Image recognition-based water bucket abrasion degree evaluation method, device and equipment

Abstract

The present application relates to the field of computer vision, and provide a kind of water bucket abrasion degree evaluation method, device and equipment based on image recognition, the method comprises: obtaining the image to be evaluated of target water bucket;Based on the first image segmentation model of pre-set, the functional area image of target water bucket is identified from the image to be evaluated;Based on the second image segmentation model of pre-set, the abrasion area image of target water bucket is identified from the image to be evaluated;Based on functional area image and abrasion area image, determine the proportion of abrasion part in functional area image;According to the comparison result of proportion and pre-set threshold, the abrasion degree of functional area is evaluated.The present application solves the problem that the prior art depends on artificial subjective judgment and cannot accurately and quantitatively evaluate the abrasion of key parts of water bucket, realizes the automatic, quantitative evaluation of the abrasion degree of specific functional area of water bucket based on double-model image segmentation.

Description

Technical Field

[0001] This invention relates to the field of computer vision, and in particular to a method, apparatus, and device for assessing the degree of water bucket abrasion based on image recognition. Background Technology

[0002] As the core power equipment of high-head hydropower stations, the impulse turbine's key component, the bucket, is subjected to the impact of high-speed, sand-laden water flow for extended periods. Surface erosion is the main cause of performance degradation and structural damage. Accurately assessing the degree of erosion in the key functional areas of the bucket is crucial for developing scientific maintenance strategies, ensuring the safe and stable operation of the unit, and extending the equipment's service life.

[0003] Currently, the industry mainly relies on two traditional methods for detecting and assessing the degree of water tank erosion: manual visual inspection and contact measurement. Manual visual inspection heavily depends on the inspector's personal experience and subjective judgment. It not only fails to provide a precise quantitative description of the eroded area, but also makes it difficult to create a unified, objective, and traceable digital record of the inspection results. Furthermore, it is prone to misjudgment or omission due to factors such as fatigue. While contact measurement can obtain some data, the operation process is extremely cumbersome, requiring prolonged unit shutdown, resulting in low detection efficiency and high costs. More importantly, this method poses safety risks to operators and is difficult to implement in structurally complex or confined spaces, thus lacking versatility.

[0004] Therefore, how to achieve accurate and quantitative assessment of the degree of wear in key functional areas of the water bucket without relying on human experience, in a non-contact and highly efficient manner has become a technical problem that urgently needs to be solved in the field of intelligent operation and maintenance of hydropower station equipment. Summary of the Invention

[0005] This invention provides a method, apparatus, and equipment for assessing the degree of abrasion of a water bucket based on image recognition. It solves the problem that the prior art relies on subjective human judgment and cannot accurately and quantitatively assess the abrasion of key parts of the water bucket. It realizes the automated and quantitative assessment of the degree of abrasion of specific functional areas of the water bucket based on dual-model image segmentation.

[0006] This invention provides a method for assessing the degree of water bucket abrasion based on image recognition, comprising the following steps: Obtain the image of the target water bucket to be evaluated; Based on a preset first image segmentation model, the functional area image of the target water bucket is identified from the image to be evaluated. The first image segmentation model is obtained by training a neural network model based on a labeled first sample image dataset. The first sample image dataset includes water bucket images and their corresponding functional area labeling masks. Based on a preset second image segmentation model, the abrasion region image of the target water bucket is identified from the image to be evaluated. The second image segmentation model is obtained by training a neural network model based on a labeled second sample image dataset. The second sample image dataset includes water bucket images and their corresponding abrasion region label masks. Based on the functional region image and the abrasion region image, determine the proportion of the abrasion portion in the functional region image; The degree of abrasion in the functional area is evaluated based on the comparison between the stated ratio and a preset threshold.

[0007] According to the present invention, a method for assessing the degree of water bucket abrasion based on image recognition is provided. The step of determining the proportion of the abraded portion in the functional region image based on the functional region image and the abraded region image includes: determining a set of pixels for the functional region based on the functional region image; determining a set of pixels for the abraded region based on the abraded region image; calculating the intersection set of the abraded region pixel set and the functional region pixel set; and determining the proportion of the abraded portion in the functional region image based on the ratio of the number of pixels in the intersection set to the number of pixels in the functional region pixel set.

[0008] According to the present invention, a method for evaluating the degree of water bucket abrasion based on image recognition is provided, which trains a first image segmentation model based on a first sample image dataset, including: acquiring a first training image including annotation information of functional areas of the water bucket, wherein the annotation information is used to distinguish at least one functional area among the back of the water bucket, the inner wall, the water divider, and the root; constructing multiple sets of training data pairs based on the first training image and its corresponding functional area annotation mask; inputting the training data pairs into a neural network model for training, and adjusting the parameters of the neural network model by optimizing the loss function to obtain the first image segmentation model.

[0009] According to the present invention, a method for evaluating the degree of water bucket abrasion based on image recognition is provided, which trains a second image segmentation model based on a second sample image dataset, including: acquiring a second training image including annotation information of water bucket abrasion areas, wherein the annotation information is used to identify abrasion areas with fish scale-like abrasion or notch-like morphology; constructing multiple sets of training data pairs based on the second training image and its corresponding abrasion area annotation mask; inputting the training data pairs into a neural network model for training, and adjusting the parameters of the neural network model by optimizing the loss function to obtain the second image segmentation model.

[0010] According to the present invention, a method for evaluating the abrasion degree of a water bucket based on image recognition is provided. The method for evaluating the abrasion degree of the target water bucket based on the comparison result of the proportion and the preset threshold includes: pre-setting multiple abrasion degree thresholds, dividing the abrasion degree into multiple levels based on the multiple abrasion degree thresholds; sequentially comparing the calculated proportion of the abraded part with the multiple abrasion degree thresholds to obtain a comparison result; determining the threshold interval to which the proportion of the abraded part belongs based on the comparison result, and outputting the abrasion degree level corresponding to the threshold interval as the final abrasion degree evaluation result of the target water bucket.

[0011] According to the present invention, a water bucket abrasion degree assessment method based on image recognition is provided, wherein the image to be assessed is preprocessed to reduce resolution before being input into the first image segmentation model and / or the second image segmentation model.

[0012] The present invention also provides an image recognition-based device for assessing the degree of water bucket abrasion, comprising the following modules: The image acquisition module is used to acquire images of the target water bucket to be evaluated. The first model processing module is used to identify the functional area image of the target water bucket from the image to be evaluated based on a preset first image segmentation model. The first image segmentation model is obtained by training a neural network model based on a labeled first sample image dataset. The first sample image dataset includes water bucket images and their corresponding functional area labeling masks. The second model processing module is used to identify the abrasion region image of the target water bucket from the image to be evaluated based on a preset second image segmentation model. The second image segmentation model is obtained by training a neural network model based on a labeled second sample image dataset. The second sample image dataset includes water bucket images and their corresponding abrasion region label masks. The proportion calculation module is used to determine the proportion of the worn portion in the functional area image based on the functional area image and the worn area image; The abrasion assessment module is used to assess the degree of abrasion in the functional area based on the comparison result between the ratio and a preset threshold.

[0013] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the image recognition-based water bucket abrasion assessment method as described above.

[0014] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the image recognition-based water bucket abrasion assessment method as described above.

[0015] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the image recognition-based water bucket abrasion assessment method as described above.

[0016] This invention provides a method, apparatus, and device for assessing the degree of erosion of a water bucket based on image recognition, which offers the following advantages: By using two pre-trained image segmentation models, the functional area and erosion area of ​​the target water bucket are accurately identified from the image to be evaluated. Then, by calculating the ratio of overlapping pixels, an automatic quantitative assessment of the degree of erosion in a specific functional area of ​​the water bucket is achieved. This technical solution directly replaces the traditional subjective judgment relying on human experience, avoiding the problems of inaccurate quantification and recording in visual inspection methods. Simultaneously, its non-contact image analysis method overcomes the shortcomings of contact measurement methods, such as cumbersome operation, low efficiency, and safety risks. Through the above steps, this method achieves an objective, efficient, and traceable quantitative assessment of the erosion condition of the water bucket, significantly improving the intelligence and accuracy of hydropower station equipment condition monitoring. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is one of the flowcharts of the water bucket abrasion assessment method based on image recognition provided by the present invention.

[0019] Figure 2 This is the second flowchart of the water bucket abrasion assessment method based on image recognition provided by the present invention.

[0020] Figure 3 This is a typical example of the back image of the water bucket provided by the present invention.

[0021] Figure 4 This is a typical example diagram of the water bucket back area recognition results provided by the present invention.

[0022] Figure 5 This is a typical example diagram of the water bucket abrasion area identification results provided by the present invention.

[0023] Figure 6This is a schematic diagram of the structure of the water bucket abrasion assessment device based on image recognition provided by the present invention.

[0024] Figure 7 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0026] The following is combined Figures 1-7 The embodiments of the present invention are described in detail.

[0027] The water bucket abrasion degree assessment method based on image recognition provided in this embodiment of the invention is executed by an image recognition-based water bucket abrasion degree assessment device. This device can be configured in a computer, which can be a local computer or a cloud computer. The local computer can be a computer, tablet, etc., and no specific limitation is made here.

[0028] Figure 1 This is a flowchart illustrating the water bucket abrasion assessment method based on image recognition provided by the present invention, as shown below. Figure 1 As shown, the method includes the following steps: S110. Obtain the image of the target water bucket to be evaluated.

[0029] S120. Based on a preset first image segmentation model, identify the functional region image of the target water bucket from the image to be evaluated. The first image segmentation model is obtained by training a neural network model based on a labeled first sample image dataset. The first sample image dataset includes water bucket images and their corresponding functional region labeled masks.

[0030] S130. Based on a preset second image segmentation model, identify the abrasion region image of the target water bucket from the image to be evaluated. The second image segmentation model is obtained by training a neural network model based on a labeled second sample image dataset, which includes water bucket images and their corresponding abrasion region labeled masks.

[0031] S140. Based on the functional region image and the abrasion region image, determine the proportion of the abrasion portion in the functional region image.

[0032] S150. Based on the comparison results between the ratio and the preset threshold, evaluate the degree of wear in the functional area.

[0033] In one embodiment of the present invention, the first image segmentation model and / or the second image segmentation model is a U-Net network model.

[0034] Specifically, an image of the target water bucket to be evaluated is acquired, for example, an image of the back of the water bucket captured by a high-speed camera. A pre-trained first image segmentation model (which may employ a U-Net network model) is used to identify specific functional regions, such as the back region, from this image; this first image segmentation model is trained on a first sample image dataset containing "water bucket image-functional region mask" pairs. A pre-trained second image segmentation model (which may employ a U-Net network model) is used to identify abrasion regions from the same image; this second image segmentation model is trained on a second sample image dataset containing "water bucket image-abrasion region mask" pairs.

[0035] Based on the identified functional region image and abrasion region image, the proportion of abrasion within the functional region is determined by calculating the intersection of the pixel set of the abrasion region and the pixel set of the functional region, and then calculating the ratio of the number of pixels in the intersection to the total number of pixels in the functional region. Finally, this proportion is compared with a preset abrasion degree threshold (e.g., 10% and 40%), and the abrasion degree assessment result for the functional region is output according to its threshold range (e.g., 0-10% is mild, 10%-40% is moderate, and above 40% is severe).

[0036] This embodiment employs two pre-trained image segmentation models to precisely extract functional and abrasion regions from the water bucket image sequentially. It then quantitatively calculates the abrasion percentage based on pixel-level set operations and automatically outputs the abrasion level according to a defined threshold rule. This method automates the entire process of assessing the abrasion status of key parts of the water bucket, from image acquisition to degree determination. It effectively overcomes the shortcomings of traditional manual visual inspection, such as high subjectivity and difficulty in quantification, as well as the low efficiency and poor safety of contact measurements. This significantly improves the objectivity, accuracy, and efficiency of abrasion assessment, providing reliable technical support for intelligent condition monitoring and maintenance decision-making for hydropower station equipment.

[0037] According to the present invention, a method for assessing the degree of water bucket abrasion based on image recognition is provided. The method determines the proportion of the abraded portion in the functional region image based on a functional region image and an abraded region image. The method includes: determining a set of pixels for the functional region based on the functional region image; determining a set of pixels for the abraded region based on the abraded region image; calculating the intersection set of pixels between the abraded region image and the functional region image; and determining the proportion of the abraded portion in the functional region image based on the ratio of the number of pixels in the intersection set to the number of pixels in the functional region image.

[0038] Specifically, after acquiring the functional region image and the abrasion region image, based on the functional region image (e.g., the identified back area of ​​the water trough is pure white pixels, and the background is black pixels), all pixels in the image are traversed, and all pixels whose pixel values ​​correspond to the functional region are included in one set, forming the functional region pixel set F. Based on the abrasion region image (e.g., the identified abrasion region is pure white pixels, and the background is black pixels), all pixels in the image are similarly traversed, and all pixels whose pixel values ​​correspond to the abrasion region are included in another set, forming the abrasion region pixel set M.

[0039] Calculate the intersection of these two pixel sets, that is, find all pixels that simultaneously belong to both the abrasion region pixel set M and the functional region pixel set F. These pixels constitute the intersection pixel set P. intersection Finally, the intersection pixel set P is calculated. intersection The total number of pixels (i.e., N(P)) intersection The ratio of the number of pixels in the functional region (P) to the total number of pixels in the pixel set F (i.e., N(F)) is used to determine the proportion R of the abraded portion in the functional region, i.e., R = N(P) intersection ) / N(F). For example, if the number of pixels in the functional area on the back of the water bucket is 6008, the number of pixels in the abrasion area is 1454, and the number of pixels at the intersection of the two is 847, then the calculated abrasion area ratio R is 847 / 6008 ≈ 14.10%.

[0040] This embodiment calculates the abrasion ratio by performing set operations based on the pixel sets of functional areas and abrasion areas, transforming image recognition results into precise quantitative indicators. This mathematical method is intuitive and rigorous, ensuring the objectivity and accuracy of abrasion area calculation and effectively avoiding errors caused by manual estimation or experience-based judgment. Based on pixel-level set intersection operations, it can accurately reflect the true spatial distribution and coverage of the abrasion area within the target functional area, providing a reliable data foundation for subsequent abrasion degree classification based on scientific thresholds, thus supporting the quantification and standardization of the entire evaluation process.

[0041] According to the present invention, a method for evaluating the degree of water bucket abrasion based on image recognition is provided, which trains a first image segmentation model based on a first sample image dataset, including: acquiring a first training image including annotation information of functional areas of the water bucket, wherein the annotation information is used to distinguish at least one functional area among the back of the water bucket, the inner wall, the water-dividing blade, and the root; constructing multiple sets of training data pairs based on the first training image and its corresponding functional area annotation mask; inputting the training data pairs into a neural network model for training, and adjusting the parameters of the neural network model by optimizing the loss function to obtain the first image segmentation model.

[0042] Specifically, a batch of original images containing water troughs are acquired as training material. Then, using image annotation tools, specific functional regions of the water troughs are manually and precisely annotated on these images. For example, the back region of the water trough is manually circled using a selection tool, and a corresponding binary mask image is generated as a functional region annotation mask. The pixels annotated as the back functional region are white, and the background is black, thus defining the "back" functional region. This process can be extended to annotating other functional regions such as the inner wall, water-dividing blade, and root. Thus, each original image and its corresponding functional region annotation mask together constitute a set of "original image-functional region mask" pairs. After collecting a sufficient number of such image pairs, the first sample image dataset can be constructed.

[0043] This dataset is input into a neural network model (e.g., U-Net model) for training. The network parameters are iteratively optimized to minimize the loss function value between the predicted mask and the real labeled mask until the model converges, thereby obtaining the first image segmentation model that can accurately segment the water bucket functional region from the new image.

[0044] This embodiment details the construction of training data and the model training process for the first image segmentation model, revealing how to build a reliable functional region recognition model based on specific and accurate labeled data. This method of using "original image-functional region mask" pairs as training data provides the model with a clear learning objective, ensuring that the model can learn the visual features and spatial positions of different functional regions of the water bucket in the image, thereby guaranteeing the accuracy and stability of subsequent functional region segmentation. The explicitness and standardization of this training process are crucial prerequisite technical guarantees for the automated and accurate assessment of water bucket erosion, enabling the model to adapt to water bucket images under different working conditions and improving the robustness and practicality of the entire assessment system.

[0045] According to the present invention, a method for evaluating the degree of water bucket abrasion based on image recognition is provided, which trains a second image segmentation model based on a second sample image dataset, including: acquiring a second training image including annotation information of water bucket abrasion areas, wherein the annotation information is used to identify abrasion areas with fish scale-like abrasion or notch-like morphology; constructing multiple sets of training data pairs based on the second training image and its corresponding abrasion area annotation mask; inputting the training data pairs into a neural network model for training, and adjusting the parameters of the neural network model by optimizing the loss function to obtain the second image segmentation model.

[0046] Specifically, a batch of original images containing water buckets are acquired as training material. Using image annotation tools, the abrasion areas of the water buckets are manually and precisely annotated on these images. For example, abrasion areas exhibiting fish-scale or notch-like patterns are manually circled using a selection tool, and a corresponding binary mask image is generated as the abrasion area annotation mask. The pixels annotated as abrasion areas are white, and the background is black. Thus, each original image and its corresponding abrasion area annotation mask together constitute a "original image-abrasion area mask" pair. After collecting a sufficient number of such image pairs, a second sample image dataset can be constructed.

[0047] This dataset is input into a neural network model (e.g., the U-Net model) for training. The network parameters are iteratively optimized to minimize the loss function value between the predicted mask and the real labeled mask until the model converges, thereby obtaining a second image segmentation model that can accurately segment the water trough abrasion area from the new image.

[0048] This embodiment details the construction of training data and the model training process for the second image segmentation model, revealing how to construct a reliable abrasion region recognition model based on the specific morphological features of water-erosion, such as fish-scale patterns and gaps. This method of using "original image - abrasion region mask" pairs as training data provides the model with a clear learning objective, ensuring that the model can learn the visual features of different abrasion regions in the image, thereby guaranteeing the accuracy and stability of subsequent abrasion region segmentation. The explicitness and standardization of this training process are the key technical foundation for this method to achieve accurate quantification of abrasion area and thus objective evaluation, effectively overcoming the drawbacks of strong subjectivity and inconsistent standards in manual recognition.

[0049] According to the present invention, a method for evaluating the abrasion degree of a water bucket based on image recognition is provided. The method evaluates the abrasion degree of a target water bucket based on the comparison results of the proportion and the preset threshold. The method includes: pre-setting multiple abrasion degree thresholds and dividing the abrasion degree into multiple levels based on the multiple abrasion degree thresholds; comparing the calculated proportion of the abraded part with the multiple abrasion degree thresholds in sequence to obtain the comparison results; determining the threshold interval to which the proportion of the abraded part belongs based on the comparison results, and outputting the abrasion degree level corresponding to the threshold interval as the final abrasion degree evaluation result of the target water bucket.

[0050] Specifically, several specific abrasion degree thresholds are pre-set, for example, a first threshold of 10% and a second threshold of 40%. Based on these thresholds, the abrasion degree is divided into three distinct levels: a proportion between 0% and 10% (excluding 10%) is "slight abrasion"; a proportion between 10% and 40% (excluding 40%) is "moderate abrasion"; and a proportion greater than or equal to 40% is "severe abrasion". Once the specific proportion of abrasion in a functional area such as the back of the water tank is obtained (e.g., 14.10%), this proportion is compared sequentially with the pre-set thresholds: first, it is determined whether the proportion is less than 10%; if not, it is further determined whether it is less than 40%. Based on the value of 14.10% in this example, it is greater than 10% and less than 40%. Therefore, based on the comparison results, the threshold range to which this proportion belongs is determined to be "10% to 40%". The abrasion degree level corresponding to this threshold range—"moderate abrasion"—is output as the final abrasion degree assessment result for the functional area on the back of the water tank.

[0051] This embodiment employs a method for automatically determining the degree of erosion based on clearly defined multi-level threshold rules, transforming the precisely calculated erosion area ratio into an intuitive and standardized erosion level conclusion. This rule-based decision-making method completely eliminates subjective human judgment during the assessment process, ensuring the consistency and objectivity of the assessment results and making assessment results from different times, different personnel, or for different water tanks comparable. By directly linking quantitative indicators with predefined level intervals, this method achieves automated and standardized output from data to decision, greatly improving assessment efficiency and reliability, and providing a clear and direct basis for equipment status classification management and maintenance decisions in hydropower stations.

[0052] According to the image recognition-based water bucket abrasion assessment method provided by the present invention, the image to be assessed is preprocessed to reduce its resolution before being input into the first image segmentation model and / or the second image segmentation model.

[0053] Specifically, before inputting the image to be evaluated into the first image segmentation model and / or the second image segmentation model, a preprocessing step to reduce the resolution of the image to be evaluated is included. For example, when the original image to be evaluated is acquired through a high-speed camera and its original resolution is 1080×1920 pixels, before the actual model processing, the resolution of the image is reduced to a preset, smaller size, such as 128×240 pixels, using an image scaling algorithm (such as bilinear interpolation or nearest neighbor interpolation). This step aims to generate a preprocessed image with a significantly reduced size. Subsequently, the image with reduced resolution, instead of the original high-resolution image, is input into the first image segmentation model and the second image segmentation model for subsequent recognition and processing.

[0054] This embodiment effectively optimizes the practicality and operational efficiency of the overall technical solution by specifically limiting the preprocessing operation of reducing the resolution of the image to be evaluated before model recognition. This preprocessing step significantly reduces the amount of data input to the model without significantly affecting the model's recognition accuracy, thereby significantly reducing the computational resources and time required for the model's forward inference process, and improving the processing speed and real-time potential of the entire evaluation process. This is particularly suitable for edge deployment environments with limited computing resources or online monitoring scenarios requiring rapid feedback, making the evaluation method proposed in this invention more valuable for engineering applications and more flexible.

[0055] To address the erosion problem of impact turbines, this invention proposes an image-based method for assessing the degree of bucket erosion, the process of which is as follows: Figure 2 As shown, the specific steps are as follows: Step 1: Acquire water bucket image data; Step 2: Train the water bucket functional area recognition model (U-Net model); Step 3: Train the water bucket abrasion area recognition model (U-Net model); Step 4: Input the newly acquired images into the two models mentioned above to identify the functional areas and abrasion areas; Step 5: Calculate the area of ​​the functional region and the area of ​​the abrasion region based on the image pixels, and calculate the percentage of the abrasion area. Step 6: Set the threshold for the percentage of abrasion area and evaluate the degree of abrasion in this functional area.

[0056] Step 2 includes the following sub-steps: Sub-step A1: Mark the functional areas and labels of the water jets in the image, such as the inner wall, water-dividing blades, and roots; Sub-step A2: Construct "original image-functional region mask" pairs as the training dataset; Sub-step A3: Input the U-Net model for training to obtain the water bucket functional area recognition model.

[0057] Step 3 includes the following sub-steps: Sub-step B1: Mark the water-sprayed areas and labels in the image, such as fish-scale abrasion, notches, etc.; Sub-step B2: Construct "original image-abrasion region mask" pairs as the training dataset; Sub-step B3: Input the U-Net model for training to obtain the water bucket abrasion area recognition model.

[0058] Step 5 includes the following sub-steps: Sub-step C1: Search for the intersection pixels between the etched region and a certain functional region. The specific formula is as follows: Where M represents the set of pixels in the identified abrasion region, F represents the set of pixels in a predefined functional region, and P intersection This represents the set of intersection pixels where the eroded area intersects with the functional area. It is the "intersection" operator in set operations, representing a pixel that belongs to both set M and set F.

[0059] Sub-step C2: Calculate the ratio of the sum of the intersection pixels to the sum of the pixels in the functional region, which is used as the percentage of the abrasion area in that functional region. The specific formula is as follows: in, Let P represent the intersection set. intersection The total number of pixels (i.e., the pixel area of ​​the abraded region within the functional area). This represents the total number of pixels in functional region F (i.e., the total pixel area of ​​that functional region). It indicates the proportion of the worn area in a specific functional region.

[0060] The following example uses an image of the back of the water bucket acquired by an impulse turbine testing platform, combined with the attached image. Figure 2 The following description will be provided in detail. It should be emphasized that the following description is merely exemplary and not intended to limit the scope of the invention or its application.

[0061] This invention proposes a method for assessing the degree of water bucket erosion based on image recognition, the process of which is as follows: Figure 2 As shown in the diagram. First, image data of the back of the water bucket is acquired. Second, U-Net models for recognizing the back of the water bucket and for recognizing the abrasion area of ​​the water bucket are trained separately. Then, newly acquired images are input into the two models respectively to identify the functional area and the abrasion area. Finally, the areas of the functional area and the abrasion area are calculated based on the image pixels, the abrasion area percentage is calculated, and a threshold for the abrasion area percentage is set to evaluate the degree of abrasion in the functional area.

[0062] Based on the method proposed in this invention, the back images of the water buckets collected by the impact turbine testing platform were experimentally verified.

[0063] Step 1: Obtain nozzle image data under historical shutdown conditions, and then obtain water bucket image data; Several images of the water buckets during the operation of the impulse turbine were captured using a high-speed camera, with a resolution of 1080×1920. Typical images are shown below. Figure 3 As shown.

[0064] Step 2: Train the U-Net model for recognizing the back region of the water bucket; Sub-step A1: Label the back region of the water bucket in the image. Use the Matlab image labeling tool to manually outline the back region of the water bucket as a mask.

[0065] Sub-step A2: Construct "original image-back region mask" pairs as the training dataset.

[0066] Sub-step A3: Input the training dataset into the U-Net model for training to obtain the water bucket back region recognition model. Here, to reduce the computational load, the resolution can be reduced, and the input layer size is set to 128×240. The first layer encoder of the U-Net model has 64 filters and an encoder depth of 4.

[0067] Step 3: Train the U-Net model for recognizing water trough abrasion areas; Sub-step B1: Label the abrasion area on the back of the water bucket in the image. Use the Matlab image labeling tool to manually outline the abrasion area of ​​the water bucket as a mask for the abrasion area.

[0068] Sub-step B2: Construct "original image-eroded region mask" pairs as the training dataset.

[0069] Sub-step B3: Input the training dataset into the U-Net model for training to obtain the water bucket back abrasion region recognition model. Here, to reduce the computational load, the resolution can be reduced, and the input layer size is set to 128×240. The first layer encoder of the U-Net model has 64 filters and an encoder depth of 4.

[0070] Step 4: Input the newly acquired images into the two models mentioned above to identify the functional areas and abrasion areas; Specifically, the results of the U-Net model for recognizing the back region of the water bucket and the U-Net model for recognizing the abrasion region of the water bucket are as follows: Figure 4 and Figure 5 As shown in the image. White represents the target area, and black represents the background area.

[0071] Step 5: Calculate the area of ​​the functional region and the area of ​​the abrasion region based on the image pixels, and calculate the percentage of the abrasion area. Sub-step C1: Search for the intersection pixels between the eroded area and a certain functional area.

[0072] Sub-step C2: Calculate the ratio of the sum of the intersection pixels to the sum of the functional region pixels, and use it as the percentage of the abrasion area of ​​the functional region.

[0073] Specifically, according to Figure 4 and Figure 5The grayscale images of the back and abrasion areas show 6008 and 1454 pixels respectively, with an intersection of 847 pixels. The abrasion area of ​​the back of the water bucket is calculated to be 14.10%.

[0074] Step 6: Set the threshold for the percentage of abrasion area and evaluate the degree of abrasion in this functional area.

[0075] Specifically, two thresholds (10% and 40%) were set to assess the degree of abrasion: 0-10% was considered mild abrasion, 10-40% was moderate abrasion, and above 40% was severe abrasion. Therefore, it can be determined that the abrasion on the back of the water bucket is moderate.

[0076] This invention addresses the problem of bucket erosion in impulse turbines by proposing an image recognition-based method for assessing the degree of bucket erosion. By automatically analyzing acquired images using a trained U-Net model, the method can identify and quantify eroded areas in the buckets, significantly improving detection efficiency and avoiding misjudgments and omissions caused by human subjectivity and fatigue. This provides a core technological foundation for online monitoring and periodic automatic assessment of the status of hydropower station units.

[0077] The image recognition-based water bucket abrasion assessment device provided by the present invention is described below. The image recognition-based water bucket abrasion assessment device described below can be referred to in correspondence with the image recognition-based water bucket abrasion assessment method described above.

[0078] like Figure 6 The image shown is a water bucket abrasion assessment device based on image recognition provided by the present invention, comprising: Image acquisition module 610 is used to acquire the image of the target water bucket to be evaluated; The first model processing module 620 is used to identify the functional area image of the target water bucket from the image to be evaluated based on a preset first image segmentation model. The first image segmentation model is obtained by training a neural network model based on a labeled first sample image dataset. The first sample image dataset includes water bucket images and their corresponding functional area labeling masks. The second model processing module 630 is used to identify the abrasion region image of the target water bucket from the image to be evaluated based on the preset second image segmentation model. The second image segmentation model is obtained by training a neural network model based on the labeled second sample image dataset. The second sample image dataset includes the water bucket image and its corresponding abrasion region label mask. The proportion calculation module 640 is used to determine the proportion of the worn part in the functional area image based on the functional area image and the worn area image; The wear assessment module 650 is used to assess the degree of wear in the functional area based on the comparison results between the ratio and the preset threshold.

[0079] Specifically, the functions of each module in the user account management system provided in this embodiment of the invention correspond one-to-one with the operation flow of each step in the above method-like embodiments, and the achieved effects are also the same. For details, please refer to the above embodiments, and this will not be repeated in this embodiment of the invention.

[0080] Figure 7 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 7 As shown, the electronic device may include: a processor 710, a communications interface 720, a memory 730, and a communications bus 740, wherein the processor 710, the communications interface 720, and the memory 730 communicate with each other through the communications bus 740. The processor 710 can call logic instructions in the memory 730 to execute an image recognition-based method for assessing the degree of water bucket abrasion. This method includes: acquiring an image of the target water bucket to be assessed; identifying functional region images of the target water bucket from the image to be assessed based on a preset first image segmentation model, wherein the first image segmentation model is obtained by training a neural network model on a labeled first sample image dataset, and the first sample image dataset includes the water bucket image and its corresponding functional region label mask; identifying abrasion region images of the target water bucket from the image to be assessed based on a preset second image segmentation model, wherein the second image segmentation model is obtained by training a neural network model on a labeled second sample image dataset, and the second sample image dataset includes the water bucket image and its corresponding abrasion region label mask; determining the proportion of abrasion in the functional region image based on the functional region image and the abrasion region image; and assessing the degree of abrasion in the functional region based on a comparison of the proportion with a preset threshold.

[0081] Furthermore, the logical instructions in the aforementioned memory 730 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, in essence, or the part that contributes to the prior art, or a part 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 to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0082] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the image recognition-based water bucket abrasion assessment method provided by the above methods. The method includes: acquiring an image of a target water bucket to be evaluated; identifying the functional area image of the target water bucket from the image to be evaluated based on a preset first image segmentation model, wherein the first image segmentation model is obtained by training a neural network model based on a labeled first sample image dataset, and the first sample image dataset includes the water bucket image and its corresponding functional area label mask; identifying the abrasion area image of the target water bucket from the image to be evaluated based on a preset second image segmentation model, wherein the second image segmentation model is obtained by training a neural network model based on a labeled second sample image dataset, and the second sample image dataset includes the water bucket image and its corresponding abrasion area label mask; determining the proportion of the abrasion portion in the functional area image based on the functional area image and the abrasion area image; and assessing the abrasion degree of the functional area based on a comparison result between the proportion and a preset threshold.

[0083] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the image recognition-based water bucket abrasion assessment method provided by the above methods. The method includes: acquiring an image of a target water bucket to be assessed; identifying functional region images of the target water bucket from the image to be assessed based on a preset first image segmentation model, wherein the first image segmentation model is obtained by training a neural network model based on a labeled first sample image dataset, and the first sample image dataset includes the water bucket image and its corresponding functional region label mask; identifying abrasion region images of the target water bucket from the image to be assessed based on a preset second image segmentation model, wherein the second image segmentation model is obtained by training a neural network model based on a labeled second sample image dataset, and the second sample image dataset includes the water bucket image and its corresponding abrasion region label mask; determining the proportion of abrasion in the functional region image based on the functional region image and the abrasion region image; and assessing the abrasion degree of the functional region based on a comparison between the proportion and a preset threshold.

[0084] 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; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0085] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part 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, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0086] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for assessing the degree of water bucket abrasion based on image recognition, characterized in that, include: Obtain the image of the target water bucket to be evaluated; Based on a preset first image segmentation model, the functional area image of the target water bucket is identified from the image to be evaluated. The first image segmentation model is obtained by training a neural network model based on a labeled first sample image dataset. The first sample image dataset includes water bucket images and their corresponding functional area labeling masks. Based on a preset second image segmentation model, the abrasion region image of the target water bucket is identified from the image to be evaluated. The second image segmentation model is obtained by training a neural network model based on a labeled second sample image dataset. The second sample image dataset includes water bucket images and their corresponding abrasion region label masks. Based on the functional region image and the abrasion region image, determine the proportion of the abrasion portion in the functional region image; The degree of abrasion in the functional area is evaluated based on the comparison between the stated ratio and a preset threshold.

2. The method for assessing the degree of water bucket abrasion based on image recognition according to claim 1, characterized in that, Determining the proportion of the worn portion in the functional region image based on the functional region image and the worn region image includes: Based on the functional region image, determine the set of pixels for the functional region; Based on the image of the abraded area, determine the pixel set of the abraded area; Calculate the intersection set of the pixel set of the abraded region and the pixel set of the functional region; The proportion of the abraded portion in the functional region image is determined based on the ratio of the number of pixels in the intersection pixel set to the number of pixels in the functional region pixel set.

3. The method for assessing the degree of water bucket abrasion based on image recognition according to claim 1, characterized in that, The first image segmentation model is trained based on the first sample image dataset, including: Acquire a first training image including annotation information of the functional areas of the water bucket, wherein the annotation information is used to distinguish at least one functional area among the back of the water bucket, the inner wall, the water-dividing blade and the root. Based on the first training image and its corresponding functional region annotation mask, multiple sets of training data pairs are constructed. The training data is input into the neural network model for training, and the parameters of the neural network model are adjusted by optimizing the loss function to obtain the first image segmentation model.

4. The method for assessing the degree of water bucket abrasion based on image recognition according to claim 1, characterized in that, The second image segmentation model is trained based on the second sample image dataset, including: A second training image is acquired, including annotation information of the water bucket abrasion area, the annotation information being used to identify abrasion areas with fish scale-like abrasion or notch-like morphology; Based on the second training image and its corresponding abrasion region labeled mask, multiple sets of training data pairs are constructed. The training data is input into the neural network model for training, and the parameters of the neural network model are adjusted by optimizing the loss function to obtain the second image segmentation model.

5. The method for assessing the degree of water bucket abrasion based on image recognition according to claim 1, characterized in that, The step of evaluating the abrasion degree of the target water bucket based on the comparison result of the ratio and the preset threshold includes: Multiple abrasion degree thresholds are preset, and the abrasion degree is divided into multiple levels based on these thresholds; The calculated proportion of the eroded portion is compared sequentially with the multiple erosion degree thresholds to obtain the comparison results; Based on the comparison results, the threshold range to which the proportion of the eroded portion belongs is determined, and the erosion degree level corresponding to the threshold range is output as the final erosion degree evaluation result of the target water bucket.

6. The method for assessing the degree of water bucket abrasion based on image recognition according to any one of claims 1-5, characterized in that, Before inputting the image to be evaluated into the first image segmentation model and / or the second image segmentation model, the image to be evaluated is preprocessed to reduce its resolution.

7. A device for assessing the degree of water bucket abrasion based on image recognition, characterized in that, include: The image acquisition module is used to acquire images of the target water bucket to be evaluated. The first model processing module is used to identify the functional area image of the target water bucket from the image to be evaluated based on a preset first image segmentation model. The first image segmentation model is obtained by training a neural network model based on a labeled first sample image dataset. The first sample image dataset includes water bucket images and their corresponding functional area labeling masks. The second model processing module is used to identify the abrasion region image of the target water bucket from the image to be evaluated based on a preset second image segmentation model. The second image segmentation model is obtained by training a neural network model based on a labeled second sample image dataset. The second sample image dataset includes water bucket images and their corresponding abrasion region label masks. The proportion calculation module is used to determine the proportion of the worn portion in the functional area image based on the functional area image and the worn area image; The abrasion assessment module is used to assess the degree of abrasion in the functional area based on the comparison result between the ratio and a preset threshold.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the image recognition-based water bucket abrasion assessment method as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the image recognition-based water bucket abrasion assessment method as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the image recognition-based water bucket abrasion assessment method as described in any one of claims 1 to 6.

Images