Underground cable radar image detection method based on curve knowledge and deep learning
By extracting the edges and calculating the curvature of the B-scan radar image, and producing edge and curvature images as neural network input, the problems of complex underground media and insufficient data are solved, and the accuracy and efficiency of underground cable detection are improved.
Patent Information
- Application Number
- CN202210215384.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-03-07
- Publication Date
- 2025-10-17
- Estimated Expiration
- 2042-03-07
AI Technical Summary
In existing technologies, the complexity of underground media makes radar image detection difficult, and deep learning model training requires a large amount of labeled data, while data collection and labeling are time-consuming, making it difficult to efficiently perform underground cable detection.
By extracting the edges of the B-scan radar image and calculating the curvature of the edge points, edge images and curvature images are produced as the input and output data of the neural network. The model parameters are adjusted using the training data until the lowest loss function solution is obtained, thereby reducing the amount of training data required.
The accuracy of radar image detection is improved, the demand for training data is reduced, and efficient underground cable detection is achieved.
Smart Images
Figure CN114581677B_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the field of radar image detection, and particularly relates to a radar image detection method for underground cables based on curve knowledge and deep learning. BACKGROUND
[0002] B-scan radar image is the most common type of radar signal, and most engineering applications are based on the information provided by the radar image. Since the underground medium is not uniform, and various possible impurities exist, the useful information of the radar image is often hidden in the noisy image background, and even manual detection is quite difficult.
[0003] Since deep learning has very strong learning and expression ability, it can theoretically express any complex relationship, and therefore, the detection target recognition in the radar image often relies on deep learning methods. However, deep learning is a universal method, and a special deep learning model needs to be trained for a specific task. In order to train the model (i.e., to obtain the model parameters), a large amount of labeled training data is required, and the collection and labeling of such data is quite time-consuming, and in practical applications, the amount of data is often limited. SUMMARY
[0004] In view of the above problems, the present application is proposed in order to provide a radar image detection method for underground cables based on curve knowledge and deep learning, which overcomes the above problems or at least partially solves the above problems.
[0005] In order to solve the above technical problems, the embodiments of the present application disclose the following technical solutions:
[0006] A radar image detection method for underground cables based on curve knowledge and deep learning, comprising:
[0007] S100. Extracting the edge of the B-scan radar image to make an edge image;
[0008] S200. Calculating the curvature of each edge point of the edge image to make a curvature image;
[0009] S300. Making label data, taking the edge image, the curvature image and the label data as the input and output data pairs of a neural network, and randomly dividing all the data into training data and test data according to a certain probability;
[0010] S400. Adjusting the parameters in the network model by using the training data until the minimum solution in the sense of the minimum loss function is obtained;
[0011] S500. Substituting the test data set into the model obtained by training to verify the performance of the model obtained by training, and repeating steps S400 to S500 until a certain condition is met to stop training;
[0012] S600. Determine the model parameter with the highest detection accuracy on the test dataset as the final result.
[0013] Further, in S100, the method for extracting the edge of the B-scan radar image is one of the first-order derivative-based method, the second-order derivative-based method, or the model-based method.
[0014] Further, in S100, the edge threshold used to extract the edge of the B-scan radar image is lower than that of the ordinary edge detection algorithm; the image edge detection method includes but is not limited to Roberts, Prewitt, and Sobel operators.
[0015] Further, in S200, the curvature of each edge point is calculated, and a curvature image is made, and the specific method is as follows:
[0016] S201: Detect the edge points of the edge image;
[0017] S202: Then, the edge points are thinned;
[0018] S203: Connect the breakpoints in the edge image;
[0019] S204: Calculate the curvature of each edge point;
[0020] S205: And let the curvature at the non-edge point be 0.
[0021] Further, in S204, for the edge curve in the digital image, the points on the edge curve are sequentially numbered from small to large from one end to the other end, the i-th point coordinate (x i ,y i ), and the formula for calculating the curvature of each edge point is:
[0022] k = Δx i Δ 2 y i - Δy i Δ 2 x i (1)
[0023] wherein,
[0024]
[0025]
[0026]
[0027]
[0028] wherein, (x i-1 ,y i-1 ), (x iy i ) and (x i+1 ,y i+1 ) represent the i-1, i and i+1 point coordinates respectively.
[0029] Further, in S500, the training is stopped until a certain condition is reached, including: the number of training times reaches a given number or the error accuracy reaches a given accuracy.
[0030] The beneficial effects of the above technical solutions provided by the embodiments of the present application at least include:
[0031] The underground cable radar image detection method based on curve knowledge and deep learning disclosed in the present application comprises the following steps: S100, edges of a B-scan radar image are extracted to make an edge image; S200, curvatures of each edge point of the edge image are calculated to make a curvature image; S300, label data are made, and the edge image, the curvature image and the label data are used as input and output data pairs of a neural network, all data are randomly divided into training data and test data according to a certain probability; S400, parameters in the network model are adjusted by using the training data until a minimum solution in the sense of a loss function is obtained; S500, the test data set is substituted into the model obtained by training to verify the performance of the model obtained by training, steps S400 to S500 are repeated until the number of training times reaches a given number or the error accuracy reaches a given accuracy; and S600, the model parameter with the highest detection accuracy on the test data set is determined as the final result. The present application first extracts edges in an image, calculates curvatures of each edge point, and uses the curvatures as input data of a neural network, so that the neural network focuses on edge signals in the image, the amount of parameters to be obtained is reduced, and the demand for the amount of training data is reduced.
[0032] The technical solutions of the present application will be further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF DRAWINGS
[0033] The accompanying drawings are used to provide a further understanding of the present application, and constitute a part of the specification, and are used to explain the present application together with embodiments of the present application, and do not constitute a limitation on the present application. In the drawings:
[0034] Figure 1 It is a flowchart of the underground cable radar image detection method based on curve knowledge and deep learning in the embodiment 1 of the present application.
[0035] Figure 2 It is a curvature calculation schematic diagram in the embodiment 1 of the present application.
[0036] Figure 3 It is an underground cable image data diagram generated by gprMax in the embodiment 1 of the present application. DETAILED DESCRIPTION
[0037] Exemplary embodiments of the present disclosure will be described in greater detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it is understood that the present disclosure can be embodied in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the present disclosure to those skilled in the art.
[0038] To solve the problems in the prior art, an underground cable radar image detection method based on curve knowledge and deep learning is provided in the embodiments of the present application.
[0039] Embodiment 1
[0040] The present embodiment discloses an underground cable radar image detection method based on curve knowledge and deep learning, as shown in Figure 1 , comprising:
[0041] S100. Extracting the edge of the B-scan radar image to make an edge image; in the present embodiment, the method of extracting the edge can be a basic first-order derivative method, a second-order derivative based method, or a model based method, but it is different from a pure edge detection algorithm: the edge threshold is lower than that of the ordinary edge detection algorithm, because the purpose of the present algorithm is to detect edge information, rather than to determine the true image edge. There are many methods for image edge detection, and in the present embodiment, the method for image edge detection is not limited, including but not limited to Roberts, Prewitt, Sobel operator, etc.
[0042] S200. Calculating the curvature of each edge point of the edge image to make a curvature image; in the present embodiment, the curvature of each edge point is calculated to make a curvature image, and the specific method is as follows:
[0043] S201: detecting the edge points of the edge image;
[0044] S202: then refining the edge points;
[0045] S203: connecting the breakpoints in the edge image;
[0046] S204: calculating the curvature of each edge point;
[0047] S205: and making the curvature at the non-edge points 0.
[0048] Specifically, the curvature is defined as shown in Figure 2 The curvature of a point on a curve in a plane is the rate of change of the relative arc length s of the tangent line at the point within a very short range. The edge curvature is proportional to the second derivative along the tangent direction of the edge, and inversely proportional to the gradient amplitude, so the curvature can be expressed as the derivative of x(s), y(s)
[0049]
[0050] wherein are the first and second derivatives of x with respect to s respectively, are the first and second derivatives of y with respect to s respectively. Replacing derivatives with differences, the curvature calculation method of the edge points of the present patent is obtained.
[0051] In S204 of the present embodiment, for the edge curve in the digital image, the points on the edge curve are numbered from small to large from one end to the other end, the i-th point coordinate (x i ,y i ), the formula for calculating the curvature of each edge point is:
[0052] k = Δx i Δ 2 y i - Δy i Δ 2 x i
[0053] wherein,
[0054]
[0055]
[0056]
[0057]
[0058] wherein, (x i-1 ,y i-1 ), (x i ,y i ) and (x i+1 ,y i+1 ) represent the i-1, i and i+1 point coordinates respectively.
[0059] S300. Making label data, taking the edge image, the curvature image and the label data as the input and output data pair of the neural network, and randomly dividing all the data into training data and test data according to a certain probability;
[0060] S400. Adjusting the parameters in the network model by using the training data until the minimum solution in the sense of the minimum loss function is obtained;
[0061] S500. Substituting the test data set into the model obtained by training to verify the performance of the model obtained by training, repeating steps S400 to S500 until the training is stopped under certain conditions, including: the training times reach a given number of times or the error precision reaches a given precision.
[0062] S600. Determining the model parameter with the highest detection accuracy on the test data set as the final result.
[0063] To verify the effectiveness of the model, the present application performs an experiment on the underground cable image data generated based on gprMax (as shown in Figure 3 The results show that the accuracy is as high as 99.8%, which is 2.5 percentage points higher than the result of directly training the Resnet network.
[0064] The underground cable radar image detection method based on curve knowledge and deep learning disclosed in the embodiment comprises the following steps: S100. Extracting the edges of the B-scan radar image to make an edge image; S200. Calculating the curvatures of the edge points of the edge image to make a curvature image; S300. Making label data, taking the edge image, the curvature image and the label data as the input and output data pairs of a neural network, and randomly dividing all the data into training data and test data according to a certain probability; S400. Adjusting the parameters in the network model by using the training data until the minimum solution in the sense of the minimum loss function is obtained; S500. Substituting the test data set into the model obtained by training to verify the performance of the model obtained by training, repeating steps S400 to S500 until the number of training times reaches a given number or the error accuracy reaches a given accuracy; and S600. Determining the model parameter with the highest detection accuracy on the test data set as the final result. The present application first extracts the edges in the image, calculates the curvatures of the edge points, and takes the curvatures as the input data of the neural network, so that the neural network focuses on the edge signals in the image, reduces the amount of parameters to be obtained, and reduces the demand for the amount of training data.
[0065] It should be understood that the specific order or hierarchy of steps in the processes disclosed is an example of exemplary methods. Based upon design preferences, it is understood that the specific order or hierarchy of steps in the processes can be re-arranged while remaining within the scope of the present disclosure. The accompanying method claims present elements of the various steps in exemplary order and are not intended to be limited to the specific order or hierarchy presented.
[0066] In the detailed description above, various features are grouped together in single embodiments for the purpose of streamlining the disclosure. This disclosed approach is not to be interpreted as reflecting an intention that the embodiments of the claimed subject matter require more features than the claim requires. On the contrary, as reflected by the appended claims, the present application is directed to embodiments substantially stripped of all features except those set forth in the claims. Accordingly, the claims are hereby expressly incorporated into this detailed description, with each claim standing on its own as a separate embodiment of this application.
[0067] Those of skill would further appreciate that the various illustrative logical blocks, modules, circuits, and algorithm steps described in connection with the embodiments disclosed herein can be implemented as electronic hardware, computer software, or combinations of both. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the overall system. Skilled artisans can implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present disclosure.
[0068] The steps of a method or algorithm described in connection with the embodiments disclosed herein can be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. A software module can reside in RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor such that the processor can read information from, and write information to, the storage medium. In the alternative, the storage medium can be integral to the processor. The processor and the storage medium can reside in an ASIC. The ASIC can reside in a user terminal. In the alternative, the processor and the storage medium can reside as discrete components in a user terminal.
[0069] For a software implementation, the techniques described herein can be implemented with modules (e.g., procedures, functions, and so on) that perform the functions described herein. The software codes can be stored in memory units and executed by processors. The memory unit can be implemented within the processor or external to the processor, in which case it can be communicatively coupled to the processor via various means as is known in the art.
[0070] The above description includes one or more examples of the embodiments. Of course, not all possible combinations of components or method steps described above can be claimed as embodiments. One of ordinary skill in the art can recognize that modifications and variations of the described embodiments are possible and are within the scope of the present disclosure. It is therefore intended that the embodiments described herein be considered in all respects as only illustrative and not restrictive. Specifically, the description of the embodiments should be considered to be illustrative and not exhaustive, and should be considered in the light of the claims. Further, the use of the term "comprise" in the specification is to be construed in the same way as the term "comprise" as explained in the description of the claims. Also, the use of the term "or" in the claims is to be construed as "non-exclusive or" as explained in the description of the claims.
Claims
1. A method for detecting underground cables using radar images based on curve knowledge and deep learning, characterized in that: include: S100. Extract the edge of the B-scan radar image and create an edge image; S200. Calculate the curvature of each edge point of the edge image and create a curvature image; S300. Create label data, use the edge image, curvature image and label data as input and output data pairs of the neural network, and randomly divide all data into training data and test data according to a certain probability; S400. Using the training data to adjust the parameters in the network model until the lowest solution in the sense of minimizing the loss function is obtained; S500. Substitute the test data set into the trained model to verify the performance of the trained model, and repeat steps S400 to S500 until a certain condition is reached to stop training; S600. Determine the model parameters with the highest detection accuracy on the test data set as the final result.
2. The underground cable radar image detection method based on curve knowledge and deep learning according to claim 1, characterized in that: In S100 , the method for extracting the edge of the B-scan radar image is a method based on a first-order derivative, a method based on a second-order derivative, or a method based on a model.
3. The underground cable radar image detection method based on curve knowledge and deep learning according to claim 1, characterized in that: In S100, the method for extracting the edge of the B-scan radar image includes at least Roberts, Prewitt, and Sobel operators.
4. The underground cable radar image detection method based on curve knowledge and deep learning according to claim 1, characterized in that: In S200, the curvature of each edge point is calculated to create a curvature image. The specific method is as follows: S201. Detect edge points of edge images; S202. Refine the edge points; S203. Connect the breakpoints in the edge image; S204. Calculate the curvature of each edge point; S205. Set the curvature at non-edge points to 0.
5. The underground cable radar image detection method based on curve knowledge and deep learning according to claim 4, characterized in that: In S204, for the edge curve in the digital image, the points on the edge curve are numbered from small to large from one end to the other, and the coordinates of the i-th point are , the formula for calculating the curvature of each edge point is: ； in, ； in, 、 and They represent the points on the edge curve, numbered from small to large from one end to the other, and the coordinates of the i-1th, i and i+1th points.
6. The underground cable radar image detection method based on curve knowledge and deep learning according to claim 1, characterized in that: In S500 , the training is stopped until certain conditions are met, including: the number of training times reaches a given number or the error accuracy reaches a given accuracy.