Cable defect detection method and system based on image processing

Abstract

The invention discloses a cable defect detection method and system based on image processing, and the method comprises the following steps: S1, multi-view image collection: carrying out the image collection of a cable from different views at the same time through a plurality of image collection devices, obtaining a plurality of groups of images containing the surface and different depth level information of the cable, and meanwhile, the spatial position and attitude parameters of each acquisition device are recorded. According to the cable defect detection method and system based on image processing, comprehensive cable information is acquired through multi-view image acquisition, the image quality is improved in combination with adaptive image enhancement, and the defects are accurately identified by using multi-scale feature fusion and deep transfer learning, so that compared with the prior art, various defects of the cable can be more accurately detected, and the detection accuracy is improved. The defects include fine defects and hidden defects, and the detection precision is effectively improved. For example, in an actual test, tiny insulation layer damage defects which are difficult to detect by some traditional methods can be accurately identified and positioned by the detection system.

Description

Technical Field

[0001] This invention relates to the field of cable inspection technology, specifically to a cable defect detection method and system based on image processing. Background Technology

[0002] In power systems, cables serve as a crucial carrier for electrical energy transmission, and their safe and stable operation directly impacts the reliability of power supply. With accelerating urbanization and continuously increasing electricity demand, the scale of cable installations is expanding, and their application environments are becoming increasingly complex and diverse, including urban underground networks, mountainous areas, and high-humidity coastal regions. Therefore, timely and accurate detection of cable defects is essential for ensuring the safe operation of power systems.

[0003] Traditional cable defect detection methods, such as manual inspection and offline electrical testing, have many limitations. Manual inspection is not only inefficient and highly susceptible to human error, but also struggles to detect hidden and subtle defects. Offline electrical testing requires power outages, disrupting power supply continuity, and can only detect some electrical performance-related defects, failing to comprehensively reflect the cable's actual condition. In recent years, image processing-based cable defect detection technology has gradually become a research hotspot. However, existing technologies still have significant shortcomings. On the one hand, in complex environments, such as those with drastic lighting changes and strong electromagnetic interference, the quality of acquired cable images is unstable. Existing detection systems lack effective image enhancement and adaptive processing capabilities, leading to difficulties in defect feature extraction and a significant decrease in detection accuracy. On the other hand, existing methods often rely on single-scale image feature analysis, making it difficult to simultaneously identify defects of different sizes and types, and exhibiting poor adaptability to complex cable structures and diverse defects. Furthermore, existing systems cannot dynamically adjust detection parameters in real time based on the cable's operating environment and its own operating status, making it difficult to maintain stable and efficient detection performance under different operating conditions. Therefore, we propose an image processing-based cable defect detection method and system. Summary of the Invention

[0004] The purpose of this invention is to provide a cable defect detection method and system based on image processing, so as to solve the problems of low detection accuracy, poor adaptability to complex environments, inability to identify defects of different sizes and types, and inability to dynamically adjust detection parameters according to operating conditions in the existing cable defect detection technology mentioned in the background.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a cable defect detection method based on image processing, comprising the following steps: S1, Multi-view image acquisition: Multiple image acquisition devices are used to simultaneously acquire images of the cable from different perspectives, obtaining multiple sets of images containing information about the cable surface and different depth levels. At the same time, the spatial position and attitude parameters of each acquisition device are recorded to establish the correspondence between the images and the actual position of the cable.

[0006] S2, Adaptive Image Enhancement: For the image acquired in step S1, an appropriate enhancement algorithm is automatically selected based on the image's noise distribution, contrast, and brightness characteristics. For example, a method based on a combination of local histogram equalization and homomorphic filtering is used to adaptively enhance the image, highlighting cable surface details and potential defect features.

[0007] S3, Multi-scale Feature Fusion and Deep Transfer Learning: For the enhanced image, a multi-scale convolutional neural network structure is used to extract image features at different scales and perform feature fusion. The fused features are then input into a deep neural network model based on transfer learning. This model is pre-trained on a large number of general image datasets and fine-tuned for the cable defect detection task to achieve accurate identification and classification of cable defects.

[0008] S4, Dynamic Parameter Adjustment: Real-time acquisition of cable operating environment parameters, such as temperature, humidity, electromagnetic interference intensity, and cable's own operating parameters, such as current or voltage. Based on changes in these parameters, using a pre-established parameter correlation model, the threshold, feature weights, and other key parameters in the defect detection process are dynamically adjusted to adapt to the detection needs under different working conditions.

[0009] S5, Result Output and Early Warning: Based on the analysis results of steps S3 and S4, output a detailed cable defect report, including defect location, type and severity information. If a defect that seriously affects the safe operation of the cable is detected, an early warning signal will be issued immediately through various means, such as audible and visual alarms and SMS notifications.

[0010] Preferably, the image acquisition device in step S1 includes a visible light camera, an ultraviolet imager, and an ultrasonic imager, which are used to acquire visible light images of the cable surface, ultraviolet images generated by corona discharge, and ultrasonic images of the internal structure of the cable, respectively, so as to realize the acquisition of multi-dimensional information of the cable.

[0011] Preferably, in step S2, when selecting an enhancement method, the adaptive image enhancement algorithm first evaluates the noise of the image. If the noise level is higher than a set threshold, homomorphic filtering with good denoising effect is preferred. If the image contrast is low, local histogram equalization is used to improve the contrast, so as to achieve targeted enhancement of images of different quality.

[0012] Preferably, in step S3, the multi-scale convolutional neural network structure adopts a pyramid-shaped convolutional layer design, gradually expanding the receptive field of the convolutional kernel from shallow to deep layers to obtain image features at different scales. Then, through feature splicing and fusion operations, the multi-scale features are input into a deep neural network model based on transfer learning. This model is an improved Faster R-CNN network that introduces an attention mechanism module to enhance attention to defect features.

[0013] Preferably, the parameter association model in step S4 is obtained by machine learning training on a large amount of historical environmental parameters, cable operation parameters and defect detection results data, and nonlinear mapping relationships between parameters are established using algorithms such as regression analysis and decision trees to achieve dynamic parameter adjustment.

[0014] The image processing-based cable defect detection system includes a multi-view acquisition module, an adaptive enhancement module, a feature fusion analysis module, a dynamic parameter adjustment module, and a result output early warning module. The multi-view acquisition module acquires images of the cable from multiple perspectives using different types of image acquisition devices and records the spatial position and orientation parameters of the acquisition devices. The adaptive enhancement module, connected to the multi-view acquisition module, performs adaptive image enhancement processing on the acquired images to highlight cable surface details and potential defect features. The feature fusion analysis module, connected to the adaptive enhancement module, extracts and fuses image features using a multi-scale convolutional neural network structure, and then uses a deep neural network model based on transfer learning to identify and classify cable defects. The dynamic parameter adjustment module, connected to the feature fusion analysis module and environmental and cable operation parameter acquisition devices, dynamically adjusts key parameters during defect detection based on real-time acquired parameters. The result output early warning module, connected to the feature fusion analysis module and the dynamic parameter adjustment module, outputs a cable defect report and issues an early warning signal when a serious defect is detected.

[0015] Preferably, the image acquisition device in the multi-view acquisition module is mounted on an adjustable robotic arm or track device, and the position and angle of the acquisition device can be precisely adjusted through remote control to obtain the best cable image acquisition perspective.

[0016] Preferably, the adaptive enhancement module, feature fusion analysis module, and dynamic parameter adjustment module all adopt a distributed computing architecture, utilizing the computing resources of the cloud computing platform to improve data processing speed and the system's parallel processing capability, thereby meeting the needs of real-time detection.

[0017] Compared with the prior art, the beneficial effects of the present invention are: This image processing-based cable defect detection method and system acquires comprehensive cable information through multi-view image acquisition, improves image quality by combining adaptive image enhancement, and accurately identifies defects by utilizing multi-scale feature fusion and deep transfer learning. Compared with existing technologies, it can more accurately detect various cable defects, including subtle and hidden defects, effectively improving detection accuracy. For example, in actual testing, the detection system of this invention can accurately identify and locate minute insulation layer damage defects that are difficult to detect using traditional methods. The system can dynamically adjust detection parameters according to changes in cable operating environment parameters and its own operating parameters, adapting to complex and ever-changing operating environments, such as scenarios with drastic changes in lighting, strong electromagnetic interference, and large differences in temperature and humidity. It maintains stable detection performance under different environmental conditions, overcoming the problem of significant decrease in detection accuracy in complex environments of existing technologies. The adjustable robotic arm or track device of the multi-view acquisition module can quickly acquire the best acquisition angle. The distributed computing architecture improves data processing speed and parallel processing capabilities, making the entire detection process efficient and fast. It can complete the detection of a large number of cables in a short time, meeting the power system's requirements for high-efficiency cable detection. It can not only accurately detect the location of cable defects, but also perform detailed analysis of defect types and severity, providing comprehensive and accurate information for cable maintenance and repair, helping to formulate reasonable maintenance strategies and improve the pertinence and effectiveness of cable maintenance. Attached Figure Description

[0018] Figure 1 This is a flowchart of the cable defect detection method of the present invention; Figure 2 This is a system architecture diagram for multi-view image acquisition according to the present invention; Figure 3 This is a system architecture diagram for adaptive image enhancement according to the present invention; Figure 4 This is a system architecture diagram of the multi-scale feature fusion and deep transfer learning of the present invention; Figure 5 This is a system architecture diagram for the dynamic parameter adjustment of this invention; Figure 6 This is a system architecture diagram for the output and early warning of the results of this invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figure 1-6This invention provides a technical solution: a cable defect detection method based on image processing, comprising the following steps: S1, Multi-view image acquisition: Multiple image acquisition devices are used to simultaneously acquire images of the cable from different perspectives, obtaining multiple sets of images containing information about the cable surface and different depth levels. At the same time, the spatial position and attitude parameters of each acquisition device are recorded to establish the correspondence between the images and the actual position of the cable.

[0021] S2, Adaptive Image Enhancement: For the image acquired in step S1, an appropriate enhancement algorithm is automatically selected based on the image's noise distribution, contrast, and brightness characteristics. For example, a method based on a combination of local histogram equalization and homomorphic filtering is used to adaptively enhance the image, highlighting cable surface details and potential defect features.

[0022] S3, Multi-scale Feature Fusion and Deep Transfer Learning: For the enhanced image, a multi-scale convolutional neural network structure is used to extract image features at different scales and perform feature fusion. The fused features are then input into a deep neural network model based on transfer learning. This model is pre-trained on a large number of general image datasets and fine-tuned for the cable defect detection task to achieve accurate identification and classification of cable defects.

[0023] S4, Dynamic Parameter Adjustment: Real-time acquisition of cable operating environment parameters, such as temperature, humidity, electromagnetic interference intensity, and cable's own operating parameters, such as current or voltage. Based on changes in these parameters, using a pre-established parameter correlation model, the threshold, feature weights, and other key parameters in the defect detection process are dynamically adjusted to adapt to the detection needs under different working conditions.

[0024] S5, Result Output and Early Warning: Based on the analysis results of steps S3 and S4, output a detailed cable defect report, including defect location, type and severity information. If a defect that seriously affects the safe operation of the cable is detected, an early warning signal will be issued immediately through various means, such as audible and visual alarms and SMS notifications.

[0025] In this invention: the image acquisition device in step S1 includes a visible light camera, an ultraviolet imager, and an ultrasonic imager, which are used to acquire visible light images of the cable surface, ultraviolet images generated by corona discharge, and ultrasonic images of the internal structure of the cable, respectively, so as to realize the acquisition of multi-dimensional information of the cable.

[0026] In this invention: In step S2, the adaptive image enhancement algorithm first evaluates the noise of the image when selecting the enhancement method. If the noise level is higher than the set threshold, homomorphic filtering with good denoising effect is preferred. If the image contrast is low, local histogram equalization is used to improve the contrast, so as to achieve targeted enhancement of images of different quality.

[0027] In this invention: the multi-scale convolutional neural network structure in step S3 adopts a pyramid-shaped convolutional layer design, gradually expanding the receptive field of the convolutional kernel from shallow to deep layers to obtain image features at different scales. Then, through feature splicing and fusion operations, the multi-scale features are input into a deep neural network model based on transfer learning. This model is an improved Faster R-CNN network that introduces an attention mechanism module to enhance attention to defect features.

[0028] In this invention: the parameter association model in step S4 is obtained by machine learning training on a large amount of historical environmental parameters, cable operation parameters and defect detection results data. The nonlinear mapping relationship between parameters is established by using algorithms such as regression analysis and decision trees to achieve dynamic parameter adjustment.

[0029] The image processing-based cable defect detection system includes a multi-view acquisition module, an adaptive enhancement module, a feature fusion analysis module, a dynamic parameter adjustment module, and a result output early warning module. The multi-view acquisition module is used to acquire images of the cable from multiple perspectives using different types of image acquisition devices, and records the spatial position and attitude parameters of the acquisition devices. In practical applications, the installation of the multi-view acquisition module needs to be rationally laid out according to the cable laying environment. For example, in urban underground cable tunnels, visible light cameras, ultraviolet imagers, and ultrasonic imagers are mounted on a track-movable device, with the track laid along the cable route. The moving speed of the track device and the angle of the imaging devices are remotely controlled to ensure comprehensive and complete acquisition of cable images. The data acquisition and control unit uses a high-performance industrial computer, equipped with a dedicated data acquisition card and control software, to achieve precise control of the imaging devices and rapid acquisition and storage of image data. During the acquisition process, the imaging devices are calibrated periodically to ensure the accuracy and consistency of the acquired images. For example, the visible light camera undergoes focal length and white balance calibration weekly, while the ultraviolet and ultrasonic imagers undergo sensitivity calibration monthly. The adaptive enhancement module, connected to the multi-view acquisition module, performs adaptive image enhancement processing on the acquired images, highlighting cable surface details and potential defect features. The adaptive image enhancement module adopts a modular design approach in its software implementation. First, a noise assessment submodule is developed, utilizing classic noise assessment algorithms, such as wavelet transform-based noise estimation methods, to evaluate the noise level of the acquired images. When the noise level exceeds a set threshold, the homomorphic filtering submodule is invoked for denoising and enhancement processing; if the image contrast is low, the local histogram equalization submodule is activated. In practical applications, based on the characteristics of different types of cable images, the parameters of homomorphic filtering and local histogram equalization are optimized through extensive experiments to achieve the best enhancement effect. For example, for cable images with complex surface textures, appropriately adjusting the window size and contrast enhancement factor of local histogram equalization makes texture details clearer while avoiding over-enhancement that could lead to image distortion. The feature fusion analysis module is connected to the adaptive enhancement module, using a multi-scale convolutional neural network structure to extract and fuse image features. Then, a deep neural network model based on transfer learning is used to identify and classify cable defects. The dynamic parameter adjustment module is connected to the feature fusion analysis module and environmental and cable operation parameter acquisition equipment. Based on the real-time acquired parameters, it dynamically adjusts key parameters in the defect detection process. The multi-scale convolutional neural network structure and the deep neural network model based on transfer learning are built and trained within a deep learning framework, such as TensorFlow or PyTorch. During training, the model is first pre-trained using a large number of general image datasets, and then fine-tuned on a cable defect detection dataset.For multi-scale convolutional neural network structures, parameters such as the number of convolutional layers, kernel size, and stride were optimized experimentally to achieve the best multi-scale feature extraction results. For example, in a defect detection task targeting a specific type of cable, after multiple experiments, a 5-layer pyramidal convolutional layer design was determined, with a shallow kernel size of 3×3 and a gradually increasing kernel size to 7×7 in deeper layers. The stride was adjusted between 1 and 3 according to the needs of different layers. For the improved Faster R-CNN network, when introducing the attention mechanism module, the weight parameters of the attention mechanism were adjusted experimentally to enable the model to pay more attention to defect features and improve the accuracy of defect identification. The parameter association model in the dynamic parameter adjustment module was trained by collecting long-term cable operating environment parameters, cable's own operating parameters, and corresponding defect detection results data. Data collection adopted a distributed sensor network to ensure the accuracy and real-time nature of parameter acquisition. During model training, various machine learning algorithms were used for comparative experiments, and the algorithm with the best performance was selected to construct the parameter association model. For example, when inspecting cables in a specific area, by comparing the performance of algorithms such as regression analysis, decision trees, and random forests in processing cable data for that area, the random forest algorithm was ultimately chosen to build the parameter association model because it has higher accuracy and stability on this dataset. After the model is trained, it is deployed in the dynamic parameter adjustment module, which adjusts key parameters such as defect detection thresholds and feature weights in real time based on the collected parameters. The result output early warning module is connected to the feature fusion analysis module and the dynamic parameter adjustment module to output cable defect reports and issue early warning signals when serious defects are detected. The result output early warning module is implemented through a specially developed software interface. The software interface displays the cable defect report in an intuitive way, including the defect location marked on the cable line diagram, a textual description of the defect type, and a severity level classification. For audible and visual alarms, specialized alarm devices, such as high-brightness LED warning lights and high-decibel sirens, are connected. When a serious defect is detected, the software controls the alarm devices to emit audible and visual signals. The SMS notification function is implemented through an SMS gateway, sending defect information to the mobile phones of relevant maintenance personnel in a preset format. Meanwhile, the result output early warning module also has data storage and query functions, which can store historical test reports and early warning information in the database for easy subsequent query and analysis, providing data support for the long-term maintenance and management of cables.

[0030] In this invention, the image acquisition device in the multi-view acquisition module is installed on an adjustable robotic arm or track device, and the position and angle of the acquisition device can be precisely adjusted through remote control to obtain the best cable image acquisition perspective.

[0031] In this invention, the adaptive enhancement module, feature fusion analysis module, and dynamic parameter adjustment module all adopt a distributed computing architecture, utilizing the computing resources of the cloud computing platform to improve data processing speed and the system's parallel processing capabilities, thus meeting the needs of real-time detection.

[0032] Working Principle: The multi-view acquisition module serves as the information entry point for the entire system. Its onboard visible light camera, ultraviolet imager, and ultrasonic imager work simultaneously from different perspectives. The visible light camera captures images of the physical state of the cable surface, the ultraviolet imager captures ultraviolet images generated by corona discharge to detect early signs of insulation degradation, and the ultrasonic imager acquires images of the cable's internal structure to detect internal defects. Simultaneously, the system records the spatial position and attitude parameters of each device, providing a spatial coordinate reference for subsequent defect localization. The acquisition devices, mounted on an adjustable robotic arm or track device, can be remotely controlled to adjust their position and angle, ensuring comprehensive and high-quality multi-dimensional image data. The acquired image data is then transmitted to an adaptive augmentation system. After the enhancement module, the image is first evaluated for noise, contrast, and brightness. If the noise level is higher than a set threshold, homomorphic filtering is used to suppress noise and enhance edge details. If the contrast is low, local histogram equalization is used to improve the contrast of local details. This adaptive processing highlights the subtle defect features on the cable surface, laying a good foundation for subsequent feature extraction. The enhanced image then enters the feature fusion analysis module. The multi-scale convolutional neural network adopts a pyramid-shaped convolutional layer design. Shallow small convolutional kernels capture subtle textures and small defect edges, while deep large convolutional kernels acquire large-scale features such as large defect contours. Multi-scale information is then fused through feature concatenation. The fused feature input is based on an improved Faster algorithm using transfer learning. The R-CNN network, leveraging pre-trained general feature recognition capabilities and fine-tuned with a cable defect dataset, can accurately identify and classify different types of cable defects. Its attention mechanism module enhances the focus on defect features, further improving recognition accuracy. The dynamic parameter adjustment module receives real-time environmental parameters such as temperature and humidity from environmental parameter acquisition devices, as well as cable operating parameters such as current and voltage. Based on a parameter association model pre-trained with extensive historical data, this model uses regression analysis and decision trees to establish a non-linear mapping between parameters and detection parameters. When real-time parameters change, the defect detection threshold and feature weights are dynamically adjusted to maintain stable detection accuracy under different operating conditions. Finally, the result output warning module integrates the defect identification results from the feature fusion analysis module and the optimized parameters from the dynamic parameter adjustment module to generate a detailed report including defect location, type, and severity. If a severe defect is detected, an immediate warning is issued via audible and visual alarms and SMS notifications. Simultaneously, relevant information is stored for subsequent query and analysis, achieving efficient and accurate detection and management of cable defects throughout the entire process.

[0033] In summary, this image processing-based cable defect detection method and system acquires comprehensive cable information through multi-view image acquisition, improves image quality by combining adaptive image enhancement, and accurately identifies defects by utilizing multi-scale feature fusion and deep transfer learning. Compared with existing technologies, it can more accurately detect various cable defects, including subtle and hidden defects, effectively improving detection accuracy. For example, in actual testing, the detection system of this invention can accurately identify and locate minute insulation layer damage defects that are difficult to detect using traditional methods. The system can dynamically adjust detection parameters according to changes in cable operating environment parameters and its own operating parameters, adapting to complex and ever-changing operating environments, such as scenarios with drastic changes in lighting, strong electromagnetic interference, and large differences in temperature and humidity. It maintains stable detection performance under different environmental conditions, overcoming the problem of significant decrease in detection accuracy in complex environments of existing technologies. The adjustable robotic arm or track device of the multi-view acquisition module can quickly acquire the best acquisition angle. The distributed computing architecture improves data processing speed and parallel processing capabilities, making the entire detection process efficient and fast. It can complete the detection of a large number of cables in a short time, meeting the power system's requirements for high-efficiency cable detection. It can not only accurately detect the location of cable defects, but also perform detailed analysis of defect types and severity, providing comprehensive and accurate information for cable maintenance and repair, helping to formulate reasonable maintenance strategies and improve the pertinence and effectiveness of cable maintenance.

[0034] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0035] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A cable defect detection method based on image processing, characterized in that: Includes the following steps: S1, Multi-view image acquisition: Multiple image acquisition devices are used to simultaneously acquire images of the cable from different perspectives, obtaining multiple sets of images containing information about the cable surface and different depth layers. At the same time, the spatial position and attitude parameters of each acquisition device are recorded to establish the correspondence between the images and the actual position of the cable. S2, Adaptive Image Enhancement: For the image acquired in step S1, based on the image's noise distribution, contrast, and brightness characteristics, an appropriate enhancement algorithm is automatically selected, such as a method based on a combination of local histogram equalization and homomorphic filtering, to adaptively enhance the image and highlight cable surface details and potential defect features. S3, Multi-scale Feature Fusion and Deep Transfer Learning: For the enhanced image, a multi-scale convolutional neural network structure is used to extract image features at different scales and perform feature fusion. The fused features are then input into a deep neural network model based on transfer learning. This model is pre-trained on a large number of general image datasets and fine-tuned for the cable defect detection task to achieve accurate identification and classification of cable defects. S4, Dynamic Parameter Adjustment: Real-time acquisition of cable operating environment parameters, such as temperature, humidity, electromagnetic interference intensity, and cable's own operating parameters, such as current or voltage. Based on the changes in these parameters, using a pre-established parameter correlation model, the threshold, feature weights, and other key parameters in the defect detection process are dynamically adjusted to adapt to the detection needs under different working conditions. S5, Result Output and Early Warning: Based on the analysis results of steps S3 and S4, output a detailed cable defect report, including defect location, type and severity information. If a defect that seriously affects the safe operation of the cable is detected, an early warning signal will be issued immediately through various means, such as audible and visual alarms and SMS notifications.

2. The cable defect detection method based on image processing according to claim 1, characterized in that, In step S1, the image acquisition equipment includes a visible light camera, an ultraviolet imager, and an ultrasonic imager, which are used to acquire visible light images of the cable surface, ultraviolet images generated by corona discharge, and ultrasonic images of the internal structure of the cable, respectively, so as to realize the acquisition of multi-dimensional information of the cable.

3. The cable defect detection method based on image processing according to claim 1, characterized in that, In step S2, when selecting an enhancement method, the adaptive image enhancement algorithm first evaluates the noise of the image. If the noise level is higher than a set threshold, homomorphic filtering with good denoising effect is preferred. If the image contrast is low, local histogram equalization is used to improve the contrast, so as to achieve targeted enhancement of images of different quality.

4. The cable defect detection method based on image processing according to claim 1, characterized in that, In step S3, the multi-scale convolutional neural network structure adopts a pyramid-shaped convolutional layer design, gradually expanding the receptive field of the convolutional kernel from shallow to deep layers to obtain image features at different scales. Then, through feature splicing and fusion operations, the multi-scale features are input into a deep neural network model based on transfer learning. This model is an improved Faster R-CNN network that introduces an attention mechanism module to enhance attention to defect features.

5. The cable defect detection method based on image processing according to claim 1, characterized in that, In step S4, the parameter association model is obtained by machine learning training on a large amount of historical environmental parameters, cable operation parameters and defect detection results data. Regression analysis, decision tree and other algorithms are used to establish nonlinear mapping relationships between parameters in order to achieve dynamic parameter adjustment.

6. A cable defect detection system based on image processing, used to implement the detection method according to any one of claims 1-5, characterized in that, The system includes a multi-view acquisition module, an adaptive enhancement module, a feature fusion analysis module, a dynamic parameter adjustment module, and a result output early warning module. The multi-view acquisition module acquires images of the cable from multiple perspectives using different types of image acquisition devices and records the spatial position and attitude parameters of the acquisition devices. The adaptive enhancement module, connected to the multi-view acquisition module, performs adaptive image enhancement processing on the acquired images to highlight cable surface details and potential defect features. The feature fusion analysis module, connected to the adaptive enhancement module, extracts and fuses image features using a multi-scale convolutional neural network structure, and then uses a deep neural network model based on transfer learning to identify and classify cable defects. The dynamic parameter adjustment module, connected to the feature fusion analysis module and environmental and cable operation parameter acquisition devices, dynamically adjusts key parameters in the defect detection process based on real-time acquired parameters. The result output early warning module, connected to the feature fusion analysis module and the dynamic parameter adjustment module, outputs a cable defect report and issues an early warning signal when a serious defect is detected.

7. The cable defect detection system based on image processing according to claim 6, characterized in that, The image acquisition device in the multi-view acquisition module is installed on an adjustable robotic arm or track device. The position and angle of the acquisition device can be precisely adjusted through remote control to obtain the best cable image acquisition perspective.

8. The cable defect detection system based on image processing according to claim 6, characterized in that, The adaptive enhancement module, feature fusion analysis module, and dynamic parameter adjustment module all adopt a distributed computing architecture, utilizing the computing resources of the cloud computing platform to improve data processing speed and the system's parallel processing capabilities, thus meeting the needs of real-time detection.

Images