Image registration method and system for EMU underbody fault detection

Abstract

The present invention provides an image registration method and system for fault detection of the underbody of an EMU, the method comprising: obtaining an original image of the underbody of the EMU; preprocessing the original image to obtain a first image; performing image registration on the first image and the corresponding template image through Fourier-Mellin transform to obtain translation, rotation and scaling parameters, and processing the first image according to the obtained translation, rotation and scaling parameters to obtain a second image for fault detection. The present invention realizes high-precision registration of the underbody image of the EMU by utilizing all the information in the image frequency domain, rather than using a rigid transformation method based on feature constraint points such as affine transformation; the image registration algorithm based on Fourier-Mellin determines the precise position of the template image relative to the image to be tested, and obtains relative translation, rotation and scaling parameters.

Description

Technical Field

[0001] The present invention relates to the technical field of fault detection, and in particular to an image registration method and system for detecting underbody faults of electric multiple unit trains. Background Art

[0002] With the rapid development of high-speed railway construction, the maintenance tasks of EMU operation stations are becoming increasingly arduous. Automatic fault detection methods such as TEDS and fault repair robots are widely accepted and adopted, gradually replacing traditional manual underbody appearance inspections to reduce labor intensity, improve the quality of maintenance operations, and ensure vehicle operation safety. The use of intelligent image analysis technology to automatically detect defects in EMU underbody components has become the basis for ensuring safe and stable vehicle operation. The EMU fault repair robot uses industrial-grade array cameras, six-axis robotic arms, RGV rail cars, etc. as hardware platforms, and uses image recognition and artificial intelligence technology to realize an intelligent underbody fault detection robot system that integrates acquisition perception, analysis and processing, and automatic control.

[0003] At present, various target detection technologies for defects in key components of EMU underbody, whether based on edge detection and morphological operations or feature extraction methods based on machine vision and deep learning, are based on difference analysis between the image to be detected and the template image. In practice, due to the relative motion of trains and equipment and dynamic control of imaging, environmental lighting conditions and insufficient local fill light, bad weather such as wind / rain / snow / sand and dust, it is very easy to cause distortion, dislocation, uneven brightness and other problems in the image to be detected, resulting in misjudgment by the intelligent identification algorithm for defect detection. The image elements of the underbody components of EMUs are relatively simple, and the local features are highly similar, which places high demands on the quality of the images to be processed by the target detection algorithm. A high-precision and robust image registration algorithm is required to pre-process the original collected images of the underbody. Summary of the invention

[0004] The purpose of the present invention is to provide an image registration method and system for EMU underbody fault detection. Some embodiments can be used to solve the defect of misjudgment produced by the defect detection intelligent identification algorithm in the prior art, realize the underbody images with complex environment dynamic imaging conditions, relatively single image elements, and highly similar local features, and propose a registration technology that meets the real-time processing requirements and has the characteristics of high precision and robustness.

[0005] The present invention provides an image registration method for detecting underbody faults of a multiple unit train. The method comprises: obtaining an original image of the underbody of the multiple unit train; preprocessing the original image to obtain a first image; performing image registration on the first image and a corresponding template image through Fourier-Mellin transform to obtain translation, rotation and scaling parameters, and processing the first image according to the obtained translation, rotation and scaling parameters to obtain a second image for fault detection.

[0006] According to an image registration method for detecting underbody faults of an EMU provided by the present invention, obtaining the original image of the underbody of the EMU comprises: setting an image acquisition device capable of moving under the EMU; the image acquisition device acquires a group of images while moving under the vehicle; and cutting and splicing the group of images to form the original image.

[0007] According to an image registration method for EMU underbody fault detection provided by the present invention, preprocessing the original image to obtain the first image includes: enhancing the contrast of the original image.

[0008] According to an image registration method for EMU underbody fault detection provided by the present invention, the contrast enhancement of the original image includes: adjusting the contrast of the original image by adopting gamma transformation.

[0009] According to an image registration method for EMU underbody fault detection provided by the present invention, the preprocessing of the original image to obtain the first image includes: performing windowing processing on the original image for reducing frequency domain noise in image registration.

[0010] According to an image registration method for EMU underbody fault detection provided by the present invention, the windowing process of the original image includes: selecting a Hanning window as a windowing function.

[0011] According to an image registration method for EMU underbody fault detection provided by the present invention, a first image and a corresponding template image are registered through Fourier-Mellin transform to obtain translation, rotation and scaling parameters, including: after Fourier transforming the first image and the template image, the amplitude spectra of the two are calculated; after high-pass filtering the amplitude spectrum, logarithmic-polar coordinate transformation is performed; and rotation and scaling parameters are obtained through a phase correlation algorithm.

[0012] According to an image registration method for EMU underbody fault detection provided by the present invention, a first image and a corresponding template image are registered through Fourier-Mellin transform to obtain translation, rotation and scaling parameters, including: performing an inverse Fourier transform on the mutual power spectrum of the first image and the template image after the rotation and scaling parameters are adjusted to obtain an impulse function, and obtaining the translation parameter from the peak of the impulse function.

[0013] According to an image registration method for EMU underbody fault detection provided by the present invention, the method also includes: partitioning and cutting the original image to obtain multiple first images; performing multiple tasks in parallel, wherein the tasks are used to perform image registration on the multiple first images and corresponding template images to obtain multiple sets of translation, rotation and scaling parameters; and combining the multiple sets of translation, rotation and scaling parameters as the translation, rotation and scaling parameters of the original image.

[0014] According to an image registration method for EMU underbody fault detection provided by the present invention, the calculation process in the method is accelerated by parallel processing using a graphics card.

[0015] According to an image registration method for EMU underbody fault detection provided by the present invention, the partitioning of the original image includes: partitioning according to a size of 512X512 pixels.

[0016] The present invention also provides an image registration system for EMU underbody fault detection, comprising: an acquisition module, which acquires an original image of the EMU underbody; a preprocessing module, which preprocesses the original image to obtain a first image; and a registration module, which performs image registration on the first image and a corresponding template image through Fourier-Mellin transform to obtain translation, rotation and scaling parameters, and processes the first image according to the obtained translation, rotation and scaling parameters to obtain a second image for fault detection.

[0017] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the image registration method for detecting underbody faults of an EMU as described above are implemented.

[0018] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the image registration method for detecting underbody faults of EMUs as described above are implemented.

[0019] The present invention also provides a computer program product, comprising a computer program, which, when executed by a processor, implements the steps of the image registration method for detecting underbody faults of EMUs as described above.

[0020] The image registration method and system for EMU underbody fault detection provided by the present invention realize high-precision registration of EMU underbody images by utilizing all information in the image frequency domain instead of using rigid transformation methods based on feature constraint points such as affine transformation; the precise position of a template image relative to an image to be tested is determined based on a Fourier-Merlin image registration algorithm to obtain relative translation, rotation and scaling parameters; a complete algorithm flow for realizing EMU underbody image registration after an EMU underbody fault detection robot collects original images is proposed, and after necessary image preprocessing and registration calculations, high-precision, reliable and stable matching of images can be completed within a reasonable time. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0022] Figure 1 A schematic flow chart of an image registration method for detecting underbody faults of a train set provided by the present invention;

[0023] Figure 2 A schematic diagram of a device for obtaining an original image of the bottom of a train set provided by the present invention;

[0024] Figure 3 A schematic diagram of the algorithm flow of the EMU vehicle bottom image registration provided by the present invention;

[0025] Figure 4 A structural schematic diagram of an image registration system for EMU underbody fault detection provided by the present invention;

[0026] Figure 5 An example physical structure diagram of an electronic device is provided.

[0027] Reference numerals:

[0028] 1: Robot guide rail; 2: EMU operation track; 3: CCD area array camera;

[0029] 4: Charging pile; 5: Pit; 6: Undercarriage inspection robot. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0031] The image registration method for EMU underbody fault detection provided by the embodiment of the present application is described in detail below with reference to the accompanying drawings through specific embodiments and their application scenarios.

[0032] Figure 1 The flowchart of the image registration method for the EMU underbody fault detection provided by the present invention is as follows: Figure 1As shown, an image registration method for detecting underbody faults of an EMU is provided, the method comprising: step 100, obtaining an original image of the underbody of the EMU; step 200, preprocessing the original image to obtain a first image; step 300, performing image registration on the first image and a corresponding template image through Fourier-Mellin transform to obtain translation, rotation and scaling parameters, and processing the first image according to the obtained translation, rotation and scaling parameters to obtain a second image for fault detection.

[0033] Optionally, the subject implementing the registration method may be a computer, which is connected to an external input device, such as a camera, which inputs the original image into the computer after acquiring it. The computer is encapsulated with a preprocessing program, which may include a judgment on whether the image meets the standard and various corresponding processing methods, such as judging whether the original image is overexposed or underexposed, and whether the contrast needs to be adjusted; further, judging whether the original image contains a picture of the bottom of the EMU; further, judging the size of the original image and whether a partition operation needs to be performed.

[0034] The computer also encapsulates the Fourier-Mellin transform program. The Fourier-Mellin transform is a method based on the Fourier transform. The Fourier transform is first performed, then converted into an amplitude spectrum, and then some processing methods are used to obtain parameters. In the image registration process, it is often necessary to deal with translation, rotation, scale transformation, occlusion, deformation and other problems. The Fourier-Mellin transform can well deal with translation, in-plane rotation, scaling and occlusion, and is a method with strong robustness. The computer also encapsulates the defect detection intelligent identification algorithm. The second image obtained after processing is used as the detection target image of the defect detection intelligent identification algorithm.

[0035] The image registration method for EMU underbody fault detection provided in this embodiment realizes high-precision registration of EMU underbody images by utilizing all information in the image frequency domain instead of using rigid transformation methods based on feature constraint points such as affine transformation; determines the precise position of the template image relative to the image to be tested based on the Fourier-Merlin image registration algorithm, and obtains relative translation, rotation, and scaling parameters; proposes a complete algorithm flow for realizing EMU underbody image registration after the EMU underbody fault detection robot collects the original image, and after necessary image preprocessing and registration calculation, can complete high-precision, reliable and stable matching of images within a reasonable time.

[0036] Furthermore, in another embodiment, obtaining the original image of the bottom of the EMU includes: setting an image acquisition device capable of moving under the EMU; the image acquisition device acquires a group of images while moving under the vehicle; and cutting and splicing the group of images to form the original image.

[0037] Figure 2A schematic diagram of a device for obtaining an original image of the bottom of a train provided by the present invention, such as Figure 2 As shown, during the EMU maintenance operation in the EMU operation center, the EMU fault detection robot performs detection operations on the guide rails laid in the track trench, and realizes position positioning, image acquisition, data transmission, and fault detection through the electromechanical control system.

[0038] The array camera on the fault detection robot takes a narrowband image each time. As the robot moves on the guide rails under the vehicle, the electromechanical control system automatically takes a series of images of the vehicle. Ideally, these images are stitched together to complete the capture of a complete image of the vehicle under a certain section of the EMU. In practice, due to the difference in the length of each vehicle sequence of the EMU and the dynamic movement of the equipment, it is generally difficult to achieve pixel-level precise stitching, and some irrelevant image content also needs to be cropped.

[0039] Assume that the length of a certain section of the EMU to be tested is L, the total time for the robot to perform uniform speed inspection on the vehicle is t, the width of a single image taken by the array camera is l, and the length of the image must cover both sides of the vehicle body. In order to obtain a complete bottom image of the vehicle, the minimum shooting frequency of the array camera can be obtained: Min{f}=L / lt. During the robot's movement, the electromechanical control system must ensure that there is no missed inspection, that is, full coverage of the local area. In addition, the overlapping parts of the continuous images are spliced; the images beyond the two ends of the vehicle are cropped.

[0040] The image registration method for EMU underbody fault detection provided in this embodiment obtains a series of continuous narrow-band images of the underbody through the robot's inspection of a certain vehicle section of the EMU, and obtains a complete underbody acquisition image of the vehicle through splicing and cropping.

[0041] Furthermore, in another embodiment, the features of the image under the EMU vehicle are single, the environment for capturing the image under the vehicle is dim, and the effect presented by the image display system is often darker than the desired image, requiring substantial contrast enhancement. Therefore, the original image is preprocessed to obtain the first image, including: enhancing the contrast of the original image. Preferably, enhancing the contrast of the original image includes: using gamma transform to adjust the contrast of the original image.

[0042] The basic form of the gamma transform is as follows:

[0043] s=cr γ

[0044] Where r is the input value of the grayscale image (the original grayscale value), and its value range is [0,1].

[0045] s is the grayscale output value after gamma transformation, c is the grayscale scaling factor, usually 1,

[0046] γ is the gamma factor, which controls the scaling of the entire transformation. c and γ are positive constants. When γ<1, the highlight part of the image is expanded and the dark part of the image is compressed; when γ>1, the highlight part of the image is compressed and the dark part of the image is expanded. Based on field experience, c=1 and γ=0.3 are set.

[0047] The image registration method for EMU underbody fault detection provided in this embodiment enhances contrast through gamma transformation, expands the highlight part of the image, and compresses the dark part of the image.

[0048] Furthermore, in another embodiment, preprocessing the original image to obtain the first image includes: performing windowing processing on the original image to reduce frequency domain noise in image registration. Preferably, performing windowing processing on the original image includes: selecting a Hanning window as a windowing function.

[0049] In order to eliminate the boundary effect of the image and smooth out the noise in the frequency domain caused by the discontinuous edges of the original image, it is generally necessary to perform windowing on the original image before executing the image registration algorithm. Common window functions include rectangular window (i.e., no windowing), triangular window, Hanning window, Hamming window, Gaussian window, etc. Except for the rectangular window, other windows are reflected in the time domain as high in the middle and low at both ends.

[0050] The Hanning window is selected based on experience, and its function expression is as follows:

[0051]

[0052] The image registration method for EMU underbody fault detection provided in this embodiment uses a Hanning window to perform windowing processing on the original image to complete the processing of the image to be tested.

[0053] Furthermore, in another embodiment, the first image and the corresponding template image are registered through Fourier-Mellin transform to obtain translation, rotation and scaling parameters, including: performing inverse Fourier transform on the mutual power spectrum of the first image and the template image to obtain an impulse function, and obtaining the translation parameter from the peak of the impulse function; after Fourier transforming the first image and the template image, the amplitude spectra of the two are calculated; after high-pass filtering the amplitude spectrum, log-polar coordinate transformation is performed; and rotation and scaling parameters are obtained through a phase correlation algorithm.

[0054] Assume that f1(x, y) and f2(x, y) represent the template image and the image to be tested (i.e., the first image), respectively, and f2(x, y) is obtained by translating f1(x, y) by (Δx, Δy), rotating by Δθ, and scaling by λ, that is, the following relationship is satisfied:

[0055] f2(x,y)=f1[λ(xcosΔθ+ysinΔθ)-Δx,λ(-xsinΔθ+ycosΔθ)

[0056] Perform Fourier transform on both sides and let F(μ,ν) be the Fourier transform of f(x,y), we can get:

[0057]

[0058] The above formula converts the image signal from the time domain to the frequency domain. According to the basic properties of the two-dimensional Fourier transform, it can be seen from the above formula that the translation (Δx, Δy) does not affect the amplitude spectrum of F(μ, ν), but only affects its phase spectrum. Performing an inverse Fourier transform on the cross power spectrum of the two image signals can obtain an impulse function (Dirac function), and the translation parameters (Δx, Δy) of the two images are obtained from the peak of the function.

[0059] Calculate the amplitude spectrum of both sides, let M(μ,ν) represent the amplitude spectrum of F(μ,ν), and we can get:

[0060]

[0061] Convert the amplitude spectrum to the log-polar coordinate space, let M(lgρ,θ) represent the log-polar coordinate transformation of M(μ,ν), and we can get:

[0062]

[0063] The Δθ and lgλ in the above log-polar coordinates correspond to the rotation and scaling between the image signals f1(x,y) and f2(x,y).

[0064] The above is the Fourier-Mellin transform. Using the phase correlation algorithm, we can obtain the sub-pixel precision translation parameters (Δx, Δy), rotation parameters Δθ, and scaling parameters λ between f1(x, y) and f2(x, y).

[0065] Figure 3 The algorithm flow diagram of the EMU vehicle bottom image registration provided by the present invention is as follows: Figure 3 As shown, extract the bottom template image f1(x, y) of the same type of EMU vehicle, start the registration calculation of the image to be tested f2(x, y); perform Fourier transform on the two image signals, and calculate the amplitude spectrum of the two images; perform high-pass filtering according to the amplitude spectrum of the two images; perform logarithmic-polar coordinate conversion on the two images, and obtain the rotation Δθ and scaling parameter λ through the phase correlation algorithm; calculate the cross-power spectrum of the first image and the template image after the rotation parameter Δθ and scaling parameter λ are adjusted, perform inverse transformation of the cross-power spectrum, and obtain the translation parameters (Δx, Δy); obtain all the registration parameters of the image to be tested with respect to the template image, and complete the image registration.

[0066] In an image registration experiment, local images of the traction motor and sand-spreading pipe under the EMU were used. Parameters were obtained through the registration algorithm, and the original acquired images were translated, rotated, and scaled to achieve registration with the template image.

[0067] For the sand-sprinkling tube image, the actual rotation angle of the image to be tested relative to the template image is 8 ° , without translation and scaling, and no high-pass filtering was performed to verify the experimental effect. The rotation parameter obtained from the log-polar coordinate transformation is 9.46 ° ; The error exceeds 1 degree, but the result is basically acceptable.

[0068] For the traction motor image, the actual translation of the image to be tested relative to the template image is (100,9), and the rotation angle is 5 ° , the zoom ratio is 3 times, and high-pass filtering has been performed in advance. Through processing and calculation, Δx=99.973, Δy=9.012, Δθ=5.14, λ=2.98 are obtained. Obviously, the image registration result is highly consistent with the actual situation.

[0069] The image registration method for EMU underbody fault detection provided in this embodiment adopts a phase correlation algorithm encapsulated in a software library to align the template image and the image to be tested. Different from the affine transformation algorithm, the Fourier-Mellin transform algorithm applies the phase correlation algorithm in the image frequency domain and uses the global information of the image to obtain image registration parameters, especially for EMU underbody images with single elements in each region and few high-curvature feature points, the method has higher accuracy and robustness, and the experimental results are consistent with the expected effects.

[0070] Furthermore, in another embodiment, the method further comprises: partitioning the original image to obtain a plurality of first images; performing a plurality of tasks in parallel, wherein the tasks are used to perform image registration between the plurality of first images and the corresponding template images to obtain a plurality of sets of translation, rotation and scaling parameters; and combining the plurality of sets of translation, rotation and scaling parameters as the translation, rotation and scaling parameters of the original image. Preferably, partitioning the original image comprises: partitioning according to a size of 512×512 pixels.

[0071] Optionally, integrating multiple groups of translation, rotation and scaling parameters specifically includes: after removing some values ​​that are obviously out of average level, averaging the remaining groups of translation, rotation and scaling parameters.

[0072] In order to test the performance of the Fourier-Mellin based image registration algorithm, this paper selects the template images of the traction motor and sand-spreading pipe mentioned above and the images to be tested, and crops images of three different sizes: 256X256 pixels, 512X512 pixels, and 1024X1024 pixels from the original images.

[0073] From the performance test results, it can be seen that the running time of the registration algorithm increases nonlinearly with the image size. In practical applications, in order to complete the registration processing of a large number of underbody images, a parallel processing strategy is adopted in combination with hardware equipment. The algorithm execution efficiency of the image with a size of 512x512 pixels is high, which can ensure the real-time performance of the on-site EMU fault detection operation. The complete underbody image is partitioned and cropped, and the parallel registration processing process is executed separately to generate a comprehensive image registration conclusion.

[0074] The image registration method for EMU underbody fault detection provided in this embodiment divides the image into multiple partitions of 512X512 pixel size, calculates translation, rotation and scaling parameters respectively, and adopts parallel processing in this process, and finally performs integration, which improves the operating speed on the one hand and the registration accuracy on the other hand.

[0075] Furthermore, in another embodiment, a high-performance graphics card is used to accelerate and parallelize the calculations.

[0076] The image registration method for EMU underbody fault detection provided in this embodiment can be seen from the algorithm steps that the time-consuming costs of operations such as Fourier transform, inverse Fourier transform, cross-power spectrum and its inverse transform, logarithmic-polar coordinate transformation, etc. are the highest. The calculation of translation parameters and the calculation of rotation and scaling parameters should be implemented through concurrent processing processes, and can also be optimized for parallel processing through single instruction multiple data streams to further improve the real-time performance of the calculation.

[0077] The image registration system for EMU underbody fault detection provided by the present invention is described below. The image registration system for EMU underbody fault detection described below and the image registration method for EMU underbody fault detection described above can be referenced to each other.

[0078] Figure 4 The structural diagram of the image registration system for the EMU underbody fault detection provided by the present invention is as follows: Figure 4 As shown, the present invention provides an image registration system for EMU underbody fault detection, comprising: an acquisition module, the acquisition module acquires the original image of the EMU underbody; a preprocessing module, the preprocessing module preprocesses the original image to obtain a first image; a registration module, the registration module performs image registration on the first image and the corresponding template image through Fourier-Mellin transform to obtain translation, rotation and scaling parameters, and processes the first image according to the obtained translation, rotation and scaling parameters to obtain a second image for fault detection.

[0079] The image registration system for EMU underbody fault detection provided in this embodiment realizes high-precision registration of EMU underbody images by utilizing all information in the image frequency domain instead of using rigid transformation methods based on feature constraint points such as affine transformation; determines the precise position of the template image relative to the image to be tested based on the Fourier-Merlin image registration algorithm, and obtains relative translation, rotation, and scaling parameters; proposes a complete algorithm flow for realizing EMU underbody image registration after the EMU underbody fault detection robot collects the original image, and after necessary image preprocessing and registration calculation, can complete high-precision, reliable and stable matching of images within a reasonable time.

[0080] Figure 5 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 5 As shown, the electronic device may include: a processor 710, a communication interface 720, a memory 730 and a communication bus 740, wherein the processor 710, the communication interface 720 and the memory 730 communicate with each other through the communication bus 740. The processor 710 may call the logic instructions in the memory 730 to execute the image registration method for the underbody fault detection of the EMU, the method comprising: obtaining the original image of the underbody of the EMU; preprocessing the original image to obtain a first image; performing image registration on the first image and the corresponding template image through Fourier-Mellin transform to obtain translation, rotation and scaling parameters, and processing the first image according to the obtained translation, rotation and scaling parameters to obtain a second image for fault detection.

[0081] In addition, the logic instructions in the above-mentioned memory 730 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0082] On the other hand, the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the image registration method for EMU underbody fault detection provided by the above-mentioned methods, and the method includes: acquiring an original image of the EMU underbody; preprocessing the original image to obtain a first image; performing image registration on the first image and the corresponding template image through Fourier-Mellin transform to obtain translation, rotation and scaling parameters, and processing the first image according to the obtained translation, rotation and scaling parameters to obtain a second image for fault detection.

[0083] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the above-mentioned image registration method for EMU underbody fault detection, the method comprising: acquiring an original image of the EMU underbody; preprocessing the original image to obtain a first image; performing image registration on the first image and a corresponding template image through Fourier-Mellin transform to obtain translation, rotation and scaling parameters, and processing the first image according to the obtained translation, rotation and scaling parameters to obtain a second image for fault detection.

[0084] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0085] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0086] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An image registration method for EMU underbody fault detection, characterized in that: The method comprises: Obtain the original image of the bottom of the EMU; Preprocessing the original image to obtain a first image; Performing image registration on the first image and the corresponding template image through Fourier-Mellin transform to obtain translation, rotation and scaling parameters, and processing the first image according to the obtained translation, rotation and scaling parameters to obtain a second image for fault detection; The obtaining of the original image of the bottom of the EMU comprises: A movable image acquisition device is arranged on the bottom of the EMU; The image acquisition device acquires a set of narrowband images while the vehicle moves under the vehicle, and the length of the narrowband images covers both sides of the vehicle body; The group of narrow-band images are trimmed and spliced ​​to form an original image, where the original image is a complete underbody image of the vehicle; Performing image registration with the first image and the corresponding template image through Fourier-Mellin transform to obtain translation, rotation and scaling parameters includes: After Fourier transforming the first image and the template image, the amplitude spectra of the two are calculated; The magnitude spectrum was high-pass filtered and then log-polar transformed; The rotation and scaling parameters are obtained through phase correlation algorithm; The method further comprises: Cut the original image into sections to obtain a plurality of first images; Multiple tasks are performed in parallel, wherein the tasks are used to perform image registration between the multiple first images and the corresponding template images to obtain multiple sets of translation, rotation and scaling parameters; Multiple sets of translation, rotation and scaling parameters are combined as the translation, rotation and scaling parameters of the original image.

2. The image registration method for EMU underbody fault detection according to claim 1 is characterized in that: The preprocessing of the original image to obtain the first image comprises: Enhance the contrast of the original image.

3. The image registration method for EMU underbody fault detection according to claim 2 is characterized in that: The step of enhancing the contrast of the original image comprises: Gamma transformation is used to adjust the contrast of the original image.

4. The image registration method for EMU underbody fault detection according to claim 1, characterized in that: The preprocessing of the original image to obtain the first image comprises: The original image is subjected to windowing processing to reduce the frequency domain noise in image registration.

5. The image registration method for EMU underbody fault detection according to claim 4 is characterized in that: The windowing process for the original image comprises: Select Hanning Window as the Windowing Function.

6. The image registration method for EMU underbody fault detection according to claim 1, characterized in that: Performing image registration with the first image and the corresponding template image through Fourier-Mellin transform to obtain translation, rotation and scaling parameters also includes: Perform inverse Fourier transform on the mutual power spectra of the first image and the template image after the rotation and scaling parameters are adjusted to obtain an impulse function, and obtain the translation parameter from the peak value of the impulse function.

7. The image registration method for EMU underbody fault detection according to claim 1, characterized in that: The computing process in the method is accelerated by parallel processing using a graphics card.

8. The image registration method for EMU underbody fault detection according to claim 1, characterized in that: The partitioning of the original image comprises: The partitions are cut according to the size of 512X512 pixels.

9. An image registration system for EMU underbody fault detection, characterized in that: include: An acquisition module, wherein the acquisition module acquires an original image of the bottom of the EMU; A preprocessing module, wherein the preprocessing module preprocesses the original image to obtain a first image; A registration module, wherein the registration module performs image registration on the first image and the corresponding template image through Fourier-Mellin transform to obtain translation, rotation and scaling parameters, and processes the first image according to the obtained translation, rotation and scaling parameters to obtain a second image for fault detection; The obtaining of the original image of the bottom of the EMU comprises: A movable image acquisition device is arranged on the bottom of the EMU; The image acquisition device acquires a set of narrowband images while the vehicle moves under the vehicle, and the length of the narrowband images covers both sides of the vehicle body; The group of narrow-band images are trimmed and spliced ​​to form an original image, where the original image is a complete underbody image of the vehicle; Performing image registration with the first image and the corresponding template image through Fourier-Mellin transform to obtain translation, rotation and scaling parameters includes: After Fourier transforming the first image and the template image, the amplitude spectra of the two are calculated; The magnitude spectrum was high-pass filtered and then log-polar transformed; The rotation and scaling parameters are obtained through phase correlation algorithm; The preprocessing module is also used to divide the original image into sections to obtain a plurality of first images; The registration module is also used to perform multiple tasks in parallel, and the tasks are used to perform image registration on multiple first images with corresponding template images to obtain multiple sets of translation, rotation and scaling parameters; and to combine the multiple sets of translation, rotation and scaling parameters as the translation, rotation and scaling parameters of the original image.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the image registration method for EMU underbody fault detection as described in any one of claims 1-8 are implemented.

11. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the image registration method for EMU underbody fault detection as described in any one of claims 1-8 are implemented.

12. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the image registration method for EMU underbody fault detection as described in any one of claims 1-8 are implemented.

Images