Underwater equipment damage intelligent detection method, device and equipment and storage medium
Through intelligent detection methods, deep learning models are used to analyze the damage characteristics of underwater equipment, solving the problems of long and subjectiveness of manual inspection, achieving efficient and accurate damage assessment of underwater facilities, and improving the scientific nature of risk management and reliability.
Patent Information
- Application Number
- CN202411271734.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-11
- Publication Date
- 2024-12-24
AI Technical Summary
In the prior art, the damage assessment of underwater facilities mainly relies on manual inspection, which has problems such as time-consuming, low efficiency, strong subjectivity and insufficient accuracy and coverage, resulting in the complexity of risk management of deep-sea aquaculture facilities.
An intelligent detection method for damage in underwater equipment is adopted. By collecting equipment images and status data, the images and features of damage areas are extracted, and the damage characteristics are analyzed using pre-trained deep learning models, the damage evaluation results are generated, and the damage level of the equipment is determined. The method includes edge detection, texture analysis, color feature extraction and comprehensive feature fusion, combined with device status data to achieve intelligent detection and evaluation.
It improves the accuracy and efficiency of damage detection, reduces the subjectivity and error of manual operations, ensures the timeliness and reliability of equipment status monitoring, and provides a scientific basis for the maintenance of underwater equipment and insurance claims.
Smart Images

Figure CN119181014A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of image detection technology, and in particular to an intelligent detection method, device, equipment and storage medium for underwater equipment damage. Background Art
[0002] Against the backdrop of the rapid development of the marine economy, modern agriculture, especially aquaculture, is expanding into deep-sea areas. As the scale of deep-sea aquaculture expands, the safety and stability of related facilities such as deep-sea cages have become the focus of attention of aquaculture operators and insurance agencies. However, the complex and changeable deep-sea environment, such as wind and waves, swells, biological erosion and other natural factors, can easily cause damage to these aquaculture facilities. Such damage will not only affect the benefits of aquaculture, but may also cause significant economic losses.
[0003] At present, damage assessment of deep-sea aquaculture facilities mainly relies on manual inspection, a traditional method with obvious limitations in practical application. Manual inspection is not only time-consuming and inefficient, but also has poor accuracy and consistency due to the subjectivity in the assessment process. In addition, the deep-sea environmental conditions are harsh, and manual inspections are difficult to fully cover all damaged areas, further increasing the difficulty and risk of assessment. These problems directly affect the efficiency and accuracy of insurance claims, making risk management of deep-sea aquaculture more complicated.
[0004] In the financial field, especially in the insurance industry, how to improve the efficiency and accuracy of damage assessment of deep-sea aquaculture facilities has become an urgent issue to be resolved. Summary of the invention
[0005] The main purpose of the present invention is to provide an intelligent detection method, device, equipment and storage medium for underwater equipment damage, aiming to solve the technical problems in the prior art that damage assessment of underwater facilities mainly relies on manual inspection, which is time-consuming, highly subjective, and lacks accuracy and coverage.
[0006] To achieve the above object, the present invention provides an intelligent detection method for underwater equipment damage, comprising:
[0007] Collect equipment images and equipment status data of underwater target equipment;
[0008] extracting a damaged area image of the underwater target device from the device image;
[0009] Extracting damage features from the damaged area image and the device status data, wherein the damage features include shape, color and texture features of the damaged area;
[0010] The damage characteristics are analyzed using a pre-trained deep learning model to generate a damage assessment result to determine the damage level of the underwater target equipment.
[0011] In one embodiment, extracting the damaged area image of the underwater target device from the device image includes:
[0012] Obtaining damage-free images of normal equipment in an undamaged state;
[0013] Calculating the color histograms of the intact image and the device image, calculating the color histogram difference between the intact image and the device image using the Bhattacharyya distance, and marking the area where the color histogram difference exceeds a color difference threshold as a damaged area;
[0014] Calculating the gray level co-occurrence matrix of the intact image and the device image, calculating the texture feature difference between the intact image and the device image based on the gray level co-occurrence matrix, and marking the area where the texture feature difference exceeds the texture difference threshold as a damaged area;
[0015] Calculating a grayscale histogram of the device image, selecting a global threshold according to the grayscale histogram, marking an area in the device image with a grayscale value higher than the threshold as a potential damage area, and marking an accurate damage area based on comparing the potential damage area with the damage-free image;
[0016] Based on the marked damaged area, a damaged area image is extracted from the device image.
[0017] In one embodiment, extracting damage features from the damaged area image and the device status data includes:
[0018] Based on the damaged area image, the Canny edge detection algorithm is used to extract the edge of the damaged area, identify and track the outline of the damaged area, and calculate the shape features of the damaged area, wherein the shape features include area, perimeter, aspect ratio and roundness;
[0019] Converting the damaged area image from RGB color space to HSV or Lab color space, generating color histograms for Hue, Saturation, Value or L, a, b channels, and extracting key color features based on the color histograms, the color features including average hue, color saturation, and concentration of color distribution;
[0020] The gray-level co-occurrence matrix of the damaged area is calculated, and the texture frequency and directional features of the damaged area are extracted using a Gabor filter. The comprehensive texture feature value is calculated by combining the gray-level co-occurrence matrix, texture frequency and directional features.
[0021] extracting temperature features, pressure features, and depth features from the device status data;
[0022] The extracted shape features, color features and texture features are fused with the temperature features, pressure features and depth features to obtain a comprehensive damage feature.
[0023] In one embodiment, the damage characteristics are analyzed using a pre-trained deep learning model to generate a damage assessment result, including:
[0024] Inputting the damage features into the deep learning model, and assigning damage weights to the shape features, color features, and texture features of the damaged area respectively;
[0025] Calculating a shape damage score based on the shape characteristics of the damage area;
[0026] Calculating a color damage score based on the color characteristics of the damaged area;
[0027] Calculating a texture damage score based on the texture features of the damaged area;
[0028] Calculate a comprehensive damage score according to the shape damage score, color damage score, texture damage score and damage weight;
[0029] A damage assessment result including a damage level is generated based on the comprehensive damage score and the damage level rule.
[0030] In one embodiment, calculating the texture damage score based on the texture features of the damaged area includes:
[0031] Obtaining an undamaged image of a normal device in an undamaged state to establish a baseline texture feature;
[0032] Extracting texture features from the damage-free image, the texture features including contrast, uniformity and roughness, and assigning a corresponding weight to each texture feature;
[0033] Comparing the texture features of the damaged area with the texture features of the undamaged image, and calculating the feature difference between each texture feature;
[0034] The texture damage score is determined based on the feature differences between each texture feature and the corresponding weights.
[0035] In one embodiment, after generating the damage assessment result, the method further includes:
[0036] Obtain the insurance terms of the underwater target equipment, including insurance coverage, compensation standards, deductibles and claims settlement procedures;
[0037] Based on the damage assessment result, determining the damaged portion of the underwater target equipment that meets the insurance compensation scope;
[0038] Calculate out-of-pocket expenses and deductibles based on the portion of damage eligible for insurance coverage, and generate repair recommendations;
[0039] A damage assessment report is generated based on the out-of-pocket expenses, deductibles, and repair recommendations.
[0040] In one embodiment, before collecting the device image and device status data of the underwater target device, the method further includes:
[0041] Installing four cameras at four corners of the underwater target device so that the horizontal distance between each camera is equal and each camera covers a viewing angle of 120 degrees;
[0042] A top camera is installed at the center of the top of the underwater target device, and the top camera shoots vertically downward;
[0043] A bottom camera is installed at the center of the bottom of the underwater target device, and the bottom camera shoots vertically upward;
[0044] Depth sensors are installed on the four corners of the underwater target device, temperature sensors are installed on preset key positions of the underwater target device, and pressure sensors are installed on the surface of the underwater target device.
[0045] Furthermore, to achieve the above-mentioned purpose, the present invention also provides an underwater equipment damage intelligent detection device, which includes a memory, a processor, and an underwater equipment damage intelligent detection program stored in the memory and executable on the processor, and the underwater equipment damage intelligent detection program, when executed by the processor, implements the steps of the underwater equipment damage intelligent detection method as described above.
[0046] Furthermore, to achieve the above-mentioned purpose, the present invention also provides a computer storage medium, on which is stored a program for intelligent detection of underwater equipment damage, and when the program for intelligent detection of underwater equipment damage is executed by a processor, the steps of the method for intelligent detection of underwater equipment damage as described above are implemented.
[0047] Beneficial effects: The present invention relates to an intelligent detection method for underwater equipment damage, which collects equipment images and equipment status data of underwater target equipment, pre-processes the equipment images, extracts damaged area images of underwater target equipment, extracts damage features from damaged area images and equipment status data, and uses a pre-trained deep learning model to analyze the damage features, generates damage assessment results, and determines the damage level of underwater target equipment. The present invention can realize intelligent detection and rapid assessment of underwater equipment damage, improve the accuracy and efficiency of damage detection, reduce the subjectivity and errors of manual operation, ensure the timeliness and reliability of equipment status monitoring, and provide a scientific basis for underwater equipment maintenance and insurance claims. BRIEF DESCRIPTION OF THE DRAWINGS
[0048] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which:
[0049] Figure 1 It is a flow chart of an embodiment of an intelligent detection method for underwater equipment damage according to the present invention;
[0050] Figure 2 It is a functional module schematic diagram of a preferred embodiment of the intelligent detection device for underwater equipment damage of the present invention;
[0051] Figure 3 It is a structural schematic diagram of the equipment hardware operating environment involved in the embodiment of the underwater equipment damage intelligent detection equipment of the present invention. DETAILED DESCRIPTION
[0052] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.
[0053] Against the backdrop of the rapid development of the marine economy, modern agriculture, especially aquaculture, is expanding into deep-sea areas. As the scale of deep-sea aquaculture expands, the safety and stability of related facilities such as deep-sea cages have become the focus of attention of aquaculture operators and insurance agencies. However, the complex and changeable deep-sea environment, such as wind and waves, swells, biological erosion and other natural factors, can easily cause damage to these aquaculture facilities. Such damage will not only affect the benefits of aquaculture, but may also cause significant economic losses.
[0054] At present, damage assessment of deep-sea aquaculture facilities mainly relies on manual inspection, a traditional method with obvious limitations in practical application. Manual inspection is not only time-consuming and inefficient, but also has poor accuracy and consistency due to the subjectivity in the assessment process. In addition, the deep-sea environmental conditions are harsh, and manual inspections are difficult to fully cover all damaged areas, further increasing the difficulty and risk of assessment. These problems directly affect the efficiency and accuracy of insurance claims, making risk management of deep-sea aquaculture more complicated.
[0055] In the financial field, especially in the insurance industry, how to improve the efficiency and accuracy of damage assessment of deep-sea aquaculture facilities has become an urgent issue to be resolved.
[0056] See also Figure 1 , Figure 1 This is a flow chart of an embodiment of an intelligent detection method for underwater equipment damage provided by the present invention. It should be noted that although a logical sequence is shown in the flow chart, in some cases, the steps shown or described may be performed in a different order than that shown here.
[0057] like Figure 1 As shown, the underwater equipment damage intelligent detection method proposed by the present invention comprises the following steps:
[0058] S10, collecting device images and device status data of underwater target devices;
[0059] In this embodiment, the underwater target device refers to a device used in a deep sea or underwater environment, such as a deep sea cage, a submersible, a seabed sensor, etc. The appearance and environment of the device are captured by a high-resolution image taken by a camera or a drone installed around the underwater device. The device operation status information collected in real time by the sensor includes but is not limited to the device's temperature, pressure, depth, inclination, vibration and other physical parameters.
[0060] High-definition cameras and sensors (such as temperature sensors, pressure sensors, depth sensors, etc.) can be installed at key locations or around underwater target equipment, and data can be transmitted to a control center via wired or wireless means for real-time monitoring and recording.
[0061] In other specific embodiments, when collecting device images and device status data of underwater target devices, a high-definition camera carried by an autonomous mobile unmanned underwater vehicle (AUV) is used to flexibly approach and photograph various angles of the target device. The device status data is collected by various sensors of the AUV, including vibration sensors and acoustic sensors, to capture the operating stability of the device and the acoustic characteristics of the surrounding environment, and the data is transmitted back to the shore through an underwater acoustic communication system.
[0062] Through the above steps, comprehensive monitoring and damage detection of underwater target equipment are achieved, which effectively improves the accuracy and efficiency of detection, reduces the subjectivity and errors of manual operation, and provides a scientific basis for equipment maintenance and risk management.
[0063] S20, extracting a damaged area image of the underwater target device from the device image;
[0064] In this embodiment, the device images refer to images collected by underwater imaging equipment (such as cameras), which reflect the appearance of underwater devices. They are used to identify and analyze potential damage areas of the equipment. Images are usually collected by cameras fixed to the equipment or by cameras carried by unmanned underwater robots. These images capture the surface conditions of the equipment when it is running underwater.
[0065] The damaged area image is a portion separated from the device image, which clearly identifies possible damage on the device surface. The process of extracting the damaged area image usually involves image processing technology. Basic edge detection methods (such as Sobel and Canny algorithms) may be used to identify sudden changes in the image, which usually correspond to damaged areas on the device surface. Image segmentation algorithms (such as threshold-based segmentation) can separate these changes from the entire image and generate an image containing only the damaged area. It is used to further analyze the specific damaged areas on the device surface, such as cracks, deformations, or other physical damage. By accurately extracting these areas, the accuracy of damage assessment can be improved.
[0066] Edge detection is an image processing technique used to detect parts of an image where pixel values change dramatically, usually indicating the edges or boundaries of an object. Common edge detection algorithms include Canny, Sobel, etc. These algorithms identify edges by calculating the gradient of image pixels. In device images, they can be used to detect parts of the device surface that change suddenly, which may be signs of damage.
[0067] The threshold segmentation method can be used to divide the image into different regions according to the gray value or color information of the image. Usually, the damaged area will show characteristics that are significantly different from the surrounding background. For the extraction of damaged areas of underwater equipment, image segmentation can effectively separate these areas from other irrelevant parts.
[0068] In another specific embodiment, a multi-scale image analysis method is used to extract the damaged area image of the underwater target device from the device image. The device image is decomposed at multiple scales by wavelet transform to identify damage features at different scales. Then, combined with texture analysis methods such as gray level co-occurrence matrix (GLCM) and Gabor filter, the damaged area in the image is enhanced, and the features at each scale are compared to automatically identify and extract the most obvious damaged area. The extracted damaged area image is subjected to subsequent texture analysis and morphological processing to ensure the accuracy of the extracted area.
[0069] By extracting the damaged area image of the underwater target equipment from the equipment image, the potential damaged parts on the equipment surface can be accurately located and identified, which improves the accuracy and efficiency of damage detection, reduces the influence of human subjective judgment, ensures the reliability and consistency of the subsequent damage assessment process, and effectively improves the maintenance management level of the equipment.
[0070] S30, extracting damage features from the damaged area image and the device status data, wherein the damage features include shape, color and texture features of the damaged area;
[0071] In this embodiment, the damaged area image is a part extracted from the device image in the previous step, specifically showing the area on the device surface where damage may exist. The damaged area image is separated from the overall device image through image processing techniques, such as edge detection and image segmentation, to identify the damaged part of the device.
[0072] Equipment status data refers to the equipment operating status information collected in real time by various sensors (such as temperature sensors, pressure sensors, acceleration sensors, etc.). These data are collected by sensors installed in key parts of the equipment and synchronized with the damaged area image to analyze the nature and possible causes of the damage.
[0073] Damage features are key indicators that describe damage conditions extracted from damaged area images and equipment status data. They mainly include the following categories:
[0074] Shape characteristics: describe the geometric shape of the damaged area, such as area, perimeter, aspect ratio, roundness, etc.
[0075] Color features: describe the color information of the damaged area, such as average hue, color uniformity, color distribution concentration, etc.
[0076] Texture features: describe the surface texture of the damaged area, such as contrast, roughness, directionality, etc.
[0077] Shape feature extraction: The geometric properties of the damaged area are calculated through image analysis algorithms such as edge detection and morphological operations.
[0078] Color feature extraction: Through color space conversion (such as conversion from RGB to HSV or Lab) combined with color histogram, the color information of the damaged area is extracted.
[0079] Texture feature extraction: The surface characteristics of the damaged area are analyzed through texture analysis methods such as gray-level co-occurrence matrix (GLCM) and Gabor filter.
[0080] In other specific embodiments, when extracting damage features from damaged area images and equipment status data, three-dimensional reconstruction technology is used to extract the shape features of the damaged area and analyze the volume and depth of the damage. Color features are extracted through spectral analysis methods to identify color changes after damage to different materials. Texture features are combined with multi-directional Gabor filtering and adaptive threshold methods to detect surface structural changes in damaged areas. Equipment status data (such as vibration frequency and acoustic features) are used to conduct a comprehensive analysis of damage, generate maintenance recommendations, and help asset management decisions.
[0081] By extracting damage features from damaged area images and equipment status data, a comprehensive and accurate assessment of damage to underwater target equipment can be achieved, significantly improving the accuracy and efficiency of damage detection.
[0082] S40, using a pre-trained deep learning model to analyze the damage characteristics and generate a damage assessment result to determine the damage level of the underwater target equipment.
[0083] In this embodiment, the pre-trained deep learning model refers to a model that has been pre-trained on a large number of data sets. These models have learned rich feature representations and can be used directly for specific tasks or adapted to new tasks after fine-tuning. Convolutional neural network (CNN) is usually selected as the basic model because it is good at processing image data. Other possible models include ResNet, Inception, VGG, etc. The model is pre-trained on a large-scale public image dataset (such as ImageNet) or an industry-specific dataset so that it can recognize and extract common image features. Before being applied to underwater equipment damage detection, the model needs to be fine-tuned on a damage image dataset related to the field. With a small amount of task-specific data, the parameters of the model are further adjusted to make it more suitable for specific damage identification tasks.
[0084] Damage features are key indicators of damage description extracted from images and device status data through the previous steps, including shape, color, and texture features. The extracted damage features (such as area, perimeter, color histogram, texture features, etc.) are converted into feature vectors, which are used as inputs to the model. In order to ensure that the numerical ranges of different features are consistent, standardization techniques such as Z-score standardization or Min-Max normalization are applied to prevent a certain feature from dominating the model analysis.
[0085] The input damage features are processed by the deep learning model to generate quantitative analysis results of the damage. The input feature vector is processed by each layer of the model, including the convolution layer, the pooling layer, and the fully connected layer, to extract a higher-level feature representation. In the middle layer of the model, features such as shape, color, and texture are effectively integrated to form a comprehensive understanding of the damaged area. The final output layer uses activation functions such as Softmax to map the integrated features into a probability distribution of damage levels. By analyzing these probabilities, the severity of the damage is determined.
[0086] The damage assessment result refers to the damage level or score output by the model after analyzing the input features, reflecting the health status of the equipment. According to the probability distribution of the model output, the classification with the highest probability is selected as the damage level. For example, the model can output multiple levels such as "minor damage", "moderate damage", and "serious damage". The damage level is combined with the corresponding equipment status information to generate a detailed damage assessment report, including the damage area, damage type, recommended maintenance measures, etc.
[0087] In one specific embodiment, for the damage assessment of deep-sea aquaculture cages, when using a pre-trained deep learning model to analyze damage features, the model focuses on identifying common types of damage in deep-sea cages. First, the damage features are converted into multi-dimensional feature vectors, including shape features (such as the degree of damage to the cage structure), color features (such as color changes due to corrosion or biological attachment), and texture features (such as wear and tear of the mesh). These features are input into a specially trained convolutional neural network (CNN) model, which contains multiple convolutional layers and pooling layers to extract and enhance the expressiveness of the features. In the middle layer, the model uses an attention mechanism to focus on the most representative damage feature areas to improve the model's adaptability to complex damage scenarios. Finally, the model outputs the damage level, generates a probability distribution of multiple damage levels through the Softmax layer, determines the most likely damage level, and generates a detailed damage assessment report, including analysis of different damage areas and priority repair recommendations.
[0088] In another specific embodiment, the deep learning model used for equipment monitoring on offshore platforms uses a multi-task learning framework when analyzing damage features. This framework not only analyzes shape, color, and texture features, but also combines sensor data (such as vibration and temperature changes) to perform multimodal feature fusion. Each type of feature is processed by a different sub-network, shape features are processed by a ResNet-based sub-network, color features are processed by an adaptive color enhancement network, and texture features are analyzed using images processed by Gabor filters as input. The multi-task learning model simultaneously outputs damage classification and regression results (such as damage severity scores), and combines the results of each task in a weighted manner to generate the final damage level and repair priority recommendations.
[0089] By using the pre-trained deep learning model to analyze damage features, accurate assessment of the damage level of underwater target equipment can be achieved. This not only improves the accuracy of damage detection, but also provides real-time and reliable damage assessment reports, providing strong technical support for equipment maintenance, risk management and insurance claims, reducing subjective errors in manual assessment, and improving overall operating efficiency and equipment safety.
[0090] The present invention relates to an intelligent detection method for underwater equipment damage, which collects equipment images and equipment status data of underwater target equipment, pre-processes the equipment images, extracts damaged area images of underwater target equipment, extracts damage features from damaged area images and equipment status data, and uses a pre-trained deep learning model to analyze the damage features, generates damage assessment results, and determines the damage level of underwater target equipment. The present invention can realize intelligent detection and rapid assessment of underwater equipment damage, improves the accuracy and efficiency of damage detection, reduces the subjectivity and errors of manual operation, ensures the timeliness and reliability of equipment status monitoring, and provides a scientific basis for maintenance and insurance claims of underwater equipment.
[0091] In one embodiment, the above S20 includes:
[0092] S201, obtaining an undamaged image of a normal device in an undamaged state;
[0093] S202, calculating the color histograms of the intact image and the device image, calculating the color histogram difference between the intact image and the device image using the Bhattacharyya distance, and marking the area where the color histogram difference exceeds the color difference threshold as a damaged area;
[0094] S203, calculating the gray level co-occurrence matrix of the intact image and the device image, calculating the texture feature difference between the intact image and the device image based on the gray level co-occurrence matrix, and marking the area where the texture feature difference exceeds the texture difference threshold as a damaged area;
[0095] S204, calculating a grayscale histogram of the device image, selecting a global threshold according to the grayscale histogram, marking an area in the device image with a grayscale value higher than the threshold as a potential damage area, and marking an accurate damage area based on comparing the potential damage area with the damage-free image;
[0096] S205, extracting a damaged area image from the device image based on the marked damaged area.
[0097] In this embodiment, the undamaged image refers to a standard image taken when the equipment is not damaged, which is used as a reference image for subsequent damage detection. The camera collects images of the equipment in an undamaged state when the equipment is initially installed or maintained. These images are stored and used to compare with the images collected later to identify possible damaged areas.
[0098] The color histogram is a statistical graph of the frequency of occurrence of different color values in an image, reflecting the overall color distribution of the image. Convert the intact image and the real-time device image to HSV or Lab color space respectively, and calculate the color histogram of each channel. This step is usually completed using an image processing library (such as OpenCV).
[0099] Bhattacharyya distance is a distance that measures the similarity between two probability distributions and is often used to compare the color histogram differences between two images. The color histogram of the intact image is compared with the color histogram of the device image to calculate the Bhattacharyya distance. The smaller the distance value, the more similar the color distributions of the two images are. Otherwise, there may be damage.
[0100] The color difference threshold is used to determine whether a part of an image may be damaged. After calculating the Bhattacharyya distance, the areas exceeding the predetermined threshold are marked as possible damaged areas. These areas have significant differences in color features from the damage-free image.
[0101] Gray-level co-occurrence matrix (GLCM) is used to describe the spatial relationship between gray values in an image and reflect the texture characteristics of the image. For damage-free images and device images, their gray-level co-occurrence matrices are calculated to extract the texture characteristics of the image. Typical texture characteristics include contrast, entropy, uniformity, etc.
[0102] Texture feature differences are differences between image texture features extracted based on the gray-level co-occurrence matrix, and are used to identify possible damaged areas in the image. The difference between texture features is calculated by comparing the gray-level co-occurrence matrix of the undamaged image and the device image. If the difference value exceeds the preset texture difference threshold, it is marked as a possible damaged area.
[0103] Grayscale histogram is a statistical graph of the frequency of occurrence of each grayscale value in an image, reflecting the brightness distribution of the image. Grayscale processing is performed on the device image, and its grayscale histogram is calculated to identify possible damaged areas.
[0104] The global threshold is a gray value threshold used to distinguish between possible damaged areas and normal areas in an image. According to the gray histogram of the device image, a suitable global threshold is selected to mark areas with gray values higher than the threshold as potential damaged areas.
[0105] After initially identifying the potential damage area, the final damage area is accurately marked by comparing it pixel by pixel with the damage-free image. The potential damage area is carefully compared with the damage-free image, combined with the previous color and texture feature differences, and the accurate damage area is finally confirmed.
[0106] After the damaged areas are identified, they are extracted from the overall device image to form separate damaged area images for further analysis. Using image segmentation technology, the marked damaged areas are extracted from the device image to form damaged area images for subsequent analysis.
[0107] In another specific embodiment, it is applied to the monitoring of deep-sea communication cables. When extracting the damaged area image from the equipment image, the standard image in the undamaged state is first obtained by hyperspectral imaging technology to capture a wider range of spectral information. Then, the color histograms of the equipment image and the undamaged image are calculated, and the Bhattacharyya distance based on spectral matching is used to further improve the detection accuracy of color differences. When calculating the grayscale co-occurrence matrix, the multi-scale local binary pattern (LBP) is combined with the grayscale co-occurrence matrix to enhance the detection ability of subtle texture differences. During the grayscale histogram calculation process, the equipment image is processed by adaptive histogram equalization (CLAHE) to enhance the contrast of the image and improve the recognition effect of potential damaged areas. When the damaged area is finally marked, a generative adversarial network (GAN) is used to generate multiple possible damaged areas, and the most accurate damaged area image is selected for extraction by comparing the real image.
[0108] Through the above steps, this embodiment can accurately extract possible damage areas from the device image, comprehensively consider multiple features such as color, texture and grayscale, improve the accuracy and reliability of damage detection, and reduce misjudgment.
[0109] In one embodiment, in the above S30, extracting damage features from the damaged area image and the device status data includes:
[0110] S301, extracting the edge of the damaged area using the Canny edge detection algorithm based on the damaged area image, identifying and tracking the outline of the damaged area, and calculating the shape features of the damaged area, wherein the shape features include area, perimeter, aspect ratio, and roundness;
[0111] S302, converting the damaged area image from the RGB color space to the HSV or Lab color space, generating a color histogram for the Hue, Saturation, Value or L, a, b channels, and extracting key color features based on the color histogram, wherein the color features include average hue, color saturation, and concentration of color distribution;
[0112] S303, calculating the gray level co-occurrence matrix of the damaged area, using a Gabor filter to extract the texture frequency and directional features of the damaged area, and combining the gray level co-occurrence matrix, texture frequency and directional features to calculate a comprehensive texture feature value;
[0113] S304, extracting temperature features, pressure features, and depth features from the device status data;
[0114] S305, fusing the extracted shape features, color features and texture features with the temperature features, pressure features and depth features to obtain a comprehensive damage feature.
[0115] In this embodiment, the damaged area image is firstly subjected to Gaussian filtering to reduce noise; then the gradient amplitude and direction of the image are calculated; then, the non-maximum suppression technique is applied to retain only the local maximum of the gradient amplitude; finally, the complete edge of the damaged area is formed through dual threshold detection and edge connection.
[0116] Shape features describe the geometric properties of the damaged area, including area, perimeter, aspect ratio, and circularity. After extracting the edges, a contour tracking algorithm (such as the findContours function in OpenCV) is used to identify the contour of the damaged area. Then, the area and perimeter of the contour are calculated, and the aspect ratio (the ratio of the width to the height of the minimum circumscribed rectangle of the area) and circularity (4π×area / perimeter2) are further calculated.
[0117] HSV and Lab are two different color spaces that are more perceptual and robust than RGB in expressing color features. The damaged area image is converted from RGB space to HSV or Lab space in order to extract color features more effectively. The HSV space consists of hue, saturation, and value, while the Lab space consists of lightness (L), a channel (representing green to red), and b channel (representing blue to yellow).
[0118] A color histogram is a statistical chart of the distribution of different color values in an image, used to describe color features. In the converted color space, color histograms are generated for the Hue, Saturation, and Value (or L, a, b) channels respectively. Through these histograms, key color features such as average hue (Hue), color saturation (Saturation), and the concentration of color distribution can be extracted.
[0119] Gray-level co-occurrence matrix (GLCM) is used to describe the spatial relationship of pixel gray values in an image, thereby extracting the texture features of the image. For the damaged area image, it is first converted into a grayscale image, and then its gray-level co-occurrence matrix is calculated to extract texture features such as contrast, entropy, and uniformity. These features reflect the surface roughness and complexity of the damaged area.
[0120] The Gabor filter is a linear filter used to detect texture features at specific frequencies and directions. The Gabor filter is applied to the damaged area image to extract its responses at multiple directions and frequencies. Through these responses, the texture frequency and direction features of the damaged area can be extracted to analyze the periodicity and directionality of the surface texture.
[0121] The comprehensive texture feature value is a combination of multiple texture features, providing a comprehensive description of the surface characteristics of the damaged area. The texture features extracted from the gray level co-occurrence matrix and Gabor filter are normalized and weighted, and then one or more comprehensive texture feature values are calculated to represent the overall texture properties of the damaged area.
[0122] The device status data includes physical quantities such as temperature, pressure, and depth collected by sensors, which provide information about the environment and device operating status. The device's temperature, pressure, and depth data are collected in real time and standardized into feature vectors for fusion with image features.
[0123] Feature fusion refers to combining features extracted from different sources or types of data to form a richer and more comprehensive description. The shape, color and texture features extracted from the image are fused with the equipment status data (such as temperature, pressure and depth) to generate comprehensive damage features using methods such as weighted averaging or feature splicing. These comprehensive features can more accurately reflect the degree of damage to the equipment.
[0124] Through the above steps, this embodiment can comprehensively and accurately extract the damage characteristics of underwater target equipment, comprehensively consider multi-dimensional factors such as shape, color, texture, and equipment operation status, and provide a scientific basis for damage assessment. It not only improves the accuracy and efficiency of damage detection, but also enhances the system's adaptability to different environments and equipment states, and effectively supports equipment maintenance and risk management.
[0125] In one embodiment, in the above S40, the damage characteristics are analyzed using a pre-trained deep learning model to generate a damage assessment result, including:
[0126] S401, inputting the damage feature into the deep learning model, and assigning damage weights to the shape feature, color feature and texture feature of the damage area respectively;
[0127] S402, calculating a shape damage score based on the shape characteristics of the damage area;
[0128] S403, calculating a color damage score based on the color features of the damaged area;
[0129] S404, calculating a texture damage score based on the texture features of the damaged area;
[0130] S405, calculating a comprehensive damage score according to the shape damage score, the color damage score, the texture damage score and the damage weight;
[0131] S406: Generate a damage assessment result including a damage level based on the comprehensive damage score and the damage level rule.
[0132] In this embodiment, the damage feature is input into the deep learning model, which means that the shape, color, texture and other features extracted from the image and sensor data are input into the deep learning model in the form of feature vectors for processing. The damage weight is a weighted weight of the importance of different features, and these weights can be automatically learned based on the training data. After the damage features such as shape, color, and texture are normalized, they are input into the pre-trained deep learning model as input vectors. Within the model, different weights are assigned to each feature, and these weights are continuously adjusted during the training process through the back-propagation algorithm, so that the model can automatically learn the optimal weight allocation strategy.
[0133] The shape damage score is a quantitative score obtained by analyzing the geometric shape features of the damaged area (such as area, perimeter, aspect ratio, roundness, etc.), which is used to measure the contribution of shape features to damage. After the shape features are input into the model, the model calculates the shape damage score through a fully connected layer or a specific regression layer. The score reflects the severity of the shape feature to the overall damage. The higher the score, the more severe the damage shown by the shape feature.
[0134] The color damage score is a score based on the color features of the damaged area (such as average hue, color saturation, concentration of color distribution, etc.) to evaluate the impact of color changes on damage. The color features are input into the model, and the color damage score is calculated using the weights and regression layer of the color features. This score is used to quantify the degree of impact of the color features on the overall damage.
[0135] The texture damage score is obtained by analyzing the texture features of the damaged area (such as contrast, uniformity, directionality, etc.), reflecting the contribution of texture features to the damage. The texture features are input into a specific layer of the model, processed using a convolutional layer or other layer suitable for texture analysis, and finally a texture damage score is generated. The score indicates the significance of the texture feature in the damaged area. The higher the score, the greater the contribution of texture changes to the damage.
[0136] The comprehensive damage score is a weighted average of the three damage scores of shape, color, and texture, which is used to indicate the severity of the overall damage. The shape, color, and texture damage scores are weighted and summed using the fully connected layer within the pre-trained deep learning model to generate a comprehensive damage score. This score combines the influence of different features to provide a more comprehensive damage assessment.
[0137] The damage level is generated based on the comprehensive damage score and the predetermined grade classification rules to classify the degree of damage to the equipment. The comprehensive damage score is compared with the preset damage level standard, and the damage assessment result is generated through the classifier layer of the deep learning model, and the corresponding damage level is attached, such as "minor damage", "moderate damage", "serious damage", etc.
[0138] This embodiment can automatically generate accurate damage assessment results by analyzing damage features using a pre-trained deep learning model. The independent scores of each damage feature (shape, color, texture) are weighted and fused to generate a comprehensive damage score, making the overall damage assessment more comprehensive and accurate. The final damage level assessment result not only improves the automation of detection, but also reduces the error of manual intervention, significantly improving the efficiency and scientificity of equipment maintenance decisions. It can adapt to different environments and equipment types, and is widely used in damage detection and assessment of underwater equipment.
[0139] In one embodiment, the above S404 includes:
[0140] S4041, obtaining an undamaged image of a normal device in an undamaged state to establish a reference texture feature;
[0141] S4042, extracting texture features from the damage-free image, the texture features including contrast, uniformity and roughness, and assigning a corresponding weight to each texture feature;
[0142] S4043, comparing the texture features of the damaged area with the texture features of the undamaged image, and calculating the feature difference between each texture feature;
[0143] S4044, determining a texture damage score based on the feature difference between each texture feature and the corresponding weight.
[0144] In this embodiment, texture features describe the pattern and regularity of pixel value changes in an image. Commonly used texture features include contrast (indicating the degree of change in the grayscale of an image), uniformity (indicating whether the grayscale distribution of an image is consistent), and roughness (indicating the coarseness and fineness of the surface texture). The contrast and uniformity of the damage-free image are extracted by using a gray level co-occurrence matrix (GLCM), and the roughness feature is extracted by using image processing technology (such as a method based on wavelet transform or Fourier transform). These features are standardized to form a reference texture feature vector.
[0145] Weights are a quantification of the importance of different texture features. Different weight assignments can reflect the relative importance of each texture feature in damage assessment. Initial weights are assigned to texture features such as contrast, uniformity, and roughness based on a pre-trained model or through expert experience. Weights can be dynamically adjusted during model training to optimize the accuracy of damage scoring.
[0146] Feature difference refers to the numerical difference between the texture feature of the damaged area and the reference texture feature, reflecting the change of texture. The difference in contrast, uniformity and roughness between the damaged area and the undamaged area is calculated by comparing pixel by pixel or based on feature vectors. The larger the difference value, the greater the deviation of the texture feature of the damaged area from the reference texture feature.
[0147] The texture damage score is a score calculated based on the difference of each texture feature and the corresponding weight, which is used to quantify the contribution of texture changes to damage. The difference value of each texture feature is multiplied by the corresponding weight to obtain a weighted difference value. Then, these weighted difference values are summed or weighted averaged to obtain the final texture damage score. The higher the score, the greater the difference between the texture features of the damaged area and the baseline texture features, that is, the more severe the damage.
[0148] In one specific embodiment, it is applied to damage detection of deep-sea aquaculture cages, extracting texture features from the undamaged images of deep-sea aquaculture cages taken at the initial stage of installation, extracting contrast and uniformity using gray-level co-occurrence matrix (GLCM), and extracting roughness features using wavelet transform. These features are stored as baseline texture features after standardization. Different texture feature weights are assigned based on the characteristics of deep-sea cage materials. Contrast and roughness are given higher weights because they have a greater impact on the detection of corrosion and breakage, while uniformity is assigned a lower weight. By comparing the current damaged area with the baseline texture features, the difference value of each texture feature is calculated. Combined with the weights, the weighted difference values of contrast and roughness have a greater impact on the texture damage score. The final generated texture damage score reflects the degree of texture abnormality in the damaged area of the cage, providing a basis for further damage assessment.
[0149] Through the above steps, this embodiment can accurately calculate the texture damage score based on the texture features of the damaged area, providing a detailed quantitative basis for the damage assessment of the equipment. Combining the baseline texture features of the undamaged image with the comparative analysis of the current damaged area effectively improves the accuracy and reliability of texture damage detection.
[0150] In one embodiment, in the above S40, after generating the damage assessment result, the following step is further included:
[0151] S501, obtaining the insurance terms of the underwater target equipment, including insurance coverage, compensation standards, deductibles and claim settlement procedures;
[0152] S502, based on the damage assessment result, determining a damaged portion of the underwater target equipment that meets the insurance compensation scope;
[0153] S503, calculating out-of-pocket expenses and deductibles based on the damaged portion that meets the insurance compensation scope, and generating repair suggestions;
[0154] S504: Generate a damage assessment report based on the out-of-pocket expenses, deductible, and repair recommendations.
[0155] In this embodiment, the insurance clauses include relevant information specified in the insurance contract, such as insurance coverage, compensation standards, deductibles, and claims settlement procedures, which are used to determine whether the damaged part is covered by the insurance. Relevant insurance clause information is extracted from the insurance company's database or contract documents to ensure that all necessary clauses are covered, such as which damages are covered, what the compensation ratio is, what the deductible is, and the specific steps for claims settlement.
[0156] According to the insurance terms, determine which damages in the damage assessment results meet the insurance compensation conditions. Compare the damage assessment results with the insurance terms and conditions, and identify the damage parts that meet the compensation scope through the rule engine or automated decision-making system. This may involve checking whether the damage type, damage degree, and cause are covered by the insurance.
[0157] Based on the compensation standards and deductibles stipulated in the insurance contract, the expenses (out-of-pocket expenses) that need to be borne by the equipment owner are calculated, and corresponding repair recommendations are made. The calculation process involves deducting the deductible from the compensation standards and generating repair recommendations based on the cost of the repair, the severity of the damage, possible repair measures, etc. The system automatically processes these calculations and presents the results to the user, including the recommended repair plan and expected costs.
[0158] The damage assessment report is a comprehensive assessment of the damage, including out-of-pocket expenses, deductibles, repair recommendations, etc., for insurance claims and subsequent repairs. The system integrates all relevant information into a report, including detailed damage assessment results, the part eligible for insurance compensation, calculated out-of-pocket expenses and deductibles, and recommended repair plans. The report format is usually a standardized document that is easy for insurance companies and equipment owners to use.
[0159] In one specific embodiment, the insurance claim for deep-sea aquaculture cages is applied, and the insurance clauses related to deep-sea aquaculture cages are extracted from the contract management system stored in the cloud. The clauses include the insurance coverage of the equipment, the compensation ratio for specific damage, the deductible, and the detailed process of claim settlement. The system automatically parses these clauses and associates them with the damage assessment results. The system automatically compares the damage assessment results with the insurance clauses to identify whether the damage to the cage belongs to the risk type covered by the insurance, such as corrosion or mechanical damage caused by the marine environment. The eligible damaged part will be marked for subsequent calculations. Based on the identified compensation scope, the cost part that the equipment owner needs to bear is calculated. The system considers the deductible, compensation ratio, and recommended maintenance plan, and automatically generates detailed maintenance suggestions, including recommended materials, estimated construction period, and cost estimates. All calculation results and suggestions are summarized to generate a damage assessment report, which includes a description of the damage, the damaged part that meets the insurance compensation, out-of-pocket expenses, deductibles, and detailed maintenance suggestions. The report format is a standardized PDF document, which is convenient for online submission to the insurance company for claims.
[0160] Through the above steps, this embodiment can systematically match the damage assessment results with the insurance terms, automatically identify the damaged parts that meet the insurance compensation scope, and generate detailed repair suggestions and damage assessment reports. This not only improves the efficiency of claims settlement, but also reduces the errors of manual judgment, provides clear and accurate claims settlement guidance for equipment owners, and ensures the scientificity and rationality of repair decisions.
[0161] In one embodiment, before the above S10, the method further includes:
[0162] S001, installing four cameras at four corners of the underwater target device, so that the horizontal distance between each camera is equal and each camera covers a viewing angle of 120 degrees;
[0163] S002, installing a top camera at the center of the top of the underwater target device, and the top camera shoots vertically downward;
[0164] S003, installing a bottom camera at the center of the bottom of the underwater target device, and the bottom camera shoots vertically upward;
[0165] S004, installing depth sensors on four corners of the underwater target device, installing temperature sensors on preset key positions of the underwater target device, and installing pressure sensors on the surface of the underwater target device.
[0166] In this embodiment, fixed cameras are installed at the four corners of the device, and the installation angle and position of the cameras are adjusted so that the horizontal distance between each camera is equal, and the camera viewing angle is configured to ensure that the shooting area of each camera can seamlessly cover the environment around the device. This configuration ensures 360-degree full coverage around the device and avoids any blind spots.
[0167] Select a suitable camera and install it in the center of the top of the equipment, ensuring that the camera can shoot the top area of the equipment vertically downward. This helps to capture the full view of the top of the equipment, especially when there is possible damage or abnormalities on the surface of the equipment, the camera can provide detailed image data.
[0168] Another camera is installed at the center of the bottom of the device, with the camera's viewpoint facing vertically upward and focusing on the area at the bottom of the device. This camera can capture a panoramic image of the bottom of the device, especially in deep water environments, which helps to monitor possible damage or abnormalities at the bottom of the device.
[0169] Select depth, temperature and pressure sensors suitable for underwater use and install them in the four corners, key parts and surfaces of the equipment. Depth sensors are used to detect the performance of the equipment at different depths; temperature sensors are placed in parts of the equipment that are susceptible to ambient temperature to monitor temperature changes; pressure sensors are installed on the outer surface of the equipment to monitor the impact of water pressure on the equipment. Sensors collect data in real time and provide continuous monitoring of the equipment status.
[0170] In one specific embodiment, the top camera is equipped with an optical zoom function, which can automatically adjust the focal length to observe key areas on the top of the cage in detail, such as the water inlet and the drain. The bottom camera adds an infrared night vision function, which enables it to capture clear images even in low-light environments, and is particularly suitable for monitoring the status of the bottom of the cage around the clock in deep-sea environments. The depth sensor uses multi-band ultrasonic depth measurement technology, which can more accurately monitor the depth changes of underwater equipment and automatically correct measurement errors caused by changes in water pressure. The temperature sensor uses a highly sensitive semiconductor temperature sensor that can quickly respond to temperature changes in extreme environments. The pressure sensor uses piezoelectric technology to monitor the water pressure on the surface of the equipment in real time, and dynamically adjusts it based on the depth data to ensure the safe operation of the equipment at different depths.
[0171] Through the above steps, this embodiment can ensure comprehensive monitoring and accurate status monitoring of underwater target equipment. Cameras and sensors installed at different locations of the equipment can collect all-round images and environmental data in real time, which helps to detect potential risks and problems early, and provide reliable data support to ensure the safe operation and maintenance decision-making of the equipment.
[0172] The present invention also provides an intelligent detection device for underwater equipment damage, referring to Figure 2 , Figure 2 This is a functional module diagram of a preferred embodiment of the underwater equipment damage intelligent detection device of the present invention. The underwater equipment damage intelligent detection device comprises:
[0173] A data acquisition module, which collects device images and device status data of underwater target devices;
[0174] An image processing module extracts a damaged area image of the underwater target device from the device image;
[0175] A feature extraction module extracts damage features from the damage area image and the device status data, wherein the damage features include shape, color and texture features of the damage area;
[0176] The damage assessment module uses a pre-trained deep learning model to analyze the damage characteristics and generate a damage assessment result to determine the damage level of the underwater target equipment.
[0177] The specific implementation of the intelligent detection device for underwater equipment damage of the present invention is basically the same as the embodiments of the above-mentioned intelligent detection method for underwater equipment damage, and will not be repeated here.
[0178] The present invention also provides an intelligent underwater equipment damage detection device, such as Figure 3As shown, the underwater equipment damage intelligent detection device may include: a processor 1001, such as a CPU, a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. Among them, the communication bus 1002 is used to realize the connection and communication between these components. The user interface 1003 may include a display screen (Display), an input unit such as a keyboard (Keyboard), and the user interface 1003 may also include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a WI-FI interface). The memory 1005 may be a high-speed RAM memory, or a stable memory (non-volatile memory), such as a disk memory. The memory 1005 may also be a storage device independent of the aforementioned processor 1001.
[0179] Those skilled in the art will understand that Figure 3 The hardware structure of the underwater equipment damage intelligent detection device shown in the figure does not constitute a limitation of the underwater equipment damage intelligent detection device, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.
[0180] like Figure 3 As shown, the memory 1005 as a storage medium may include an operating system, a network communication module, a user interface module, and an underwater equipment damage intelligent detection program. The operating system is a program for managing and controlling the underwater equipment damage intelligent detection device and software resources, and supports the operation of the network communication module, the user interface module, the underwater equipment damage intelligent detection program, and other programs or software; the network communication module is used to manage and control the network interface 1004; the user interface module is used to manage and control the user interface 1003.
[0181] exist Figure 3 In the hardware structure of the underwater equipment damage intelligent detection device shown, the network interface 1004 is mainly used to connect to the background server and communicate data with the background server; the user interface 1003 is mainly used to connect to the client and communicate data with the client; the processor 1001 can call the underwater equipment damage intelligent detection program stored in the memory 1005, and perform the same operations as the underwater equipment damage intelligent detection method.
[0182] The specific implementation of the underwater equipment damage intelligent detection device of the present invention is basically the same as the embodiments of the underwater equipment damage intelligent detection method described above, and will not be repeated here.
[0183] In addition, an embodiment of the present invention further proposes a computer storage medium, on which is stored an underwater equipment damage intelligent detection program, which, when executed by a processor, implements the steps of the underwater equipment damage intelligent detection method described above.
[0184] The specific implementation of the computer storage medium of the present invention is basically the same as the embodiments of the above-mentioned underwater equipment damage intelligent detection method, and will not be repeated here.
Claims
1. An intelligent detection method for underwater equipment damage, characterized in that: The following steps are involved: Collect equipment images and equipment status data of underwater target equipment; extracting a damaged area image of the underwater target device from the device image; Extracting damage features from the damaged area image and the device status data, wherein the damage features include shape, color and texture features of the damaged area; The damage characteristics are analyzed using a pre-trained deep learning model to generate a damage assessment result to determine the damage level of the underwater target equipment.
2. The intelligent detection method for underwater equipment damage according to claim 1, characterized in that: Extracting a damaged area image of the underwater target device from the device image includes: Obtaining damage-free images of normal equipment in an undamaged state; Calculating the color histograms of the intact image and the device image, calculating the color histogram difference between the intact image and the device image using the Bhattacharyya distance, and marking the area where the color histogram difference exceeds a color difference threshold as a damaged area; Calculating the gray level co-occurrence matrix of the intact image and the device image, calculating the texture feature difference between the intact image and the device image based on the gray level co-occurrence matrix, and marking the area where the texture feature difference exceeds the texture difference threshold as a damaged area; Calculating a grayscale histogram of the device image, selecting a global threshold according to the grayscale histogram, marking an area in the device image with a grayscale value higher than the threshold as a potential damage area, and marking an accurate damage area based on comparing the potential damage area with the damage-free image; Based on the marked damaged area, a damaged area image is extracted from the device image.
3. The intelligent detection method for underwater equipment damage according to claim 1, characterized in that: Extracting damage features from the damaged area image and the device status data includes: Based on the damaged area image, the Canny edge detection algorithm is used to extract the edge of the damaged area, identify and track the outline of the damaged area, and calculate the shape features of the damaged area, wherein the shape features include area, perimeter, aspect ratio and roundness; Converting the damaged area image from RGB color space to HSV or Lab color space, generating color histograms for Hue, Saturation, Value or L, a, b channels, and extracting key color features based on the color histograms, the color features including average hue, color saturation, and concentration of color distribution; The gray-level co-occurrence matrix of the damaged area is calculated, and the texture frequency and directional features of the damaged area are extracted using a Gabor filter. The comprehensive texture feature value is calculated by combining the gray-level co-occurrence matrix, texture frequency and directional features. extracting temperature features, pressure features, and depth features from the device status data; The extracted shape features, color features and texture features are fused with the temperature features, pressure features and depth features to obtain a comprehensive damage feature.
4. The intelligent detection method for underwater equipment damage according to claim 1, characterized in that: The damage characteristics are analyzed using a pre-trained deep learning model to generate damage assessment results, including: Inputting the damage features into the deep learning model, and assigning damage weights to the shape features, color features, and texture features of the damaged area respectively; Calculating a shape damage score based on the shape characteristics of the damage area; Calculating a color damage score based on the color characteristics of the damaged area; Calculating a texture damage score based on the texture features of the damaged area; Calculate a comprehensive damage score according to the shape damage score, color damage score, texture damage score and damage weight; A damage assessment result including a damage level is generated based on the comprehensive damage score and the damage level rule.
5. The intelligent detection method for underwater equipment damage according to claim 4, characterized in that: Calculating a texture damage score based on the texture features of the damaged area includes: Obtaining an undamaged image of a normal device in an undamaged state to establish a baseline texture feature; Extracting texture features from the damage-free image, the texture features including contrast, uniformity and roughness, and assigning a corresponding weight to each texture feature; Comparing the texture features of the damaged area with the texture features of the undamaged image, and calculating the feature difference between each texture feature; The texture damage score is determined based on the feature differences between each texture feature and the corresponding weights.
6. The intelligent detection method for underwater equipment damage according to claim 1, characterized in that: After the damage assessment results are generated, they also include: Obtain the insurance terms of the underwater target equipment, including insurance coverage, compensation standards, deductibles and claims settlement procedures; Based on the damage assessment result, determining the damaged portion of the underwater target equipment that meets the insurance compensation scope; Calculate out-of-pocket expenses and deductibles based on the portion of damage eligible for insurance coverage, and generate repair recommendations; A damage assessment report is generated based on the out-of-pocket expenses, deductibles, and repair recommendations.
7. The intelligent detection method for underwater equipment damage as claimed in claim 1, characterized in that: Before collecting the device image and device status data of the underwater target device, it also includes: Installing four cameras at four corners of the underwater target device so that the horizontal distance between each camera is equal and each camera covers a viewing angle of 120 degrees; A top camera is installed at the center of the top of the underwater target device, and the top camera shoots vertically downward; A bottom camera is installed at the center of the bottom of the underwater target device, and the bottom camera shoots vertically upward; Depth sensors are installed on the four corners of the underwater target device, temperature sensors are installed on preset key positions of the underwater target device, and pressure sensors are installed on the surface of the underwater target device.
8. An intelligent detection device for underwater equipment damage, characterized in that: The underwater equipment damage intelligent detection device comprises: A data acquisition module, which collects device images and device status data of underwater target devices; An image processing module extracts a damaged area image of the underwater target device from the device image; A feature extraction module extracts damage features from the damaged area image and the device status data, wherein the damage features include shape, color and texture features of the damaged area; The damage assessment module uses a pre-trained deep learning model to analyze the damage characteristics and generate a damage assessment result to determine the damage level of the underwater target equipment.
9. An intelligent underwater equipment damage detection device, characterized in that: The underwater equipment damage intelligent detection device includes a memory, a processor, and an underwater equipment damage intelligent detection program stored in the memory and executable on the processor. When the underwater equipment damage intelligent detection program is executed by the processor, the steps of the underwater equipment damage intelligent detection method as described in any one of claims 1 to 7 are implemented.
10. A computer storage medium, characterized in that: The storage medium stores an underwater equipment damage intelligent detection program, and when the underwater equipment damage intelligent detection program is executed by the processor, the steps of the underwater equipment damage intelligent detection method as described in any one of claims 1-7 are implemented.
Citation Information
Cited By
Image credibility detection method, system and device and storage medium
CN119600426A
Steel wire rope surface damage intelligent identification and diameter measurement synchronous detection system
CN119667105A
A synchronous detection system for intelligent identification of wire rope surface damage and diameter measurement
CN119667105B
Toughness-driven underwater production system on-condition maintenance decision optimization method and system
CN119919124A
Cross guide rail wear state monitoring method based on machine vision
CN120672761A
