Submarine cable detection method and system based on underwater robot
By combining underwater imaging physical models and deep learning technology, and using image enhancement functions and deblurring convolutional neural network models, the problem of blurred submarine cable images in turbulent and silty waters was solved, achieving high-precision submarine cable detection.
Patent Information
- Application Number
- CN202510826998.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-19
- Publication Date
- 2025-09-26
AI Technical Summary
In turbulent and turbid waters, the images of submarine cables captured by underwater robot cameras are insufficiently bright, have color shifts, blurred edges, and low contrast. Existing image enhancement technologies are not designed for this purpose, resulting in unrecognizable targets.
An image enhancement function based on the underwater imaging physical degradation model and transmittance estimation is adopted, combined with an image deblurring enhanced convolutional neural network model. The cable target is restored through image correction and texture refinement, and the residual structure and channel attention mechanism are used to improve image clarity and recognition accuracy.
It significantly improves the image clarity and recognition accuracy of the submarine cable target area, reduces the false detection rate, and improves the detection stability and robustness of underwater robots in complex waters.
Smart Images
Figure CN120707527A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of submarine cable detection, and in particular to a submarine cable detection method and system based on an underwater robot. Background Art
[0002] In fields such as marine engineering, power communications, and energy transmission, submarine cables serve as critical information and energy transmission vehicles, and their operational status directly impacts the stable operation of offshore platforms, submarine base stations, and land-based nodes. To ensure the long-term safe operation of submarine cables after installation, regular testing and assessment of their physical condition, installation orientation, and environmental adaptability are necessary. With the development of intelligent underwater equipment, the use of underwater robots (including remotely operated vehicles (ROVs) and autonomous underwater vehicles (AUVs) equipped with cameras to perform visual inspection of submarine cables has become a mainstream technology. This method uses cameras to capture cable images and combines manual or AI algorithms to identify cable location and status, enabling remote, unmanned submarine inspections.
[0003] In turbid water environments with high turbulence and severe sediment suspension, such as port areas, estuary intersections and offshore sediment-rich areas, a large number of suspended particles in the seawater cause severe scattering and absorption of visible light, resulting in the original images collected by underwater robot cameras generally having problems such as insufficient brightness, color shift, blurred edges and low contrast. In addition, existing image enhancement technologies are mostly general processing models that fail to combine the physical characteristics of underwater imaging and the characteristics of cable structures for targeted design, resulting in insufficient image restoration effects and unrecognizable targets. Summary of the Invention
[0004] The purpose of the present invention is to overcome the shortcomings of the above-mentioned prior art and provide a submarine cable detection method and system based on an underwater robot to solve the problem in the prior art that a large number of suspended particles in seawater produce severe scattering and absorption of visible light, resulting in the original images collected by the underwater robot camera generally having insufficient brightness, color shift, blurred edges and low contrast.
[0005] In order to achieve the above object, the present invention adopts the following technical solutions: A submarine cable detection method based on an underwater robot comprises the following steps: Collect original turbid image frames in time sequence to obtain an original turbid image sequence; Solve the transmittance and ambient light parameters of each original turbid image frame in the original turbid image sequence; based on the transmittance and ambient light parameters, correct the original turbid image frame through the image enhancement function to obtain an enhanced image frame; The enhanced image frame is subjected to texture refinement and detail restoration by an image deblurring and enhancing convolutional neural network model to obtain a sequence of clear image frames; the image deblurring and enhancing convolutional neural network model includes an encoder, a decoder, and a residual enhancement module; The cable position, cable direction and cable edge contour are extracted from the clear image frame sequence, and the obtained cable target recognition results are output.
[0006] A further improvement of the present invention is: Preferably, the transmittance is calculated by a dark channel priori algorithm, and the specific process is: calculating the local scattering degree of each original turbid image channel, and obtaining the transmittance of the original turbid image according to the local scattering degree.
[0007] Preferably, the ambient light parameters are calculated by extracting the maximum brightness region of the image globally.
[0008] Preferably, the process of correcting the original turbid image frame by the image enhancement function is: performing descattering restoration processing on the original turbid image frame by the image enhancement function to obtain an enhanced image frame; the image enhancement function is obtained by an underwater imaging physical degradation model.
[0009] Preferably, the imaging physical degradation model is: ; in, is a pixel; is the pixel point in the turbid image frame observed from the camera perspective The actual color intensity; It is the direct radiation component of the target real scene under ideal conditions of no scattering and no absorption; is the transmittance of the original turbid image sequence; is the ambient light parameter of the original turbid image sequence.
[0010] Preferably, in the image deblurring enhancement convolutional neural network model, the low-level features of the enhanced image frame are obtained through the encoder, and the image resolution is restored through the decoder to generate an output clear image frame.
[0011] Preferably, in the process of restoring the image resolution and generating the output clear image frame, an intermediate residual enhancement module is used to perform deep compensation and blur removal on the image details.
[0012] Preferably, during the processing of the residual enhancement module, a channel attention module is added after the output of each residual enhancement module to perform weighted adjustment on the feature responses of different channels.
[0013] Preferably, based on the cable target recognition result, the underwater robot control system updates the robot route in real time.
[0014] A submarine cable detection system based on an underwater robot, comprising: An acquisition unit, configured to acquire original turbid image frames in chronological order to obtain an original turbid image sequence; An enhancement unit is used to solve the transmittance and ambient light parameters of each original turbid image frame in the original turbid image sequence; based on the transmittance and ambient light parameters, the original turbid image frame is corrected by an image enhancement function to obtain an enhanced image frame; A deblurring unit is configured to perform texture refinement and detail restoration on the enhanced image frame using an image deblurring enhancement convolutional neural network model to obtain a sequence of clear image frames; the image deblurring enhancement convolutional neural network model includes an encoder, a decoder, and a residual enhancement module; The output unit is used to extract the cable position, cable direction and cable edge contour from the clear image frame sequence and output the obtained cable target recognition result.
[0015] Compared with the prior art, the present invention has the following beneficial effects: This invention discloses a submarine cable detection method based on an underwater robot. The method first acquires raw turbid image frames through camera capture, obtains a time-dependent raw turbid image sequence, and obtains the transmittance and ambient light parameters of the raw turbid image sequence. The raw turbid image frames are then corrected using an image enhancement function to obtain enhanced image frames. Based on the enhanced image frames, a clear image frame sequence is obtained, and cable identification results are obtained from the clear image frame sequence. In this invention, an image enhancement function, based on a combination of an underwater imaging physical degradation model and transmittance estimation, eliminates image blur and color distortion caused by suspended particles and light scattering, significantly improving image clarity in the target cable area. An image deblurring enhancement convolutional neural network model with a residual structure and channel attention mechanism is introduced to restore image details, improving the accuracy and stability of submarine cable detection by the underwater robot. This invention organically integrates physical models with deep learning, first eliminating environmental interference through physical model correction, and then using a deep learning model to enhance structural features, forming a cascaded processing chain from "environmental adaptation to detail enhancement." Compared with the single-stage general model, a complete technical closed loop from raw data correction to structured target recognition is constructed. This architecture shows stronger robustness in waters with dynamically changing sediment concentrations (such as estuary confluence areas), effectively reducing the false detection rate. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] Figure 1 This is a flow chart of the submarine cable detection method based on underwater robot proposed in the present invention.
[0017] Figure 2This is a flow chart of the method for solving the transmittance and ambient light parameters of the original turbid image sequence proposed by the present invention.
[0018] Figure 3 This is a block diagram of the image deblurring and enhancement convolutional neural network model proposed in this invention. DETAILED DESCRIPTION
[0019] Hereinafter, the terms "first," "second," "third," and "fourth" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the quantity of the technical features indicated. Thus, a feature identified as "first," "second," "third," or "fourth" may explicitly or implicitly include one or more of such features.
[0020] The co-shooting method provided in the embodiments of the present application can be applied to terminal devices such as mobile phones, tablet computers, wearable devices, vehicle-mounted devices, augmented reality (AR) / virtual reality (VR) devices, laptop computers, ultra-mobile personal computers (UMPCs), netbooks, and personal digital assistants (PDAs). The embodiments of the present application do not impose any restrictions on the specific types of terminal devices.
[0021] It should be noted that the terms "first," "second," and the like in the description and drawings of the present invention are used to distinguish similar objects and are not necessarily used to describe a specific order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having," as well as any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus comprising a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to these processes, methods, products, or apparatus.
[0022] like Figure 1 As shown, the first aspect of the present invention discloses a submarine cable detection method based on an underwater robot, and its specific implementation steps are as follows: S1, using an underwater robot to collect original turbid image frames in an underwater environment on the seabed along a preset robot route to obtain an original turbid image sequence; In this embodiment S1, the underwater robot uses a camera to collect original turbid image frames in the seabed underwater environment through a preset trajectory; Among them, the cameras include a forward-looking high-definition low-light camera and a downward-looking high-definition low-light camera; the forward-looking high-definition low-light camera is used to shoot submarine cables in the front field of view, and the downward-looking high-definition low-light camera is used to shoot submarine cables from a vertical direction.
[0023] In this embodiment, the original turbid image sequence is composed of original turbid image frames that are sequentially numbered in a timestamp manner and stored in a storage space of the underwater robot; The collected image frames are acquired at fixed time intervals, and timestamp information is attached to each frame to identify the order of their acquisition time. The original turbidity image sequence is an image frame sequence composed of a batch of original turbidity image frames sequentially numbered and sorted by timestamps, and stored in real time in the local storage space of the underwater robot body, providing a continuous and time-consistent data basis for subsequent image enhancement and cable identification processing.
[0024] S2, solving the transmittance and ambient light parameters of the original turbid image sequence, and constructing an image enhancement function to correct the original turbid image frame to obtain an enhanced image frame; See also Figure 2 In this embodiment S2, the method for solving the transmittance and ambient light parameters of the original turbid image sequence is as follows: S201, calculating the transmittance of the original turbid image sequence based on a dark channel prior algorithm; during the calculation process of the dark channel prior algorithm, the local scattering degree is approximated by calculating the local minimum brightness and the minimum value within the minimum filtering area in each image channel, and the transmittance of the image is calculated based on the local scattering degree; specifically, the estimated local scattering degree (i.e., the local dark channel value) is combined with an estimate of the global atmospheric light A to directly calculate an estimated value of the transmittance.
[0025] S202, using the image global brightness maximum area extraction method to calculate the ambient light parameters of the original turbid image sequence, select the brightest area from the original turbid image frame and use it as the area with the strongest light reflection to approximate the ambient light background; it should be understood that the light reflection area here is to set a threshold, and the pixels in the area whose light reflection is greater than the threshold belong to the area. Exemplarily, it can be whether the threshold is 95% of the image brightness distribution.
[0026] In this embodiment S2, the original turbid image frame is corrected using an image enhancement function. The specific method is as follows: the transmittance and ambient light parameters of the original turbid image sequence are input, and the image enhancement function performs descattering restoration processing on each original turbid image frame in a pixel-level manner to generate an enhanced image frame; The image enhancement function is constructed based on the underwater imaging physical degradation model. The underwater imaging physical degradation model is used to separate the direct radiation component of the target scene from the original turbid image frame as the enhanced image frame. In this process, the original image is corrected based on the underwater optical imaging physical model by solving the transmittance and ambient light parameters, effectively compensating for the light attenuation and scattering effects caused by suspended particles.
[0027] The underwater imaging physical degradation model is a mathematical model used to describe the image degradation process caused by absorption and scattering of light during propagation in water. It is built based on the Jaffe-McGlamery model. The Jaffe-McGlamery model decomposes the color intensity of each pixel in the original turbid image frame into two parts: ; in, is a pixel; is the pixel point in the turbid image frame observed from the camera perspective The actual color intensity; It is the direct radiation component of the target real scene under ideal conditions of no scattering and no absorption; is the transmittance of the original turbid image sequence, which indicates the light transmission rate of the water body between the camera and the scene, and reflects the attenuation degree of each pixel in the image; is the ambient light parameter of the original turbid image sequence, which is used to describe the background scattering brightness caused by suspended particles; In turbid water, the optical signal from the submarine cable is partially retained as The remaining part is reflected by the suspended particles in the water and forms the ambient light interference item. .
[0028] S3, using the image deblurring enhancement convolutional neural network model to perform image texture refinement and detail restoration operations on all enhanced image frames to obtain a clear image frame sequence; In this embodiment S3, the image deblurring and enhancement convolutional neural network model is a multi-layer neural network model based on a residual structure and a channel attention mechanism. The input of the image deblurring and enhancement convolutional neural network model is an enhanced image frame, and the output is a clear image frame, as follows: The image deblurring and enhancement convolutional neural network model contains an encoder-decoder structure and nested multiple residual enhancement modules. In the residual enhancement module of the image deblurring and enhancement convolutional neural network model, multiple convolutional layers are combined to implement feature residual learning to retain local detail information of the enhanced image frame. The residual enhancement module is connected to the channel attention module to calculate the feature response weights of each channel of the image. The network output generates a clear image frame with the same size as the enhanced image frame at the pixel level.
[0029] Among them, the image deblurring enhancement convolutional neural network model is configured to run in the computing platform of the underwater robot.
[0030] See also Figure 3 In this embodiment, the image deblurring and enhancement convolutional neural network model adopts a symmetrical structure of Encoder-Residual-Decoder, including an image encoder, multiple residual enhancement modules and an image decoder, wherein: the image encoder is used to extract low-level features of the input enhanced image frame through gradual downsampling (such as pooling or strided convolution); the decoder is used to restore the image resolution through upsampling (such as transposed convolution or interpolation) and generate an output clear image frame; the intermediate residual enhancement module is used to achieve deep compensation of image details and blur removal.
[0031] The residual architecture is designed as follows: Each residual enhancement module consists of two or more convolutional layers, connected by an identity mapping. This residual architecture is nested at the end of the encoder and before each upsampling stage of the decoder, forming a cascade of encoder → residual enhancement → decoder.
[0032] Furthermore, the channel attention mechanism is designed based on the SE-Block structure, as follows: after the output of each residual enhancement module, a channel attention module is added to perform weighted adjustment on the feature responses of different channels; by learning the weights of each channel (such as global average pooling + fully connected layer), channels important for the deblurring task (such as edges, textures) are enhanced, noise or irrelevant features are suppressed, and the discriminability of feature expression is improved.
[0033] The image deblurring and enhancement convolutional neural network model ultimately outputs a clear image frame with the same size as the input image, and the output result is jointly optimized through structural similarity loss, mean square error and edge preservation loss.
[0034] The training method of the image deblurring enhancement convolutional neural network model is as follows: The training is conducted in a supervised learning manner, using a paired training set consisting of simulated turbid images and corresponding clear images; In this embodiment, the simulated turbidity image is obtained by superimposing a scattering noise model and a color degradation function on a real clear image to simulate underwater turbidity conditions; The loss function of the image deblurring enhancement convolutional neural network includes structural similarity loss, peak signal-to-noise ratio loss and edge preservation loss terms.
[0035] S4. Input the clear image frame sequence into the cable recognition model to extract the cable position, cable direction and cable edge contour in the clear image frame sequence, output the cable target recognition result, and feed back the cable target recognition result to the underwater robot control system to update the robot route in real time.
[0036] In this embodiment S4, the cable recognition model is a target detection model based on deep learning (such as YOLOv7, MaskR-CNN), and a clear image frame sequence is input into the cable recognition model to obtain a cable target recognition result including the cable position, cable direction and cable edge contour.
[0037] In this embodiment S4, the cable target recognition result is fed back to the underwater robot control system to update the robot route in real time, as follows: The identified cable centerline trajectory is converted into a navigation path in the local coordinate system; the offset and heading difference between the robot's current position and the target trajectory are calculated based on the navigation path; the path tracking control algorithm is called to generate robot posture correction instructions and the robot route is updated in real time.
[0038] A second aspect of the present invention discloses a submarine cable detection system based on an underwater robot, comprising: An acquisition unit, configured to acquire original turbid image frames in chronological order to obtain an original turbid image sequence; An enhancement unit is used to solve the transmittance and ambient light parameters of each original turbid image frame in the original turbid image sequence; based on the transmittance and ambient light parameters, the original turbid image frame is corrected by an image enhancement function to obtain an enhanced image frame; A deblurring unit is configured to perform texture refinement and detail restoration on the enhanced image frame using an image deblurring enhancement convolutional neural network model to obtain a sequence of clear image frames; the image deblurring enhancement convolutional neural network model includes an encoder, a decoder, and a residual enhancement module; The output unit is used to extract the cable position, cable direction and cable edge contour from the clear image frame sequence and output the obtained cable target recognition result.
[0039] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0040] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.
[0041] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.
[0042] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the scope of protection of the claims of the present invention.
[0043] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.
Claims
1. A submarine cable detection method based on an underwater robot, characterized in that: The following steps are involved: Collect original turbid image frames in time sequence to obtain an original turbid image sequence; Solve the transmittance and ambient light parameters of each original turbid image frame in the original turbid image sequence; based on the transmittance and ambient light parameters, correct the original turbid image frame through the image enhancement function to obtain an enhanced image frame; The enhanced image frame is subjected to texture refinement and detail restoration by an image deblurring and enhancing convolutional neural network model to obtain a sequence of clear image frames; the image deblurring and enhancing convolutional neural network model includes an encoder, a decoder, and a residual enhancement module; The cable position, cable direction and cable edge contour are extracted from the clear image frame sequence, and the obtained cable target recognition results are output.
2. The submarine cable detection method based on an underwater robot according to claim 1, characterized in that: The transmittance is obtained by calculating the dark channel priori algorithm. The specific process is: calculating the local scattering degree of each original turbid image channel, and obtaining the transmittance of the original turbid image according to the local scattering degree.
3. The submarine cable detection method based on an underwater robot according to claim 1, characterized in that: The ambient light parameters are calculated by using an image global brightness maximum area extraction method.
4. The submarine cable detection method based on an underwater robot according to claim 1, characterized in that: The process of correcting the original turbid image frame by using the image enhancement function is as follows: performing descattering restoration processing on the original turbid image frame by using the image enhancement function to obtain an enhanced image frame; the image enhancement function is obtained by using an underwater imaging physical degradation model.
5. The submarine cable detection method based on an underwater robot according to claim 4, characterized in that: The imaging physical degradation model is: ； in, is a pixel; is the pixel point in the turbid image frame observed from the camera perspective The actual color intensity; It is the direct radiation component of the target real scene under ideal conditions of no scattering and no absorption; is the transmittance of the original turbid image sequence; is the ambient light parameter of the original turbid image sequence.
6. The submarine cable detection method based on an underwater robot according to claim 1, characterized in that: In the image deblurring and enhancement convolutional neural network model, the encoder obtains low-level features of the enhanced image frame, and the decoder restores the image resolution and generates an output clear image frame.
7. The submarine cable detection method based on an underwater robot according to claim 6, characterized in that: In the process of restoring the image resolution and generating an output clear image frame, the intermediate residual enhancement module is used to perform deep compensation and blur removal on the image details.
8. The submarine cable detection method based on an underwater robot according to claim 7, characterized in that: During the processing of the residual enhancement module, after the output of each residual enhancement module, a channel attention module is added to perform weighted adjustment on the feature responses of different channels.
9. The submarine cable detection method based on an underwater robot according to claim 1, characterized in that: Based on the cable target recognition results, the underwater robot control system updates the robot route in real time.
10. A submarine cable detection system based on an underwater robot, characterized in that: include: An acquisition unit, configured to acquire original turbid image frames in chronological order to obtain an original turbid image sequence; an enhancement unit, for solving the transmittance and ambient light parameters of each original turbid image frame in the original turbid image sequence; Based on the transmittance and ambient light parameters, the original turbid image frame is corrected by the image enhancement function to obtain an enhanced image frame; A deblurring unit is configured to perform texture refinement and detail restoration on the enhanced image frame using an image deblurring enhancement convolutional neural network model to obtain a sequence of clear image frames; the image deblurring enhancement convolutional neural network model includes an encoder, a decoder, and a residual enhancement module; The output unit is used to extract the cable position, cable direction and cable edge contour from the clear image frame sequence and output the obtained cable target recognition result.