A data acquisition method for track defect detection

Abstract

The application provides a data acquisition method for track defect detection, comprising: creating a buffer area for temporarily storing track image frames and a memory area for permanently storing predefined files; wherein the buffer area and the memory area have a storage mapping relationship; the predefined files comprise a file header and image files; storing the acquired track image frames in the buffer area at a current time and synchronously taking out the track image frames stored before the current time from the buffer area; taking the track image frames taken out from the buffer area as image files and storing all the image files in the memory area according to the storage mapping relationship. The technical solution of the application can effectively solve the technical problem in the prior art that a large number of clear track images cannot be acquired in real time for defect detection due to low acquisition efficiency.

Description

Technical Field

[0001] This invention belongs to the field of track defect detection data acquisition technology, specifically relating to a data acquisition method for track defect detection. Background Technology

[0002] High-speed rail is a major transportation infrastructure in China, and its safety is of paramount importance. Defect detection on railway tracks is a crucial link in ensuring the smooth operation of high-speed rail. With the rapid development of artificial intelligence and deep learning, visual defect detection has gradually become the mainstream detection method. Visual inspection requires a large number of track images for model training or defect prediction. Due to the long length of railways, relying on manual photography is not only inefficient but also cannot guarantee the real-time nature of track defect detection. Therefore, rapid image acquisition methods are a very important part of track defect detection.

[0003] Current methods for capturing railway track images involve mounting cameras on a trolley, which continuously acquire a sequence of images of the track as the trolley moves, and then directly storing the images on a memory card. This method suffers from slow acquisition speed, unclear images, and the possibility of defects in the acquired images, making it unsuitable for detecting subsequent track defects. Summary of the Invention

[0004] To address the aforementioned problems in the existing technology, this invention provides a data acquisition method for track defect detection. The technical problem to be solved by this invention is achieved through the following technical solution:

[0005] In a first aspect, the present invention provides a data acquisition method for track defect detection, comprising:

[0006] S100, a buffer area for temporarily storing track image frames and a memory area for permanently storing predefined files are created; wherein, the buffer area and the memory area have a storage mapping relationship; the predefined files include a file header and an image file;

[0007] S200, store the acquired orbital image frames in the buffer area at the current time and simultaneously retrieve the orbital image frames stored before the current time from the buffer area;

[0008] S300, the track image frames taken from the buffer area are used as image files, and all image files are stored in the memory area according to the storage mapping relationship.

[0009] Secondly, the present invention provides a method for detecting track defects, the method comprising:

[0010] Step 1: Obtain the memory area in the data acquisition method for track defect detection that has stored all files, and read the files from the memory area;

[0011] Step 2: Parse the file read in Step 1 to obtain multiple image files;

[0012] Step 3: Use bilinear interpolation to interpolate the multiple image files to obtain an interpolated image with a resolution that is a multiple of that before interpolation.

[0013] Step 4: Use the trained defect detection model to detect whether there is a defect in the track by detecting the interpolated image.

[0014] Beneficial effects:

[0015] This invention ensures the real-time performance and integrity of image data by establishing an efficient storage mapping mechanism between the buffer region and the memory region. Simultaneously, the application of bilinear interpolation technology significantly enhances image resolution, making defect detection more precise and effectively improving the data processing efficiency and accuracy of the track defect detection system. This is of great value for the early detection and prevention of track problems, effectively improving rail transit safety and maintenance efficiency.

[0016] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a data acquisition method for track defect detection provided by the present invention;

[0018] Figure 2 This is a schematic flowchart of a track defect detection method provided by the present invention;

[0019] Figure 3 This is a schematic diagram of the bilinear interpolation processing principle provided by the present invention;

[0020] Figure 4 This is a schematic diagram of the result of processing an orbital image frame using the bilinear interpolation method provided by the present invention. Detailed Implementation

[0021] The present invention will be further described in detail below with reference to specific embodiments, but the implementation of the present invention is not limited thereto.

[0022] Firstly, such as Figure 1 As shown, this embodiment of the invention provides a data acquisition method for track defect detection, including:

[0023] S100, a buffer area for temporarily storing track image frames and a memory area for permanently storing predefined files are created; wherein, the buffer area and the memory area have a storage mapping relationship; the predefined files include a file header and an image file; the file header specifies the size and location of the stored file, and the size and depth of the image file.

[0024] It's worth noting that images captured by the camera are primarily saved as single frames. Traditionally, stored image data includes a file header, bitmap information, and bitmap data. For different image files, the file header and bitmap information are identical, leading to redundant storage, wasted memory, and slowed storage speed. This invention defines a novel file format: a predefined file. Images acquired by the camera over a period of time are stored in this file, which consists of a file header and an image file. The file header stores information such as file size, location, image dimensions, and depth, while the image file stores the image data, i.e., track data frames. This not only reduces overall image memory usage but also avoids the slow storage speed caused by frequent file creation.

[0025] Technically, the file header design standardizes the storage format of image files, facilitating subsequent data parsing and processing. In principle, the file header contains crucial information about the image file, such as dimensions and depth, which are essential for image analysis. Effectively, the standardized file header design simplifies image file management and retrieval, improving data processing efficiency. This embodiment can also address the challenges of rapid image file location and time synchronization under large data volumes by expanding the file header information, such as adding timestamps or location information.

[0026] S200, store the acquired orbital image frames in the buffer area at the current time and simultaneously retrieve the orbital image frames stored before the current time from the buffer area;

[0027] Because high-speed cameras have a high acquisition speed, this invention uses a buffer to balance data acquisition and storage. In this invention, the acquisition and retrieval of track image frames are completed synchronously, with the buffer area serving as a temporary storage area for sequentially storing and retrieving track image frames.

[0028] S300, the track image frames taken from the buffer area are used as image files, and all image files are stored in the memory area according to the storage mapping relationship.

[0029] Because storage devices (such as hard drives) have limited memory expansion and are expensive, while hard drives not only have large storage capacity but are also inexpensive, this invention uses a method of directly storing files to the hard drive, which can effectively avoid the limitations of track segment measurement caused by the limited computer memory.

[0030] Technically, by defining a new image file format and employing a multi-threaded parallel processing mechanism, the inefficiency and slow storage speed of traditional image acquisition methods are effectively solved. In principle, the new image file format reduces the storage of redundant information, and the multi-threaded parallel processing of acquisition and reading fully utilizes hardware resources, improving the concurrency of data processing. In terms of effectiveness, this invention enables high-speed, clear image acquisition and storage, improving the real-time performance and accuracy of track defect detection.

[0031] In a specific embodiment of the present invention, S100 includes:

[0032] S110 uses std::queue as the underlying data structure storage container to create a buffer area;

[0033] This invention uses std::queue as the underlying data structure storage container, and std::mutex and std::condition_variable to achieve thread safety, ensure no data loss, and enable fast storage.

[0034] `std::queue` is a container adapter used to provide queue functionality, i.e., a First-In-First-Out (FIFO) data structure. `std::mutex` is the basic mutex class used to synchronize thread access. A mutex is a synchronization mechanism used to control access to shared resources by multiple threads. `std::mutex` provides exclusive ownership; only one thread can hold the mutex's lock state at a time. `std::condition_variable` is used to implement condition wait and notify mechanisms between threads, allowing a thread to suspend when certain conditions are not met until other threads notify that the conditions have been met.

[0035] S120, a memory area is set up on the storage device;

[0036] S130, Create a file using the CreateFile function and obtain the handle of the file;

[0037] S140, under the handle instruction, a memory-mapped object is created using the CreateFileMapping function;

[0038] S150, map the memory-mapped object to the memory region and return a pointer to the memory region;

[0039] This embodiment proposes to allocate a memory region on the computer's hard drive and directly store the acquired track image frames from the buffer region into the memory region. First, a file is created using the CreateFile function, and a file handle is obtained; second, a memory-mapped object is created using the CreateFileMapping function, and this memory-mapped object is associated with the virtual address of the memory region; then, a portion of the file is mapped to the virtual address using the MapViewOfFile function, and a pointer to the memory region is returned.

[0040] S160, establish a storage mapping relationship between the pointer and the buffer area.

[0041] This embodiment establishes a mapping relationship between a pointer and a cache address within the buffer area. The track image frames read from the cache area are directly written to a file through this pointer, thereby obtaining a railway track image file collected for the entire section.

[0042] Technically, this embodiment uses `std::queue` as the underlying data structure for the buffer, combined with memory mapping technology, to ensure thread safety and efficient data storage. In principle, the memory mapping mechanism allows mapping a file or a portion of a file into the address space of a process, thereby directly manipulating data on the hard drive and avoiding frequent file read / write operations. Effectively, this embodiment improves the efficiency of data acquisition and storage, ensuring the integrity and real-time nature of image data.

[0043] In a specific embodiment of the present invention, the pointer points to the real-time storage location of the file within the memory region, and there is a storage mapping relationship between the storage location of the current track image frame within the buffer region and the real-time storage location.

[0044] Technically, this embodiment establishes a mapping between the real-time storage location of the pointer and the storage location of the image frame in the buffer, ensuring accurate data transmission and storage. In principle, the storage mapping relationship guarantees rapid data movement from the buffer to the hard drive, avoiding redundant data storage in memory. Effectively, the method in this embodiment makes image data storage more efficient and orderly, reducing data processing latency. This invention can also introduce a data verification mechanism to ensure data integrity during transmission, solving the problems of errors and data loss during data transmission.

[0045] In a specific embodiment of the present invention, S200 includes:

[0046] S210, At the current time, the acquired orbital image frames are stored in the buffer area using the acquisition thread proThread;

[0047] S220, the storage thread conThread, which runs parallel to the acquisition thread proThread, retrieves the track image frame stored before the current time from the buffer area.

[0048] This embodiment uses the std::thread thread class to create the acquisition thread proThread and the storage thread conThread; within the acquisition thread, std::unique_lock is used... <std::mutex>To ensure data security, after the camera acquires data, the queue::push function is used to put the data into the buffer. At the same time, after the storage thread finishes storing a set of data, the queue::pop function is used to remove the data from the buffer, freeing up memory to receive new image data. Data acquisition and storage are carried out uninterruptedly through parallel threading.

[0049] It's worth noting that `std::thread` is a class for creating and managing threads in a standard way. Its constructor has several different forms, each with its specific purpose. `proThread` is a running program instance with independent functionality. `conThread` is a thread used in concurrent programming. `std::unique_lock` <std::mutex>`std::queue` is a lock management class used for more flexible and secure management of mutex locks. `queue::push` is a member function of `std::queue`, used to insert an element at the tail of the queue (i.e., the enqueue operation). Queues follow the FIFO (First-In, First-Out) principle; new elements are always added to the tail, and elements are removed from the head. `queue::pop` is a member function of `std::queue`, used to remove the head element of the queue (i.e., the dequeue operation). Queues also follow the FIFO rule; `pop()` always removes the earliest inserted element (the head of the queue), but does not return the value of that element.

[0050] Technically, this embodiment achieves real-time acquisition and storage of image data through the parallel operation of acquisition and storage threads. In principle, the thread parallelism mechanism fully utilizes the computing power of multi-core processors, improving the concurrency of data processing. In terms of effectiveness, the technology in this embodiment ensures the real-time nature and integrity of image data, improving the efficiency of track defect detection. This embodiment can also address performance issues in data processing under different hardware environments by adjusting the number of threads or optimizing the thread scheduling strategy.

[0051] In a specific embodiment of the present invention, prior to S200, the data acquisition method for track defect detection further includes:

[0052] Use the std::thread thread class to create the acquisition thread proThread and the storage thread conThread.

[0053] Technically, this embodiment uses the `std::thread` class to create threads, ensuring flexibility in thread creation and management. In principle, the thread class provides APIs for creating and managing threads, making multithreaded programming simpler and more efficient. In terms of effectiveness, the solution in this embodiment simplifies multithreaded programming and improves the concurrency of data acquisition and storage. This embodiment can also address performance and resource consumption issues in thread management under high load environments by using other thread libraries or optimizing thread creation and destruction mechanisms.

[0054] In a specific embodiment of the present invention, S300 includes:

[0055] S310, determine the storage location of the track image frame taken from the buffer area in the buffer area, and determine the real-time storage location in the memory area that has a storage mapping relationship with the storage location;

[0056] S320, the orbital image frame taken from the buffer area is used as an image file, and the image file is stored in the real-time storage location in S310 according to the pointer.

[0057] Technically, this embodiment ensures accurate storage of image data by defining a storage mapping relationship. In principle, the storage mapping relationship, based on memory mapping technology, ensures rapid data movement from the buffer to the hard drive. Effectively, the solution in this embodiment simplifies the image data storage process and improves data storage efficiency and reliability. This embodiment can also address the risk of data loss during storage by introducing a data backup mechanism.

[0058] Secondly, such as Figure 2 As shown, an embodiment of the present invention provides a method for detecting track defects, including:

[0059] Step 1: Obtain the memory area where all files have been stored in the data acquisition method for track defect detection in the first aspect, and read the files from the memory area;

[0060] Since the acquired orbital image sequence is not a single image file, all image files are stored in a single file. However, visual defect detection, whether for model training or defect estimation, requires a single image file. Therefore, it is necessary to parse the large image file into several independent image files. The parsing method still uses memory mapping. First, the CreateFile function is called to open the image file and obtain the file handle. Then, the CreateFileMapping function is used to create a memory-mapped object. Next, the MapViewOfFile function is used to obtain the file pointer. By using the pointer offset, the file header and image files are parsed one by one to obtain independent orbital image frames.

[0061] Step 2: Parse the file read in Step 1 to obtain multiple image files;

[0062] Step 3: Use bilinear interpolation to interpolate the multiple image files to obtain an interpolated image with a resolution that is a multiple of that before interpolation.

[0063] Step 4: Use the trained defect detection model to detect whether there is a defect in the track by detecting the interpolated image.

[0064] The defect detection model of the present invention can adopt an existing neural network model. The training process of the neural network model can be referred to in existing relevant literature, and will not be described in detail here.

[0065] To ensure acquisition speed, the selected camera resolution will not be too high. To ensure the accuracy of defect detection, this solution proposes an optimized bilinear interpolation method to increase the image resolution by at least twice the size of the original track image frame.

[0066] Technically, this embodiment uses bilinear interpolation to perform super-resolution processing on the image, improving image clarity and facilitating defect detection. In principle, bilinear interpolation estimates the value of missing pixels by calculating a weighted average of surrounding pixels, thereby increasing image resolution. Effectively, the technique in this embodiment ensures image clarity and improves the accuracy and reliability of defect detection. This embodiment can further enhance image quality and address image clarity issues in low-light or complex environments by introducing other image enhancement techniques, such as contrast enhancement or sharpening.

[0067] In a specific embodiment of the present invention, step 3 includes:

[0068] Step 31: For any image file, obtain the target image by applying the coordinate transformation formula to the original pixels in the image file; the coordinate transformation formula is expressed as:

[0069]

[0070] In the formula, and Let x and y be the x and y coordinates of a pixel in the target image. and The length and width of the target image; and This represents the width and height of the original image file. , For the target image at this pixel point ( , The coordinates of the original image file corresponding to the given coordinates.

[0071] Step 32: Determine the insertion pixel point for every two pixels in the target image, and adjust the pixel value of the insertion pixel point using the interpolation of the target image on the x-axis and y-axis respectively and their corresponding weights.

[0072] Bilinear interpolation involves inserting a new pixel between every two pixels, where the grayscale value of the new pixel is determined by the surrounding pixels. For example... Figure 3 As shown, it is known , , , Given the coordinates and corresponding pixel values, solve for the unknown function f at the green point. The pixel value.

[0073] First, calculate the linear interpolation in the X direction:

[0074]

[0075]

[0076] Then perform linear interpolation in the y direction to obtain :

[0077]

[0078] Further simplification of the formula yields:

[0079]

[0080] Since it is the closest point, The coordinates of the four points differ by 1, so the result can be further simplified to:

[0081]

[0082] Finally, by weighting and preserving the high-frequency components of the image, the image becomes sharper at the edges. This can be achieved by inserting adjusted pixel values, represented as follows:

[0083]

[0084]

[0085]

[0086]

[0087] in, This represents the pixel value of the inserted pixel. The weight of the horizontal distance between the inserted pixel and the left boundary line. This represents the weight of the vertical distance between the inserted pixel and the upper boundary line. Represents the pixels in the target image pixel values, Represents the pixels in the target image pixel values, Represents the pixels in the target image pixel values, Represents the pixels in the target image pixel value, pixel The coordinates are pixel The coordinates are pixel The coordinates are pixel The coordinates are The coordinates of the inserted pixel are .

[0088] Technically, this embodiment calculates the pixel value of the inserted pixel using a bilinear interpolation formula, achieving image super-resolution processing. In principle, this formula estimates the value of the inserted pixel based on the distance weights between pixels, using a weighted average of surrounding pixels, thereby improving image resolution. In terms of effect, the method in this embodiment improves image sharpness, making image edges sharper, which is beneficial for accurate defect detection. This embodiment can also address the accuracy issue of interpolation processing under different image features by optimizing the weight calculation formula, such as introducing Gaussian weights or adaptive weight adjustment.

[0089] Step 33: Insert the adjusted pixel value between the corresponding two pixels in the target image to obtain the interpolated image.

[0090] Technically, this embodiment uses bilinear interpolation to perform super-resolution processing on the image, improving image clarity. In principle, bilinear interpolation estimates the value of missing pixels using a weighted average of surrounding pixels, ensuring sharpness of image edges through weight adjustment. In effect, the technique in this embodiment improves image resolution, making the image clearer and facilitating accurate defect detection. This embodiment can also address the applicability of image interpolation processing in different scenarios by introducing other interpolation algorithms, such as cubic spline interpolation or nearest neighbor interpolation.

[0091] refer to Figure 4 , Figure 4 This is a schematic diagram illustrating the results of processing a track image frame using bilinear interpolation. Figure 4 The left image is the original track image frame, and the right image is the interpolated image. From Figure 4 As can be seen, the resolution of the image processed using the bilinear interpolation method of this invention is significantly improved.

[0092] The technical solution of this application relates to a workflow, specifically described as follows: First, a novel image file format is defined to reduce the storage of redundant information and improve storage efficiency. Then, a data buffer area and a memory-mapped object are created. Through the parallel operation of the acquisition thread (proThread) and the storage thread (conThread), real-time image data acquisition, storage, and processing are achieved. After the image data is stored on the hard disk, the image file is parsed using memory mapping technology to obtain multiple independent image files. Then, an optimized bilinear interpolation algorithm is used to perform super-resolution processing on the image files to improve image clarity. Finally, a trained defect detection model is used to detect defects in the processed image to obtain the detection result of whether there are defects in the track. The entire workflow is efficient and orderly, ensuring real-time acquisition, storage, and processing of image data, and improving the accuracy and real-time performance of track defect detection.

[0093] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0094] Although this application has been described herein in conjunction with various embodiments, those skilled in the art will understand and implement other variations of the disclosed embodiments by reviewing the accompanying drawings, the disclosure, and the appended claims in carrying out the claimed application. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude a plurality.

[0095] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.< / std::mutex> < / std::mutex>

Claims

1. A data acquisition method for track defect detection, characterized in that, Comprising: S100, creating a buffer area for temporarily storing track image frames and a memory area for permanently storing predefined files; wherein the buffer area and the memory area have a storage mapping relationship; the predefined files include a file header and image files; S200, storing the track image frames collected at the current time into the buffer area and synchronously taking out the track image frames stored before the current time from the buffer area by using a storage thread conThread parallel to the collection thread proThread; S300, taking the track image frames taken out from the buffer area as image files, and storing all the image files into the memory area according to the storage mapping relationship by using the pointer pointing to the real-time storage location in the memory area.

2. The data acquisition method for track defect detection according to claim 1, characterized in that, S100 comprises: S110, creating a buffer area by using std::queue as an underlying data structure storage container; S120, setting a memory area on a storage device; S130, creating a file by using the CreateFile function and obtaining the handle of the file; S140, creating a memory mapping object by using the CreateFileMapping function under the indication of the handle; S150, mapping the memory mapping object and the memory area to each other and returning a pointer pointing to the memory area; S160, establishing a storage mapping relationship between the pointer and the buffer area.

3. The data acquisition method for track defect detection according to claim 2, characterized in that, The pointer points to the real-time storage location in the memory area for storing files, and the storage location of the current track image frames in the buffer area has a storage mapping relationship with the real-time storage location.

4. The data acquisition method for track defect detection of claim 1, wherein, The file header specifies the size, position, size and depth of the image files of the storage file.

5. The data acquisition method for rail defect detection of claim 1, wherein, S200 comprises: S210, storing the track image frames collected at the current time into the buffer area by using the collection thread proThread; S220, taking out the track image frames stored before the current time from the buffer area by using the storage thread conThread parallel to the collection thread proThread.

6. The data acquisition method for rail defect detection of claim 1, wherein, Before S200, the data collection method for track defect detection further comprises: Creating the collection thread proThread and the storage thread conThread by using the std::thread thread class.

7. The data acquisition method for rail defect detection of claim 3, wherein, S300 comprises: S310, determining the storage location of the track image frames taken out from the buffer area in the buffer area, and determining the real-time storage location in the memory area having a storage mapping relationship with the storage location; S320, taking the track image frames taken out from the buffer area as image files, and storing the image files into the real-time storage location in S310 according to the pointer.

8. A method of rail defect detection, characterized in that, The track defect detection method comprises: Step 1, obtaining the memory area in which all the files have been stored in the data collection method for track defect detection according to any one of claims 1 to 7, and reading the files from the memory area; Step 2, parsing the file read in step 1 to obtain a plurality of image files based on the size, location, size and depth information of the image file of the storage file specified in the file header of the predefined file; Step 3, using a bilinear interpolation method to perform interpolation processing on the plurality of image files to obtain an interpolation processing image with a resolution multiple of the resolution before interpolation processing; Step 4, using a trained defect detection model to detect the interpolation processing image to obtain a detection result of whether the track has defects.

9. The rail defect detection method according to claim 8, characterized in that, Step 3 includes: Step 31, for any image file, obtaining a target image from the original pixels in the image file according to a coordinate conversion formula; Step 32, determining an inserted pixel point to be inserted between each two pixel points in the target image, and adjusting the pixel value of the inserted pixel point using the interpolation of the target image on the x-axis and the y-axis and the respective corresponding weight; Step 33, inserting the inserted pixel point with the adjusted pixel value between the corresponding two pixel points in the target image to obtain an interpolation processing image.

10. The rail defect detection method of claim 8, wherein, The pixel value of the inserted pixel point adjusted in step 33 is represented by the formula: wherein, represents a pixel value of the inserted pixel point, represents a distance weight of the inserted pixel point from the left boundary line in the horizontal direction, represents a distance weight of the inserted pixel point from the upper boundary point line in the vertical direction, represents a pixel value of the pixel point in the target image, represents a pixel value of the pixel point in the target image, represents a pixel value of the pixel point in the target image, represents a pixel value of the pixel point in the target image, the coordinate of the pixel point is , the coordinate of the pixel point is , the coordinate of the pixel point is , the coordinate of the pixel point is , and the coordinate of the inserted pixel point is .

Images