Wind power plant construction quality detection method, device, equipment, medium and product

Abstract

The invention provides a wind power plant construction quality detection method, device, equipment, medium and product, and relates to the technical field of construction detection.The method comprises the steps that a scene image set corresponding to a to-be-detected wind power plant is input into a scene image recognition model, and a classified scene image set output by the scene image recognition model is obtained; the scene image recognition model is obtained based on a general scene image sample set of the wind power plant and corresponding label training; inputting the classification scene image set into a construction quality detection model to obtain a construction quality detection result corresponding to the to-be-detected wind power plant output by the construction quality detection model; the construction quality detection model is obtained by training based on the classification scene sample set and the construction quality problem data set. According to the technical scheme, the construction quality of the wind power plant is detected through the classified image set and the construction quality detection model, the link of manual detection is removed, the labor cost is reduced, and the detection efficiency and accuracy of construction quality detection of the wind power plant are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of construction detection, and in particular to a wind farm construction quality detection method, device, equipment, medium and product. Background Art

[0002] With the rapid development of the wind power industry, the control and management of wind farm construction quality has become increasingly important. Control and management of wind farm construction quality is a key factor in ensuring project safety, improving project efficiency, and guaranteeing economic benefits, and wind farm construction quality inspection is a key link in this process.

[0003] In traditional wind farm construction quality control and management methods, wind farm construction quality inspection relies on manual inspection and sampling inspection, which has problems such as low inspection efficiency, high labor costs, and low inspection accuracy due to the subjectivity of human judgment. Summary of the Invention

[0004] The present invention provides a wind farm construction quality inspection method, device, equipment, medium and product to solve the defects of low inspection efficiency, high labor cost and low inspection accuracy caused by subjectivity of human judgment in the prior art. The technical solution of the present invention classifies the scene image set based on the scene image recognition model, and then inspects the construction quality of the wind farm through the classified image set and the construction quality inspection model, eliminating the manual inspection link, reducing labor costs, and improving the inspection efficiency and accuracy of the wind farm construction quality inspection.

[0005] The present invention provides a method for detecting the construction quality of a wind farm, comprising the following steps.

[0006] Inputting a scene image set corresponding to the wind farm to be detected into a scene image recognition model to obtain a classified scene image set output by the scene image recognition model; the scene image recognition model is trained based on a general scene image sample set of wind farms and corresponding labels; The classified scene image set is input into a construction quality inspection model to obtain the construction quality inspection result corresponding to the wind farm to be inspected output by the construction quality inspection model; the construction quality inspection model is trained based on the classified scene sample set and the construction quality problem data set; the classified scene sample set and the construction quality problem data set both correspond to the general scene image sample set.

[0007] According to a wind farm construction quality detection method provided by the present invention, the general scene image sample set is constructed in the following manner: Construct an initial general scene image sample set based on past image data corresponding to the wind farm; Performing denoising processing on the initial general scene image sample set to obtain a denoised image sample set corresponding to the initial general scene image sample set; Performing image enhancement processing on the denoised image sample set to obtain a general scene image sample set corresponding to the denoised image sample set; the image enhancement processing includes at least one of cropping processing, flipping processing, rotation processing, color transformation processing, blurred image processing, random erasing of partial areas, and multi-image mixing processing.

[0008] According to a wind farm construction quality detection method provided by the present invention, the scene image recognition model is obtained by the following method: Construct the initial scene image recognition model in the order of convolutional layers, activation function layers, pooling layers, and multiple fully connected layers; The initial scene image recognition model is iteratively trained based on the general scene image sample set and the corresponding labels to obtain the scene image recognition model; wherein, during the iterative training process, the parameters of the initial scene image recognition model are optimized based on a preset loss function.

[0009] According to a wind farm construction quality detection method provided by the present invention, the construction quality detection model is obtained by the following method: Iteratively training an initial construction quality detection model based on the classification scene sample set and the construction quality problem dataset, and obtaining the construction quality detection model when the model accuracy of the initial construction quality detection model reaches an accuracy threshold; Among them, the initial construction quality detection model is a general deep convolutional neural network architecture based on computer vision; the model accuracy of the initial construction quality detection model is determined based on the output result of the initial construction quality detection model in the current iteration round; the classified scene sample set is obtained by inputting the general scene image sample set into the scene image recognition model.

[0010] According to a wind farm construction quality inspection method provided by the present invention, the method further includes: Determining a digital twin image set corresponding to the scene image set based on digital twin technology; Inputting the scene image set and the digital twin image set into an image comparison model, and obtaining a target difference between the scene image set and the digital twin image set output by the image comparison model; In a case where the target difference is greater than or equal to a difference threshold, inputting the classified scene image set into the construction quality detection model; Among them, the image contrast model is trained based on the general scene image sample set and the digital twin image sample set corresponding to the general scene image sample set.

[0011] According to a wind farm construction quality inspection method provided by the present invention, the image contrast model is used to extract scene image features of the scene image set and digital twin image features of the digital twin image set; it is also used to map the scene image features and the digital twin image features to feature spaces respectively to obtain a first feature vector corresponding to the scene image features and a second feature vector corresponding to the digital twin image features; it is also used to determine the target difference based on the first feature vector and the second feature vector.

[0012] The present invention also provides a wind farm construction quality detection device, comprising the following modules: A classification module is configured to input a scene image set corresponding to the wind farm to be detected into a scene image recognition model to obtain a classified scene image set output by the scene image recognition model; the scene image recognition model is trained based on a general scene image sample set of wind farms and corresponding labels; A detection module is used to input the classified scene image set into a construction quality detection model to obtain the construction quality detection results corresponding to the wind farm to be detected output by the construction quality detection model; the construction quality detection model is trained based on the classified scene sample set and the construction quality problem data set; the classified scene sample set and the construction quality problem data set both correspond to the general scene image sample set.

[0013] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, any of the above-described wind farm construction quality detection methods is implemented.

[0014] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for detecting the construction quality of a wind farm as described above is implemented.

[0015] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements any of the above-mentioned methods for detecting the construction quality of a wind farm.

[0016] The wind farm construction quality inspection method, device, equipment, medium and product provided by the present invention are as follows: a scene image set corresponding to the wind farm to be inspected is input into a scene image recognition model to obtain a classified scene image set output by the scene image recognition model; the scene image recognition model is trained based on a general scene image sample set of the wind farm and corresponding labels; the classified scene image set is input into a construction quality inspection model to obtain a construction quality inspection result corresponding to the wind farm to be inspected output by the construction quality inspection model; the construction quality inspection model is trained based on a classified scene sample set and a construction quality problem data set; the classified scene sample set and the construction quality problem data set both correspond to the general scene image sample set. The technical solution of the present invention classifies the scene image set based on the scene image recognition model, and then inspects the construction quality of the wind farm using the classified image set and the construction quality inspection model, eliminating the manual inspection link, reducing labor costs, and improving the inspection efficiency and accuracy of the wind farm construction quality inspection. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0018] Figure 1 This is one of the flow charts of the wind farm construction quality detection method provided by the present invention.

[0019] Figure 2 Schematic diagram of scene types of a general scene image sample set provided by the present invention.

[0020] Figure 3 It is a data schematic diagram of the construction quality problem data set provided by the present invention.

[0021] Figure 4 This is the second flow chart of the wind farm construction quality detection method provided by the present invention.

[0022] Figure 5 It is a structural schematic diagram of the wind farm construction quality detection device provided by the present invention.

[0023] Figure 6 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0024] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0025] In view of the above problems in the prior art, the present invention provides a wind farm construction quality detection method. Figure 1 This is one of the flow charts of the wind farm construction quality inspection method provided by the present invention, such as Figure 1 As shown, the method includes the following steps 110 to 120.

[0026] Step 110: Input the scene image set corresponding to the wind farm to be detected into a scene image recognition model to obtain a classified scene image set output by the scene image recognition model; the scene image recognition model is trained based on a general scene image sample set of wind farms and corresponding labels.

[0027] Specifically, the scene image recognition model may be pre-trained based on a general scene image sample set of a wind farm and labels corresponding to the general scene image sample set. The general scene image sample set may include images corresponding to multiple general construction scenes of a wind farm. Figure 2 is a schematic diagram of scene types of a general scene image sample set provided by the present invention, such as Figure 2 As shown, the general scene image sample set may include images corresponding to wind turbine foundations, images corresponding to construction roads, images corresponding to collector lines, images corresponding to booster stations, and images corresponding to other scenes.

[0028] Wind farms requiring quality inspection during construction can be identified as wind farms to be inspected, and a scene image set corresponding to the wind farms to be inspected can be obtained. This scene image set can include, for example, multiple photos of the construction site. This scene image set can be acquired at the wind farm construction site using methods such as drone aerial photography, fixed cameras, wearable recorders, or camera helmets.

[0029] Furthermore, the scene image set corresponding to the wind farm to be detected can be input into the scene image recognition model, and the scene image recognition model can classify the images in the scene image set, thereby obtaining a classified scene image set output by the scene image recognition model.

[0030] Step 120: Input the classified scene image set into the construction quality inspection model to obtain the construction quality inspection result corresponding to the wind farm to be inspected output by the construction quality inspection model; the construction quality inspection model is trained based on the classified scene sample set and the construction quality problem data set; the classified scene sample set and the construction quality problem data set both correspond to the general scene image sample set.

[0031] Specifically, the construction quality detection model can be pre-trained based on the classification scene sample set and the construction quality problem data set. The classification scene sample set can be the output of the scene image recognition model after the general scene image sample set is input into the scene image recognition model. The construction quality problem data set can be pre-constructed based on experience. The construction quality problem data set is essentially a label for model training. Therefore, the construction quality problem data set corresponds to the classification scene sample set and also corresponds to the general scene image sample set. The construction quality problem data set is a data set that describes the images of the general scene image sample set or the classification scene sample set. Figure 3 This is a data diagram of the construction quality problem dataset provided by the present invention. Figure 3 As shown, the Construction Quality Issues Dataset can indicate whether images in a general scene image sample set or a classified scene sample set are related to personnel issues, machine operation issues, technical method issues, external environment issues, or other quality issues, and further indicate whether each image is a safe image or a problem image. For example, the Construction Quality Issues Dataset indicates that Image A in the general scene image sample set is related to personnel issues and is a problem image. It can also indicate that Image B in the general scene image sample set is related to external environment issues and is a safe image.

[0032] Furthermore, by inputting the classified scene image set into the construction quality inspection model, the model outputs a construction quality inspection result corresponding to the wind farm to be inspected. This construction quality inspection result indicates whether each image in the scene image set corresponding to the wind farm to be inspected is a safe image or a problematic image (with the specific problem annotated for problematic images).

[0033] The wind farm construction quality inspection method provided by the present invention comprises the following steps: inputting a scene image set corresponding to the wind farm to be inspected into a scene image recognition model to obtain a classified scene image set output by the scene image recognition model; the scene image recognition model is trained based on a general scene image sample set of the wind farm and corresponding labels; inputting the classified scene image set into a construction quality inspection model to obtain a construction quality inspection result corresponding to the wind farm to be inspected output by the construction quality inspection model; the construction quality inspection model is trained based on a classified scene sample set and a construction quality problem data set; the classified scene sample set and the construction quality problem data set both correspond to the general scene image sample set. The technical solution of the present invention classifies the scene image set based on the scene image recognition model, and then inspects the construction quality of the wind farm using the classified image set and the construction quality inspection model, thereby eliminating the manual inspection link, reducing labor costs, and improving the inspection efficiency and accuracy of the wind farm construction quality inspection.

[0034] In one embodiment, the general scene image sample set is constructed in the following manner: Construct an initial general scene image sample set based on past image data corresponding to the wind farm; Performing denoising processing on the initial general scene image sample set to obtain a denoised image sample set corresponding to the initial general scene image sample set; Performing image enhancement processing on the denoised image sample set to obtain a general scene image sample set corresponding to the denoised image sample set; the image enhancement processing includes at least one of cropping processing, flipping processing, rotation processing, color transformation processing, blurred image processing, random erasing of partial areas, and multi-image mixing processing.

[0035] Specifically, an initial general scene image sample set can be constructed based on past image data of the wind farm. The past image data can be historically collected data or data from open source datasets such as the CIFAR-100 dataset and the MIT Places2 dataset. After obtaining the initial general scene image sample set, the initial general scene image sample set can be subjected to denoising processing to obtain a denoised image sample set. The denoising processing may include (1) removing non-keyword images; (2) removing images that are occluded by key scene content (images with a relatively low degree of occlusion may be retained); (3) directly deleting images that do not match the format or converting them into a preset format (for example, JPG format or JPEG format); (4) deleting or retaining duplicate images for image enhancement processing; and (5) removing damaged images.

[0036] Furthermore, in order to generate more sample images, the denoised image sample set can be enhanced to obtain a general scene image sample set corresponding to the denoised image sample set. The image enhancement process can include one or more of the following processes: (1) Cropping: extracting a sub-region from the original image to generate more samples. Cropping can simulate the change of camera perspective and improve the model's ability to understand local features; (2) Flipping: performing horizontal and vertical mirroring operations on the image can increase the diversity of the data without changing the image content. Flipping does not affect the image semantics, but can greatly increase the number of samples and improve overfitting; (3) Rotation: rotating the image at a certain angle to provide multiple perspectives to help the model recognize the rotated object; (4) Color transformation: generating diverse images by adjusting the color properties of the image (such as brightness, contrast, saturation and hue) to simulate different lighting conditions and color configurations. Color transformation processing helps to improve the robustness of the model under different lighting conditions; (5) Other image enhancement processing: blurred image processing, random erasing of partial areas and multi-image mixing processing, etc. It is easy to understand that the image enhancement processing in this embodiment can also include other image enhancement processing methods in addition to the above-mentioned processing methods, for example, it can also include synthetic data processing (synthesizing new images and data through computer graphics technology to expand the scope and diversity of the data set).

[0037] In the above embodiment, the initial general scene image sample set is denoised to provide a basis for subsequent image classification, and further image enhancement processing is performed to increase the diversity of the general scene image sample set, thereby improving the robustness of the model when subsequently training the scene image recognition model.

[0038] In one embodiment, the scene image recognition model is obtained by: Construct the initial scene image recognition model in the order of convolutional layers, activation function layers, pooling layers, and multiple fully connected layers; The initial scene image recognition model is iteratively trained based on the general scene image sample set and the corresponding labels to obtain the scene image recognition model; wherein, during the iterative training process, the parameters of the initial scene image recognition model are optimized based on a preset loss function.

[0039] Specifically, the initial scene image recognition model can be constructed in the order of convolutional layers, activation function layers, pooling layers, and multiple fully connected layers. For example, the initial scene image recognition model can include six layers, with the first layer being a convolutional layer, the second layer being an activation function layer, the third layer being a pooling layer, and the fourth to sixth layers being fully connected layers. Dropout layers can be added to the fourth and fifth fully connected layers.

[0040] Furthermore, the initial scene image recognition model can be iteratively trained based on the general scene image sample set and the corresponding labels. When the general scene image sample set is input into the initial scene image recognition model, the images in the general scene image sample set can also be adjusted to a three-way color format of 100 by 100 pixels and randomly shuffled. When the model accuracy meets the requirements or the number of iterative training rounds reaches the preset number, a scene image recognition model can be obtained. During the iterative training process, the parameters of the initial scene image recognition model are optimized based on a preset loss function. A penalty term proportional to the weight size can be added to the preset loss function to avoid overfitting of the model. In addition, the backpropagation algorithm (BP), L2 regularization, and Dropout technology can be combined in the process of optimizing the parameters of the initial scene image recognition model.

[0041] In the above embodiment, a scene image recognition model is obtained through iterative training, which lays a foundation for the classification of scene image sets.

[0042] In one embodiment, the construction quality detection model is obtained by: Iteratively training an initial construction quality detection model based on the classification scene sample set and the construction quality problem dataset, and obtaining the construction quality detection model when the model accuracy of the initial construction quality detection model reaches an accuracy threshold; Among them, the initial construction quality detection model is a general deep convolutional neural network architecture based on computer vision; the model accuracy of the initial construction quality detection model is determined based on the output result of the initial construction quality detection model in the current iteration round; the classified scene sample set is obtained by inputting the general scene image sample set into the scene image recognition model.

[0043] Specifically, an initial construction quality inspection model can be pre-built. The initial construction quality inspection model can be a general computer vision-based deep convolutional neural network architecture, such as the Inception-V3 convolutional neural network architecture or YOLOv5 (You Only Look Once version 5, an object detection algorithm). The initial construction quality inspection model can then be iteratively trained based on a classified scene sample set (the classified scene sample set can be obtained by inputting a general scene image sample set into a scene image recognition model) and a construction quality problem dataset. When the model accuracy of the initial construction quality inspection model reaches an accuracy threshold, a construction quality inspection model is obtained. The accuracy threshold can be pre-set as needed and is not specifically limited in this embodiment of the present invention.

[0044] The model accuracy of the initial construction quality detection model is determined based on the output results of the initial construction quality detection model in the current iteration round. For example, the model accuracy is It can be determined by the following formula: in, Indicates that there is a problem with the output result. Indicates that there is more likely to be a problem with the output results. Indicates that the output is more likely to be free of problems. Indicates that there is no problem with the output result. 、 、 、 Both are constants, corresponding to the number of values ​​returned by the model.

[0045] In the above embodiment, the initial construction quality detection model is iteratively trained based on the classification scene sample set and the construction quality problem data set. When the model accuracy of the initial construction quality detection model reaches the accuracy threshold, the construction quality detection model is obtained, thereby ensuring the accuracy of the construction quality detection model, so that the construction quality detection results obtained according to the construction quality detection model have higher accuracy.

[0046] In one embodiment, the method further comprises: Determining a digital twin image set corresponding to the scene image set based on digital twin technology; Inputting the scene image set and the digital twin image set into an image comparison model, and obtaining a target difference between the scene image set and the digital twin image set output by the image comparison model; In a case where the target difference is greater than or equal to a difference threshold, inputting the classified scene image set into the construction quality detection model; Among them, the image contrast model is trained based on the general scene image sample set and the digital twin image sample set corresponding to the general scene image sample set.

[0047] Specifically, the initial image contrast model can be pre-trained based on a general scene image sample set and a digital twin image sample set corresponding to the general scene image sample set to obtain an image contrast model, wherein the initial image contrast model can be constructed based on a Siamese Network.

[0048] Furthermore, the digital twin image set corresponding to the scene image set can be determined based on the digital twin technology, and the scene image set and the digital twin image set can be input into the image comparison model respectively to obtain the target difference between the scene image set and the digital twin image set output by the image comparison model. The target difference can characterize the difference between the scene image set and the digital twin image set. Figure 4 This is the second flow chart of the wind farm construction quality inspection method provided by the present invention, such as Figure 4 As shown, when the target difference is greater than or equal to the difference threshold, the classified scene image set can be input into the construction quality detection model, and then the construction quality detection results output by the construction quality detection model can be obtained. When the target difference is less than the difference threshold, no detection is performed.

[0049] In the above embodiment, detection is performed only when the target difference between the scene image set and the digital twin image set is greater than the difference threshold, which can further improve the accuracy of construction quality detection.

[0050] In one embodiment, the image contrast model is used to extract scene image features of the scene image set and digital twin image features of the digital twin image set; it is also used to map the scene image features and the digital twin image features to feature spaces respectively to obtain a first feature vector corresponding to the scene image features and a second feature vector corresponding to the digital twin image features; it is also used to determine the target difference based on the first feature vector and the second feature vector.

[0051] Specifically, in order to compare the scene image set and the digital twin image set, the image comparison model first needs to extract the scene image features of the scene image set and the digital twin image features of the digital twin image set. Then, the scene image features and the digital twin image features can be mapped to the feature space respectively, and the first eigenvector corresponding to the scene image features and the second eigenvector corresponding to the digital twin image features can be obtained. The target difference can be further calculated based on the first eigenvector and the second eigenvector. It is easy to understand that when training the image comparison model, the image comparison model also performs similar processing as described above in this embodiment on the general scene image sample set and the digital twin image sample set corresponding to the general scene image sample set.

[0052] In the above embodiment, the specific role of the image contrast model is defined. The image contrast model extracts features, maps the features to a feature space to obtain vectors, and finally calculates the difference based on the vectors, thereby improving the accuracy of target similarity.

[0053] The wind farm construction quality detection device provided by the present invention is described below. The wind farm construction quality detection device described below and the wind farm construction quality detection method described above can be referenced to each other.

[0054] Figure 5 This is a schematic diagram of the structure of the wind farm construction quality detection device provided by the present invention. Figure 5 As shown, the wind farm construction quality detection device 500 includes the following modules: The classification module 510 is configured to input a scene image set corresponding to the wind farm to be detected into a scene image recognition model to obtain a classified scene image set output by the scene image recognition model; the scene image recognition model is trained based on a general scene image sample set of wind farms and corresponding labels; The detection module 520 is used to input the classified scene image set into the construction quality detection model to obtain the construction quality detection result corresponding to the wind farm to be detected output by the construction quality detection model; the construction quality detection model is trained based on the classified scene sample set and the construction quality problem data set; the classified scene sample set and the construction quality problem data set both correspond to the general scene image sample set.

[0055] The wind farm construction quality inspection device provided by the present invention inputs a scene image set corresponding to the wind farm to be inspected into a scene image recognition model to obtain a classified scene image set output by the scene image recognition model; the scene image recognition model is trained based on a general scene image sample set of the wind farm and corresponding labels; the classified scene image set is input into a construction quality inspection model to obtain a construction quality inspection result corresponding to the wind farm to be inspected output by the construction quality inspection model; the construction quality inspection model is trained based on a classified scene sample set and a construction quality problem data set; the classified scene sample set and the construction quality problem data set both correspond to the general scene image sample set. The technical solution of the present invention classifies the scene image set based on the scene image recognition model, and then inspects the construction quality of the wind farm using the classified image set and the construction quality inspection model, eliminating the manual inspection link, reducing labor costs, and improving the inspection efficiency and accuracy of the wind farm construction quality inspection.

[0056] In one embodiment, the wind farm construction quality detection device further includes a construction module, which is specifically configured to: Construct an initial general scene image sample set based on past image data corresponding to the wind farm; Performing denoising processing on the initial general scene image sample set to obtain a denoised image sample set corresponding to the initial general scene image sample set; Performing image enhancement processing on the denoised image sample set to obtain a general scene image sample set corresponding to the denoised image sample set; the image enhancement processing includes at least one of cropping processing, flipping processing, rotation processing, color transformation processing, blurred image processing, random erasing of partial areas, and multi-image mixing processing.

[0057] In one embodiment, the wind farm construction quality detection device further includes a first iterative training module, which is specifically configured to: Construct the initial scene image recognition model in the order of convolutional layers, activation function layers, pooling layers, and multiple fully connected layers; The initial scene image recognition model is iteratively trained based on the general scene image sample set and the corresponding labels to obtain the scene image recognition model; wherein, during the iterative training process, the parameters of the initial scene image recognition model are optimized based on a preset loss function.

[0058] In one embodiment, the wind farm construction quality detection device further includes a second iterative training module, which is specifically configured to: Iteratively training an initial construction quality detection model based on the classification scene sample set and the construction quality problem dataset, and obtaining the construction quality detection model when the model accuracy of the initial construction quality detection model reaches an accuracy threshold; Among them, the initial construction quality detection model is a general deep convolutional neural network architecture based on computer vision; the model accuracy of the initial construction quality detection model is determined based on the output result of the initial construction quality detection model in the current iteration round; the classified scene sample set is obtained by inputting the general scene image sample set into the scene image recognition model.

[0059] In one embodiment, the wind farm construction quality inspection device further includes an image comparison module, which is specifically used to: Determining a digital twin image set corresponding to the scene image set based on digital twin technology; Inputting the scene image set and the digital twin image set into an image comparison model, and obtaining a target difference between the scene image set and the digital twin image set output by the image comparison model; In a case where the target difference is greater than or equal to a difference threshold, inputting the classified scene image set into the construction quality detection model; Among them, the image contrast model is trained based on the general scene image sample set and the digital twin image sample set corresponding to the general scene image sample set.

[0060] In one embodiment, the image contrast model is used to extract scene image features of the scene image set and digital twin image features of the digital twin image set; it is also used to map the scene image features and the digital twin image features to feature spaces respectively to obtain a first feature vector corresponding to the scene image features and a second feature vector corresponding to the digital twin image features; it is also used to determine the target difference based on the first feature vector and the second feature vector.

[0061] Figure 6 An example of a physical structure diagram of an electronic device is shown below. Figure 6 As shown, the electronic device may include: a processor 610, a communications interface 620, a memory 630, and a communication bus 640, wherein the processor 610, the communications interface 620, and the memory 630 communicate with each other via the communication bus 640. The processor 610 may call the logic instructions in the memory 630 to execute the wind farm construction quality inspection method, which includes: Inputting a scene image set corresponding to the wind farm to be detected into a scene image recognition model to obtain a classified scene image set output by the scene image recognition model; the scene image recognition model is trained based on a general scene image sample set of wind farms and corresponding labels; The classified scene image set is input into a construction quality inspection model to obtain the construction quality inspection result corresponding to the wind farm to be inspected output by the construction quality inspection model; the construction quality inspection model is trained based on the classified scene sample set and the construction quality problem data set; the classified scene sample set and the construction quality problem data set both correspond to the general scene image sample set.

[0062] Furthermore, the logic instructions in the aforementioned memory 630 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0063] On the other hand, the present invention further provides a computer program product, comprising a computer program, which may be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the wind farm construction quality inspection method provided by the above methods, which includes: Inputting a scene image set corresponding to the wind farm to be detected into a scene image recognition model to obtain a classified scene image set output by the scene image recognition model; the scene image recognition model is trained based on a general scene image sample set of wind farms and corresponding labels; The classified scene image set is input into a construction quality inspection model to obtain the construction quality inspection result corresponding to the wind farm to be inspected output by the construction quality inspection model; the construction quality inspection model is trained based on the classified scene sample set and the construction quality problem data set; the classified scene sample set and the construction quality problem data set both correspond to the general scene image sample set.

[0064] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for detecting the construction quality of a wind farm provided by the above methods is implemented. The method comprises: Inputting a scene image set corresponding to the wind farm to be detected into a scene image recognition model to obtain a classified scene image set output by the scene image recognition model; the scene image recognition model is trained based on a general scene image sample set of wind farms and corresponding labels; The classified scene image set is input into a construction quality inspection model to obtain the construction quality inspection result corresponding to the wind farm to be inspected output by the construction quality inspection model; the construction quality inspection model is trained based on the classified scene sample set and the construction quality problem data set; the classified scene sample set and the construction quality problem data set both correspond to the general scene image sample set.

[0065] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0066] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0067] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A wind farm construction quality inspection method, characterized in that: include: Inputting a scene image set corresponding to the wind farm to be detected into a scene image recognition model to obtain a classified scene image set output by the scene image recognition model; The scene image recognition model is obtained by training based on a general scene image sample set of a wind farm and corresponding labels; The classified scene image set is input into a construction quality inspection model to obtain the construction quality inspection result corresponding to the wind farm to be inspected output by the construction quality inspection model; the construction quality inspection model is trained based on the classified scene sample set and the construction quality problem data set; the classified scene sample set and the construction quality problem data set both correspond to the general scene image sample set.

2. The wind farm construction quality detection method according to claim 1, characterized in that: The general scene image sample set is constructed in the following way: Construct an initial general scene image sample set based on past image data corresponding to the wind farm; Performing denoising processing on the initial general scene image sample set to obtain a denoised image sample set corresponding to the initial general scene image sample set; Performing image enhancement processing on the denoised image sample set to obtain a general scene image sample set corresponding to the denoised image sample set; the image enhancement processing includes at least one of cropping processing, flipping processing, rotation processing, color transformation processing, blurred image processing, random erasing of partial areas, and multi-image mixing processing.

3. The wind farm construction quality detection method according to claim 1, characterized in that: The scene image recognition model is obtained by: Construct the initial scene image recognition model in the order of convolutional layers, activation function layers, pooling layers, and multiple fully connected layers; The initial scene image recognition model is iteratively trained based on the general scene image sample set and the corresponding labels to obtain the scene image recognition model; wherein, during the iterative training process, the parameters of the initial scene image recognition model are optimized based on a preset loss function.

4. The wind farm construction quality inspection method according to claim 1, characterized in that: The construction quality detection model is obtained by the following method: Iteratively training an initial construction quality detection model based on the classification scene sample set and the construction quality problem dataset, and obtaining the construction quality detection model when the model accuracy of the initial construction quality detection model reaches an accuracy threshold; Among them, the initial construction quality detection model is a general deep convolutional neural network architecture based on computer vision; the model accuracy of the initial construction quality detection model is determined based on the output result of the initial construction quality detection model in the current iteration round; the classified scene sample set is obtained by inputting the general scene image sample set into the scene image recognition model.

5. The wind farm construction quality inspection method according to any one of claims 1 to 4, characterized in that: The method further comprises: Determining a digital twin image set corresponding to the scene image set based on digital twin technology; Inputting the scene image set and the digital twin image set into an image comparison model, and obtaining a target difference between the scene image set and the digital twin image set output by the image comparison model; In a case where the target difference is greater than or equal to a difference threshold, inputting the classified scene image set into the construction quality detection model; Among them, the image contrast model is trained based on the general scene image sample set and the digital twin image sample set corresponding to the general scene image sample set.

6. The wind farm construction quality inspection method according to claim 5, characterized in that: The image contrast model is used to extract the scene image features of the scene image set and the digital twin image features of the digital twin image set; it is also used to map the scene image features and the digital twin image features to the feature space respectively to obtain the first feature vector corresponding to the scene image features and the second feature vector corresponding to the digital twin image features; it is also used to determine the target difference based on the first feature vector and the second feature vector.

7. A wind farm construction quality inspection device, characterized in that: include: A classification module, configured to input a scene image set corresponding to the wind farm to be detected into a scene image recognition model, and obtain a classified scene image set output by the scene image recognition model; The scene image recognition model is obtained by training based on a general scene image sample set of a wind farm and corresponding labels; A detection module is used to input the classified scene image set into a construction quality detection model to obtain the construction quality detection results corresponding to the wind farm to be detected output by the construction quality detection model; the construction quality detection model is trained based on the classified scene sample set and the construction quality problem data set; the classified scene sample set and the construction quality problem data set both correspond to the general scene image sample set.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the wind farm construction quality detection method according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the wind farm construction quality detection method according to any one of claims 1 to 6 is implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the wind farm construction quality detection method according to any one of claims 1 to 6 is implemented.

Images