Engineering cableway stockyard point space data clustering method and system based on road feature enhancement

Abstract

The present application belongs to the technical field of power transmission line engineering cableway design and construction, and particularly relates to an engineering cableway stockyard point spatial data clustering method and system based on road feature enhancement. The method comprises the following steps: S1, road feature extraction: extracting road center line feature data, road average width feature data, and road unit direction feature data; S2, road feature enhancement processing: calculating boundary minimum distance feature data, transverse offset feature data based on cross product, and longitudinal projection position feature data; S3, spatial data clustering: based on the enhanced road features, using the K-means clustering algorithm, the stockyard points are automatically divided into left stockyard point set and right stockyard point set; S4, spatial consistency optimization: based on the principle of geographical spatial continuity, the outlier point label is corrected through the neighbor point voting mechanism. The accuracy and generalization of the spatial data clustering of the stockyard points of the power transmission line engineering cableway are improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of power transmission line engineering cableway design and construction, and particularly relates to an engineering cableway stockyard point spatial data clustering method and system based on road feature enhancement. BACKGROUND

[0002] In the mountainous construction of a power transmission line, a cableway is a core channel for material transportation. Stockyards (i.e., temporary material stacking points) are distributed along both sides of a construction road, and the left and right stockyards need to correspond to independent cableway systems. Incorrect partitioning will lead to secondary material transportation, extend the construction period by more than 30%, and seriously reduce the transportation efficiency.

[0003] The existing partitioning methods include the following:

[0004] The first method is a manual surveying and mapping method that relies on experience and judgment. The error rate of a curved road section is greater than 40%, and it takes a long time (2 people per day for a 1 km road section). In particular, for S-shaped curved stockyards, the left and right stockyards are easily confused.

[0005] The second method is a traditional K-means clustering method that ignores the road direction and only uses Euclidean distance clustering. It cannot perceive the orientation attribute and easily causes the same-side stockyards of a curved section to be split.

[0006] The third method is a road centerline constraint method that requires a high-precision centerline model (unmanned aerial vehicle aerial surveying and fitting). The cost is high and not suitable for temporary roads. In particular, the centerline of a gravel road is shaken, which causes the partitioning to deviate. SUMMARY

[0007] The technical problem to be solved by the application is to provide an engineering cableway stockyard point spatial data clustering method and system based on road feature enhancement. The accuracy and generalization of the power transmission line engineering cableway stockyard point spatial data clustering are improved.

[0008] The technical solution adopted by the application to solve the technical problem is: an engineering cableway stockyard point spatial data clustering method based on road feature enhancement, including the following steps:

[0009] S1, road feature extraction, including extracting road centerline feature data, road average width feature data, and road unit direction feature data;

[0010] S2, road feature enhancement processing, including calculating boundary minimum distance feature data based on the road centerline feature data, calculating transverse offset feature data based on the cross product based on the road average width feature data, and calculating longitudinal projection position feature data based on the road unit direction feature data;

[0011] S3, spatial data clustering: based on the minimum distance feature data of the boundary, the lateral offset feature data based on the cross product, and the longitudinal projection position feature data, the K-means clustering algorithm is used to automatically divide the stockyard points into left stockyard point set and right stockyard point set;

[0012] S4, spatial consistency optimization: based on the principle of geographical spatial continuity, the real stockyard points are continuously distributed in the extension direction of the road, and the outlier point label is corrected through the neighbor point voting mechanism.

[0013] Further, the step S1 of extracting the road center line feature data comprises the steps of: based on the linear interpolation method of the road boundary point pair, the center point coordinates of the corresponding boundary points are calculated to construct a continuous center line.

[0014] Further, the step S1 of calculating the center point coordinates of the corresponding boundary points comprises the steps of:

[0015] S111, the road boundary point set on both sides is respectively left boundary point set and right boundary point set;

[0016] S112, boundary sorting and point pair matching;

[0017] S113, the center point coordinates are calculated by using the center point calculation formula.

[0018] Further, the road average width feature data in step S1 is extracted by using the geometric mean method, comprising the steps of:

[0019] S121, the road boundary point set on both sides is respectively left boundary point set and right boundary point set;

[0020] S122, calculate the sectional width;

[0021] S123, calculate the average width of the sectional width.

[0022] Further, the road unit direction feature data in step S1 is extracted, comprising the steps of:

[0023] S131, collect the road start point coordinates and the road end point coordinates;

[0024] S132, calculate the road start-end point vector difference according to the road start point coordinates and the road end point coordinates;

[0025] S133, calculate the road length according to the road start point coordinates and the road end point coordinates;

[0026] S134, calculate the road unit direction vector according to the road start-end point vector difference and the road length.

[0027] Further, the step S2 of calculating the boundary minimum distance feature data according to the road center line feature data comprises the steps of: calculating the nearest neighbor distance of the stockyard point to the road boundary point on both sides according to the input stockyard point coordinates and road boundary coordinates, and obtaining the boundary minimum distance feature data after normalization processing.

[0028] Further, the step S2 of calculating the cross product based lateral offset feature data according to the road average width feature data comprises the steps of: determining the orientation of the stockyard point relative to the road segment by the right-hand rule of vector cross product, and obtaining the cross product based lateral offset feature data after sign distance normalization processing.

[0029] Further, the step S2 of calculating the longitudinal projection position feature data according to the road unit direction feature data comprises the steps of:

[0030] S201, determining a road direction reference, taking the road starting point as the origin and the road unit direction vector as the road direction reference axis;

[0031] S202, projection operation, calculating the projection length of the stockyard point vector on the road direction reference axis;

[0032] S203, normalization processing, dividing the projection length by the total length of the road to eliminate the scale effect.

[0033] Further, the step S3 of automatically dividing the stockyard points into a left stockyard point set and a right stockyard point set comprises the steps of:

[0034] S301, using a deterministic initialization method, calculating the weighted feature mean of the left candidate point set and the right candidate point set based on the signed lateral offset feature to pre-classify the stockyard points, when , the candidate point is a left candidate point; when , the candidate point is a right candidate point; calculating the weighted feature mean of the left candidate point set and the right candidate point set respectively to determine the left cluster center and the right cluster center;

[0035] S302, adjusting the feature weight according to the road curvature;

[0036] S303, distributing the stockyard points to the nearest cluster according to the dynamic weight, the feature matrix, and the cluster center, and updating the cluster center; wherein the determination basis for updating the cluster center is whether the clustering label converges, and the determination steps of whether the clustering label converges are:

[0037] S3031, setting the convergence threshold to 0.000001;

[0038] S3032, ensuring that it is not the first iteration;

[0039] S3033, calculating the number of points with different labels.

[0040] S3034, calculate the proportion of label change points;

[0041] S3035, judge whether the change proportion is less than a convergence threshold value;

[0042] When the classification label change rate of the stockyard point is less than 0.001%, it is considered that the clustering result has been stable.

[0043] Further, the spatial consistency optimization in step S4 includes the steps of:

[0044] S401, definition of spatial neighbors, introducing triple geometric constraints through spatial neighbors;

[0045] S402, spatial neighbor search.

[0046] The system for clustering spatial data of engineering cableway stockyard points based on road feature enhancement adopts a method for clustering spatial data of engineering cableway stockyard points based on road feature enhancement, which includes a road feature extraction module, a road feature enhancement processing module, a stockyard point spatial data clustering module, and a stockyard point spatial consistency optimization module.

[0047] The road feature extraction module is configured to extract road center line feature data, road average width feature data, and road direction feature data.

[0048] The road feature enhancement processing module calculates boundary minimum distance feature data based on the extracted road center line feature data, calculates cross product-based lateral offset feature data according to the road average width feature data, and calculates longitudinal projection position feature data according to the road unit direction feature data.

[0049] The stockyard point spatial data clustering module is configured to automatically divide stockyard points into a left stockyard point set and a right stockyard point set.

[0050] The stockyard point spatial consistency optimization module corrects outlier point labels based on a geographical spatial continuity principle through a neighbor point voting mechanism.

[0051] Compared with the prior art, the present application has the beneficial effects that: the present application provides an engineering cableway yard point spatial data clustering method and system based on road feature enhancement. The accuracy and generalization of the spatial data clustering of the cableway yard points of the power transmission line engineering are improved. In terms of curved road section classification, the error rate of the manual surveying and mapping method is greater than 40%, and the error rate of the present method is less than 5%, and the accuracy is greatly improved. The processing time of 10,000 points is less than 10 minutes, which meets the 24-hour construction period requirement and improves the efficiency. The present method supports variable-width roads (3-15 m) and sparse yard points (density <0.5 points / 10 m), and has strong engineering adaptability. The present method can also avoid the risk of cableway collision and reduce the secondary transportation cost.

[0052] The present application also has the following advantages:

[0053] Firstly, the present application solves the problem of accurate determination of the orientation (left / right side) of the yard points on the curved road section, breaks through the limitations of the traditional Euclidean distance clustering in the curved road section, and realizes three-dimensional spatial orientation perception.

[0054] Secondly, the present application calculates the position of the yard points along the road extension direction through projection, provides quantitative basis for cableway transportation path optimization and material scheduling sequence planning, for example, quantifies the position of the yard points in the road extension direction, provides position reference for cableway segmented erection, guides the material transportation sequence, and solves the technical bottleneck that the traditional method cannot represent the longitudinal distribution of the yard points.

[0055] Thirdly, the present application solves the aliasing problem of the traditional clustering method in the curved road section through feature space reconstruction, and realizes high-precision classification of the yard orientation.

[0056] Fourthly, the present application solves the problems of misclassification of individual yard points due to measurement error or terrain mutation, and incorrect allocation of a small number of points due to feature space overlap in sharp curved road sections. BRIEF DESCRIPTION OF DRAWINGS

[0057] Figure 1 The flowchart of the present application is shown in the figure;

[0058] Figure 2 The flowchart of the extraction of the road centerline is shown in the figure;

[0059] Figure 3 The flowchart of the calculation of the average width of the road is shown in the figure;

[0060] Figure 4 The flowchart of the calculation of the unit direction vector of the road is shown in the figure;

[0061] Figure 5 The flowchart of the calculation of the minimum distance of the boundary is shown in the figure;

[0062] Figure 6 The flowchart of the calculation of the transverse offset based on the cross product is shown in the figure;

[0063] Figure 7 Flow chart for calculating longitudinal projection position;

[0064] Figure 8 Flow chart for spatial data clustering;

[0065] Figure 9 Flow chart for spatial consistency optimization;

[0066] Figure 10 System structure diagram. DETAILED DESCRIPTION

[0067] The application is further illustrated below in conjunction with the accompanying drawings Figure 1 , 2 , 3, 4, 5, 6, 7, 8, 9 and 10, and the embodiments.

[0068] The engineering cableway stockyard point spatial data clustering method based on road feature enhancement comprises the following steps:

[0069] S1, road feature extraction: including extracting road center line feature data, road average width feature data, and road unit direction feature data. Road feature refers to road geometric feature.

[0070] Specifically, the step of extracting road center line feature data in step S1 comprises the following steps: based on the linear interpolation method of road two-side boundary point pairs, the center point coordinates of the corresponding boundary points are calculated to construct a continuous center line.

[0071] Further, the step of calculating the center point coordinates of the corresponding boundary points in step S1 comprises the following steps:

[0072] S111, assuming that the road two-side boundary point sets are left-side boundary point set and right-side boundary point set;

[0073] Left-side boundary point set: ;

[0074] Right-side boundary point set: ;

[0075] Wherein:

[0076] : number of boundary points, = ;

[0077] : left-side boundary point coordinates, input by survey data;

[0078] : right-side boundary point coordinates, input by survey data; S112, boundary sorting and point pair matching;

[0079] S113, the center point coordinates are calculated by using the center point calculation formula;

[0080] Center point calculation formula:

[0081] Wherein:

[0082] : the first center point coordinates;

[0083] : the corresponding position left and right boundary point coordinates;

[0084] Specifically, the geometric mean method is used to extract the road average width feature data in step S1, including the following steps:

[0085] S121, the road boundary point set on both sides is respectively set as the left boundary point set and the right boundary point set;

[0086] Left boundary point set: ;

[0087] Right boundary point set: ;

[0088] The road has four boundary point pairs;

[0089] S122, the sub-width is calculated;

[0090]

[0091] S123, the average width of the sub-width is calculated;

[0092]

[0093] The direction vector normalization method based on the road head and tail boundary points is used to realize the overall trend representation, the head and tail point connection line represents the road macro trend, three-dimensional direction perception, including plane direction and elevation change, unitization processing, eliminating the influence of scale, adapting to different length road sections.

[0094] Specifically, the road unit direction feature data in step S1 is extracted, including the following steps:

[0095] S131, the road head point coordinates and the road tail point coordinates are collected;

[0096] Road head point coordinates: ;

[0097] Road tail point coordinates: ;

[0098] The road start point coordinate is the first data after the data matrix is formed by scanning the road boundary by cell, and the road end point coordinate is the last data after the data matrix is formed by scanning the road boundary by cell.

[0099] S132, calculating a road start-end point vector difference according to the road start point coordinate and the road end point coordinate ;

[0100] S133, calculating a road length according to the road start point coordinate and the road end point coordinate ;

[0101] S134, calculating a road unit direction vector according to the road start-end point vector difference and the road length . .

[0102] S2, road feature enhancement processing: including calculating boundary minimum distance feature data according to road center line feature data, calculating cross product based lateral offset feature data according to road average width feature data, and calculating longitudinal projection position feature data according to road unit direction feature data.

[0103] Specifically, the boundary minimum distance feature data calculated according to the road center line feature data in step S2 includes the following steps: using KD tree search to calculate the nearest neighbor distance of the stockyard point to the boundary points on both sides of the road according to the input stockyard point coordinate and road boundary coordinate, and obtaining the boundary minimum distance feature data after normalization processing.

[0104] Based on the nearest neighbor distance mapping of the boundaries on both sides of the road, the three-dimensional spatial relationship between the stockyard point and the boundary is quantified into three independent features, including: 1) boundary directivity, separate left / right side boundary distance calculation; 2) spatial symmetry, complementary distance features on both sides; 3) scale invariance, width normalization to adapt to different specifications of roads.

[0105] Let:

[0106] , stockyard point coordinate;

[0107] , road boundary coordinate;

[0108] , road boundary coordinate;

[0109] , road average width;

[0110] Boundary minimum distance feature calculation formula:

[0111]

[0112] wherein,

[0113] is the Euclidean norm:

[0114]

[0115] : stockyard point coordinate;

[0116] : road left boundary coordinate;

[0117] : road right boundary coordinate;

[0118] : road average width;

[0119] : minimum distance of stockyard point to road left boundary;

[0120] : minimum distance of stockyard point to road right boundary;

[0121] : minimum distance of stockyard point to road left boundary road width normalized value to adapt to different specifications of road;

[0122] : minimum distance of stockyard point to road right boundary road width normalized value to adapt to different specifications of road;

[0123] Specifically, the cross product-based lateral deviation feature data calculated according to the road average width feature data in step S2 includes the steps of: determining the orientation of the stockyard point relative to the road segment by using the right-hand rule of vector cross product, and obtaining the cross product-based lateral deviation feature data after symbol distance normalization processing. The problem of accurate determination of the orientation (left / right side) of the stockyard point on the curved road segment is solved, the limitation of traditional Euclidean distance clustering on curved road segments is broken through, and three-dimensional spatial orientation perception is realized.

[0124] Let the stockyard point coordinate be , and the calculation formula is as follows:

[0125]

[0126] wherein,

[0127] : three-dimensional coordinate of the stockyard point to be classified;

[0128] : three-dimensional coordinate of the road lower right boundary;

[0129] : 3D coordinate of the left upper boundary of the road;

[0130] : Road width direction vector (vector from the lower right to the upper left);

[0131] : Vector from the stockpile point to the left upper boundary of the road;

[0132] : Cross product vector of

[0133] : Z component of the cross product vector of , the sign determines the orientation, negative value: the stockpile point is on the right side of the road, positive value: the stockpile point is on the left side of the road;

[0134] : Sign piecewise function;

[0135] : Modulus of the cross product vector of

[0136] : Modulus of vector

[0137] : Perpendicular distance between the modulus of the cross product vector of and the modulus of the road width direction vector;

[0138] : Output characteristic value normalized lateral offset, value range [-1, 1], negative value = left side, positive value = right side;

[0139] : Floating point calculation tolerance;

[0140] : Road average width;

[0141] Specifically, the longitudinal projection position characteristic data is calculated according to the road unit direction characteristic data in step S2, including the steps of: mapping the stockpile point to the road direction reference axis based on vector projection theory, and obtaining the longitudinal projection position characteristic data after normalization processing. The position of the stockpile point in the extension direction of the road is calculated by projection, which provides a quantitative basis for cableway transportation path optimization and material scheduling sequence planning, for example, quantifying the position of the stockpile point in the extension direction of the road, providing a position reference for cableway segmented erection, guiding the material transportation sequence, and solving the technical bottleneck that the traditional method cannot represent the longitudinal distribution of the stockpile point.​​​

[0142] Specifically, the calculation of the longitudinal projection position feature data according to the road unit direction feature data in step S2 comprises the steps of:

[0143] S201, determining a road direction reference, taking a road starting point as an origin and a road unit direction vector as a road direction reference axis;

[0144] S202, projection operation, calculating a projection length of a stockyard point vector on the road direction reference axis;

[0145] The stockyard point vector is a stockyard point vector matrix to be input, which is a raw data set to be processed and analyzed. The stockyard point vector matrix is formed by a data matrix formed by the last process through the road boundary and the cell scanning.

[0146] S203, normalization processing, the projection length is divided by the total length of the road to eliminate the scale effect.

[0147] Let the stockyard point coordinates be The calculation formula is as follows:

[0148]

[0149] Wherein,

[0150] : the coordinate of the starting point of the road;

[0151] : the unit direction vector of the road;

[0152] : the total length of the road;

[0153] : the direction vector of the coordinates of any stockyard site to the starting point;

[0154] : the dot product of the direction vector of the coordinates of any stockyard site to the starting point and the unit direction vector of the road is projected to obtain the projection length;

[0155] : the output feature value, the projection length is divided by the total length of the road to eliminate the scale effect;

[0156] Formula explanation:

[0157] : 0 represents the starting point of the road, and 1 represents the end point of the road;

[0158] When or , it means that the stockyard point is outside the road extension line;

[0159] S3, spatial data clustering: based on the minimum distance feature data of the boundary, the lateral offset feature data based on the cross product, and the longitudinal projection position feature data, the K-means clustering algorithm is used to automatically divide the stockyard points into a left stockyard point set and a right stockyard point set. Through feature space reconstruction, the aliasing problem of traditional clustering methods in curved road sections is solved, and high-precision classification of the stockyard orientation is realized.

[0160] Specifically, the automatic division of the stockyard points into the left stockyard point set and the right stockyard point set in step S3 includes the steps of:

[0161] S301, using a deterministic initialization method, based on signed lateral offset features Pre-classify the stockyard points, when , it is a left candidate point; when , it is a right candidate point; calculate the weighted feature mean of the left candidate point set and the right candidate point set respectively, and determine the left cluster center and the right cluster center;

[0162] S302, adjust the feature weight according to the road curvature;

[0163] S303, according to the dynamic weight, the feature matrix, and the cluster center, assign the stockyard points to the nearest cluster, and update the cluster center; wherein the determination basis for updating the cluster center is whether the clustering label converges, and the determination steps of whether the clustering label converges are:

[0164] S3031, set the convergence threshold to 0.000001;

[0165] S3032, ensure that it is not the first iteration; because there is no old label to compare at the first iteration;

[0166] S3033, calculate the number of points with different new and old labels; that is, count the number of points with changed labels;

[0167] S3034, calculate the proportion of points with changed labels;

[0168] S3035, judge whether the change proportion is less than the convergence threshold;

[0169] When the change rate of the classification label of the stockyard point is less than 0.001%, that is, , it is considered that the clustering result has stabilized.

[0170] Let the stockyard point set , each point has 4-dimensional enhanced features:

[0171]

[0172] Clustering objective function:

[0173]

[0174] in,

[0175] Cluster center vector ( Indicates the left side. (Indicates the right side);

[0176] :point Does it belong to a cluster? ;

[0177] Weighted Euclidean distance;

[0178] Total number of points in the material yard;

[0179] Weighted Euclidean distance (improved distance calculation):

[0180]

[0181] in,

[0182] Weighting coefficients are dynamically adjusted based on road morphology.

[0183] Input feature value, material yard point Normalized distance to boundary 1;

[0184] Input feature value, material yard point Normalized distance to boundary 2;

[0185] Input feature value, signed horizontal offset, i.e., material yard point. Location relative to a road segment;

[0186] Input feature value, longitudinal projection location, i.e., material yard point. Location along the direction of road extension;

[0187] Material yard For input feature values The cluster center vector;

[0188] Material yard For input feature values The cluster center vector;

[0189] Material yard For input feature values The cluster center vector;

[0190] : stockpile point For input feature values of cluster center vectors;

[0191] S4, spatial consistency optimization: based on the principle of geographical spatial continuity, the real stockpile points show continuous distribution in the direction of road extension, and the outlier point label is corrected through the neighbor point voting mechanism. Solve the problem that individual stockpile points are misclassified due to measurement error or terrain mutation, and a small number of points are misallocated due to feature space overlap in sharp curve road sections.

[0192] Specifically, the spatial consistency optimization in step S4 includes the following steps:

[0193] S401, definition of spatial neighbors, introduce triple geometric constraints through spatial neighbors; Ensure that the neighbor points and target points have reliable spatial correlation;

[0194] S402, spatial neighbor search.

[0195] Let the input clustering label set ( ), the optimized clustering label set ;

[0196] For each point :

[0197]

[0198] Wherein,

[0199] : the spatial neighbor set of point ;

[0200] : indicator function (true = 1, false = 0);

[0201] : distance weight ( );

[0202] : clustering confidence.

[0203] The system for clustering the spatial data of the stockpile points of the engineering cableway based on road feature enhancement adopts a method for clustering the spatial data of the stockpile points of the engineering cableway based on road feature enhancement, which includes a road feature extraction module, a road feature enhancement processing module, a stockpile point spatial data clustering module, and a stockpile point spatial consistency optimization module.

[0204] The road feature extraction module is configured to extract road center line feature data, road average width feature data, and road direction feature data, thereby providing basic parameters for road feature enhancement.

[0205] The road feature enhancement processing module is configured to calculate boundary minimum distance feature data based on the extracted road center line feature data, calculate cross product-based lateral offset feature data based on the road average width feature data, and calculate longitudinal projection position feature data based on the road unit direction feature data, thereby providing feature parameters for stockyard point spatial clustering.

[0206] The stockyard point spatial data clustering module is configured to automatically divide stockyard points into a left stockyard point set and a right stockyard point set, thereby achieving high-precision classification of stockyard point data.

[0207] The stockyard point spatial consistency optimization module is configured to correct outlier point labels through a neighbor point voting mechanism based on the principle of geographic spatial continuity, thereby optimizing the classification accuracy of the stockyard.

[0208] The embodiments of the specific implementation are preferred embodiments of the present application, and are not intended to limit the protection scope of the present application. Equivalent changes made in the structure, shape, and principle of the present application should be included in the protection scope of the present application.

Claims

1. A method for clustering of engineering ropeway stockyard point spatial data based on road feature enhancement, characterized in that, The method comprises the steps of: S1, road feature extraction: including extracting road centerline feature data, road average width feature data, and road unit direction feature data; S2, road feature enhancement processing: including calculating boundary minimum distance feature data according to road centerline feature data, determining the orientation of the stockyard point relative to the road section by using the right-hand rule of vector cross product, obtaining cross product-based lateral offset feature data after symbol distance normalization processing, and calculating longitudinal projection position feature data according to road unit direction feature data; S3, spatial data clustering: based on the boundary minimum distance feature data, the cross product-based lateral offset feature data, and the longitudinal projection position feature data, the K-means clustering algorithm is used to automatically divide the stockyard points into left stockyard point set and right stockyard point set; S4, spatial consistency optimization: based on the principle of geographical spatial continuity, the real stockyard points are continuously distributed in the extension direction of the road, and the outlier point label is corrected through the neighbor point voting mechanism.

2. The road feature enhancement based engineering ropeway stock yard point spatial data clustering method as claimed in claim 1, wherein, The road centerline feature data extraction in step S1 comprises the steps of:

3. The road feature enhancement based engineering ropeway stock yard point spatial data clustering method of claim 2, wherein, S111, setting the road two-side boundary point set as left-side boundary point set and right-side boundary point set; S112, boundary sorting and point pair matching; S113, using the center point calculation formula to obtain the center point coordinates. The road average width feature data extraction in step S1 comprises the steps of:

4. The road feature enhancement based engineering ropeway stock yard point spatial data clustering method as claimed in claim 1, wherein, S121, setting the road two-side boundary point set as left-side boundary point set and right-side boundary point set; S122, calculating the segmented width; S123, calculating the average width of the segmented width. The road unit direction feature data extraction in step S1 comprises the steps of:

5. The road feature enhancement based engineering ropeway stock yard point spatial data clustering method as claimed in claim 1, wherein, S131, collecting the road start point coordinates and the road end point coordinates; S132, calculating the road start-end point vector difference according to the road start point coordinates and the road end point coordinates; S133, calculating the road length according to the road start point coordinates and the road end point coordinates; S134, calculating the road unit direction vector according to the road start-end point vector difference and the road length. The boundary minimum distance feature data calculation in step S2 comprises the steps of: using KD tree search to calculate the nearest neighbor distance of the stockyard point to the road two-side boundary points according to the input stockyard point coordinates and road boundary coordinates, and obtaining the boundary minimum distance feature data after normalization processing.

6. The road feature enhancement based engineering ropeway stock yard point spatial data clustering method as claimed in claim 1, wherein, The longitudinal projection position feature data calculation in step S2 comprises the steps of:

7. The road feature enhancement based engineering ropeway stock yard point spatial data clustering method as claimed in claim 1, wherein, S201, determining the road direction reference, taking the road start point as the origin and the road unit direction vector as the road direction reference axis; S202, projection operation, calculating the projection length of the stockyard point vector on the road direction reference axis; S203, normalization processing, dividing the projection length by the total road length to eliminate the scale effect. ​ 8. The road feature enhancement based engineering ropeway stock yard point spatial data clustering method as claimed in claim 1, wherein, The automatic division of the stockyard points into the left stockyard point set and the right stockyard point set in step S3 comprises the steps of: S301, using a deterministic initialization method, based on signed lateral offset features Carry out stockyard point pre-classification, when is a left candidate point; when is a right candidate point; respectively calculate the weighted feature mean of the left candidate point set and the right candidate point set, and determine the left cluster center and the right cluster center; S302, adaptively adjusting the feature weight according to the road curvature; S303, distributing the stockyard points to the nearest cluster and updating the cluster center according to the dynamic weight, the feature matrix, and the cluster center; wherein the judgment basis for updating the cluster center is whether the clustering label converges, and the judgment steps for whether the clustering label converges are: S3031, setting the convergence threshold to 0.000001; S3032, ensuring that it is not the first iteration; S3033, calculating the number of points with different labels between the new and old labels; S3034, calculating the proportion of the label change points; S3035, judging whether the change proportion is less than the convergence threshold; When the classification label change rate of the stockyard points is less than 0.001%, it is considered that the clustering result has stabilized.

9. The road feature enhancement based engineering ropeway stock yard point spatial data clustering method as claimed in claim 1, wherein, The spatial consistency optimization in step S4 comprises the steps of: S401, definition of spatial neighbors, introducing triple geometric constraints through spatial neighbors; S402, spatial neighbor searching.

10. A system for road feature enhancement based clustering of ropeway stockyard point spatial data, characterized in that, The engineering cableway stockyard point spatial data clustering method based on road feature enhancement according to any one of claims 1-9 comprises a road feature extraction module, a road feature enhancement processing module, a stockyard point spatial data clustering module, and a stockyard point spatial consistency optimization module; The road feature extraction module is used for extracting road center line feature data, road average width feature data, and road direction feature data; The road feature enhancement processing module calculates boundary minimum distance feature data based on the extracted road center line feature data, judges the orientation of the stockyard points relative to the road section by using the right-hand rule of vector cross product, obtains transverse offset feature data based on cross product after symbol distance normalization processing, and calculates longitudinal projection position feature data according to road unit direction feature data; The stockyard point spatial data clustering module is used for automatically dividing the stockyard points into the left stockyard point set and the right stockyard point set; The stockyard point spatial consistency optimization module corrects the outlier label based on the geographical spatial continuity principle and the neighbor point voting mechanism.

Images