image of the photovoltaic power station based on the recognition result; The photovoltaic panel recognition model includes a feature extraction network, a feature fusion network, and a detection network, and the feature fusion network performs feature fusion based on an SPPF network and an FPN network.

[0006] In a possible implementation, the preprocessing of the real-time image of the photovoltaic power station to obtain the target real-time image of the photovoltaic power station includes: The real-time image of the photovoltaic power station is sequentially subjected to image enhancement processing, noise removal processing, illumination compensation processing and region segmentation processing to obtain a target real-time image of the photovoltaic power station.

[0007] In a possible implementation, the image enhancement processing is implemented based on a contrast stretching algorithm, the noise removal processing is implemented based on a median filtering algorithm, the illumination compensation processing is implemented based on histogram equalization and a cumulative distribution function, and the region segmentation processing is implemented based on a Canny edge detection algorithm.

[0008] In a possible implementation, the photovoltaic panel recognition model determines model parameters based on an SGD optimizer in a training process, and a learning rate of the SGD optimizer is determined based on a cosine annealing strategy.

[0009] In a possible implementation, the photovoltaic panel recognition model further includes a post-processing network configured to eliminate repeated photovoltaic panel bounding boxes based on an NMS algorithm.

[0010] In a possible implementation, the determining, based on the recognition result, of a photovoltaic panel counting result corresponding to the target real-time image of the photovoltaic power station includes: determining, based on the recognition result, a photovoltaic panel bounding box with a confidence value greater than a confidence threshold value; performing filtering processing on the photovoltaic panel bounding box with the confidence value greater than the confidence threshold value, and determining a number of photovoltaic panel bounding boxes with the confidence value greater than the confidence threshold value after filtering as the photovoltaic panel counting result corresponding to the target real-time image of the photovoltaic power station.

[0011] In a possible implementation, the filtering processing on the photovoltaic panel bounding box with the confidence value greater than the confidence threshold value includes: merging overlapping photovoltaic panel bounding boxes based on a DBSCAN clustering algorithm, and filtering misrecognized photovoltaic panel bounding boxes based on a bounding box area threshold value and an aspect ratio threshold value.

[0012] In another aspect, the present application also provides a photovoltaic panel counting device, including: a processing module configured to acquire a real-time image of a photovoltaic power station, and perform preprocessing on the real-time image of the photovoltaic power station to obtain a target real-time image of the photovoltaic power station; an identification module configured to input the target real-time image of the photovoltaic power station into a photovoltaic panel recognition model to obtain a recognition result output by the photovoltaic panel recognition model, the recognition result including a bounding box of each photovoltaic panel in the target real-time image of the photovoltaic power station and a confidence value of each photovoltaic panel bounding box; a counting module configured to determine, based on the recognition result, a photovoltaic panel counting result corresponding to the target real-time image of the photovoltaic power station; The photovoltaic panel recognition model comprises a feature extraction network, a feature fusion network and a detection network, and the feature fusion network performs feature fusion based on an SPPF network and an FPN network.

[0013] In a second aspect, the present application further provides a counting device comprising a memory and a processor, wherein, The memory is configured to store a program. The processor is coupled to the memory and configured to execute the program stored in the memory to implement the steps of the photovoltaic panel counting method in any of the above implementation manners.

[0014] In a third aspect, the present application further provides a computer readable storage medium for storing computer readable programs or instructions, which can implement the steps of the photovoltaic panel counting method in any of the above implementation manners when executed by a processor.

[0015] The photovoltaic panel counting method, device, counting device and storage medium provided by the present application have the following beneficial effects: first, the real-time image of the photovoltaic power station is acquired, then the image quality is improved and the environmental interference is reduced through preprocessing, the target real-time image of the photovoltaic power station is input into the photovoltaic panel recognition model, the feature fusion is performed through the combination of the SPPF network and the FPN network to strengthen the recognition accuracy of the photovoltaic panel recognition model for photovoltaic panels of different sizes, the recognition result containing the bounding box of each photovoltaic panel in the target real-time image of the photovoltaic power station and the confidence of each photovoltaic panel bounding box is obtained, the photovoltaic panel counting result is determined through the recognition result, the whole photovoltaic panel counting process can be automatically run, thereby improving the efficiency of photovoltaic panel counting, and the recognition accuracy of photovoltaic panel counting is effectively improved through image preprocessing, feature fusion for photovoltaic panels of different sizes, etc., and the efficiency and recognition accuracy of photovoltaic panel counting are effectively improved. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 An embodiment flowchart of the photovoltaic panel counting method provided by the present application is shown in the figure; Figure 2 An embodiment flowchart of the image preprocessing process provided by the present application is shown in the figure; Figure 3 An embodiment flowchart of the photovoltaic panel detection and counting process provided by the present application is shown in the figure; Figure 4 An embodiment structure diagram of the photovoltaic panel counting device provided by the present application is shown in the figure; Figure 5 An embodiment structure diagram of the counting device provided by the present application is shown in the figure. DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the present application will be clearly and completely described in connection with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0018] In the description of the embodiments of the present application, unless otherwise specified, the meaning of "a plurality of" is two or more. The association relationship of the associated objects is described by "and / or", which means that there can be three relationships, for example: A and / or B can represent the following three cases: A exists alone, A and B exist simultaneously, and B exists alone.

[0019] The "first", "second", and the like described in the embodiments of the present application are only for the purpose of description, and cannot be understood as indicating or implying the relative importance of the indicated technical features or implicitly indicating the number of the indicated technical features. Therefore, the technical features limited by "first" and "second" can explicitly or implicitly include at least one of the features.

[0020] In this document, referring to "embodiments" means that the specific features, structures, or properties described in connection with the embodiments can be included in at least one embodiment of the present application. The phrase appears at various places in the specification does not necessarily all refer to the same embodiment, nor is it necessarily mutually exclusive of other embodiments. It is explicitly and implicitly understood by a person of ordinary skill in the art that the embodiments described herein can be combined with other embodiments.

[0021] The present application provides a photovoltaic panel counting method, device, counting equipment and storage medium, which are described below respectively.

[0022] It should be noted that the photovoltaic panel counting method provided by the present application can be executed by a desktop computer, a portable computer, a handheld terminal, etc. The photovoltaic panel counting method provided by the present application can be applied to the photovoltaic panel counting scene of a photovoltaic power station, and can also be applied to other scenes that need to count equipment through images, which is not limited by the present application.

[0023] Figure 1 An embodiment flowchart of the photovoltaic panel counting method provided by the present application is shown in FIG. 1. Figure 1 As shown in FIG. 1, the photovoltaic panel counting method comprises the following steps. S101, acquiring a real-time image of a photovoltaic power station, and preprocessing the real-time image of the photovoltaic power station to obtain a target real-time image of the photovoltaic power station.

[0024] It should be noted that: since the photovoltaic panel counting method of the present application is based on images, the real-time image of the photovoltaic power station needs to be obtained first, for example, the real-time image can be collected by a high-resolution industrial camera (resolution not less than 1920*1080) equipped with a wide-angle lens, the camera can be deployed at a key position of the photovoltaic power station (such as the top of the support or a drone), to realize full-area coverage, in addition to real-time image acquisition, if in order to reduce the data transmission pressure, images can also be collected at regular intervals (for example, every hour). After obtaining the real-time image of the photovoltaic power station, in order to improve the image quality and reduce environmental interference, and improve the accuracy of subsequent image recognition, the real-time image of the target photovoltaic power station can be preprocessed to obtain the target photovoltaic power station real-time image.

[0025] S102, input the target photovoltaic power station real-time image into the photovoltaic panel recognition model, obtain the recognition result output by the photovoltaic panel recognition model, the recognition result includes the boundary box of each photovoltaic panel in the target photovoltaic power station real-time image and the confidence of each photovoltaic panel boundary box.

[0026] Among them, the photovoltaic panel recognition model includes a feature extraction network, a feature fusion network and a detection network, the feature fusion network is based on SPPF network and FPN network for feature fusion.

[0027] It should be noted that: the present application identifies the target photovoltaic power station real-time image through the photovoltaic panel recognition model, inputs the target photovoltaic power station real-time image into the photovoltaic panel recognition model during identification, and can obtain the recognition result output by the photovoltaic panel recognition model. The recognition result contains the boundary box of each photovoltaic panel in the target photovoltaic power station real-time image and the confidence of each photovoltaic panel boundary box, thereby providing a basis for subsequent photovoltaic panel counting. The photovoltaic panel recognition model is composed of a feature extraction network, a feature fusion network and a detection network, the feature fusion network performs feature fusion through a spatial pyramid fast pooling (Spatial Pyramid Pooling Fast, SPPF) network and a feature pyramid network (Feature Pyramid Network, FPN), so as to adapt to photovoltaic panels of different sizes and improve the recognition accuracy of high-level semantic features and low-level detail features.

[0028] S103, determining the photovoltaic panel counting result corresponding to the target photovoltaic power station real-time image based on the recognition result.

[0029] It should be noted that: finally, the counting result of the photovoltaic panel corresponding to the target photovoltaic power station real-time image can be determined through the recognition result. Generally, the counting result of the photovoltaic panel can be obtained according to the number of photovoltaic panel bounding boxes with a confidence higher than a confidence threshold in the recognition result, however, there may still be some errors in the recognition result, therefore, the recognition result can be further filtered to improve the recognition accuracy of counting, such as filtering the bounding boxes with obviously abnormal size, etc.

[0030] In summary, the photovoltaic panel counting method provided by the embodiment of the application first acquires a photovoltaic power station real-time image, then improves the image quality and reduces environmental interference through preprocessing, inputs the target photovoltaic power station real-time image into a photovoltaic panel recognition model, combines the SPPF network and the FPN network to perform feature fusion to strengthen the recognition accuracy of the photovoltaic panel recognition model on photovoltaic panels of different sizes, obtains a recognition result containing the bounding box of each photovoltaic panel in the target photovoltaic power station real-time image and the confidence of each photovoltaic panel bounding box, and determines the counting result of the photovoltaic panel through the recognition result. The entire photovoltaic panel counting process can be automatically run, thereby improving the efficiency of photovoltaic panel counting, and the recognition accuracy of photovoltaic panel counting is effectively improved through image preprocessing, feature fusion for photovoltaic panels of different sizes, etc. The efficiency and recognition accuracy of photovoltaic panel counting are effectively improved.

[0031] In some embodiments of the application, the preprocessing of the photovoltaic power station real-time image to obtain the target photovoltaic power station real-time image comprises: sequentially performing image enhancement processing, noise removal processing, illumination compensation processing and region segmentation processing on the photovoltaic power station real-time image to obtain the target photovoltaic power station real-time image.

[0032] It should be noted that: when the photovoltaic power station real-time image is preprocessed, the photovoltaic power station real-time image can be sequentially subjected to image enhancement processing, noise removal processing, illumination compensation processing and region segmentation processing to obtain the target photovoltaic power station real-time image, thereby improving the image quality and reducing environmental interference.

[0033] In some embodiments of the application, the image enhancement processing is implemented based on a contrast stretching algorithm, the noise removal processing is implemented based on a median filter algorithm, the illumination compensation processing is implemented based on histogram equalization and cumulative distribution function, and the region segmentation processing is implemented based on a Canny edge detection algorithm.

[0034] It should be noted that: the image enhancement processing can be implemented through a contrast stretching algorithm, and the specific formula is as follows:

[0035] wherein, is the original pixel value, and are the minimum and maximum gray values of the image.

[0036] The noise removal processing can be implemented by a median filtering algorithm, and the specific window size can be set to 3x3.

[0037] The illumination compensation processing can be implemented by histogram equalization and a cumulative distribution function, and the specific formula is as follows:

[0038] wherein, represents a cumulative distribution function, is an original pixel value which can be obtained by histogram equalization.

[0039] The region segmentation processing can be implemented by a Canny edge detection algorithm, for example, a low threshold value of 50 and a high threshold value of 150 can be set to extract the photovoltaic panel contour and reduce background interference.

[0040] In some embodiments of the present application, the photovoltaic panel recognition model determines the model parameters based on an SGD optimizer during the training process, and the learning rate of the SGD optimizer is determined based on a cosine annealing strategy.

[0041] It should be noted that when training the photovoltaic panel recognition model, the model parameters can be determined by a Stochastic Gradient Descent (SGD) optimizer, the initial learning rate can be set to 0.01, the momentum can be set to 0.937, the batch size can be set to 32, and the training iteration can be 200 times. In order to better optimize the model parameters, the learning rate of the SGD optimizer can be determined by a cosine annealing strategy.

[0042] In some embodiments of the present application, the photovoltaic panel recognition model further includes a post-processing network for eliminating duplicate photovoltaic panel bounding boxes based on an NMS algorithm.

[0043] It should be noted that in order to further improve the recognition accuracy, a post-processing network can also be set in the photovoltaic panel recognition model to eliminate duplicate photovoltaic panel bounding boxes by a Non-Maximum Suppression (NMS) algorithm.

[0044] In some embodiments of the present application, the determination of the photovoltaic panel count result corresponding to the real-time image of the target photovoltaic power station based on the recognition result comprises: determining a photovoltaic panel bounding box with a confidence greater than a confidence threshold based on the recognition result; The bounding boxes of photovoltaic panels with a confidence level greater than the confidence level threshold are filtered, and the number of the filtered bounding boxes with a confidence level greater than the confidence level threshold is determined as the photovoltaic panel count result corresponding to the real-time image of the target photovoltaic power station.

[0045] It should be noted that when determining the photovoltaic panel count result corresponding to the real-time image of the target photovoltaic power station based on the recognition results, the bounding boxes of photovoltaic panels with a confidence level greater than the confidence threshold (e.g., 0.8) in the recognition results can be identified first. Then, the bounding boxes of photovoltaic panels with a confidence level greater than the confidence threshold can be filtered to further improve the recognition accuracy and obtain the photovoltaic panel count result corresponding to the real-time image of the target photovoltaic power station.

[0046] In some embodiments of the present invention, the filtering process for photovoltaic panel bounding boxes with confidence scores greater than a confidence threshold includes: The DBSCAN clustering algorithm is used to merge overlapping photovoltaic panel bounding boxes, and misidentified photovoltaic panel bounding boxes are filtered out based on bounding box area thresholds and aspect ratio thresholds.

[0047] It should be noted that when filtering photovoltaic panel bounding boxes with confidence scores greater than the confidence threshold, overlapping photovoltaic panel bounding boxes can first be merged using the DBSCAN clustering algorithm (for example, the distance threshold can be set to 20 pixels and the minimum number of samples can be set to 2). Then, misidentified photovoltaic panel bounding boxes can be filtered out using a bounding box area threshold (e.g., 1000 pixels) and an aspect ratio threshold (which can be set to 0.5-2) to further improve recognition accuracy.

[0048] Finally, after obtaining the real-time count results of the photovoltaic panels, time series analysis (with a sliding window size of 24 hours) can be used to compare with historical data to detect whether there are any abnormalities in the number of photovoltaic panels.

[0049] To address the problems of low efficiency, insufficient recognition accuracy, and poor real-time performance in traditional manual inspection methods, and to achieve automated, efficient, and accurate counting of photovoltaic panels, adapting to complex and ever-changing application scenarios, this invention provides a photovoltaic panel counting method, which specifically includes the following modules: 1. Image acquisition module: Equipment: High-resolution industrial cameras (resolution no less than 1920×1080) are used, equipped with wide-angle lenses, and deployed in key locations of photovoltaic power plants (such as the top of the support structure or drones) to achieve full area coverage.

[0050] Data collection method: Supports timed photo capture (once per hour) or real-time video stream capture, with data transmitted to the server via 5G network or fiber optic cable.

[0051] 2. Data Preprocessing Module: Combination Figure 2To enhance image quality and reduce environmental interference, the preprocessing steps include: Image Enhancement: Apply contrast stretching algorithm, formula is , where is the original pixel value, and are the minimum and maximum gray values of the image.

[0052] Noise Removal: Use median filter algorithm, window size is 3x3, calculation formula is .

[0053] Illumination Compensation: Apply histogram equalization, adjust pixel values based on cumulative distribution function (CDF), formula is .

[0054] Region Segmentation: Use Canny edge detection algorithm, set low threshold 50 and high threshold 150, extract photovoltaic panel outline, reduce background interference.

[0055] 3、DeepSeek Large Model Recognition Module: In combination Figure 3 , this module is based on deep convolutional neural network (CNN) to build target detection model, specific implementation as follows: 3.1、Model Architecture: 3.1.1、Basic Network: Use YOLOv5s model, including Backbone (CSPDarknet53), Neck (SPPF and FPN) and Head (prediction layer).

[0056] 3.1.2、Multi-scale detection: Through feature pyramid network (FPN) to fuse high-level semantic features and low-level detailed features, adapt to photovoltaic panels of different sizes.

[0057] 3.2、Model Training: 3.2.1、Dataset: Collect 10000 photovoltaic panel images, manually label bounding boxes.

[0058] 3.2.2、Data Augmentation: Apply random rotation (±15°), scaling (0.8-1.2 times), horizontal flip and brightness adjustment (±20%), expand to 50000 images.

[0059] 3.2.3、Transfer Learning: Initialize Backbone with ImageNet pre-trained CSPDarknet53 weights.

[0060] 3.2.4、Loss Function: Use CIoU loss, formula is , where is the intersection over union, is the center point distance, For diagonal distance, Measure the consistency of the aspect ratio.

[0061] 3.2.5, Optimization algorithm: use SGD optimizer, initial learning rate 0.01, momentum 0.937, batch size 32, train 200 epochs.

[0062] 3.2.6, Hyperparameter adjustment: learning rate uses cosine annealing strategy, formula is Where is the current iteration step, is the total number of steps.

[0063] 3.2.7, Evaluation index: on the validation set, mAP reaches 96.5%, recall rate is 98%.

[0064] 3.3, Model inference: 3.3.1, Input preprocessed image, size adjustment to 640x640.

[0065] 3.3.2, Extract features through Backbone, fuse multi-scale features through Neck, and predict photovoltaic panel bounding box and confidence through Head.

[0066] 3.3.3, Post-processing uses non-maximum suppression (NMS), IoU threshold is 0.5, confidence threshold is 0.8, to ensure that each photovoltaic panel is detected only once.

[0067] 4, Counting and data statistics module: Counting method: count the number of detection boxes with confidence higher than 0.8.

[0068] Anti-repetition counting: use DBSCAN clustering algorithm (distance threshold 20 pixels, minimum sample size 2) to merge overlapping boxes.

[0069] Abnormal filtering: set area threshold (>1000 pixels) and aspect ratio threshold (0.5-2.0) to remove misidentified targets.

[0070] Data analysis: based on time series analysis (sliding window size 24 hours), compare with historical data, detect number anomalies.

[0071] 5, Result output and feedback module: Interface design: develop Web interface, use React framework to show real-time counting results and photovoltaic panel position heat map.

[0072] Alarm mechanism: if the number of photovoltaic panels decreases by more than 5%, send email to operation and maintenance personnel, or push to platform.

[0073] Cloud support: Support model updates through Over-the-Air Technology (OTA) to optimize performance.

[0074] The photovoltaic panel counting method provided by the application has the following advantages: High precision: The model mAP reaches 96.5%, significantly improving the recognition accuracy.

[0075] High efficiency: The processing time of a single image is less than 0.8 seconds, meeting the real-time requirements.

[0076] Robustness: Adapt to complex scenes such as light changes and occlusions.

[0077] Automation: The whole process does not require human intervention, reducing costs.

[0078] At the same time, the photovoltaic panel counting method provided by the application is tested in a 100MW photovoltaic power station, and the system accuracy reaches 97%, about 120 images are processed per day, and the workload of manual inspection is reduced by 80%. The application realizes efficient and accurate counting of photovoltaic panels through the DeepSeek large model, overcoming the shortcomings of traditional methods. The model mAP reaches 96.5%, the processing speed is fast, and the adaptability is strong, which can be widely applied to intelligent management of photovoltaic power stations, reducing costs and improving efficiency.

[0079] In order to better implement the photovoltaic panel counting method in the embodiments of the application, on the basis of the photovoltaic panel counting method, as shown in Figure 4 The application also provides a photovoltaic panel counting device, which comprises: The processing module 401 is used for acquiring real-time images of a photovoltaic power station and preprocessing the real-time images of the photovoltaic power station to obtain target real-time images of the photovoltaic power station; The recognition module 402 is used for inputting the target real-time images of the photovoltaic power station into a photovoltaic panel recognition model to obtain a recognition result output by the photovoltaic panel recognition model, wherein the recognition result comprises a bounding box of each photovoltaic panel in the target real-time images of the photovoltaic power station and a confidence degree of each photovoltaic panel bounding box; The counting module 403 is used for determining a photovoltaic panel counting result corresponding to the target real-time images of the photovoltaic power station based on the recognition result. The photovoltaic panel recognition model comprises a feature extraction network, a feature fusion network and a detection network, and the feature fusion network performs feature fusion based on an SPPF network and an FPN network.

[0080] The photovoltaic panel counting device 400 provided in the above embodiments can realize the technical solutions described in the above photovoltaic panel counting method embodiments, and the principles of the implementation of the above modules or units can be referred to the corresponding content in the above photovoltaic panel counting method embodiments, which will not be described here.

[0081] like Figure 5 As shown, the present invention also provides a counting device 500. The counting device 500 includes a processor 501, a memory 502, and a display 503. Figure 5 Only a portion of the components of the counting device 500 are shown; however, it should be understood that implementation of all shown components is not required, and more or fewer components may be implemented instead.

[0082] In some embodiments, processor 501 may be a central processing unit (CPU), microprocessor, or other data processing chip, used to run program code stored in memory 502 or process data, such as the magnetic resonance image optimization method of the present invention.

[0083] In some embodiments, processor 501 may be a single server or a group of servers. The server group may be centralized or distributed. In some embodiments, processor 501 may be local or remote. In some embodiments, processor 501 may be implemented on a cloud platform. In one embodiment, the cloud platform may include a private cloud, public cloud, hybrid cloud, community cloud, distributed cloud, internal cloud, multi-cloud, etc., or any combination thereof.

[0084] In some embodiments, memory 502 may be an internal storage unit of the counting device 500, such as a hard disk or memory of the counting device 500. In other embodiments, memory 502 may also be an external storage device of the counting device 500, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., provided on the counting device 500.

[0085] Furthermore, the memory 502 may include both internal storage units of the counting device 500 and external storage devices. The memory 502 is used to store the application software and various types of data installed on the counting device 500.

[0086] In some embodiments, display 503 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an organic light-emitting diode (OLED) touchscreen. Display 503 is used to display information from the counting device 500 and to display a visual user interface. Components 501-503 of the counting device 500 communicate with each other via a system bus.

[0087] In one embodiment, when processor 501 executes the photovoltaic panel counting program in memory 502, the following steps can be implemented: acquire a real-time image of a photovoltaic power station, and pre-process the real-time image of the photovoltaic power station to obtain a target real-time image of the photovoltaic power station; input the target real-time image of the photovoltaic power station into a photovoltaic panel recognition model to obtain a recognition result output by the photovoltaic panel recognition model, the recognition result including a bounding box of each photovoltaic panel in the target real-time image of the photovoltaic power station and a confidence of each photovoltaic panel bounding box; determine a photovoltaic panel counting result corresponding to the target real-time image of the photovoltaic power station based on the recognition result; The photovoltaic panel recognition model includes a feature extraction network, a feature fusion network, and a detection network, and the feature fusion network performs feature fusion based on an SPPF network and an FPN network.

[0088] It should be understood that, in addition to the above functions, the processor 501 can also implement other functions when executing the photovoltaic panel counting program in the memory 502, and specific implementations can be referred to the description of the corresponding method embodiments.

[0089] Further, the type of the counting device 500 is not specifically limited in the embodiments of the present application, and the counting device 500 can be a portable electronic device such as a mobile phone, a tablet computer, a personal digital assistant (PDA), a wearable device, a laptop, etc. Exemplary embodiments of the portable electronic device include but are not limited to a portable electronic device running an IOS, android, microsoft or other operating system. The above portable electronic device can also be other portable electronic devices, such as a laptop having a touch-sensitive surface (e.g. a touch panel). It should also be understood that in some other embodiments of the present application, the counting device 500 can also be a desktop computer having a touch-sensitive surface (e.g. a touch panel).

[0090] Correspondingly, the embodiments of the present application also provide a computer readable storage medium for storing computer readable programs or instructions, which can implement the steps or functions of the photovoltaic panel counting method provided by the above method embodiments when executed by a processor.

[0091] Those skilled in the art can understand that all or part of the processes of the above-mentioned embodiments can be completed by a computer program instructing related hardware (such as a processor, a controller, etc.) to complete, and the computer program can be stored in a computer readable storage medium. The computer readable storage medium is a disk, an optical disk, a read-only memory or a random access memory, etc.

[0092] The photovoltaic panel counting method, the device, the counting equipment and the storage medium provided by the application are described in detail above, the principle and the implementation mode of the application are described by applying specific examples in this paper, and the above example is only used to help understand the method and the core idea of the application; at the same time, for the skilled in the art, according to the idea of the application, the specific implementation mode and the application range will be changed, and the above description should not be understood as the limitation of the application.

Claims

1. A method for counting photovoltaic panels, characterized in that, include: Acquire real-time images of a photovoltaic power station and preprocess the real-time images of the photovoltaic power station to obtain a real-time image of the target photovoltaic power station; The real-time image of the target photovoltaic power station is input into the photovoltaic panel recognition model to obtain the recognition result output by the photovoltaic panel recognition model. The recognition result includes the bounding box of each photovoltaic panel in the real-time image of the target photovoltaic power station and the confidence level of each photovoltaic panel bounding box. Based on the recognition results, the photovoltaic panel count results corresponding to the real-time image of the target photovoltaic power station are determined; The photovoltaic panel recognition model includes a feature extraction network, a feature fusion network, and a detection network. The feature fusion network performs feature fusion based on the SPPF network and the FPN network.

2. The photovoltaic panel counting method according to claim 1, characterized in that, The preprocessing of the real-time image of the photovoltaic power station to obtain the real-time image of the target photovoltaic power station includes: The real-time image of the photovoltaic power station is sequentially processed by image enhancement, noise removal, illumination compensation, and region segmentation to obtain the real-time image of the target photovoltaic power station.

3. The photovoltaic panel counting method according to claim 2, characterized in that, The image enhancement processing is based on the contrast stretching algorithm, the noise removal processing is based on the median filtering algorithm, the illumination compensation processing is based on histogram equalization and cumulative distribution function, and the region segmentation processing is based on the Canny edge detection algorithm.

4. The photovoltaic panel counting method according to claim 1, characterized in that, The photovoltaic panel recognition model determines its parameters based on the SGD optimizer during training, and the learning rate of the SGD optimizer is determined based on the cosine annealing strategy.

5. The photovoltaic panel counting method according to claim 1, characterized in that, The photovoltaic panel recognition model also includes a post-processing network for removing duplicate photovoltaic panel bounding boxes based on the NMS algorithm.

6. The photovoltaic panel counting method according to claim 1, characterized in that, The step of determining the photovoltaic panel count result corresponding to the real-time image of the target photovoltaic power station based on the recognition result includes: Based on the identification results, photovoltaic panel bounding boxes with confidence scores greater than the confidence threshold are determined; The bounding boxes of photovoltaic panels with a confidence level greater than the confidence level threshold are filtered, and the number of the filtered bounding boxes with a confidence level greater than the confidence level threshold is determined as the photovoltaic panel count result corresponding to the real-time image of the target photovoltaic power station.

7. The photovoltaic panel counting method according to claim 6, characterized in that, The filtering process for photovoltaic panel bounding boxes with confidence scores greater than the confidence threshold includes: The DBSCAN clustering algorithm is used to merge overlapping photovoltaic panel bounding boxes, and misidentified photovoltaic panel bounding boxes are filtered out based on bounding box area thresholds and aspect ratio thresholds.

8. A photovoltaic panel counting device, characterized in that, include: The processing module is used to acquire real-time images of the photovoltaic power station and preprocess the real-time images of the photovoltaic power station to obtain the real-time image of the target photovoltaic power station. The recognition module is used to input the real-time image of the target photovoltaic power station into the photovoltaic panel recognition model and obtain the recognition result output by the photovoltaic panel recognition model. The recognition result includes the bounding box of each photovoltaic panel in the real-time image of the target photovoltaic power station and the confidence level of each photovoltaic panel bounding box. The counting module is used to determine the photovoltaic panel count result corresponding to the real-time image of the target photovoltaic power station based on the recognition result; The photovoltaic panel recognition model includes a feature extraction network, a feature fusion network, and a detection network. The feature fusion network performs feature fusion based on the SPPF network and the FPN network.

9. A counting device, characterized in that, Including memory and processor, among which, The memory is used to store programs; The processor, coupled to the memory, is used to execute the program stored in the memory to implement the steps in the photovoltaic panel counting method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, Used to store computer-readable programs or instructions, which, when executed by a processor, can implement the steps in the photovoltaic panel counting method according to any one of claims 1 to 7.

Images