Method and system for monitoring snow depth data of high-speed rail

Abstract

The invention provides a high-speed rail track snow depth data monitoring method and system, and relates to the technical field of railway safety, and the method comprises the steps: obtaining a first space angle image and a second space angle image; obtaining snowflake landing parameters; obtaining a balance thickness; obtaining a first snow depth monitoring result; acquiring snow depth sensing data; performing data conversion on the snow depth sensing data after the abnormal data is removed to obtain a second snow depth monitoring result; and carrying out confidence analysis on the first snow depth monitoring result and the second snow depth monitoring result, and carrying out weighted calculation on the first snow depth monitoring result and the second snow depth monitoring result according to the confidence to obtain a snow depth monitoring result, thereby solving the problems that the monitoring equipment is relatively single and the snow depth monitoring result cannot be monitored in the prior art. The technical problem that the accuracy of the snow depth detection result is insufficient due to the fact that the analysis of the monitoring data is not detailed enough is solved, and the technical effects of improving the accuracy of the snow depth monitoring result and improving the monitoring precision are achieved.

Description

Technical Field

[0001] This application relates to the field of railway safety technology, specifically to a method and system for monitoring snow depth data on high-speed railway tracks. Background Technology

[0002] After snowfall, thick snow often accumulates on high-speed rail tracks. If high-speed trains continue to run at their original speed, the probability of traffic accidents will increase significantly, seriously threatening passenger safety. Therefore, accurate monitoring of snow depth is of great importance.

[0003] Traditional snow depth monitoring uses laser measurement, but laser measurement is greatly affected by changes in light and external environment. It often results in situations where there is no actual snowfall but the sensor collects a snow depth value that is not 0. Therefore, existing technologies suffer from the technical problem of insufficient accuracy in snow depth detection results due to the relatively simple monitoring equipment and the lack of detailed analysis of monitoring data. Summary of the Invention

[0004] This application provides a method and system for monitoring snow depth data on high-speed railway tracks, which solves the technical problem that the accuracy of snow depth detection results is insufficient due to the relatively simple monitoring equipment and the lack of detailed analysis of monitoring data in the prior art.

[0005] According to a first aspect of this application, a method for monitoring snow depth data on high-speed railway tracks is provided, comprising: acquiring images of a first spatial angle and a second spatial angle of a monitoring point on a target high-speed railway line using an image sensor, thereby obtaining a first spatial angle image and a second spatial angle image, wherein the first spatial angle is a spatial angle perpendicular to the track, and the second spatial angle is parallel to the track plane; identifying snowflake unit size and snowflake falling speed based on the first spatial angle image to obtain snowfall falling parameters; performing snowfall thickness uniformity analysis within the track using the second spatial angle image to obtain a uniform thickness; performing snow depth analysis using the snowflake falling parameters and the uniform thickness to obtain a first snow depth monitoring result; monitoring snow depth at the monitoring point using a laser displacement sensor to obtain snow depth sensing data; performing anomaly analysis on the snow depth sensing data, removing abnormal data, and performing data transformation on the snow depth sensing data after removing abnormal data to obtain a second snow depth monitoring result; performing confidence analysis on the first snow depth monitoring result and the second snow depth monitoring result, and performing a weighted calculation on the first snow depth monitoring result and the second snow depth monitoring result using the confidence level to obtain a snow depth monitoring result.

[0006] According to a second aspect of this application, a snow depth data monitoring system for high-speed railway tracks is provided, comprising: an image acquisition module, wherein the image acquisition module is used to acquire images of a monitoring point on a target high-speed railway line at a first spatial angle and a second spatial angle respectively using an image sensor, to obtain a first spatial angle image and a second spatial angle image, wherein the first spatial angle is a spatial angle perpendicular to the track, and the second spatial angle is parallel to the track plane; a first image analysis module, wherein the first image analysis module is used to identify snowflake unit size and snowflake falling speed based on the first spatial angle image to obtain snowflake falling parameters; and a second image analysis module, wherein the second image analysis module is used to perform snowfall thickness uniformity analysis within the track based on the second spatial angle image to obtain uniform thickness; and a first snow depth data acquisition module. The system comprises: a first snow depth monitoring module, which analyzes snow depth using the snowflake falling parameters and the uniform thickness to obtain a first snow depth monitoring result; a laser monitoring module, which monitors snow depth at the monitoring point using a laser displacement sensor to obtain snow depth sensing data; a second snow depth monitoring module, which performs anomaly analysis on the snow depth sensing data, removes abnormal data, and performs data transformation on the snow depth sensing data after removing abnormal data to obtain a second snow depth monitoring result; and a confidence analysis module, which performs confidence analysis on the first and second snow depth monitoring results, and performs a weighted calculation on the first and second snow depth monitoring results using confidence levels to obtain a final snow depth monitoring result.

[0007] The beneficial effects that can be achieved by adopting one or more technical solutions in this application are as follows: Image sensors are used to acquire images of monitoring points along the target high-speed railway line from a first spatial angle and a second spatial angle, respectively. The first spatial angle is perpendicular to the track, while the second spatial angle is parallel to the track plane. Based on the first spatial angle image, snowflake unit size and snowfall speed are identified to obtain snowfall parameters. The second spatial angle image is used to analyze the snowfall thickness uniformity within the track, obtaining the uniform thickness. Snow depth analysis is then performed using the snowfall parameters and uniform thickness to obtain the first snow depth monitoring result. A laser displacement sensor is used to monitor snow depth at the monitoring points, acquiring snow depth sensing data. Anomaly analysis is performed on the snow depth sensing data, and outliers are removed. The snow depth sensing data after anomaly removal is then transformed to obtain the second snow depth monitoring result. Confidence analysis is performed on the first and second snow depth monitoring results, and a weighted calculation is performed based on the confidence levels to obtain the final snow depth monitoring result. Therefore, by using image sensors and laser displacement sensors for joint snow depth analysis, the accuracy and precision of snow depth monitoring are improved. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in this application or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. The accompanying drawings, which constitute a part of this application, are used to provide a further understanding of this application. The illustrative embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0009] Figure 1 A flowchart illustrating the method for monitoring snow depth data on high-speed railway tracks, provided in an embodiment of this application; Figure 2 This is a schematic diagram of the structure of a snow depth data monitoring system for high-speed railway tracks, provided in an embodiment of this application.

[0010] Explanation of reference numerals in the attached figures: Image acquisition module 11, First image analysis module 12, Second image analysis module 13, First snow depth monitoring module 14, Laser monitoring module 15, Second snow depth monitoring module 16, Confidence analysis module 17. Detailed Implementation

[0011] To make the objectives, technical solutions, and advantages of this application more apparent, exemplary embodiments of this application will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0012] The terminology used in this specification is for describing embodiments and not for limiting the application. As used in the specification, the singular terms “a,” “an,” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. When used in the specification, the terms “comprising” and / or “including” specify the presence of a step, operation, element, and / or component, but do not preclude the presence or addition of one or more other steps, operations, elements, components, and / or groups thereof.

[0013] Unless otherwise defined, all terms used in this specification (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. Terms, as defined in common dictionaries, shall not be interpreted in an idealized or overly formal sense unless expressly defined herein. Throughout this specification, the same reference numerals denote the same elements.

[0014] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for display, data used for analysis, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0015] Example 1 Figure 1 This application provides a method for monitoring snow depth data on high-speed railway tracks, the method comprising: The first spatial angle and the second spatial angle of the monitoring point of the target high-speed railway line are captured by the image sensor to obtain the first spatial angle image and the second spatial angle image, wherein the first spatial angle is the spatial angle perpendicular to the track and the second spatial angle is parallel to the track plane. The image sensor refers to existing high-precision cameras, drones, smart cameras, and other devices used for image acquisition. The target high-speed rail line refers to the line corresponding to the high-speed rail track for which snow depth data monitoring is to be conducted. The monitoring point refers to the sampling point set up based on existing technology for snow depth monitoring. For example, the monitoring points can be selected and divided based on the different terrain and meteorological environment of the target high-speed rail line, so that one monitoring point can represent the snow depth monitoring results of a section of the line. The image sensor acquires images of the monitoring points of the target high-speed rail line at a first spatial angle and a second spatial angle, respectively, to obtain a first spatial angle image and a second spatial angle image. The first spatial angle is a spatial angle perpendicular to the track, and the second spatial angle is parallel to the track plane. In simple terms, the first spatial angle and the second spatial angle refer to the plane to which the acquired first spatial angle image and the second spatial angle image belong. Specifically, the first spatial angle refers to the plane located above the track and perpendicular to the track plane, that is, the image of the space where snowflakes fall is acquired, and the first spatial angle image is the image of the snowflake falling process. The second spatial angle is parallel to the track plane, that is, the image of the track after the snowflakes fall into the track is acquired as the second spatial angle image.

[0016] Based on the first spatial angle image, snowflake unit size and snowflake falling speed are identified to obtain snowflake falling parameters; Based on the first spatial angle image, snowflake unit size and snowflake falling speed are identified to obtain snowflake falling parameters. Understandably, snowflake falling parameters include snowflake unit size and snowflake falling speed. The specific acquisition process is detailed below.

[0017] In a preferred embodiment, it further includes: The first spatial angle image is segmented using a snowflake segmentation channel to obtain multiple segmented images. The snowflake segmentation channel is a fully convolutional neural network model, including an encoder and a decoder, which are constructed through training on snowflake segmentation samples. Snowflake contour recognition is performed on the multiple segmented images to obtain multiple contour recognition results. The acquisition parameters of the image sensor are obtained, and the snowflake unit size is identified using the acquisition parameters and the multiple contour recognition results. Snowflake position recognition analysis is performed on the first spatial angle image to obtain the snowflake falling speed. The snowflake falling parameters are composed of the snowflake unit size and the snowflake falling speed.

[0018] In a preferred embodiment, it further includes: Snowflake position recognition is performed on the first time-series image in the first spatial angle image to obtain a first position identification result; snowflake position recognition is performed on the second time-series image in the first spatial angle image to obtain a second position identification result, wherein the first time-series image and the second time-series image are snowflake images acquired in adjacent consecutive time periods; the distance between the first position identification result and the second position identification result is calculated in the vertical direction to obtain a first vertical adjacent distance; the snowflake falling speed is calculated based on the first vertical adjacent distance.

[0019] The first spatial angle image is segmented using a snowflake segmentation channel to obtain multiple segmented images. The snowflake segmentation channel is a fully convolutional neural network model, including an encoder and a decoder. The encoder and decoder are constructed through training on snowflake segmentation samples, which are snowflake falling image samples acquired using existing techniques, and image segmentation samples where the snowflake positions in the snowflake falling image samples are marked and segmented. Using the snowflake falling image samples as input to the fully convolutional neural network model (encoder and decoder), and using the image segmentation samples to supervise and adjust the output, a snowflake segmentation channel trained to convergence is obtained. The first spatial angle image is input into the snowflake segmentation channel for image segmentation to obtain multiple segmented images. Snowflake contour recognition is performed on the multiple segmented images to obtain multiple contour recognition results. That is, edge recognition is performed on the snowflakes in each segmented image using existing edge recognition techniques to obtain snowflake contours as multiple contour recognition results. Further, the acquisition parameters of the image sensor are obtained, including parameters such as the image sensor's pixels and focal length. The snowflake unit size is then identified using these acquisition parameters and the multiple contour recognition results. Simply put, based on the multiple contour recognition results, the unit size of the snowflake in the image can be calculated, such as its area and thickness. However, when the image sensor captures an image, it scales the target, meaning the size in the image is not the actual size. The size of the acquired image is related to pixels, focal length, etc. The relationship between the acquisition parameters and the image scaling ratio can be read from the image sensor's user manual. The acquisition parameters are also read from the image sensor's control terminal. Then, based on the relationship between the acquisition parameters and the image scaling ratio, and the unit size of the snowflake in the image, the actual snowflake unit size is obtained as the snowflake unit size.

[0020] Further analysis of snowflake positions in the first spatial angle image is performed to obtain the snowflake falling speed. The specific process is as follows: Snowflake positions are identified in the first time-series image within the first spatial angle image to obtain a first position identification result. That is, the first spatial angle image contains multiple images acquired in consecutive time intervals. The first time-series image refers to any one of these images, and the second time-series image is an image acquired at a time adjacent to the first time-series image. Snowflake position identification in the first time-series image is performed using existing target recognition methods to identify and mark the snowflake positions in the first time-series image, obtaining the first position identification result. Target recognition is a common technique used by those skilled in the art, and therefore will not be elaborated upon here. Snowflake positions are further identified in the second time-series image within the first spatial angle image to obtain a second position identification result. Here, the first time-series image and the second time-series image are snowflake images acquired in consecutive time intervals, meaning that the first position identification result and the second position identification result have a corresponding relationship. That is, the same snowflake has different positions in the two images. The corresponding identification positions are obtained from the first and second position identification results, and the vertical distance between the corresponding identification positions is calculated to obtain the first vertical adjacent distance. The snowflake falling speed is calculated based on the first vertical adjacent distance, that is, by dividing the first vertical adjacent distance by the time difference between the first time series image and the second time series image. The result is the snowflake falling speed. This allows for analysis of the snowflake falling space, providing support for subsequent snow depth monitoring and improving the accuracy of snow depth monitoring.

[0021] Finally, the snowflake unit size and the snowflake falling speed are used to form the snowflake falling parameters, which provide support for subsequent snow depth monitoring and improve the accuracy of snow depth monitoring.

[0022] The snowfall thickness uniformity analysis within the track is performed using the second spatial angle image to obtain the uniform thickness; Using the second spatial angle image, a snowfall thickness uniformity analysis is performed within the track to obtain the uniform thickness. This means that after snowflakes fall onto the track, the thickness of the snowfall deposition within the track will deviate due to factors such as the track's temperature, wind direction, and wind speed. The snowfall thickness uniformity analysis analyzes the dispersion of this deviation data, thus facilitating snow depth monitoring. The specific acquisition process is as follows: In a preferred embodiment, it further includes: A first reference line is set, and the first reference line has a first reference thickness identifier; the reference deviation distance of the second spatial angle image is identified at multiple points using the first reference thickness identifier of the first reference line to obtain a set of reference deviation distances; the thickness dispersion analysis of the set of reference deviation distances is performed at multiple points to obtain a dispersion coefficient; if the dispersion coefficients meet a predetermined dispersion threshold, the set of reference deviation distances is averaged to generate the uniform thickness using the averaged result.

[0023] In a preferred embodiment, it further includes: If the dispersion coefficient does not meet the predetermined dispersion threshold, cluster analysis is performed on the benchmark deviation distance set to obtain multiple deviation distance clusters, each with a location identifier; based on the multiple deviation distance clusters, multi-location balanced thickness analysis is performed to obtain the multi-location balanced thickness, which is then used as the balanced thickness.

[0024] A first reference line is established, which has a first reference thickness marker. This first reference line is set by those skilled in the art; it involves randomly selecting a horizontal line position on the track and recording its distance from the lowest plane of the track as the first reference thickness marker, providing a basis for subsequent thickness equalization analysis. The first reference thickness marker of the first reference line is used to identify the reference deviation distance at multiple points in the second spatial angle image, obtaining a set of reference deviation distances. That is, the first reference line and the first reference thickness marker are marked in the second spatial angle image, and then the snowflake coverage thickness at different locations, i.e., the highest plane of the snowflake, is obtained. The distance between the highest plane of the snowflake and the first reference line is extracted to form the set of reference deviation distances. Multi-point thickness dispersion analysis is performed using the set of reference deviation distances. The dispersion coefficient is used to represent the magnitude of the difference in the data within the set of reference deviation distances. Existing methods such as variance and standard deviation can be used for identification, without limitation. A predetermined dispersion threshold is set by those skilled in the art, without limitation. The allowable snow depth error can be obtained based on practical experience as the predetermined dispersion threshold. If the dispersion coefficients meet the predetermined dispersion threshold, the reference deviation distance set is averaged, and the averaged result is used to generate the balanced thickness. That is, an averaged result can be calculated for each second spatial angle image. The averaged result can be understood as the distance between the actual thickness of the snowflake and the first reference line. Based on the first reference thickness identifier, the actual thickness of the snowflake can be obtained through simple mathematical calculation. The ratio of the actual thickness of two consecutive second spatial angle images to the difference in image acquisition time is used as the balanced thickness.

[0025] If the dispersion coefficient does not meet the predetermined dispersion threshold, cluster analysis is performed on the benchmark deviation distance set. This involves aggregating data with deviations within the predetermined dispersion threshold range to obtain multiple deviation distance clusters. These clusters are identified by location markers, indicating the positions corresponding to the distance data within each cluster. This facilitates subsequent monitoring of different locations and improves monitoring accuracy. Based on these multiple deviation distance clusters, a multi-point balanced thickness analysis is performed to obtain the multi-point balanced thickness. This balanced thickness is then used as the average thickness for each deviation distance cluster, yielding the balanced thickness corresponding to each cluster. Therefore, by performing a balanced analysis of snowflake thickness within a monitoring point, it prevents variations in snowflake thickness at different locations due to external environmental factors such as track temperature and wind. This multi-point thickness balanced analysis improves the accuracy of snow depth monitoring results.

[0026] Snow depth analysis was performed using the snowflake falling parameters and the uniform thickness to obtain the first snow depth monitoring result; In simple terms, it involves establishing a correlation between the snowflake falling parameters and the uniform thickness. Specifically, the snowflake falling parameters can be used as the horizontal axis and the uniform thickness as the vertical axis to obtain the relationship between the uniform thickness and the snowflake falling parameters. This allows for real-time updates of the snowflake falling parameters based on the monitoring time, and prediction of the snow melting thickness based on the relationship between the parameters. The prediction result is then used as the first snow depth monitoring result.

[0027] Snow depth is monitored at the monitoring point using a laser displacement sensor to obtain snow depth sensing data; A laser displacement sensor is an existing device that monitors snow depth by emitting laser light. It measures snow depth using the principle of a phase-detection laser rangefinder. Phase-detection laser rangefinders utilize radio frequency to modulate the laser beam and measure the phase delay of the modulated light traveling back and forth along the measuring line. Based on the wavelength of the modulated light, the distance represented by this phase delay is calculated, thus indirectly determining the time required for the light to travel back and forth along the measuring line, and consequently, the snow depth. Snow depth sensing data refers to the returned modulated light received by the laser displacement sensor.

[0028] Anomaly analysis is performed on the snow depth sensing data to remove abnormal data. The snow depth sensing data after removing abnormal data is then used for data transformation to obtain the second snow depth monitoring result. Laser displacement sensors are prone to errors when monitoring snow depth, i.e., false alarms. For example, they may detect snow even when there is no snow. Therefore, it is necessary to remove abnormal data and convert the snow depth sensing data after removing abnormal data. That is, the returned modulated light is converted into snow depth. The conversion of sensing data of laser displacement sensors is a common technique used by those skilled in the art, so it will not be elaborated here. The converted snow depth is used as the second snow depth monitoring result.

[0029] In a preferred embodiment, it further includes: The snow depth sensing data is subjected to frequency domain analysis and transformation to obtain first-order differential peak data of the spectrum; peak periodicity analysis is performed on the first-order differential peak data of the spectrum, and abnormal data is identified based on periodicity characteristics, wherein the abnormal data is sensing data whose periodicity characteristics do not meet the predetermined periodicity characteristics; the abnormal data is removed from the snow depth sensing data.

[0030] Specifically, the snow depth sensing data is subjected to frequency domain analysis and transformation to obtain first-order differential peak data of the spectrum. Specifically, the snow depth sensing data is subjected to Fourier transform using existing technology to convert it from the time domain to the frequency domain. A first-order differential operation is then performed on the converted frequency domain data to obtain multiple frequency values. The first-order differential operation is a common technique used by those skilled in the art and will not be elaborated upon here. A curve is plotted for the multiple frequency values ​​according to the acquisition time. The peak values ​​are extracted from the curves to obtain the first-order differential peak data of the spectrum. Peak periodicity analysis is performed on the first-order differential peak data of the spectrum. Anomaly data is identified based on the periodicity characteristics. The anomaly data refers to sensing data whose periodicity characteristics do not meet the predetermined periodicity characteristics. In simpler terms, multiple time intervals and corresponding frequency values ​​between each peak and the next peak are calculated, and the variance of these multiple time intervals and frequency values ​​is obtained as the periodicity characteristic. Snowfall is a natural phenomenon and does not possess strong periodicity characteristics; therefore, the anomaly data is data with strong periodicity characteristics. Based on practical experience, predetermined periodic characteristics are set. These characteristics include the variance corresponding to the time interval and frequency value. This can be achieved by professionals in the field through periodic analysis of snow depth sensor data samples without anomalies, obtaining the periodic characteristics of these samples as the predetermined periodic characteristics. Sensor data whose periodic characteristics do not meet the predetermined periodic characteristics are then discarded as outliers, thereby improving the accuracy of snow depth sensor data and ultimately enhancing the accuracy of snow depth monitoring.

[0031] A confidence analysis was performed on the first snow depth monitoring results and the second snow depth monitoring results. The first snow depth monitoring results and the second snow depth monitoring results were weighted by confidence to obtain the snow depth monitoring results.

[0032] A confidence analysis is performed on the first snow depth monitoring results and the second snow depth monitoring results. The first snow depth monitoring results and the second snow depth monitoring results are weighted and calculated using the confidence scores to obtain the snow depth monitoring results. Here, the confidence score refers to the credibility of the first snow depth monitoring results and the second snow depth monitoring results, which is set by professionals in the art according to the actual situation. Specifically, the method described in this embodiment can be followed to obtain the first snow depth monitoring result sample, the second snow depth monitoring result sample, and the corresponding actual snow depth sample, respectively. The error between the first snow depth monitoring result sample and the second snow depth monitoring result sample and the actual snow depth sample is calculated as their respective confidence scores. A weighting coefficient is set according to the proportion of the error. Then, the first snow depth monitoring results and the second snow depth monitoring results are weighted and calculated to obtain the snow depth monitoring results.

[0033] In a preferred embodiment, it further includes: The train operating speed of the target high-speed railway line is obtained; track heat generation analysis is performed based on the train operating speed to obtain track heat generation information; a heat-snow depth correlation relationship is constructed, and snow depth impact analysis is performed on the track heat generation information to obtain a snow depth attenuation index; the snow depth monitoring results are compensated and corrected using the snow depth attenuation index.

[0034] Specifically, when a train runs, it rubs against the track surface, generating heat. This heat causes snow to melt, reducing the accuracy of snow depth monitoring results. Therefore, the first step is to obtain the train's operating speed on the target high-speed rail line, specifically based on actual conditions. Track heat generation analysis is then performed using this train speed. This can be achieved through train operation tests, i.e., installing heat sensors on the train to collect track heat generation information. A heat-snow depth correlation is then established, meaning that heat affects the rate of snow depth descent. This can be achieved through actual testing, obtaining the snow depth descent under different heat levels as the heat-snow depth correlation. Based on the track heat generation information and the heat-snow depth correlation, a snow depth impact analysis is performed, obtaining the descent rate corresponding to the track heat generation information as a snow depth attenuation index. This snow depth attenuation index is used to compensate and correct the snow depth monitoring results. Specifically, the time required for the train to pass over the track in actual conditions is multiplied by the snow depth attenuation index to obtain a calculation result. This calculated result is then subtracted from the snow depth monitoring results to compensate and correct the snow depth monitoring results, improving their accuracy.

[0035] Based on the above analysis, the beneficial effects that one or more technical solutions provided in this application can achieve are as follows: Image sensors are used to acquire images of monitoring points along the target high-speed railway line from a first spatial angle and a second spatial angle, respectively. The first spatial angle is perpendicular to the track, while the second spatial angle is parallel to the track plane. Based on the first spatial angle image, snowflake unit size and snowfall speed are identified to obtain snowfall parameters. The second spatial angle image is used to analyze the snowfall thickness uniformity within the track, obtaining the uniform thickness. Snow depth analysis is then performed using the snowfall parameters and uniform thickness to obtain the first snow depth monitoring result. A laser displacement sensor is used to monitor snow depth at the monitoring points, acquiring snow depth sensing data. Anomaly analysis is performed on the snow depth sensing data, and outliers are removed. The snow depth sensing data after anomaly removal is then transformed to obtain the second snow depth monitoring result. Confidence analysis is performed on the first and second snow depth monitoring results, and a weighted calculation is performed based on the confidence levels to obtain the final snow depth monitoring result. Therefore, by using image sensors and laser displacement sensors for joint snow depth analysis, the accuracy and precision of snow depth monitoring are improved.

[0036] Example 2 Based on the same inventive concept as the method for monitoring snow depth data on high-speed railway tracks in the foregoing embodiments, such as Figure 2 As shown, this application also provides a snow depth data monitoring system for high-speed railway tracks, the system comprising: Image acquisition module 11 is used to acquire images of the first spatial angle and the second spatial angle of the monitoring point of the target high-speed railway line through an image sensor, and obtain the first spatial angle image and the second spatial angle image, wherein the first spatial angle is a spatial angle perpendicular to the track, and the second spatial angle is parallel to the track plane. The first image analysis module 12 is used to identify the unit size of snowflakes and the falling speed of snowflakes based on the first spatial angle image, and to obtain snowflake falling parameters. The second image analysis module 13 is used to perform snowfall thickness uniformity analysis within the track using the second spatial angle image to obtain the uniform thickness. The first snow depth monitoring module 14 is used to perform snow depth analysis based on the snowflake falling parameters and the uniform thickness to obtain the first snow depth monitoring result. Laser monitoring module 15, the laser monitoring module 15 is used to monitor snow depth at the monitoring point through laser displacement sensor and acquire snow depth sensing data; The second snow depth monitoring module 16 is used to perform anomaly analysis on the snow depth sensing data, remove abnormal data, and perform data conversion on the snow depth sensing data after removing abnormal data to obtain the second snow depth monitoring result. The confidence analysis module 17 is used to perform confidence analysis on the first snow depth monitoring result and the second snow depth monitoring result, and to perform weighted calculation on the first snow depth monitoring result and the second snow depth monitoring result with confidence to obtain the snow depth monitoring result.

[0037] Furthermore, the first image analysis module 12 also includes: The first spatial angle image is segmented using a snowflake segmentation channel to obtain multiple segmented images. The snowflake segmentation channel is a fully convolutional neural network model, including an encoder and a decoder. The encoder and the decoder are constructed by training snowflake segmentation samples. Snowflake contour recognition is performed on the multiple segmented images to obtain multiple contour recognition results; The acquisition parameters of the image sensor are obtained, and the unit size of the snowflake is identified using the acquisition parameters and the multiple contour recognition results; The snowflake position is identified and analyzed in the first spatial angle image to obtain the snowflake falling speed; The snowflake falling parameters are composed of the snowflake unit size and the snowflake falling speed.

[0038] Furthermore, the first image analysis module 12 also includes: Perform snowflake position recognition on the first temporal image in the first spatial angle image to obtain the first position identification result; Snowflake position recognition is performed on the second time-series image in the first spatial angle image to obtain a second position identification result, wherein the first time-series image and the second time-series image are snowflake images acquired in adjacent consecutive time periods; The distance between the first location identification result and the second location identification result is calculated in the vertical direction to obtain the first vertical adjacent distance; The falling speed of the snowflakes is calculated based on the first vertical adjacent distance.

[0039] Furthermore, the second image analysis module 13 also includes: A first reference line is set, and the first reference line has a first reference thickness mark. Using the first reference thickness marker of the first reference reference line, the second spatial angle image is used to identify the reference deviation distance at multiple points to obtain a set of reference deviation distances; The thickness dispersion analysis at multiple points is performed using the aforementioned set of reference deviation distances to obtain the dispersion coefficients; If the discrete coefficients meet a predetermined discreteness threshold, the reference deviation distance set is averaged, and the averaged result is used to generate the uniform thickness.

[0040] Furthermore, the second image analysis module 13 also includes: If the dispersion coefficient does not meet the predetermined dispersion threshold, cluster analysis is performed on the benchmark deviation distance set to obtain multiple deviation distance clusters, and the multiple deviation distance clusters have location identifiers. Based on the multiple deviation distance clusters, a multi-point equilibrium thickness analysis is performed to obtain the multi-point equilibrium thickness, which is then used as the equilibrium thickness.

[0041] Furthermore, the second snow depth monitoring module 16 also includes: The snow depth sensing data is subjected to frequency domain analysis and transformation to obtain first-order differential peak data of the spectrum; Peak periodicity analysis is performed on the first-order difference peak data of the spectrum, and abnormal data is identified based on the periodicity characteristics, wherein the abnormal data is sensor data whose periodicity characteristics do not meet the predetermined periodicity characteristics. Remove the abnormal data from the snow depth sensing data.

[0042] Furthermore, the system also includes a monitoring and compensation module, which comprises: Obtain the train speed of the target high-speed rail line; Track heat generation analysis is performed at the train's operating speed to obtain track heat generation information; The correlation between heat and snow depth was established, and the impact of orbital heat generation information on snow depth was analyzed to obtain the snow depth attenuation index. The snow depth monitoring results are compensated and corrected using the snow depth attenuation index.

[0043] The specific example of the snow depth data monitoring method for high-speed railway tracks in the aforementioned Embodiment 1 is also applicable to the snow depth data monitoring system for high-speed railway tracks in this embodiment. Through the foregoing detailed description of the snow depth data monitoring method for high-speed railway tracks, those skilled in the art can clearly understand the snow depth data monitoring system for high-speed railway tracks in this embodiment. Therefore, for the sake of brevity, it will not be described in detail here.

[0044] It should be understood that various forms of processes shown above can be used, with steps rearranged, added, or deleted, as long as the desired result of the technical solution disclosed in this application can be achieved, and this document does not impose any restrictions.

[0045] Note that the above are merely preferred embodiments and the technical principles employed in this application. Those skilled in the art will understand that this application is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of this application. Therefore, although this application has been described in detail through the above embodiments, this application is not limited to the above embodiments, and may include many other equivalent embodiments without departing from the concept of this application, the scope of which is determined by the scope of the appended claims.

Claims

1. A method for monitoring snow depth data of a high-speed rail track, characterized in that, The method comprises: The first spatial angle and the second spatial angle of the monitoring point of the target high-speed railway line are respectively image collected by an image sensor to obtain a first spatial angle image and a second spatial angle image, wherein the first spatial angle is a spatial angle perpendicular to the track, and the second spatial angle is parallel to the track plane; Snowflake unit size identification and snowflake falling speed identification are performed based on the first spatial angle image to obtain snowflake falling parameters; The second spatial angle image is used for track snowfall thickness balance analysis to obtain a balanced thickness; Snow depth analysis is performed based on the snowflake falling parameters and the balanced thickness to obtain a first snow depth monitoring result; Snow depth monitoring is performed at the monitoring point by a laser displacement sensor to obtain snow depth sensing data; Abnormal analysis is performed on the snow depth sensing data to eliminate abnormal data, and data conversion is performed on the snow depth sensing data after the abnormal data are eliminated to obtain a second snow depth monitoring result; Confidence analysis is performed on the first snow depth monitoring result and the second snow depth monitoring result, and the first snow depth monitoring result and the second snow depth monitoring result are weighted calculated based on the confidence to obtain a snow depth monitoring result.

2. The method of claim 1, wherein, The snowflake unit size identification and the snowflake falling speed identification based on the first spatial angle image to obtain the snowflake falling parameters comprise: An image segmentation is performed on the first spatial angle image by using a snowflake segmentation channel to obtain a plurality of segmented images, wherein the snowflake segmentation channel is a full convolutional neural network model comprising an encoder and a decoder, and the encoder and the decoder are constructed by snowflake segmentation sample training; Snowflake contour identification is performed on the plurality of segmented images to obtain a plurality of contour identification results; The acquisition parameters of the image sensor are obtained, and snowflake unit size identification is performed based on the acquisition parameters and the plurality of contour identification results; Snowflake position identification analysis is performed on the first spatial angle image to obtain a snowflake falling speed; The snowflake unit size and the snowflake falling speed constitute the snowflake falling parameters.

3. The method of claim 2, wherein, The snowflake position identification analysis performed on the first spatial angle image to obtain the snowflake falling speed comprises: Snowflake position identification is performed on a first time sequence image in the first spatial angle image to obtain a first position identification result; Snowflake position identification is performed on a second time sequence image in the first spatial angle image to obtain a second position identification result, wherein the first time sequence image and the second time sequence image are snowflake images collected at adjacent continuous times; A first vertical adjacent distance is obtained by performing distance calculation on the first position identification result and the second position identification result in a vertical direction; The snowflake falling speed is calculated based on the first vertical adjacent distance.

4. The method of claim 1, wherein, The track snowfall thickness balance analysis based on the second spatial angle image to obtain the balanced thickness comprises: A first reference line is set, and the first reference line has a first reference thickness identification; A plurality of point reference deviation distance identifications are performed on the second spatial angle image based on the first reference thickness identification of the first reference line to obtain a reference deviation distance set; Performing thickness dispersion analysis on the multiple points with the reference deviation distance set to obtain a dispersion coefficient; If the dispersion coefficient meets a predetermined dispersion threshold, performing mean value processing on the reference deviation distance set to generate the balanced thickness based on the mean value processing result.

5. The method of claim 4, wherein, The method further includes: If the dispersion coefficient does not meet the predetermined dispersion threshold, performing clustering analysis on the reference deviation distance set to obtain a plurality of deviation distance aggregated clusters, the plurality of deviation distance aggregated clusters having point location identifiers; Performing balanced thickness analysis on the multiple points based on the plurality of deviation distance aggregated clusters to obtain a multiple point balanced thickness, and taking the multiple point balanced thickness as the balanced thickness.

6. The method of claim 1, wherein, The method further includes: Performing frequency domain analysis and conversion on the snow depth sensing data to obtain frequency spectrum first-order difference peak value data; Performing peak periodicity analysis on the frequency spectrum first-order difference peak value data, and identifying abnormal data based on periodicity characteristics, wherein the abnormal data is sensing data that does not meet a predetermined periodicity characteristic; Removing the abnormal data from the snow depth sensing data.

7. The method of claim 1, wherein, The method further includes: Obtaining train running speed of the target high-speed rail line; Performing track heat generation analysis based on the train running speed to obtain track heat generation information; Constructing a heat-snow depth correlation relationship, and performing snow depth influence analysis on the track heat generation information to obtain a snow depth attenuation index; Compensating and correcting the snow depth monitoring result based on the snow depth attenuation index.

8. A system for monitoring snow depth data of a high-speed rail track, characterized in that, The system for performing the steps of the method of any one of claims 1 to 7 includes: An image acquisition module configured to acquire first and second space angle images of a monitoring point of a target high-speed rail line by an image sensor, the first space angle being perpendicular to the track, and the second space angle being parallel to the track plane; A first image analysis module configured to identify snowflake unit size and snowflake falling speed based on the first space angle image to obtain snowflake falling parameters; A second image analysis module configured to perform track snow thickness balancing analysis based on the second space angle image to obtain a balanced thickness; A first snow depth monitoring module configured to perform snow depth analysis based on the snowflake falling parameters and the balanced thickness to obtain a first snow depth monitoring result; A laser monitoring module configured to monitor snow depth at the monitoring point by a laser displacement sensor to obtain snow depth sensing data; A second snow depth monitoring module configured to perform abnormal analysis on the snow depth sensing data to remove abnormal data, and to perform data conversion on the snow depth sensing data after removing the abnormal data to obtain a second snow depth monitoring result. A confidence analysis module is configured to perform confidence analysis on the first snow depth monitoring result and the second snow depth monitoring result, and to perform weighted calculation on the first snow depth monitoring result and the second snow depth monitoring result with confidence to obtain a snow depth monitoring result.

Images