Low-speed train image acquisition system and method

Abstract

The application relates to a low-speed train image acquisition system and method, and belongs to the technical field of image acquisition. The low-speed train image acquisition system comprises a radar module, a magnetic steel sensor module, an image acquisition module and a controller. When it is determined that a train is in a moving state and enters an image acquisition area based on radar signals and magnetic steel signals, the image acquisition time of each carriage of the train is determined based on the magnetic steel signals, and the carriages successively passing through the image acquisition area are subjected to image acquisition. When it is determined that the train stops in the image acquisition area based on the radar signals and the magnetic steel signals, the image acquisition of the train is stopped, and the carriage information of the acquired images is recorded. When it is determined that the train starts in the image acquisition area based on the radar signals, the remaining carriages of the train are subjected to image acquisition based on the carriage information of the acquired images, the carriage images of the acquired carriages are spliced, and the train image is obtained. The application can realize continuous and accurate acquisition of train carriage images.

Description

Technical Field

[0001] This invention relates to the field of image acquisition technology, and in particular to an image acquisition system and method for low-speed trains. Background Technology

[0002] In the field of railway safety monitoring, image acquisition of running trains is an important means of fault detection and status monitoring.

[0003] Existing technologies mainly rely on fixed-position magnets or photoelectric sensors to trigger linear array cameras for image capture. However, this system has significant limitations. When the train is moving at low speed or creeping, the traditional triggering method is unstable and prone to missed triggers. When the train stops and restarts, the system struggles to achieve continuous and seamless follow-up checks, resulting in data interruptions.

[0004] It is evident that existing technologies consistently miss images of trains moving at low speeds and are prone to interruption. Summary of the Invention

[0005] In view of this, it is necessary to provide a low-speed train image acquisition system and method to solve the problems of missed acquisition and easy interruption in the acquisition of train images under low speed in the existing technology.

[0006] To address the aforementioned problems, in a first aspect, the present invention provides a low-speed train image acquisition system, comprising a radar module, a magnetic sensor module, an image acquisition module, and a controller, wherein... The radar module is used to collect radar signals when the train passes by, and the magnet sensor module is used to collect magnet signals when the train passes by. The controller is used to determine the image acquisition time of each carriage of the train based on the magnetic signal when the train is determined to be in a moving state and enters the image acquisition area based on the radar signal and the magnetic signal, and to send the image acquisition command to the image acquisition module based on the image acquisition time. The controller is also used to send an image recording instruction to the image acquisition module when it is determined that the train has stopped in the image acquisition area based on radar signals and magnetic steel signals, and to send an image continuation acquisition instruction to the image acquisition module based on the carriage information of the already acquired images when it is determined that the train has started in the image acquisition area based on radar signals. The image acquisition module is used to acquire images of the carriages that pass through the image acquisition area in sequence based on the image acquisition command, stop acquiring images of the train based on the image recording command and record the carriage information of the acquired images, and acquire images of the remaining carriages of the train based on the image continuation acquisition command. The controller is also used to stitch together the images of each carriage to obtain a train image.

[0007] In one possible implementation, when the controller determines that the train is moving and has entered the image acquisition area based on radar and magnetic signals, it is configured to: Radar speed measurement is used to determine the train's speed based on radar signals; The train's magnetic speed measurement is determined based on the time interval between the first magnetic signal of the train passing through the image acquisition area (initial acquisition by the first magnetic signal of the train passing through the image acquisition area) and the second magnetic signal of the train passing through the image acquisition area (initial acquisition by the second magnetic signal of the train passing through the image acquisition area), and the distance between the first and second magnetic signals. When the difference between the radar speed measurement and the magnet speed measurement is within a preset range, the train is determined to be in a moving state and has entered the image acquisition area.

[0008] In one possible implementation, when determining the image acquisition time for each carriage of the train based on the magnetic steel signal, the controller is used to: The image acquisition time of each carriage of the train is determined based on the acquisition time of the third magnet signal of each carriage passing the third magnet, which is deployed in the image acquisition area, wherein the third magnet is after the first magnet and the second magnet.

[0009] In one possible implementation, when the controller determines the image acquisition time for each carriage of the train based on the magnetic steel signal and sends an image acquisition command to the image acquisition module based on the image acquisition time, it is used to: The real-time speed of each carriage as it passes through the image acquisition area is calculated based on the radar signal and the magnet signal. Based on the real-time vehicle speed and preset calibration coefficients, the image acquisition line frequency for the carriage is determined, and an image acquisition command is sent to the image acquisition module. The image acquisition command includes the time for image acquisition and the image acquisition line frequency.

[0010] In one possible implementation, when the controller sends an image continuation acquisition command to the image acquisition module based on the carriage information of the already acquired images, it is used to: The image acquisition time of the first carriage among the remaining carriages is determined based on the carriage number of the last carriage that has been captured in the image, the number of axles that the last carriage has passed through the image acquisition area, and the real-time speed of the train. Based on the image acquisition time of the first carriage and the magnetic signal of the remaining carriage passing through the image acquisition area, the image acquisition time of the remaining carriage is determined, and an image continuation acquisition command is sent to the image acquisition module. The image continuation acquisition command includes the image acquisition time of the remaining carriage.

[0011] In one possible implementation, when the controller stitches together the acquired images of each carriage to obtain a train image, it is used to: A deep learning model is used to extract the foreground mask from the carriage image; The foreground mask is fused with a preset background model to obtain the foreground image of the carriage; The foreground images of each carriage are stitched together according to their carriage numbers to obtain the train image.

[0012] In one possible implementation, when the controller stitches together the foreground images of each carriage according to the carriage number to obtain the train image, it is used to: Determine the acquisition time of the carriage image corresponding to the foreground image, and determine the geometric distortion correction parameters of the foreground image based on the train speed corresponding to the acquisition time and the average train speed in the entire image acquisition area; The foreground image is corrected based on the geometric distortion correction parameters to obtain the corrected foreground image of the carriage; The corrected foreground images of each carriage are stitched together according to the carriage number to obtain the train image.

[0013] In one possible implementation, the magnetic sensor module includes multiple magnetic sensors, which are sequentially distributed on the rails in the image acquisition area at preset intervals.

[0014] In one possible implementation, the radar module includes multiple millimeter-wave radars, each of which is arranged on both sides of the track at a preset height and angle.

[0015] Secondly, the present invention also provides a method for acquiring images of low-speed trains, comprising: When the train is determined to be moving and enters the image acquisition area based on radar and magnetic signals, the image acquisition time of each carriage of the train is determined based on the magnetic signals, and the images of the carriages that pass through the image acquisition area in sequence are acquired based on the image acquisition time. When the radar and magnetic steel signals determine that the train has stopped within the image acquisition area, the image acquisition of the train is stopped, and the carriage information of the acquired images is recorded. When the radar signal determines that the train has started in the image acquisition area, the remaining carriages of the train are imaged based on the carriage information of the already acquired images, and the images of each carriage are stitched together to obtain the train image.

[0016] The beneficial effects of this invention are as follows: The low-speed train image acquisition system provided by this invention includes a radar module, a magnetic sensor module, an image acquisition module, and a controller. The radar module is used to acquire radar signals when a train passes by and send these signals to the controller. The magnetic sensor module is used to acquire magnetic signals when a train passes by and send these signals to the controller. When the radar signal and magnetic signal determine that the train is moving and has entered the image acquisition area, the image acquisition time for each carriage of the train is determined based on the magnetic signal. Based on these image acquisition times, images of the carriages that sequentially pass through the image acquisition area are acquired. By using both radar and magnetic signals to determine whether the train is moving and has entered the image acquisition area, the determination of train arrival at the station is more accurate. This prevents the omission of images of the first carriage due to a single magnetic signal. Determining the acquisition time of each carriage using magnetic signals allows for precise positioning of each carriage, enabling image acquisition of the carriages at the exact time and ensuring the accuracy of image acquisition. When the radar and magnetic signals determine that the train has stopped within the image acquisition area, image acquisition of the train is stopped, and the carriage information of the acquired images is recorded. When the radar signals determine that the train has started moving within the image acquisition area, images of the remaining carriages of the train are acquired based on the carriage information of the acquired images, and the acquired carriage images of each carriage are stitched together to obtain the train image. When it is determined that the train has stopped in the station, the acquisition of carriage images is stopped, and the corresponding information is recorded. When the train is detected to start moving again, images of subsequent carriages are acquired based on the recorded information. This effectively prevents discontinuity in carriage images and ensures the integrity of carriage image acquisition. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the structure of a low-speed train image acquisition system provided in an embodiment of the present invention; Figure 2 This is a flowchart illustrating a train status determination method provided in an embodiment of the present invention; Figure 3 This is a flowchart illustrating an image acquisition method provided in an embodiment of the present invention; Figure 4 A flowchart illustrating another image acquisition method provided in an embodiment of the present invention; Figure 5A schematic flowchart of a train image stitching method provided in an embodiment of the present invention; Figure 6 This is a schematic flowchart of an image correction method provided in an embodiment of the present invention; Figure 7 This is a flowchart illustrating a low-speed train image acquisition method provided in an embodiment of the present invention. Detailed Implementation

[0019] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0020] In the description of the embodiments of the present invention, unless otherwise stated, "multiple" means two or more. "And / or" describes the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone.

[0021] The terms "first," "second," etc., used in the embodiments of this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a technical feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature.

[0022] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0023] A specific embodiment of the present invention, such as Figure 1 As shown, a low-speed train image acquisition system is disclosed, including a radar module 101, a magnetic sensor module 102, an image acquisition module 103, and a controller 104, wherein... Radar module 101 is used to collect radar signals when a train passes by, and magnet sensor module 102 is used to collect magnet signals when a train passes by. The controller 103 is used to determine the image acquisition time of each carriage of the train based on the magnetic signal when the train is determined to be in a moving state and enters the image acquisition area based on the radar signal and the magnetic signal, and to send an image acquisition command to the image acquisition module based on the image acquisition time. The controller 103 is also used to send an image recording instruction to the image acquisition module when it is determined that the train has stopped in the image acquisition area based on radar signals and magnetic steel signals; and to send an image continuation acquisition instruction to the image acquisition module based on the carriage information of the already acquired images when it is determined that the train has started in the image acquisition area based on radar signals. The image acquisition module 104 is used to acquire images of the carriages that pass through the image acquisition area in sequence based on the image acquisition command, stop acquiring images of the train based on the image recording command and record the carriage information of the acquired images, and acquire images of the remaining carriages of the train based on the image continuation acquisition command. The controller 103 is also used to stitch together the images of each carriage to obtain a train image.

[0024] In this embodiment of the invention, the provided low-speed train image acquisition system is a system for acquiring images of trains traveling at low speeds. Low speed generally refers to a speed of 5 km / h or less, typically when the train is stopping at a station. The image acquisition area refers to the area that the train needs to pass through at low speed to enter the station, or it can be a portion of the area where the train is stopping. Optionally, a radar is deployed in or near the image acquisition area to measure the train's speed. Magnets are installed on the track; multiple magnets are deployed in and around the image acquisition area to determine whether a vehicle is passing by, and simultaneously, the magnets can be used for speed measurement.

[0025] In this embodiment of the invention, serial port data from a millimeter-wave radar, including target distance, radial velocity, and azimuth information; digital level trigger signals from a magnetic sensor array; and analog distance signals from an ultrasonic sensor are acquired. The data is timestamped and preprocessed at a frequency of at least 10Hz. When the millimeter-wave radar continuously detects an object with a speed greater than 0.5 km / h within 10 meters ahead and magnet 1 is not triggered, a train approach warning is established. When magnets 1 and 2 are triggered, and the radar speed signal matches the speed measured by magnets 1 and 2, the train is considered to have entered the warning state, and image acquisition of the train carriages begins. The specific method for determining the image acquisition time of each carriage based on the magnetic signals will be described in detail later in this invention.

[0026] In this embodiment of the invention, when radar signals detect that the train speed has dropped to 0 km / h for more than 30 seconds and the magnetic steel signal is no longer triggered, the train is determined to be in a stopped state, and images of the carriages are no longer collected. Simultaneously, the carriage information of the already collected images is recorded for subsequent continuous acquisition of carriage images. The carriage information includes at least the number of carriages already captured, and the characteristics of the last captured carriage, such as its car number, car type, or position relative to the acquisition point.

[0027] In this embodiment of the invention, the radar first detects that a stationary target has begun to move, with its speed increasing from 0. At this time, since the train is starting slowly, the first wheel may not have yet run over the magnet, so there is no magnet signal output. Based solely on the radar signal, it can be determined that the train has started moving, and accurate acquisition of carriage images begins. The previously recorded number of carriages, N, is read, indicating that the Nth carriage has been photographed, and the next carriage to be photographed is the (N+1)th carriage. Initially, due to the possible temporary absence of the magnet signal, the system can briefly switch to a radar-based speed measurement mode. Based on the real-time radar speed and the known carriage length, the time it takes for the (N+1)th carriage to reach the photographing point is estimated, and acquisition is triggered. As the train speed increases and the magnet resumes generating a regular pulse sequence, the program can switch back to the more precise magnet triggering mode to continue acquiring images of carriages N+2, N+3, and so on, until the rear of the train completely leaves the image acquisition area and radar detection range. After the entire train has passed, all the acquired carriage images are automatically stitched together in the acquisition order to generate a complete and comprehensive image of the entire train length for subsequent system use.

[0028] The low-speed train image acquisition system provided by this invention includes a radar module, a magnetic sensor module, an image acquisition module, and a controller. The radar module collects radar signals as the train passes and sends these signals to the controller. The magnetic sensor module collects magnetic signals as the train passes and sends these signals to the controller. When the radar and magnetic signals indicate that the train is moving and has entered the image acquisition area, the system determines the image acquisition time for each carriage based on the magnetic signals. Based on these acquisition times, images are acquired sequentially from the carriages that pass through the image acquisition area. By using both radar and magnetic signals to determine whether the train is moving and has entered the image acquisition area, the system provides more accurate judgment of train arrival at the station. This prevents the first carriage from being missed due to a single magnetic signal. Determining the acquisition time for each carriage using magnetic signals allows for precise location of each carriage, enabling image acquisition at the exact moment and ensuring the accuracy of the image acquisition. When the radar and magnetic signals determine that the train has stopped within the image acquisition area, image acquisition of the train is stopped, and the carriage information of the acquired images is recorded. When the radar signals determine that the train has started moving within the image acquisition area, images of the remaining carriages of the train are acquired based on the carriage information of the acquired images, and the acquired carriage images of each carriage are stitched together to obtain the train image. When it is determined that the train has stopped in the station, the acquisition of carriage images is stopped, and the corresponding information is recorded. When the train is detected to start moving again, images of subsequent carriages are acquired based on the recorded information. This effectively prevents discontinuity in carriage images and ensures the integrity of carriage image acquisition.

[0029] In some possible embodiments of the present invention, such as Figure 2As shown, when the controller determines that the train is moving and has entered the image acquisition area based on radar and magnetic signals, it is used to: S201, radar speed measurement based on radar signals to determine train speed; S202, the train's magnetic speed measurement is determined based on the time interval between the first magnetic signal of the train passing through the first magnetic signal initially collected by the first magnetic signal deployed in the image acquisition area and the second magnetic signal of the train passing through the second magnetic signal initially collected by the second magnetic signal deployed in the image acquisition area, and the deployment distance between the first magnetic signal and the second magnetic signal. S203: When the difference between radar speed measurement and magnetic speed measurement is within a preset range, the train is determined to be in a moving state and has entered the image acquisition area.

[0030] In this embodiment of the invention, based on the continuous signal transmitted by the radar installed in front of the image acquisition area, the radial velocity of the target relative to the radar is calculated in real time or directly read as the radar speed measurement of the train. The radar provides a continuous and direct measurement of the target's motion state. At the starting position of the image acquisition area, a first magnet and a second magnet are arranged at intervals along the track direction, with a fixed distance between them. When the train head passes the first magnet and the second magnet in sequence, a first magnet signal and a second magnet signal are generated respectively. The time interval between these two signals is recorded. According to the speed formula, the average speed of the train passing through this fixed baseline is calculated as the train's magnet speed measurement. This speed is based on physical contact triggering and has extremely high instantaneous accuracy and anti-electromagnetic interference capability. The difference between the radar speed measurement and the magnet speed measurement is calculated, and the absolute value of this difference is compared with a preset threshold range. If and only if the difference falls within the preset range, it is determined that the following two conditions are met: the moving target detected by the radar and the object that triggered the magnet are the same entity, i.e., the train, thus eliminating the possibility of the radar mistakenly tracking other targets. The entity is in a stable state of motion, and its speed has been verified by measurements based on two different principles, confirming the reliability of the results. Therefore, it is ultimately determined that the train is moving and has entered the image acquisition area.

[0031] Furthermore, the preset range can be set according to the actual application scenario. For example, it can be set so that the absolute value of the difference between the two speed measurements does not exceed 0.5 km / h, or the relative error does not exceed 10%. The setting of this threshold needs to take into account factors such as the speed measurement error of the radar, the measurement accuracy of the magnet installation distance, and the timing accuracy of the time interval. Its fundamental purpose is to allow reasonable sensor error while effectively filtering out abnormal or mismatched readings.

[0032] The embodiments of the present invention use cross-verification of information from two heterogeneous sensors, radar and magnet, to form a simple AND logic. The system only confirms the train's entry when both sensors simultaneously and in a coordinated manner indicate the same motion event. This fundamentally avoids false starts or missed starts caused by a single sensor failure or interference, and significantly improves the robustness of the system.

[0033] In some possible embodiments of the present invention, when the controller determines the image acquisition time of each carriage of the train based on the magnetic steel signal, it is used to: The image acquisition time of each carriage of the train is determined based on the acquisition time of the signal of the third magnet that each carriage passes by the third magnet, which is deployed in the image acquisition area. The third magnet is after the first magnet and the second magnet.

[0034] In this embodiment of the invention, a third magnet is also provided in the image acquisition area. This third magnet follows the first and second magnets and is specifically used to determine the image acquisition time for each carriage. Specifically, when a train wheel passes over the third magnet, a third magnetic signal is generated. Based on this third magnet signal, a precise segmented trigger pulse signal is generated. This segmented trigger pulse signal controls the image acquisition module to acquire images of each carriage, thus achieving precise image acquisition for each carriage.

[0035] This invention uses specialized magnets to determine the passing time of each train carriage, thereby determining the image acquisition time for each carriage and ensuring the accuracy of image acquisition.

[0036] In some possible embodiments of the present invention, such as Figure 3 As shown, when the controller determines the image acquisition time for each carriage of the train based on the magnetic steel signal, and sends the image acquisition command to the image acquisition module based on the image acquisition time, it is used for: S301 calculates the real-time speed of each carriage as it passes through the image acquisition area based on radar and magnetic signals. S302 determines the image acquisition line frequency of the acquisition carriage based on the real-time vehicle speed and preset calibration coefficients, and sends an image acquisition command to the image acquisition module. The image acquisition command includes the time for image acquisition and the image acquisition line frequency.

[0037] In this embodiment of the invention, to ensure the accuracy of carriage image acquisition and prevent image distortion, the real-time train speed at each image acquisition point is continuously calculated based on radar and magnetic signals throughout the entire process of the train passing through the image acquisition area. While radar provides continuous, real-time speed measurement data, its instantaneous data may be affected by noise. A discrete, high-precision instantaneous speed can be calculated using the time interval between two adjacent magnetic pulses. Through a data fusion algorithm, the high-frequency but potentially noisy radar speed measurement is combined with the low-frequency but high-precision magnetic speed measurement to obtain a smooth, accurate, and real-time updated train speed curve. This speed corresponds to the instantaneous train speed at the center line position of the image acquisition area. A calibration coefficient K is preset, obtained through calibration during system installation and debugging. Its physical meaning is the proportional relationship between the camera frequency F and the train speed V while ensuring the image target achieves the expected pixel resolution in the scanning direction, i.e., F = K × V. The expected pixel resolution can be set according to actual detection requirements. Therefore, when acquiring each carriage, the system dynamically calculates the required image acquisition frequency based on the current real-time train speed using a formula. At the image acquisition time determined based on the magnetic steel signal, the image acquisition control system sends the calculated dynamic line frequency to the line scan camera, which then starts acquiring images at this line frequency until it is determined that the acquisition of the carriage is complete.

[0038] This invention ensures that the pixel scale of the acquired image in the direction of travel remains constant relative to the actual physical scale, regardless of the train's speed, by matching the acquisition frequency with the real-time speed of the train. This fundamentally eliminates image stretching or compression caused by speed changes, resulting in consistent and accurate appearance proportions for the images of each carriage, greatly improving the usefulness of the images and the accuracy of subsequent analysis.

[0039] In some possible embodiments of the present invention, such as Figure 4 As shown, when the controller sends an image continuation acquisition command to the image acquisition module based on the already acquired image carriage information, it is used for: S401, based on the carriage number of the last carriage in the acquired images, the number of axles that the last carriage has passed through the image acquisition area, and the real-time speed of the train, determine the image acquisition time of the first carriage among the remaining carriages; S402 determines the image acquisition time of the remaining carriages based on the image acquisition time of the first carriage and the magnetic signal of the remaining carriages passing through the image acquisition area, and sends an image continuation acquisition command to the image acquisition module. The image continuation acquisition command includes the image acquisition time of the remaining carriages.

[0040] In this embodiment of the invention, when the train stops, the acquisition of carriage images ceases. However, at this point, only images of the front carriages may have been acquired, and images of the rear carriages still need to be acquired. When the train starts moving, based on the carriage information recorded in the aforementioned embodiment, the image acquisition time for the first carriage requiring image acquisition after the train's start can be determined, and the image acquisition time for each subsequent carriage can be determined based on this acquisition time. Specifically, the carriage number of the last carriage already acquired needs to be obtained first. This information is obtained from the acquired images through image recognition, or it is provided separately by the carriage number recognition system and bound to the acquisition sequence when the train enters. It is the unique identifier that determines the acquired boundary. The system continuously counts the number of axles that the last carriage has passed through the image acquisition area before the train stops, using magnetic signals. For example, for a four-axle carriage, if the record shows that two axles have passed, it means that part of the carriage has not yet completely passed through the acquisition area. The current real-time speed is obtained through radar or a radar / magnetic fusion algorithm when the train starts moving.

[0041] Furthermore, based on the carriage number, the total number of axles and wheelbase layout of the carriage can be determined by querying the vehicle model database or using known rules. Combining this with the number of axles that have already passed, the theoretical distance corresponding to the remaining portion of the last carriage that has not yet passed the acquisition point can be calculated. For example, if the wheelbase of each pair is known, the lengths between the pairs of axles that have not yet passed can be accumulated. Based on the real-time vehicle speed at startup, the time required for the remaining portion of the last carriage to completely pass the acquisition point can be estimated. Therefore, the predicted start time for image acquisition of the remaining carriages can be calculated by adding the time required for the remaining portion of the last carriage to completely pass the acquisition point, starting from the current moment.

[0042] Then, the system monitors the nearest magnet signal after the predicted start time. When the first valid magnet pulse is received, the system compares the actual pulse time with the predicted time. If it is within the allowable error range, the accuracy of the prediction is verified; if there is a deviation, the system can immediately use this actual signal to fine-tune subsequent predictions. Using this first magnet pulse from the remaining carriages as the new reference point, the system resumes the normal acquisition logic as described in the previous embodiment. That is, based on this pulse and its subsequent pulse sequence, combined with the continuously updated real-time train speed, the system dynamically determines the precise image acquisition time and frequency for each subsequent remaining carriage until the rear of the train has completely passed, completing the image acquisition of all remaining carriages.

[0043] This invention integrates three key pieces of information: carriage number, number of axles passed, and real-time train speed. It can intelligently calculate the precise starting point for subsequent image acquisition, effectively distinguishing between the rear half of the carriage that has already been acquired and the next carriage that has not yet been acquired. This fundamentally avoids duplicate and missed acquisitions, achieving seamless connection both logically and physically. The acquired image sequence is continuous, complete, and in the correct order, providing a perfect data foundation for stitching together a complete train image.

[0044] In some possible embodiments of the present invention, such as Figure 5 As shown, when the controller stitches together the images of each carriage to obtain the train image, it is used for: S501 uses a deep learning model to extract the foreground mask from the carriage image; S502, the foreground mask is fused with the preset background model to obtain the foreground image of the carriage; S503: The foreground images of each carriage are stitched together according to the carriage number to obtain the train image.

[0045] In this embodiment of the invention, each acquired original train carriage image is input into a pre-trained deep learning segmentation model, such as a semantic segmentation model based on architectures like U-Net or DeepLabV3+. The training data for this model consists of a large number of train images acquired at different times, in different weather conditions, and under different lighting conditions, along with manually annotated pixel-level foreground carriage masks. This deep learning model can understand the abstract visual features of the train carriages, rather than simple color or brightness differences. It can accurately determine at the pixel level whether each position in the image belongs to the foreground or background of the carriage, and output a binarized or probabilistic foreground mask with the same resolution as the input image. In this mask, pixels belonging to the carriage are marked as white, and background pixels are marked as black. Multiple pure background images are acquired, covering different lighting periods, and a stable and pure preset background model image is generated using statistical methods, such as mean, median, or more complex Gaussian mixture models. This model represents the background of the acquisition area under ideal conditions without train interference. For each original image of the train carriage to be processed, the following fusion process is performed: Foreground extraction: Using the foreground mask obtained in step one as the selection area, the pixel region covered by the mask is precisely extracted from the original train carriage image. This part constitutes the initial pure foreground pixel set. Background restoration: Using the corresponding preset background model image as a reference, detailed fusion processing is performed on the mask edge region. For example, pixels at the foreground edge in the original image can be fused with corresponding pixels in the background model to eliminate jagged edges that may be caused by insufficient smoothness of the segmentation model boundaries, making the transition between the foreground and background more natural. This step ensures the edge quality and visual consistency of the final foreground image.

[0046] Furthermore, based on the carriage numbers recorded and bound to each image during the acquisition process, all extracted foreground images are arranged in numerical order. Following the train's direction of travel, the sorted foreground images are then stitched together end-to-end. Since each foreground image has had its cluttered background removed and its edges optimized, simple alignment is all that's needed to generate a complete panoramic image of the train with a clean background, intact carriages, and a correct sequence. In the stitched image, the train appears as a coherent whole against a uniform, clean background.

[0047] The deep learning model in this invention can penetrate complex environmental interference and directly identify the semantic information of the carriage. Its segmentation accuracy far exceeds that of traditional algorithms based on color, brightness, or simple motion, effectively avoiding background interference and foreground loss.

[0048] In some possible embodiments of the present invention, such as Figure 6 As shown, when the controller stitches together the foreground images of each carriage according to their carriage numbers to obtain the train image, it is used for: S601, determine the acquisition time of the carriage image corresponding to the foreground image, and determine the geometric distortion correction parameters of the foreground image based on the train speed corresponding to the acquisition time and the average train speed in the entire image acquisition area. S602, the foreground image is corrected based on geometric distortion correction parameters to obtain the corrected foreground image of the carriage; S603, the corrected foreground images of each carriage are stitched together according to the carriage number to obtain the train image.

[0049] In this embodiment of the invention, for each extracted foreground image to be stitched, the precise acquisition time of the original carriage image corresponding to the foreground image is obtained. Based on the acquisition time, the real-time train speed at that time is determined, and the average train speed from the first carriage to the last carriage across the entire image acquisition area is obtained. This average speed can be obtained by dividing the total train length by the total travel time, or by the average value calculated from radar and magnetic signals throughout the process. The core idea of ​​calculating the distortion correction parameter is to uniformly correct the current image from its instantaneous speed scale at the time of acquisition to an ideal scale based on the average speed of the entire train. For example, if the real-time train speed is greater than the average train speed, it means that the train speed was too fast when the carriage was acquired, and theoretically the image may be slightly compressed. In this case, the geometric distortion correction parameter is less than 1, and the image needs to be stretched along the train's direction of travel during correction. If the real-time train speed is less than the average train speed, it means that the speed was too slow when the acquisition was made, and the image may be slightly stretched. In this case, the geometric distortion correction parameter is greater than 1, and the image needs to be compressed along the train's direction of travel during correction. If the two are equal, no correction is needed. The specific value of the geometric distortion correction parameter can be determined according to the actual situation.

[0050] Furthermore, using the calculated geometric distortion correction parameters, the corresponding carriage foreground images are geometrically transformed to generate corrected foreground images. Specifically, a one-dimensional scaling transformation is mainly performed, scaling the image width by a preset factor along the train's direction of travel, while the image height remains unchanged, generally unaffected by train speed. During the scaling process, a high-quality resampling interpolation algorithm is used to maintain the clarity of carriage textures and edges in the image. This results in corrected foreground images with geometric scales unified to the average speed reference, where the actual physical size represented by each pixel is consistent across all carriages. All carriage corrected foreground images are then stitched together along the train's direction of travel according to their carriage numbers. Since the geometric scales of each image have been uniformly corrected, simple alignment is sufficient during stitching to achieve seamless and precise joining, ultimately resulting in a complete train image with accurate geometric dimensions and visual coherence.

[0051] The embodiments of the present invention eliminate image scale errors caused by instantaneous speed differences through precise speed-based correction, enabling perfect alignment of features at the seam of adjacent carriage images, achieving a near-seamless stitching effect.

[0052] In some possible embodiments of the present invention, the magnetic sensor module includes multiple magnetic sensors, which are distributed sequentially on the rails of the image acquisition area at preset intervals.

[0053] In this embodiment of the invention, the magnetic sensor module is a magnetic sensor array, which includes multiple magnetic sensors that are sequentially distributed on the rails in the image acquisition area at preset intervals.

[0054] Furthermore, the radar module includes multiple millimeter-wave radars, each arranged on both sides of the track at a preset height and angle.

[0055] In this embodiment of the invention, the radar module includes multiple millimeter-wave radars, each millimeter-wave radar is arranged on both sides of the track according to a preset height and angle. For example, the height of the millimeter-wave radar above the ground is 1.8 meters, the distance from the track is about 2.5 meters, and the angle of the millimeter-wave radar is facing the direction of the train.

[0056] This invention also provides a method for acquiring images of low-speed trains, such as... Figure 7 As shown, the method includes: S701, when the train is determined to be in a moving state and enters the image acquisition area based on radar signals and magnetic steel signals, the image acquisition time of each carriage of the train is determined based on the magnetic steel signals, and the images of the carriages that pass through the image acquisition area in sequence are acquired based on the image acquisition time. S702: When it is determined that the train has stopped within the image acquisition area based on radar and magnetic signals, image acquisition of the train is stopped, and the carriage information of the acquired images is recorded. S703: When the radar signal determines that the train has started in the image acquisition area, the remaining carriages of the train are acquired based on the carriage information of the already acquired images, and the acquired carriage images are stitched together to obtain the train image.

[0057] The low-speed train image acquisition method provided in this embodiment can realize the image acquisition of low-speed trains based on the low-speed train image acquisition system in any of the foregoing embodiments. The specific principle has been explained in the foregoing embodiments and will not be repeated here.

[0058] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A low-speed train image acquisition system, characterized in that, It includes a radar module, a magnetic sensor module, an image acquisition module, and a controller, among which, The radar module is used to collect radar signals when the train passes by, and the magnet sensor module is used to collect magnet signals when the train passes by. The controller is used to determine the image acquisition time of each carriage of the train based on the magnetic signal when the train is determined to be in a moving state and enters the image acquisition area based on the radar signal and the magnetic signal, and to send an image acquisition command to the image acquisition module based on the image acquisition time. The controller is also used to send an image recording instruction to the image acquisition module when it is determined, based on radar signals and magnetic steel signals, that the train has stopped in the image acquisition area, and to send an image continuation acquisition instruction to the image acquisition module based on the carriage information of the already acquired images when it is determined, based on radar signals, that the train has started in the image acquisition area. The image acquisition module is used to acquire images of carriages that pass through the image acquisition area in sequence based on the image acquisition command, stop acquiring images of the train based on the image recording command, record the carriage information of the acquired images, and acquire images of the remaining carriages of the train based on the image continuation acquisition command. The controller is also used to stitch together the images of each carriage to obtain a train image.

2. The low-speed train image acquisition system according to claim 1, characterized in that, When the controller determines that the train is moving and has entered the image acquisition area based on radar and magnetic signals, it is used to: Radar speed measurement is used to determine the train's speed based on radar signals; The train's magnetic speed measurement is determined based on the time interval between the first magnetic signal of the train passing through the image acquisition area (initial acquisition by the first magnetic signal of the train passing through the image acquisition area) and the second magnetic signal of the train passing through the image acquisition area (initial acquisition by the second magnetic signal of the train passing through the image acquisition area), and the distance between the first and second magnetic signals. When the difference between the radar speed measurement and the magnet speed measurement is within a preset range, the train is determined to be in a moving state and has entered the image acquisition area.

3. The low-speed train image acquisition system according to claim 2, characterized in that, When determining the image acquisition time for each carriage of the train based on the magnetic steel signal, the controller is used to: The image acquisition time of each carriage of the train is determined based on the acquisition time of the third magnet signal of each carriage passing the third magnet, which is deployed in the image acquisition area, wherein the third magnet is after the first magnet and the second magnet.

4. The low-speed train image acquisition system according to claim 1, characterized in that, When the controller determines the image acquisition time for each carriage of the train based on the magnetic steel signal and sends an image acquisition command to the image acquisition module based on the image acquisition time, it is used to: The real-time speed of each carriage as it passes through the image acquisition area is calculated based on the radar signal and the magnet signal. Based on the real-time vehicle speed and preset calibration coefficients, the image acquisition line frequency for the carriage is determined, and an image acquisition command is sent to the image acquisition module. The image acquisition command includes the time for image acquisition and the image acquisition line frequency.

5. The low-speed train image acquisition system according to claim 4, characterized in that, When the controller sends an image continuation acquisition command to the image acquisition module based on the carriage information of the already acquired images, it is used to: The image acquisition time of the first carriage among the remaining carriages is determined based on the carriage number of the last carriage that has been captured in the image, the number of axles that the last carriage has passed through the image acquisition area, and the real-time speed of the train. Based on the image acquisition time of the first carriage and the magnetic signal of the remaining carriage passing through the image acquisition area, the image acquisition time of the remaining carriage is determined, and an image continuation acquisition command is sent to the image acquisition module. The image continuation acquisition command includes the image acquisition time of the remaining carriage.

6. The low-speed train image acquisition system according to claim 1, characterized in that, When the controller stitches together the images of each carriage to obtain a train image, it is used to: A deep learning model is used to extract the foreground mask from the carriage image; The foreground mask is fused with a preset background model to obtain the foreground image of the carriage; The foreground images of each carriage are stitched together according to their carriage numbers to obtain the train image.

7. The low-speed train image acquisition system according to claim 6, characterized in that, When the controller stitches together the foreground images of each carriage according to their carriage numbers to obtain a train image, it is used for: Determine the acquisition time of the carriage image corresponding to the foreground image, and determine the geometric distortion correction parameters of the foreground image based on the train speed corresponding to the acquisition time and the average train speed in the entire image acquisition area; The foreground image is corrected based on the geometric distortion correction parameters to obtain the corrected foreground image of the carriage; The corrected foreground images of each carriage are stitched together according to the carriage number to obtain the train image.

8. The low-speed train image acquisition system according to claim 1, characterized in that, The magnetic sensor module includes multiple magnetic sensors, which are distributed sequentially on the rails in the image acquisition area at preset intervals.

9. The low-speed train image acquisition system according to claim 1, characterized in that, The radar module includes multiple millimeter-wave radars, each of which is arranged on both sides of the track at a preset height and angle.

10. A method for acquiring images of low-speed trains, characterized in that, include: When the train is determined to be moving and enters the image acquisition area based on radar and magnetic signals, the image acquisition time of each carriage of the train is determined based on the magnetic signals, and the images of the carriages that pass through the image acquisition area in sequence are acquired based on the image acquisition time. When it is determined based on radar and magnetic signals that the train has stopped within the image acquisition area, image acquisition of the train is stopped, and the carriage information of the acquired images is recorded. When the radar signal determines that the train has started in the image acquisition area, the remaining carriages of the train are imaged based on the carriage information of the already acquired images, and the images of each carriage are stitched together to obtain the train image.

Images