Turnout apparent disease automatic inspection and intelligent identification method and system

Abstract

The invention discloses a turnout apparent disease automatic inspection and intelligent identification method and system. Security of the detection precision of the geometric state of a railway turnout area track is crucial for operation safety and maintenance efficiency. The invention provides a brand new real-time and constant track geometric data matching method. According to the method, a global coordinate system and a local curve coordinate system are utilized to improve the turnout area defect detection positioning precision. The real-time data is collected when the track detection vehicle passes through the turnout, and the constant data is collected when manual detection is carried out on the turnout under the condition that no train passes through. According to the method, constant data is converted into a space curve which can be compared with real-time data, so that accurate defect positioning is realized. In addition, the invention further introduces an algorithm based on a decision threshold value, and the algorithm is used for processing the specific structure irregularity of the turnout curve track. According to the algorithm, the real geometric irregularity and the structure abnormity in the turnout curve track direction can be effectively separated, and the track quality index is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of inspection and identification of apparent defects of turnouts, and in particular to a method and system for automatic inspection and intelligent identification of apparent defects of turnouts. Background Art

[0002] Turnouts are among the most complex pieces of equipment in a railway system, typically consisting of three parts: the turnout, the connecting section (straight strands, guide rails), the frog, and the guardrails. Because the guiderail section of a turnout is relatively short, there's insufficient distance for gradual superelevation reduction, and therefore, no superelevation is typically required. Structural irregularities include gauge reductions and gaps at the point rails and frog rails. The wheel-rail contact relationship when a vehicle passes through a turnout is more complex than on a mainline. The impact angles of the point rails, guard rails, and wing rails are much greater than those on curved tracks, and the vertical and lateral stiffness of the turnout area is also greater than that of conventional track. The wheel-rail forces on a locomotive passing through a turnout are higher than those on conventional track, resulting in a greater maintenance workload for turnouts. Turnouts are a weak link on a line, a critical and challenging area for maintenance and repair, and a key piece of equipment that affects train speed and safety.

[0003] However, the turnout area is short, and the point rail and frog rail are variable-section structures with decreasing values, resulting in drastic changes in track geometry. Currently, relying solely on real-time inspection data is insufficient to effectively locate track geometry defects detected in the turnout area. To rapidly locate defects detected by real-time track geometry, constant-time inspection and re-check positioning are typically performed using stringers or track inspection vehicles. Turnout guide curves are typically small-radius curves, and structural irregularities exist at the point rail. When track geometry irregularities are superimposed on structural irregularities, the structural irregularities lead to distortion in the output structure, making it impossible to assess using current large-value overrun and track quality indices. These challenges highlight the particularities and complexities that must be considered when conducting track geometry inspection in the turnout area. In particular, it is difficult to effectively combine real-time inspection data with constant-time track geometry inspection data to accurately locate track geometry defects. Furthermore, due to the structural irregularities of the turnout tracks, inspection results cannot truly reflect track geometry defects in the turnout area. Summary of the Invention

[0004] In view of this, an embodiment of the present invention provides a method for automatic inspection and intelligent identification of apparent defects of turnouts, comprising the following steps: Step 1: Collect real-time track geometry data and constant track geometry data, respectively. The real-time track geometry data is obtained when a track geometry inspection vehicle passes through a turnout, and the constant track geometry data is obtained by manually inspecting the turnout when no train passes. Step 2: convert the constant track geometry data from the global coordinate system to the local curve coordinate system to obtain a space curve comparable to the real-time track geometry data; Step 3: Filtering the spatial curve using a linear phase FIR digital high-pass filter, wherein the linear phase FIR digital high-pass filter is designed using three trapezoidal windows in parallel, wherein one trapezoidal window serves as a basic window and the remaining two trapezoidal windows are used to adjust the passband of the basic window; by adjusting the coefficient and window length parameters of each trapezoidal window, smaller passband fluctuations and faster transition band attenuation are achieved; Step 4: Match the filtered constant track geometry data with the real-time track geometry data, and determine the specific location of the track defect based on the matching difference to achieve accurate positioning of the track geometry defect in the turnout area; In step 5, the alignment data of the turnout curve track is processed using an algorithm based on a decision threshold to eliminate structural irregularities and truly reflect the track geometry defects.

[0005] Preferably, the constant track geometry data is converted from the global coordinate system to the local curvilinear coordinate system as follows: Step 2.1, determining a reference point, wherein the reference point is the point closest to the point to be measured in the constant track geometry data; Step 2.2, calculate the longitudinal displacement and lateral displacement between the measured point and the reference point; Step 2.3. Determine the direction of the lateral displacement based on the cross product of the normal vector and the tangent vector of the reference point; if the result of the cross product points to the left of the reference line, the direction is positive; if it points to the right, the direction is negative.

[0006] Preferably, the algorithm based on the decision threshold is specifically: Step 5.1. Construct a turnout matching template for the line: Based on the planar layout of the turnout, extract the lateral offset of the turnout branch according to the preset spatial interval; perform second-order difference on the lateral offset and perform digital high-pass filtering to construct a spatial matching template; perform amplitude normalization on the spatial matching template and symmetrically fill zero values at both ends of the template to ensure that the lengths of different types of turnout templates are the same; store the processed turnout templates in a database to construct a turnout standard template library.

[0007] Step 5.2: Locate the decision threshold in the turnout area: Based on the unique characteristics of the turnout area, combined with information from ground markings and the line turnout ledger, set an optimized threshold to automatically and efficiently determine whether the data corresponds to a turnout; Step 5.3: Determine the branch decision threshold of the switch area: Use an automatic positioning device to distinguish the incidental curve; mark the switch branch line and incidental curve to ensure the accuracy of the distinction; Step 5.4: Calculate the Pearson correlation coefficient between the turnout template and the screened turnout data; determine the opening direction of the turnout based on the positive or negative value of the Pearson correlation coefficient; compare the absolute value of the Pearson correlation coefficient to determine the passing direction of the turnout; use the Manhattan distance as a secondary matching parameter to match the track geometry data with the template to eliminate structural irregularities in the turnout area.

[0008] Preferably, a positive Pearson correlation coefficient indicates a positive correlation with the matching template, and the opening directions are the same; a negative Pearson correlation coefficient indicates a negative correlation with the matching template, and the opening directions are opposite; from the point rail to the frog rail is a forward passage, and the corresponding template is the forward template; from the frog rail to the point rail is a reverse passage, and the corresponding template is the reverse template; if the maximum absolute value is provided by the forward template, the turnout passes in the forward direction; if the maximum absolute value is provided by the reverse template, the turnout passes in the reverse direction.

[0009] On the other hand, the present invention also provides a system for automatic inspection and intelligent identification of apparent defects of turnouts, the device comprising: At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can perform the above-mentioned method for automatic inspection and intelligent identification of apparent defects of a turnout.

[0010] This invention proposes an automated inspection and intelligent identification method for apparent defects in turnouts. This method is based on a real-time and constant matching method for turnout track geometry, mapping a global coordinate system to a local curvilinear coordinate system. Real-time data is collected when a track geometry vehicle passes through the turnout, while constant data is collected during manual inspection of the turnout without a train passing through. Constant track geometry data in the turnout area is mapped onto a spatial curve derived from real-time track geometry inspection. By comparing the real-time and constant track geometry inspection data for the turnout area, defects in the turnout area can be quickly located. Because the real-time inspection output for the turnout curve track direction includes small-radius curves in the turnout guide curve, structural irregularities at the turnout point rail center, and actual track geometry defects, this invention proposes a turnout sliding matching curve track alignment processing algorithm based on decision threshold judgment to eliminate structural irregularities in the turnout curve track direction. Experimental results demonstrate that both algorithms scan and output the actual track geometry defects in the turnout curve track direction and eliminate the impact of turnout structural irregularities on the track quality index, providing guidance for the maintenance and repair of lines in the turnout area. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The above and other objects, features and advantages of the present invention will become more apparent through the following description of the embodiments of the present invention with reference to the accompanying drawings, in which: Figure 1 This is a flow chart of a method for automatic inspection and intelligent identification of apparent defects of turnouts proposed in this application; Figure 2 It is the relationship diagram between the global coordinate system and the local curvilinear coordinate system; Figure 3 It is a schematic diagram of the coordinate system conversion algorithm. DETAILED DESCRIPTION

[0012] The present application is described below based on the following embodiments, but the present application is not limited to these embodiments. In the detailed description of the present application below, certain specific details are described in detail. Those skilled in the art can fully understand the present application without the description of these details. To avoid obscuring the essence of the present application, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0013] Furthermore, persons of ordinary skill in the art will appreciate that the figures provided herein are for illustration purposes only and are not necessarily drawn to scale.

[0014] Unless the context clearly requires otherwise, words like “include”, “comprising” and the like throughout this application should be interpreted as including rather than exclusive or exhaustive; that is, as meaning “including but not limited to”.

[0015] In the description of this application, it should be understood that the terms "first", "second", etc. are used for descriptive purposes only and should not be understood to indicate or imply relative importance. In addition, in the description of this application, unless otherwise specified, "plurality" means two or more.

[0016] The inertial reference method can meet the accuracy requirements for real-time detection of the longitudinal level and track orientation of straight and curved tracks in the turnout area. However, due to structural irregularities in the turnout area, such as the small radius of the guide curve and the "straight-into-curve" phenomenon at the point rail, the geometric inspection data results of the turnout curved track are large, ranging from -10 to 9 mm. Therefore, it is impossible to effectively determine the true defects of the track geometry, nor can the true track quality index in the turnout area be accurately calculated. In addition, based on the inertial reference method, the 95% repeatability difference of the real-time detection of the longitudinal level and track orientation of straight and curved tracks in the turnout area was calculated.

[0017] Real-time longitudinal horizontal and track orientation inspections of straight and curved track in the turnout area demonstrate excellent repeatability, with a 95% repeatability variance of less than 0.4 mm. The 95% repeatability variances for longitudinal horizontal real-time inspections of curved track in the turnout area are significantly higher, at 0.22 and 0.33, respectively. The inertial reference method incorporates input from the laser system. This system outputs the lateral distance change between the inspection beam and the gauge point, as well as the vertical distance change between the inspection beam and the rail vertex. Consequently, when the system's inspection locations are concentrated at locations with varying cross-sections, such as the point rail and frog rail, outliers may occur due to displacement of the laser camera assembly.

[0018] In response to the above problems, Figure 1 As shown, this application proposes a method for automatic inspection and intelligent identification of apparent defects of turnouts, which specifically includes: Step 101: collecting real-time track geometry data and constant track geometry data, respectively. The real-time track geometry data is obtained when a track geometry inspection vehicle passes through a switch, and the constant track geometry data is obtained by manually inspecting the switch when no train passes. Step 102: convert the constant track geometry data from the global coordinate system to the local curve coordinate system to obtain a space curve comparable to the real-time track geometry data; Step 103: Filter the spatial curve using a linear phase FIR digital high-pass filter. The linear phase FIR digital high-pass filter is designed using three trapezoidal windows in parallel, where one trapezoidal window serves as a basic window and the remaining two trapezoidal windows are used to adjust the passband of the basic window. By adjusting the coefficients and window length parameters of each trapezoidal window, smaller passband fluctuations and faster transition band attenuation are achieved. Step 104: Match the filtered constant track geometry data with the real-time track geometry data, and determine the specific location of the track defect based on the matching difference to achieve accurate positioning of the track geometry defect in the turnout area; Step 105 , using an algorithm based on a decision threshold to process the alignment data of the turnout curve track to eliminate structural irregularities and truly reflect track geometry defects.

[0019] 1. Real-time and constant orbit geometry data detection and matching algorithm Based on the conversion relationship between the global coordinate system and the local curve coordinate system, the global spatial coordinates of the track geometry detected by the switch area constant are mapped to a spatial curve with one degree of freedom, and a linear phase finite impulse response (FIR) digital high-pass filter is used for spatial filtering. Compared with the rectangular window and the triangular window, the passband of the trapezoidal window is smoother, and its amplitude-frequency characteristics can be changed by flexibly adjusting the relative sizes of its two rectangular base windows. Therefore, the present invention uses three trapezoidal windows in parallel to design a high-pass filter, in which one trapezoidal window is used as the basic window. The basic window determines the approximate shape of the amplitude-frequency characteristics of the new filter, and the other two trapezoidal windows are used to adjust the passband of the basic window. By adjusting the coefficients of each trapezoidal window and its window length parameters, the transition band characteristics of the filter can be optimized to achieve smaller passband fluctuations and faster transition band attenuation. The expression of the high-pass filter is as follows: are the coefficients of the three trapezoidal windows, satisfying ; are half the window length of the rectangular base window of each trapezoidal window.

[0020] After spatial filtering using a linear-phase FIR digital high-pass filter, the track geometry output from real-time detection is compared laterally. Leveraging the precise spatial positioning characteristics of constant track geometry detection, defects in the turnout area can be effectively located due to the distortion-free detection results. Constructing a local curvilinear coordinate system requires a given reference line, which can be any curve in space. This method defines lateral displacement as the perpendicular distance relative to the base path, and the reference line as the track centerline.

[0021] like Figure 2 As shown, the coordinates of point P on the center line of the rail top surface are defined as Project point P onto the track centerline, with the projection point F. The distance between points P and F is the lateral distance D in the local curvilinear coordinate system. The curvilinear distance from the reference line start point to the projection point F is the longitudinal displacement S. The coordinates of the rail top centerline in the local curvilinear coordinate system are described by (S, D, Z). The resulting mapping is shown in the following equation: in, is the differential arc length along the reference line; in is the coordinate of the projection point F; The above mapping relationship maps the spatial coordinates of the constant track geometry measurement in the turnout area to the longitudinal displacement S and the track amplitude D. This is consistent with the physical meaning of the spatial curves used in real-time track geometry measurement in the turnout area. The track's longitudinal horizontal plane is perpendicular to the plane formed by x and y, so z is equal to the track's longitudinal horizontal amplitude Z.

[0022] like Figure 3 As shown, the steps of the real-time and constant conversion algorithm are as follows: First, determine the reference point r that is closest to a point x on the centerline of the rail top surface. In the global coordinate system, the vector formed by x and r can be linearly expressed as the following formula. In this formula, Is a scalar, representing the vertical distance from the reference point r to the point x on the measured curve: in, is the normal vector; Secondly, according to the above mapping relationship definition, the longitudinal displacement S of the reference point is the longitudinal displacement S of a point on the centerline of the rail top surface. The transverse distance D is the distance between a point on the centerline of the rail top surface and the reference point; in, is the coordinate of the reference point r; Finally, determine the direction. The left direction along the reference line S is positive, and the right direction is negative. The normal vector of the reference point in the global coordinate system is , the tangent vector is The direction is determined by the vector The cross product of .

[0023] The purpose of this step is to determine the direction of the transverse distance D from a point P on the centerline of the rail top surface to its projection point F on the reference line. The direction is determined based on the normal vector of the reference point r and tangent vector The cross product of If the cross product results in a positive direction to the left of the reference line S, and a negative direction to the right, this direction ensures that the sign of the lateral distance D accurately reflects the position of point P relative to the reference line.

[0024] 2. Algorithm for track alignment and matching of turnout curves When track geometric irregularities are superimposed on structural irregularities, the structural irregularities distort the output structure, making it impossible to assess the track quality index using the current maximum value limit and the track quality index. The amplitude of the detection waveform cannot truly reflect the track geometric irregularities in the turnout area, nor can it guide on-site maintenance and repair work, thus affecting the authenticity of the track quality index statistics.

[0025] This paper proposes a sliding matching algorithm for processing track data in the turnout area based on decision threshold judgment. This algorithm eliminates structural irregularities in the turnout guide curve, outputs the difference between the measured turnout track geometry and the designed turnout track geometry, and accurately reflects the track geometry irregularities in the turnout area. This algorithm complies with existing methods for statistically evaluating single-point maximum overrun and section data. The algorithm specifically involves the following steps: 1. Build the turnout matching template for the line According to the plane layout of the turnout, the lateral offset of the turnout branch is extracted according to the preset spatial interval; Performing a second-order difference on the lateral offset and performing digital high-pass filtering to construct a spatial matching template; Normalizing the amplitude of the spatial matching template and symmetrically filling zero values at both ends of the template to ensure that the lengths of different types of turnout templates are the same; The processed turnout templates are stored in the database to build a turnout standard template library.

[0026] Among them, the preset space interval is 0.25 meters.

[0027] 2. Positioning the decision threshold in the turnout area Based on the unique characteristics of the turnout area, combined with information from ground markings and line turnout records, an optimized threshold is set to automatically and efficiently determine whether the data corresponds to a turnout; The data initially screened were further screened to exclude interfering data at the beginning and end of the curve.

[0028] Among them, the optimization thresholds include fixed guide curve radius, insufficient or excessive superelevation, and the difference between the lateral passing speed and the main line.

[0029] 3. Determine the branch decision threshold of the fork area Use an automatic positioning device to distinguish incidental curves; Mark the turnout branch lines and incidental curves to ensure accurate distinction.

[0030] 4. Calculate the Pearson correlation coefficient between the turnout template and the filtered turnout data Calculate the Pearson correlation coefficient between the turnout template and the screened turnout data to further screen and confirm the turnout data; The opening direction of the turnout is determined based on the positive and negative values of the Pearson correlation coefficient; Compare the absolute values of the Pearson correlation coefficient to determine the passing direction of the turnout; Manhattan distance is used as a secondary matching parameter to further filter out noisy data.

[0031] The calculation formula of Pearson correlation coefficient is: The Pearson correlation coefficient in the above formula is a value between -1 and 1, calculated by dividing the covariance by the standard deviation of the two variables; when the linear relationship between the two variables is strengthened, the correlation coefficient tends to 1 or -1. In the formula, given a pair of random variables and ,in, yes and covariance of yes The standard deviation of yes The mean of yes The standard deviation of yes The mean of ; E is the expectation; Positive correlation coefficient: A positive Pearson correlation coefficient indicates a positive correlation with the matching template, which means the opening directions are the same; Negative correlation coefficient: A negative Pearson correlation coefficient indicates a negative correlation with the matching template, which means the opening directions are opposite.

[0032] The turnout opening direction is determined by the Pearson correlation coefficient, and the absolute values of the Pearson correlation coefficients are further compared. Forward travel from the point rail to the frog rail corresponds to the forward template; reverse travel from the frog rail to the point rail corresponds to the reverse template. The reverse template is a mirror image of the forward template. If the maximum absolute value is obtained from the forward template, the turnout is forward; if the maximum absolute value is obtained from the reverse template, the turnout is reverse.

[0033] The track geometry detection system is equipped with an automatic positioning device that can distinguish data from switch and non-switch areas. Based on this, the proposed decision threshold-based method for eliminating structural irregularities in switch areas uses the Pearson coefficient as the primary matching parameter and the Manhattan distance as the secondary matching parameter. The threshold for the Pearson coefficient is set at 0.8. When abnormal data is detected, the measured Pearson coefficient is typically between 0.3 and 0.5, and only data with extremely high similarity can be matched. Therefore, the Pearson coefficient can filter out most noisy data.

[0034] The Manhattan distance elimination algorithm uses the following formula: The Pearson correlation coefficient is used to normalize the data, reducing the weight of the one-dimensional waveform data amplitude. Similar interfering waveforms are then filtered out. The Manhattan distance is then used to compare the waveforms before and after processing, serving as a strong judgment criterion to eliminate interfering waveforms.

[0035] in, Represents the coordinates of two points on a two-dimensional plane.

[0036] To assess the algorithm's matching performance under varying noise levels or with missing data, we calculated the Manhattan distance between the data before and after removing structural irregularities. Based on the physical meaning of the structural irregularity removal algorithm, the average value of the detection data after removal should be significantly smaller than the average value of the detection data before removal. Therefore, calculating the Manhattan distance can eliminate false matches.

[0037] The present invention also provides a system for automatic inspection and intelligent identification of apparent defects of turnouts, the system comprising: At least one processor; and a memory in communication with the at least one processor; wherein the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to perform a method for automatic inspection and intelligent identification of apparent defects of a turnout.

[0038] This invention proposes an automated inspection and intelligent identification method for apparent defects in turnouts. This method is based on a real-time and constant matching method for turnout track geometry, mapping a global coordinate system to a local curvilinear coordinate system. Real-time data is collected when a track geometry vehicle passes through the turnout, while constant data is collected during manual inspection of the turnout without a train passing through. Constant track geometry data in the turnout area is mapped onto a spatial curve derived from real-time track geometry inspection. By comparing the real-time and constant track geometry inspection data for the turnout area, defects in the turnout area can be quickly located. Because the real-time inspection output for the turnout curve track direction includes small-radius curves in the turnout guide curve, structural irregularities at the turnout point rail center, and actual track geometry defects, this invention proposes a turnout sliding matching curve track alignment processing algorithm based on decision threshold judgment to eliminate structural irregularities in the turnout curve track direction. Experimental results demonstrate that both algorithms scan and output the actual track geometry defects in the turnout curve track direction and eliminate the impact of turnout structural irregularities on the track quality index, providing guidance for the maintenance and repair of lines in the turnout area.

[0039] The foregoing is merely a preferred embodiment of the present application and is not intended to limit the present application. Persons skilled in the art will readily appreciate that various modifications and variations are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present application are intended to be within the scope of protection of the present application.

Claims

1. A method for automatic inspection and intelligent identification of apparent defects of turnouts, characterized in that: The following steps are involved: Step 1: Collect real-time track geometry data and constant track geometry data, respectively. The real-time track geometry data is obtained when a track geometry inspection vehicle passes through a turnout, and the constant track geometry data is obtained by manually inspecting the turnout when no train passes. Step 2: convert the constant track geometry data from the global coordinate system to the local curve coordinate system to obtain a space curve comparable to the real-time track geometry data; Step 3: Filtering the spatial curve using a linear phase FIR digital high-pass filter, wherein the linear phase FIR digital high-pass filter is designed using three trapezoidal windows in parallel, wherein one trapezoidal window serves as a basic window and the remaining two trapezoidal windows are used to adjust the passband of the basic window; by adjusting the coefficient and window length parameters of each trapezoidal window, smaller passband fluctuations and faster transition band attenuation are achieved; Step 4: Match the filtered constant track geometry data with the real-time track geometry data, and determine the specific location of the track defect based on the matching difference to achieve accurate positioning of the track geometry defect in the turnout area; In step 5, the alignment data of the turnout curve track is processed using an algorithm based on a decision threshold to eliminate structural irregularities and truly reflect the track geometry defects.

2. The method according to claim 1, characterized in that The specific steps for converting constant track geometry data from the global coordinate system to the local curvilinear coordinate system are: Step 2.1, determining a reference point, wherein the reference point is the point closest to the point to be measured in the constant track geometry data; Step 2.2, calculate the longitudinal displacement and lateral displacement between the measured point and the reference point; Step 2.

3. Determine the direction of the lateral displacement based on the cross product of the normal vector and the tangent vector of the reference point; if the result of the cross product points to the left of the reference line, the direction is positive; if it points to the right, the direction is negative.

3. The method according to claim 1, characterized in that The specific algorithm based on decision threshold is: Step 5.1, constructing a turnout matching template for the line: extracting the lateral offsets of the turnout branches according to the preset spatial intervals based on the plane layout of the turnout; performing a second-order difference on the lateral offsets and performing digital high-pass filtering to construct a spatial matching template; Amplitude normalization is performed on the spatial matching template, and zero values are symmetrically filled at both ends of the template to ensure that the lengths of different types of turnout templates are the same; the processed turnout templates are stored in a database to construct a turnout standard template library; Step 5.2: Locate the decision threshold in the turnout area: Based on the unique characteristics of the turnout area, combined with information from ground markings and the line turnout ledger, set an optimized threshold to automatically and efficiently determine whether the data corresponds to a turnout; Step 5.3: Determine the branch decision threshold of the switch area: Use an automatic positioning device to distinguish the incidental curve; mark the switch branch line and incidental curve to ensure the accuracy of the distinction; Step 5.4, calculate the Pearson correlation coefficient between the turnout template and the screened turnout data; The opening direction of the turnout is determined based on the positive and negative values of the Pearson correlation coefficient; The absolute values of the Pearson correlation coefficients are compared to determine the passing direction of the turnout. The track geometry data is matched with the template using Manhattan distance as a secondary matching parameter to eliminate structural irregularities in the turnout area.

4. The method according to claim 3, characterized in that A positive Pearson correlation coefficient indicates a positive correlation with the matching template, and the opening direction is the same; a negative Pearson correlation coefficient indicates a negative correlation with the matching template, and the opening direction is opposite; passing from the point rail to the frog rail is a positive direction, and the corresponding template is a positive template; passing from the frog rail to the point rail is a reverse direction, and the corresponding template is a reverse template; If the maximum absolute value is provided by the forward template, the turnout passes in the forward direction; if the maximum absolute value is provided by the reverse template, the turnout passes in the reverse direction.

5. A system for automatic inspection and intelligent identification of apparent defects of turnouts, characterized by: The system comprises: At least one processor; and a memory in communication with the at least one processor; wherein the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute a method for automatic inspection and intelligent identification of apparent defects of a turnout as described in any one of claims 1 to 4.

Images