Intelligent identification method and identification system

Abstract

The embodiment of the present application relates to the technical field of intelligent identification, and discloses an intelligent identification method and an identification system, which comprises the following steps: setting the horizontal coordinate of the lowest longitudinal pixel point in the to-be-identified region frame and the pixel point value corresponding thereto as the target point horizontal coordinate of the to-be-identified region frame, and setting the central point vertical coordinate of the to-be-identified region frame as the target point vertical coordinate of the to-be-identified region frame, so as to determine the rail preset passing point of the to-be-identified region frame; carrying out fitting processing on all the rail preset passing points located in the left area of the rail image, so as to obtain a left rail shape curve, and carrying out fitting processing on all the rail preset passing points located in the right area of the rail image, so as to obtain a right rail shape curve; after determining the target dynamic identification region according to the left rail shape curve and the right rail shape curve, carrying out obstacle identification processing on the target dynamic identification region by using an obstacle identification model. The embodiment of the present application can improve the identification accuracy.

Description

Technical Field

[0001] This invention relates to the field of intelligent recognition technology, and in particular to an intelligent recognition method and recognition system. Background Technology

[0002] Currently, due to the long distances and complex conditions during railway transportation, there is a possibility that people, vehicles, and animals may wander on the tracks and be unable to avoid high-speed trains. Therefore, intelligent identification systems for recognizing railway tracks and obstacles within them are particularly important.

[0003] However, in practice, it has been found that most of the intelligent recognition systems currently used in railway transportation are based on the analysis of high-definition images. These systems generally have high hardware requirements, resulting in high investment costs and making it difficult to achieve widespread adoption. Furthermore, the recognition accuracy of these intelligent recognition systems is generally low in low light conditions, making them unsuitable for all-weather recognition tasks. Summary of the Invention

[0004] This invention discloses an intelligent recognition method and system that can improve recognition accuracy.

[0005] The first aspect of this invention discloses an intelligent recognition method, the method comprising:

[0006] The x-coordinate of the lowest vertical pixel and the corresponding pixel value within the region to be identified is set as the x-coordinate of the target point of the region to be identified, and the y-coordinate of the center point of the region to be identified is set as the y-coordinate of the target point of the region to be identified, so as to determine the preset railway track passing point of the region to be identified; wherein, a railway track image includes multiple regions to be identified, and each region to be identified contains a railway track.

[0007] The preset points of the rail located in the left area of ​​the rail image are fitted to obtain the left rail shape curve, and the preset points of the rail located in the right area of ​​the rail image are fitted to obtain the right rail shape curve.

[0008] After determining the target dynamic recognition area based on the shape curves of the left and right rails, the obstacle recognition model is used to perform obstacle recognition processing on the target dynamic recognition area.

[0009] As another optional implementation, in the first aspect of the present invention, before setting the abscissa of the lowest vertical pixel point and the pixel corresponding to the value within the region to be identified as the abscissa of the target point of the region to be identified, and setting the ordinate of the center point of the region to be identified as the ordinate of the target point of the region to be identified, to determine the preset rail passage point of the region to be identified, the method further includes:

[0010] The target working area in the railway track image is determined; wherein the proportion of railway track in the target working area is greater than a first specified threshold.

[0011] The target working area is divided to obtain multiple regions to be identified.

[0012] As another optional implementation, in the first aspect of the present invention, after dividing the target working area to obtain multiple regions to be identified, and before setting the x-coordinate of the pixel corresponding to the lowest vertical pixel value within the region to be identified as the x-coordinate of the target point of the region to be identified, and setting the y-coordinate of the center point of the region to be identified as the y-coordinate of the target point of the region to be identified, to determine the preset rail passage point of the region to be identified, the method further includes:

[0013] The sum of each vertical pixel within the region to be identified is calculated from left to right to obtain the lowest vertical pixel sum.

[0014] As another optional implementation, in the first aspect of the present invention, after determining the preset rail transit points of the region to be identified by setting the abscissa of the lowest vertical pixel point and the pixel point corresponding to the value within the region to be identified as the abscissa of the target point of the region to be identified, and setting the ordinate of the center point of the region to be identified as the ordinate of the target point of the region to be identified, and after performing fitting processing on all the preset rail transit points located in the left region of the rail image to obtain the left rail shape curve, and before performing fitting processing on all the preset rail transit points located in the right region of the rail image to obtain the right rail shape curve, the method further includes:

[0015] Calculate the slope between the preset points through which two adjacent rails pass;

[0016] Based on the slope, the coordinates of the next target rail point are calculated.

[0017] Detect whether the distance between the target next rail preset passing point coordinate position and the actual next rail preset passing point coordinate position is greater than a second specified threshold; if so, set the target next rail preset passing point as the final next rail preset passing point.

[0018] If the distance between the target next rail preset passing point coordinate position and the actual next rail preset passing point coordinate position is not greater than the second specified threshold, the actual next rail preset passing point is set as the final next rail preset passing point.

[0019] As another optional implementation, in the first aspect of the present invention,

[0020] After determining the target dynamic recognition region based on the shape curves of the left and right rails, the obstacle recognition model is used to perform obstacle recognition processing on the target dynamic recognition region, including:

[0021] Obtain the object to be identified in the target dynamic recognition area;

[0022] The obstacle recognition model is used to perform obstacle matching processing on the object to be identified in order to obtain the obstacle matching degree;

[0023] Detect whether the obstacle matching degree is higher than a third specified threshold; if so, mark the object to be identified and issue a prompt message.

[0024] As another optional implementation, in the first aspect of the present invention,

[0025] The method further includes determining the target working area in the railway track image; wherein, before the proportion of railway track in the target working area exceeds a first specified threshold, the method further includes:

[0026] The obstacle recognition model is generated by iteratively training obstacle data in an infrared thermal imaging obstacle database.

[0027] As another optional implementation, in the first aspect of the present invention, the target dynamic recognition area is polygonal in shape, and the proportion of rails in the target dynamic recognition area is greater than a fourth specified threshold.

[0028] A second aspect of this invention discloses an identification system, the identification system comprising:

[0029] The first setting unit is used to set the horizontal coordinate of the lowest vertical pixel point and the corresponding pixel point in the region to be identified as the horizontal coordinate of the target point of the region to be identified, and to set the vertical coordinate of the center point of the region to be identified as the vertical coordinate of the target point of the region to be identified, so as to determine the preset railway track passing point of the region to be identified; wherein, a railway track image includes multiple regions to be identified, and each region to be identified contains a railway track.

[0030] The fitting unit is used to fit all the preset points of the rail located in the left area of ​​the rail image to obtain the left rail shape curve, and to fit all the preset points of the rail located in the right area of ​​the rail image to obtain the right rail shape curve.

[0031] The recognition unit is used to perform obstacle recognition processing on the target dynamic recognition area using an obstacle recognition model after determining the target dynamic recognition area based on the shape curve of the left rail and the shape curve of the right rail.

[0032] As another optional implementation, in a second aspect of the present invention, the identification system further includes:

[0033] The determining unit is used to set the x-coordinate of the lowest vertical pixel point and the pixel corresponding to the value within the region to be identified as the x-coordinate of the target point of the region to be identified, and to set the y-coordinate of the center point of the region to be identified as the y-coordinate of the target point of the region to be identified, so as to determine the target working area in the rail image before determining the preset rail passage point of the region to be identified; wherein the proportion of rail in the target working area is greater than a first specified threshold.

[0034] A segmentation unit is used to segment the target working area to obtain multiple regions to be identified.

[0035] A third aspect of this invention discloses an identification system, the identification system comprising:

[0036] Memory containing executable program code;

[0037] A processor coupled to the memory;

[0038] The processor calls the executable program code stored in the memory to execute an intelligent recognition method disclosed in the first aspect of the present invention.

[0039] A fourth aspect of the present invention discloses a computer-readable storage medium storing a computer program, wherein the computer program causes a computer to execute an intelligent identification method disclosed in the first aspect of the present invention.

[0040] The fifth aspect of this invention discloses a computer program product that, when run on a computer, causes the computer to execute some or all of the steps of any of the intelligent recognition methods of the first aspect.

[0041] The sixth aspect of this invention discloses an application publishing platform for publishing computer program products, wherein when the computer program products are run on a computer, the computer performs some or all of the steps of any of the intelligent recognition methods of the first aspect.

[0042] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0043] In this embodiment of the invention, the abscissa of the lowest vertical pixel point and the pixel corresponding to the value within the region to be identified is set as the abscissa of the target point of the region to be identified, and the ordinate of the center point of the region to be identified is set as the ordinate of the target point of the region to be identified, thereby determining the preset rail passage points of the region to be identified; wherein, a rail image includes multiple regions to be identified, and each region contains rails; all the preset rail passage points located in the left region of the rail image are fitted to obtain a left rail shape curve, and all the preset rail passage points located in the right region of the rail image are fitted to obtain a right rail shape curve; after determining the target dynamic identification region based on the left and right rail shape curves, an obstacle recognition model is used to perform obstacle recognition processing on the target dynamic identification region. Therefore, this embodiment of the invention can improve the recognition accuracy. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the 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.

[0045] Figure 1 This is a flowchart illustrating an intelligent recognition method disclosed in an embodiment of the present invention;

[0046] Figure 2 This is a flowchart illustrating another intelligent recognition method disclosed in an embodiment of the present invention;

[0047] Figure 3 This is a schematic diagram of the structure of an identification system disclosed in an embodiment of the present invention;

[0048] Figure 4 This is a schematic diagram of another identification system disclosed in an embodiment of the present invention;

[0049] Figure 5 This is a schematic diagram of another identification system disclosed in an embodiment of the present invention. Detailed Implementation

[0050] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] It should be noted that the terms "first," "second," "third," "fourth," etc., used in the specification and claims of this invention are used to distinguish different objects, not to describe a specific order. The terms "comprising" and "having," and any variations thereof, in the embodiments of this invention are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products, or devices.

[0052] This invention discloses an intelligent recognition method and system that can improve recognition accuracy.

[0053] The following is a detailed description in conjunction with the accompanying drawings.

[0054] Example 1

[0055] Please see Figure 1 , Figure 1 This is a flowchart illustrating an intelligent recognition method disclosed in an embodiment of the present invention. Figure 1 As shown, the intelligent recognition method may include the following steps.

[0056] 101. The recognition system sets the x-coordinate of the lowest vertical pixel point and the corresponding pixel point within the region to be recognized as the x-coordinate of the target point of the region to be recognized, and sets the y-coordinate of the center point of the region to be recognized as the y-coordinate of the target point of the region to be recognized, so as to determine the preset points through which the railway tracks pass in the region to be recognized; wherein, a railway track image includes multiple regions to be recognized, and each region to be recognized contains railway tracks.

[0057] As an optional implementation, in this embodiment of the invention, the region to be identified in this application should include most of the railway track to ensure the correctness of the selected base point. The system only needs to determine whether there are obstacles in and around the railway track, and other areas can be ignored.

[0058] 102. The recognition system performs fitting processing on all the preset points of the railway track located in the left area of ​​the railway track image to obtain the shape curve of the left railway track, and performs fitting processing on all the preset points of the railway track located in the right area of ​​the railway track image to obtain the shape curve of the right railway track.

[0059] As an optional implementation, in this embodiment of the invention, a spline curve can be fitted using the target points in the left and right columns respectively, and this curve represents the shape of the railway track. A suitable fitting method is selected, and during the fitting process, appropriate target points are carefully selected; target points with too large a difference are actively discarded.

[0060] As an optional implementation, in this embodiment of the invention, the target points obtained in the left and right columns are mostly scattered. These points need to be smoothly connected to properly fit the shape of the railway track. Since the fitting is performed using a CPU, it is necessary to ensure the fitting effect while avoiding excessive computational load and time consumption. Therefore, a suitable fitting method must be selected. After selecting the fitting method, two curves, the left and right, are finally obtained, which are the railway track curves.

[0061] 103. After determining the target dynamic recognition area based on the shape curves of the left and right rails, the recognition system uses the obstacle recognition model to perform obstacle recognition processing on the target dynamic recognition area.

[0062] As an optional implementation, in this embodiment of the invention, the target dynamic recognition area set according to the shape of the railway track can be a polygon that includes most of the railway track.

[0063] As an optional implementation, in this embodiment of the invention, after obtaining the rail shape fitting curve, the present application can define a dynamically identified polygon based on the two curves and the points on the curves, ensuring that the subsequent system identification is only performed within this polygon, while ignoring other risk-free areas.

[0064] As an optional implementation, in this embodiment of the invention, after obtaining the railway track curve and its two columns of target points, a dynamic identification zone can be defined based on the coordinates of these points and the shape of the curve. This step can be understood as finding the smallest bounding polygon of the two fitted curves and the two columns of coordinate points. This polygon is the railway track area updated in real time, and only the presence of obstacles exceeding a threshold in this area needs to be identified. This ensures that the railway track area is defined in real time according to the different shapes of the railway track ahead during movement, obstacles within the area are identified, and timely alarms are issued to remind the driver to operate, reducing the probability of accidents.

[0065] As an optional implementation, in this embodiment of the invention, if the surface temperature of an object exceeds absolute zero, it will radiate electromagnetic waves. Infrared thermal imaging uses photoelectric technology to detect the infrared signal of a specific band emitted by the object's thermal radiation, converting the signal into an image that can be distinguished by human vision. Infrared thermal imaging can be performed day or night to ensure normal operation. However, unlike high-definition images, infrared thermal imaging is not as detailed; it is more like a blurred grayscale image. In the image, the imaging effects of railway tracks and obstacles differ. Obstacles are mostly moving living objects with high heat, while railway tracks are static objects with low heat. This application can use an obstacle recognition model to identify living obstacles and dynamically delineate an intelligent recognition area for the region where the railway tracks are located using pixel analysis and curve fitting methods. Combined with an alarm system that sets thresholds, an alarm is triggered when obstacles exceeding the threshold are identified in the dynamic intelligent analysis area, reducing the possibility of accidents.

[0066] exist Figure 1 In the intelligent recognition method, the recognition system is used as the execution subject for description. It should be noted that... Figure 1 The subject executing the intelligent recognition method can also be an independent device associated with the recognition system, which is not limited in the embodiments of the present invention.

[0067] It is evident that implementation Figure 1 The described intelligent recognition method can improve recognition accuracy.

[0068] In addition, implementation Figure 1 The described intelligent identification method is capable of real-time monitoring of railway tracks.

[0069] Example 2

[0070] Please see Figure 2 , Figure 2 This is a flowchart illustrating another intelligent recognition method disclosed in an embodiment of the present invention. Figure 2 The intelligent recognition method may include the following steps:

[0071] 201. The recognition system iteratively trains obstacle data from the infrared thermal imaging obstacle database to generate an obstacle recognition model.

[0072] As an optional implementation, in this embodiment of the invention, the obstacle database includes people, vehicles, animals, etc., and this application does not impose any limitations.

[0073] As an optional implementation, in this embodiment of the invention, the recognition system can prepare an infrared thermal imaging dataset of obstacles in advance, input this dataset into a neural network for training, and generate an obstacle recognition model after training to prepare for subsequent obstacle recognition.

[0074] As an optional implementation, in this embodiment of the invention, the model used is a model trained on a pre-prepared dataset, which is then deployed to a front-end intelligent device. The dataset is an infrared thermal imaging dataset of obstacles, which generally include people, vehicles, etc. After collecting a large amount of infrared data on obstacles, the data is labeled, specifying the coordinates and category names of the obstacles. The labels and images are then matched to form a dataset. This dataset is used for training, with training parameters set and iterated continuously until the accuracy meets the requirements, generating an AI model on the PC. However, this model cannot be directly deployed to an embedded device, so it needs to be converted using a tool before being deployed to the embedded device. The device starts up, enables intelligent recognition, waits for image input, and prepares to perform intelligent analysis and provide analysis results.

[0075] 202. The recognition system determines the target working area in the railway track image; wherein the proportion of railway track in the target working area is greater than a first specified threshold.

[0076] As an optional implementation, in this embodiment of the invention, the working area box is defined for the railway track image, and the box includes most of the railway track. This can reduce the amount of computation while ensuring that both railway track recognition and obstacle recognition are within the working area, thereby improving the accuracy of the recognition work.

[0077] 203. The recognition system divides the target working area to obtain multiple regions to be recognized.

[0078] As an optional implementation, in this embodiment of the invention, the present application may first select a rectangular frame of appropriate size on the left and right sides of the target working area, and this frame should contain a section of rail, because the rails have similar inclination and are clearer closer to the bottom of the screen, so the judgment should be more accurate. The two frames here are used to determine the base point and slope of the rail.

[0079] 204. The recognition system calculates the sum of each vertical pixel within the region to be recognized from left to right to obtain the lowest vertical pixel sum.

[0080] As an optional implementation, in this embodiment of the invention, the present application can calculate the sum of the pixel values ​​of each pixel point in the vertical direction from left to right within the left and right rectangular frames. The x-coordinate of the point with the smallest sum is the x-coordinate of the target point, and the y-coordinate of the center point of the rectangular frame is the y-coordinate of the target point.

[0081] As an optional implementation, in this embodiment of the invention, the operation of step 204 can be repeated until the top of the working area, that is, rectangles are continuously drawn and the minimum value of the sum of vertical pixel values ​​is calculated in each rectangle to obtain the target point of each rectangle, which is the preset railway track passing point.

[0082] 205. The recognition system sets the x-coordinate of the lowest vertical pixel point and the corresponding pixel point within the region to be recognized as the x-coordinate of the target point of the region to be recognized, and sets the y-coordinate of the center point of the region to be recognized as the y-coordinate of the target point of the region to be recognized, so as to determine the preset points through which the railway tracks pass in the region to be recognized; wherein, a railway track image includes multiple regions to be recognized, and each region to be recognized contains railway tracks.

[0083] 206. The identification system calculates the slope between the preset points of passage of two adjacent railway tracks.

[0084] 207. The recognition system calculates the coordinates of the next preset point that the target rail will pass through based on the slope.

[0085] 208. The identification system detects whether the distance between the preset coordinates of the next rail passing point of the target and the actual preset coordinates of the next rail passing point is greater than the second specified threshold. If yes, proceed to steps 209 and 211 to 214. If no, proceed to steps 210 to 214.

[0086] As an optional implementation, in this embodiment of the invention, the preset rail crossing point determined by this application may deviate too much from the actual rail. In this case, some limiting conditions need to be set to prevent the track from deviating. This application can determine the slope by calculating the slope of two preset points and obtain the next preset point based on this slope. If the distance between the point obtained by the calculated rectangle and this preset point is too large, this point is abandoned, and the point calculated using the slope is taken as the final target point.

[0087] 209. The identification system sets the target next rail preset passing point as the final next rail preset passing point.

[0088] 210. The identification system sets the actual next rail preset passing point as the final next rail preset passing point.

[0089] 211. The recognition system performs fitting processing on all the preset points of the railway track located in the left area of ​​the railway track image to obtain the shape curve of the left railway track, and performs fitting processing on all the preset points of the railway track located in the right area of ​​the railway track image to obtain the shape curve of the right railway track.

[0090] 212. The recognition system acquires the objects to be recognized in the target dynamic recognition area.

[0091] 213. The recognition system uses an obstacle recognition model to perform obstacle matching processing on the object to be recognized in order to obtain the obstacle matching degree.

[0092] 214. The recognition system detects whether the obstacle matching degree is higher than the third specified threshold. If yes, proceed to step 215. If no, end the process.

[0093] 215. The recognition system marks the object to be recognized and issues a prompt message, ending the current process.

[0094] As an optional implementation, in this embodiment of the invention, the present application can use the previously trained obstacle recognition model to identify whether there are obstacles in the target dynamic recognition area. If an obstacle is identified and the threshold is higher than a set value, a recognition box is drawn and an alarm is issued.

[0095] It is evident that implementation Figure 2 The described alternative intelligent recognition method can improve recognition accuracy.

[0096] In addition, implementation Figure 2The described alternative intelligent recognition method can reduce the probability of accidents.

[0097] Example 3

[0098] Please see Figure 3 , Figure 3 This is a schematic diagram of the structure of an identification system disclosed in an embodiment of the present invention. Figure 3 The recognition system 300 may include a first setting unit 301, a fitting unit 302, and a recognition unit 303, wherein:

[0099] The first setting unit 301 is used to set the horizontal coordinate of the lowest vertical pixel point and the corresponding pixel point in the region to be identified as the horizontal coordinate of the target point of the region to be identified, and to set the vertical coordinate of the center point of the region to be identified as the vertical coordinate of the target point of the region to be identified, so as to determine the preset passing point of the railway track of the region to be identified; wherein, a railway track image includes multiple regions to be identified, and the regions to be identified contain railway tracks.

[0100] The fitting unit 302 is used to fit all the preset points of the rail located in the left area of ​​the rail image to obtain the left rail shape curve, and to fit all the preset points of the rail located in the right area of ​​the rail image to obtain the right rail shape curve.

[0101] The recognition unit 303 is used to perform obstacle recognition processing on the target dynamic recognition area after determining the target dynamic recognition area based on the shape curves of the left and right rails.

[0102] As an optional implementation, in this embodiment of the invention, the region to be identified in this application should include most of the railway track to ensure the correctness of the selected base point. The identification unit 303 only needs to determine whether there are obstacles in and around the railway track, and other areas can be ignored.

[0103] As an optional implementation, in this embodiment of the invention, the fitting unit 302 can fit a spline curve using the target points in the left and right columns respectively, and this curve represents the shape of the railway track. A suitable fitting method is selected, and during the fitting process, appropriate target points are carefully selected; target points with too large a difference are actively discarded.

[0104] As an optional implementation, in this embodiment of the invention, the target points obtained in the left and right columns are mostly scattered. These points need to be smoothly connected to properly fit the shape of the railway track. Since the fitting is performed using a CPU, it is necessary to ensure the fitting effect while avoiding excessive computational load and time consumption. Therefore, a suitable fitting method must be selected. After selecting the fitting method, two curves, the left and right, are finally obtained, which are the railway track curves.

[0105] As an optional implementation, in this embodiment of the invention, the target dynamic recognition area set according to the shape of the railway track can be a polygon that includes most of the railway track.

[0106] As an optional implementation, in this embodiment of the invention, after obtaining the rail shape fitting curve, the present application can define a dynamically identified polygon based on the two curves and the points on the curves, ensuring that the subsequent system identification is only performed within this polygon, while ignoring other risk-free areas.

[0107] As an optional implementation, in this embodiment of the invention, after obtaining the railway track curve and its two columns of target points, a dynamic identification zone can be defined based on the coordinates of these points and the shape of the curve. This step can be understood as finding the smallest bounding polygon of the two fitted curves and the two columns of coordinate points. This polygon is the railway track area updated in real time, and only the presence of obstacles exceeding a threshold in this area needs to be identified. This ensures that the railway track area is defined in real time according to the different shapes of the railway track ahead during movement, obstacles within the area are identified, and timely alarms are issued to remind the driver to operate, reducing the probability of accidents.

[0108] As an optional implementation, in this embodiment of the invention, if the surface temperature of an object exceeds absolute zero, it will radiate electromagnetic waves. Infrared thermal imaging uses photoelectric technology to detect the infrared signal of a specific band emitted by the object's thermal radiation and converts this signal into an image that can be distinguished by human vision. Infrared thermal imaging can be performed day or night to ensure normal operation. However, unlike high-definition images, infrared thermal imaging is not as detailed; it is more like a blurred grayscale image. In the image, the imaging effect of railway tracks and obstacles differs. Obstacles are mostly moving living objects with high heat, while railway tracks are static objects with low heat. The recognition unit 303 can use an obstacle recognition model to detect living obstacles and dynamically delineate the intelligent recognition area where the railway tracks are located using pixel analysis and curve fitting methods. Combined with the alarm system setting a threshold, an alarm is triggered when obstacles exceeding the threshold are detected in the dynamic intelligent analysis area, reducing the possibility of accidents.

[0109] It is evident that implementation Figure 3 The described recognition system can improve recognition accuracy.

[0110] In addition, implementation Figure 3 The described identification system is capable of real-time monitoring of railway tracks.

[0111] Example 4

[0112] Please see Figure 4 , Figure 4 This is a schematic diagram of another identification system disclosed in an embodiment of the present invention. Figure 4 The identification system is composed of Figure 3 This was obtained by optimizing the recognition system. (Compared to...) Figure 3 Compared to other recognition systems, Figure 4 The identification system also includes:

[0113] The determining unit 304 is used by the first setting unit 301 to set the horizontal coordinate of the lowest vertical pixel point and the corresponding pixel point in the region to be identified as the horizontal coordinate of the target point of the region to be identified, and to set the vertical coordinate of the center point of the region to be identified as the vertical coordinate of the target point of the region to be identified, so as to determine the target working area in the rail image before determining the preset rail passage point of the region to be identified; wherein the proportion of rail in the target working area is greater than a first specified threshold.

[0114] The segmentation unit 305 is used to segment the target working area to obtain multiple regions to be identified.

[0115] As an optional implementation, in this embodiment of the invention, the partitioning unit 305 delineates the working area frame of the railway track image, and the frame includes most of the railway track. This can reduce the amount of computation while ensuring that both railway track recognition and obstacle recognition are within the working area, thereby improving the accuracy of the recognition work.

[0116] As an optional implementation, in this embodiment of the invention, the dividing unit 305 can first select a rectangular frame of appropriate size on the left and right sides of the target working area, and this frame should contain a section of rail, because the rails have similar inclination and are clearer closer to the bottom of the screen, so the judgment should be more accurate. The two frames here are used to determine the base point and slope of the rail.

[0117] and Figure 3 Compared to the recognition system, Figure 4 The identification system includes:

[0118] The first calculation unit 306 is used to divide the target working area by the division unit 305 to obtain multiple regions to be identified, and the first setting unit 301 sets the horizontal coordinate of the pixel corresponding to the lowest vertical pixel sum value in the region to be identified as the horizontal coordinate of the target point of the region to be identified, and sets the vertical coordinate of the center point of the region to be identified as the vertical coordinate of the target point of the region to be identified, so as to determine the preset railway passing point of the region to be identified, and calculates the sum value of each vertical pixel in the region to be identified from left to right to obtain the lowest vertical pixel sum value.

[0119] As an optional implementation, in this embodiment of the invention, the first calculation unit 306 can calculate the sum of the pixel values ​​of each pixel point in the vertical direction from left to right within the rectangular frames on both sides. The x-coordinate of the point with the smallest sum is the x-coordinate of the target point, and the y-coordinate of the center point of the rectangular frame is the y-coordinate of the target point.

[0120] As an optional implementation, in this embodiment of the invention, the first calculation unit 306 can be cyclically operated to the top of the working area, that is, continuously draw rectangular frames and calculate the minimum value of the sum of vertical pixel values ​​in each frame to obtain the target point of each rectangular frame, which is the preset railway track passing point.

[0121] and Figure 3 Compared to other recognition systems, Figure 4 The identification system includes:

[0122] The second calculation unit 307 is used to calculate the slope between two adjacent preset rail passing points after the first setting unit 301 sets the lowest vertical pixel point and the corresponding pixel point horizontal coordinate of the value in the region to be identified as the horizontal coordinate of the target point of the region to be identified, and sets the center point vertical coordinate of the region to be identified as the vertical coordinate of the target point of the region to be identified, so as to determine the preset rail passing points of the region to be identified; and before the fitting unit 302 performs fitting processing on all preset rail passing points located in the left area of ​​the rail image to obtain the left rail shape curve, and performs fitting processing on all preset rail passing points located in the right area of ​​the rail image to obtain the right rail shape curve.

[0123] The third calculation unit 308 is used to calculate the coordinates of the preset point that the next rail will pass through based on the slope.

[0124] The detection unit 309 is used to detect whether the distance between the preset coordinate position of the next rail passing point of the target and the actual preset coordinate position of the next rail passing point is greater than a second specified threshold.

[0125] The second setting unit 310 is used to set the target next rail preset passing point as the final next rail preset passing point when the detection unit 309 detects that the distance between the target next rail preset passing point coordinate position and the actual next rail preset passing point coordinate position is greater than the second specified threshold.

[0126] As an optional implementation, in this embodiment of the invention, the second setting unit 310 is further configured to detect that the distance between the target next rail preset passing point coordinate position and the actual next rail preset passing point coordinate position is not greater than the second specified threshold, and set the actual next rail preset passing point as the final next rail preset passing point.

[0127] As an optional implementation, in this embodiment of the invention, the preset rail crossing point determined by this application may deviate too much from the actual rail. In this case, some limiting conditions need to be set to prevent the track from deviating. The second calculation unit 307 can determine the slope by calculating the two preset points before and after, and obtain the next preset point based on the slope. If the distance between the point obtained by the calculation rectangle and this preset point is too large, the point is abandoned, and the point calculated by the slope is taken as the final target point.

[0128] and Figure 3 Compared to other recognition systems, Figure 4 The identification unit 303 includes:

[0129] Acquisition subunit 3031 is used to acquire the object to be identified in the target dynamic recognition area.

[0130] The matching subunit 3032 is used to perform obstacle matching processing on the object to be identified using the obstacle recognition model in order to obtain the obstacle matching degree.

[0131] The detection subunit 3033 is used to detect whether the obstacle matching degree is higher than the third specified threshold.

[0132] The marking and prompting unit 3034 is used to mark the object to be identified and issue a prompt message when the detection subunit 3033 detects that the obstacle matching degree is higher than the third specified threshold.

[0133] As an optional implementation, in this embodiment of the invention, the detection subunit 3033 can use the previously trained obstacle recognition model to identify whether there are obstacles in the target dynamic recognition area. If an obstacle is identified and the threshold is higher than the set value, a recognition box is drawn and an alarm is issued.

[0134] and Figure 3 Compared to other recognition systems, Figure 4 The identification system also includes:

[0135] The generation unit 311 is used to determine the target working area in the railway track image by the determining unit 304; wherein, before the proportion of railway track in the target working area is greater than a first specified threshold, the obstacle data in the infrared thermal imaging obstacle database is iteratively trained to generate an obstacle recognition model.

[0136] As an optional implementation, in this embodiment of the invention, the obstacle database includes people, vehicles, animals, etc., and this application does not impose any limitations.

[0137] As an optional implementation, in this embodiment of the invention, the generation unit 311 can prepare an infrared thermal imaging dataset of obstacles in advance, input this dataset into a neural network for training, and generate an obstacle recognition model after training to prepare for subsequent obstacle recognition.

[0138] As an optional implementation, in this embodiment of the invention, the model used is a model trained on a pre-prepared dataset, which is then deployed to a front-end intelligent device. The dataset is an infrared thermal imaging dataset of obstacles, which generally include people, vehicles, etc. After collecting a large amount of infrared data on obstacles, the data is labeled, specifying the coordinates and category names of the obstacles. The labels and images are then matched to form a dataset. This dataset is used for training, with training parameters set and iterated continuously until the accuracy meets the requirements, generating an AI model on the PC. However, this model cannot be directly deployed to an embedded device, so it needs to be converted using a tool before being deployed to the embedded device. The device starts up, enables intelligent recognition, waits for image input, and prepares to perform intelligent analysis and provide analysis results.

[0139] It is evident that implementation Figure 4 The other identification system described can improve identification accuracy.

[0140] In addition, implementation Figure 4 The described identification system can reduce the probability of accidents.

[0141] Example 5

[0142] Please see Figure 5 , Figure 5 This is a schematic diagram of another identification system disclosed in an embodiment of the present invention.

[0143] like Figure 5 The identification system may include:

[0144] Memory 501 storing executable program code;

[0145] Processor 502 coupled to memory 501;

[0146] Specifically, processor 502 calls the executable program code stored in memory 501 and executes it. Figures 1-2 Any intelligent recognition method.

[0147] This invention discloses a computer-readable storage medium storing a computer program, wherein the computer program causes a computer to execute... Figures 1-2 Any intelligent recognition method.

[0148] This invention also discloses a computer program product, wherein when the computer program product is run on a computer, the computer performs some or all of the steps of the methods described in the above method embodiments.

[0149] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-Erasable Programmable Read-Only Memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium capable of carrying or storing data.

[0150] The above provides a detailed description of an intelligent recognition method and system disclosed in the embodiments of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. An intelligent recognition method, characterized in that, include: The target working area in the railway track image is determined; wherein the proportion of railway track in the target working area is greater than a first specified threshold. The target working area is divided to obtain multiple regions to be identified; Calculate the sum of each vertical pixel within the region to be identified from left to right to obtain the lowest vertical pixel sum. The x-coordinate of the lowest vertical pixel and the corresponding pixel value within the region to be identified is set as the x-coordinate of the target point of the region to be identified, and the y-coordinate of the center point of the region to be identified is set as the y-coordinate of the target point of the region to be identified, so as to determine the preset railway track passing point of the region to be identified; wherein, a railway track image includes multiple regions to be identified, and each region to be identified contains a railway track. Calculate the slope between the preset points through which two adjacent rails pass; Based on the slope, the coordinates of the next target rail point are calculated. Detect whether the distance between the target next rail preset passing point coordinate position and the actual next rail preset passing point coordinate position is greater than a second specified threshold; if so, set the target next rail preset passing point as the final next rail preset passing point. If the distance between the target next rail preset passing point coordinate position and the actual next rail preset passing point coordinate position is not greater than the second specified threshold, the actual next rail preset passing point is set as the final next rail preset passing point. The preset points of the rail located in the left area of ​​the rail image are fitted to obtain the left rail shape curve, and the preset points of the rail located in the right area of ​​the rail image are fitted to obtain the right rail shape curve. After determining the target dynamic recognition area based on the shape curves of the left and right rails, the obstacle recognition model is used to perform obstacle recognition processing on the target dynamic recognition area.

2. The method according to claim 1, characterized in that, After determining the target dynamic recognition region based on the shape curves of the left and right rails, the obstacle recognition model is used to perform obstacle recognition processing on the target dynamic recognition region, including: Obtain the object to be identified in the target dynamic recognition area; The obstacle recognition model is used to perform obstacle matching processing on the object to be identified in order to obtain the obstacle matching degree; Detect whether the obstacle matching degree is higher than a third specified threshold; if so, mark the object to be identified and issue a prompt message.

3. The method according to claim 1, characterized in that, The method further includes determining the target working area in the railway track image; wherein, before the proportion of railway track in the target working area exceeds a first specified threshold, the method further includes: The obstacle recognition model is generated by iteratively training obstacle data in an infrared thermal imaging obstacle database.

4. The method according to any one of claims 1 to 3, characterized in that, The target dynamic recognition area is polygonal in shape, and the proportion of railway tracks in the target dynamic recognition area is greater than the fourth specified threshold.

5. An identification system, characterized in that, The identification system includes: A determining unit is used to determine a target working area in a railway track image; wherein the proportion of railway track in the target working area is greater than a first specified threshold. A segmentation unit is used to segment the target working area to obtain multiple regions to be identified. The first calculation unit is used to calculate the sum of each vertical pixel in the region box to be identified from left to right, so as to obtain the lowest vertical pixel sum. The first setting unit is used to set the horizontal coordinate of the lowest vertical pixel point and the corresponding pixel point in the region to be identified as the horizontal coordinate of the target point of the region to be identified, and to set the vertical coordinate of the center point of the region to be identified as the vertical coordinate of the target point of the region to be identified, so as to determine the preset railway track passing point of the region to be identified; wherein, a railway track image includes multiple regions to be identified, and each region to be identified contains a railway track. The second calculation unit is used to calculate the slope between two adjacent preset points of the rails; The third calculation unit is used to calculate the coordinates of the next preset point of the target rail based on the slope. The detection unit is used to detect whether the distance between the target next rail preset passing point coordinate position and the actual next rail preset passing point coordinate position is greater than a second specified threshold; if so, the target next rail preset passing point is set as the final next rail preset passing point. The second setting unit is used to set the actual next rail preset passing point as the final next rail preset passing point if the distance between the target next rail preset passing point coordinate position and the actual next rail preset passing point coordinate position is not greater than the second specified threshold. The fitting unit is used to fit all the preset points of the rail located in the left area of ​​the rail image to obtain the left rail shape curve, and to fit all the preset points of the rail located in the right area of ​​the rail image to obtain the right rail shape curve. The recognition unit is used to perform obstacle recognition processing on the target dynamic recognition area using an obstacle recognition model after determining the target dynamic recognition area based on the shape curve of the left rail and the shape curve of the right rail.

6. The identification system according to claim 5, characterized in that, The identification system also includes: The determining unit is used to set the x-coordinate of the lowest vertical pixel point and the pixel corresponding to the value within the region to be identified as the x-coordinate of the target point of the region to be identified, and to set the y-coordinate of the center point of the region to be identified as the y-coordinate of the target point of the region to be identified, so as to determine the target working area in the rail image before determining the preset rail passage point of the region to be identified; wherein the proportion of rail in the target working area is greater than a first specified threshold. A segmentation unit is used to segment the target working area to obtain multiple regions to be identified.

7. An identification system, characterized in that, The identification system includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the intelligent recognition method according to any one of claims 1-4.

Images