Method for automatically generating height measurement point of cargo loading end view based on laser radar

Abstract

The invention discloses a method for automatically generating a height measurement point of a cargo loading end view based on a laser radar. The method is characterized in that the algorithm for automatically generating the height measurement point of the cargo loading end view based on the laser radar comprises the following steps: step 1, slicing an end view point cloud in an XOY plane along a Y axis; 2, solving the maximum and minimum X coordinates of the point cloud in each slice, solving the average X value, and generating a point according to the Y value of the current slice; and step 3, carrying out RANSAC straight line fitting on all the generated points, wherein the fitted straight line is the center line of the object. According to the method, the height measurement points can be automatically generated according to the cargo end view point cloud, irregular cargos exceeding the range of the carrier or the protruding part of the carrier can be accurately recognized, the height measurement points and the rectangular frame can be generated, and the polygon described by the height measurement points can accurately summarize the cargo polygon; and the polygon of the cargo does not need to be summarized by using cargo contour data of tens of thousands of points, so that a worker can record height measurement data of the cargo very conveniently.

Description

Technical Field

[0001] The present invention relates to the technical field of methods for automatically generating end-view measurement height points for cargo loading using laser radar, and more specifically to a method for automatically generating end-view measurement height points for cargo loading based on laser radar. Background Art

[0002] Accurately obtaining the dimensions of railway cargo is of great significance for improving transportation safety, improving transportation efficiency, complying with transportation regulations, and optimizing resource allocation. The height measurement point provides the key dimensions, loading positions, and contour shapes of cargo during transportation, which facilitates the planning and implementation of railway transportation. With the modernization and intelligent development of railway freight, the measurement of dimensions will increasingly become the core data in railway transportation, promoting the establishment of a more refined, automated, and efficient freight management system.

[0003] The traditional height measurement method mainly includes the use of handheld laser radar to quickly and accurately collect the point cloud data of the measured object and calculate the point cloud of the cargo end view. It has the advantages of high-resolution collection, high precision, and strong anti-interference ability, but it still requires manual point selection in the map, which consumes a lot of manpower and wastes unnecessary time. Therefore, a new technical solution is needed to solve this problem. Summary of the invention

[0004] The purpose of the present invention is to provide a method for automatically generating a cargo loading end view measurement high point based on laser radar, which solves the problem that manual points need to be taken in the map later, consuming a lot of manpower and wasting unnecessary time.

[0005] To achieve the above object, the present invention provides the following technical solution: a method for automatically generating a cargo loading end view measurement high point based on a laser radar, and the algorithm for automatically generating a cargo loading end view measurement high point based on a laser radar comprises the following steps:

[0006] Step 1. Slice the end view point cloud in the XOY plane along the Y axis;

[0007] Step 2. Find the maximum and minimum X coordinates of the point cloud in each slice, find the average X value, and generate a point based on the Y value of the current slice;

[0008] Step 3. Perform RANSAC straight line fitting on all generated points, and the fitted straight line is the center line of the object;

[0009] Step 4. Fold the object point cloud to the left along the center line of the object, and then process the folded point cloud F;

[0010] Step 5. Find the convex hull point cloud and concave hull point cloud of point cloud F;

[0011] Step 6. Remove the convex hull points whose distance from the center line is less than 100 mm from the convex hull point cloud;

[0012] Step 7. Cluster the convex hull point cloud and sort it from high to low according to the Y value;

[0013] Step 8. Sort the point clouds in each result class of the clustering from high to low according to the Y value;

[0014] Step 9. Determine whether it is an inflection point;

[0015] Step 10. Determine whether the first and last points of each class are inflection points;

[0016] Step 11. Generate a focus for each class based on the x value of the object's centerline and the average of the maximum and minimum Y values ​​of each class retained in step 10;

[0017] Step 12. In each class, find the point with the largest distance from the focus of that class as the inflection point;

[0018] Step 13. Add the inflection points in step 10 and step 12 to the key point cloud set as candidate points for height measurement

[0019] Step 14. Segment the point cloud L of the leftmost 1 / 6 of the object;

[0020] Step 15. Filter out the first 40% points with the smallest X value in the point cloud L as the new point cloud TL, and sort the points in TL from high to low according to the Y value;

[0021] Step 16. Cluster the sorted TLs and sort each class and the points in the class from high to low according to the Y value;

[0022] Step 17. Find the class with the smallest point X value, which is the most prominent class LM;

[0023] Step 18. Determine whether the class is located at the bottom of point cloud F according to the minimum Y value of the class and the minimum Y value of point cloud F. If yes, jump to step 20. If not, determine whether the class is the class with the smallest Y value. If yes, find the minimum enclosing rectangle AABB of the point cloud below the class. If not, find the class with the smallest X value below the class and execute step 18 again.

[0024] Step 19. If there is no class above LM, jump to step 20, otherwise find the minimum enclosing rectangle AABB of all classes and the middle point clouds above LM;

[0025] Step 20. Calculate the minimum enclosing rectangle AABB for all the classes found in step 16 and add it to the rectangle set as a candidate rectangle together with the minimum enclosing rectangle AABB calculated in steps 18 and 19;

[0026] Step 21. Determine whether to merge two adjacent rectangles into one rectangle based on whether the X difference between adjacent candidate rectangles is greater than 20 mm, and obtain the final rectangle set;

[0027] Step 22. Delete the candidate height measurement points whose Y values ​​are lower than the maximum Y value in the rectangle set;

[0028] Step 23. If the number of candidate points is greater than 2, determine whether to merge the adjacent points into one point based on whether the X and Y differences of the adjacent candidate points are greater than 10 mm;

[0029] Step 24. Add the highest point of point cloud F to the candidate points, and sort the candidate points from high to low by Y value;

[0030] Step 25. Merge the point with the highest Y value and the second highest Y value among the candidate points according to the merging rule in step 23 to obtain the final set of height measurement points.

[0031] As a preferred implementation manner of the present invention, the concave hull point cloud of step 5 is the point cloud contour.

[0032] As a preferred embodiment of the present invention, the rule for determining whether it is an inflection point in step 9 is: whether the angle between the line connecting the point and the point after it and the line connecting the point and the point before it is greater than 14 degrees, if so, it is regarded as an inflection point, otherwise it is not an inflection point.

[0033] As a preferred embodiment of the present invention, the judgment rule in step 10 is: if yes, retain the class; if not, determine whether there is an inflection point inside the class. The judgment rule is the same as step 9. If there is an inflection point inside, retain the class; otherwise, do not retain it.

[0034] As a preferred embodiment of the present invention, in step 23, only all points after the candidate point with the highest Y value are processed. If merged, they are merged according to the following rules: adjacent candidate points a, b (Y values ​​from high to low), the previous point c of a, the next point d of b, find the line L1 connecting a and c and the line L2 connecting b and d, and find the intersection of L1 and L2, which is the point after a and b are merged.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] In order to improve the detection accuracy and efficiency, the present invention adopts a handheld 3D laser scanner to quickly obtain the point cloud data of the object to be measured after loading, and compares the projection of the flatbed truck and the cargo with the limit to obtain the over-limit level by intelligently identifying the flatbed truck and calculating the projection surface, thereby realizing accurate and efficient over-limit detection. Its fully automatic recognition and calculation do not require workers to climb on top of the cargo for manual measurement, which greatly reduces the labor intensity and safety risks. The handheld 3D laser scanner can be used on demand, has strong flexibility and strong environmental adaptability, and can meet the over-limit detection needs in different scenarios. In order to realize the visualization of the over-limit detection results, the limit polygon and the projection point cloud are visualized, a picture is generated, and the over-limit level is displayed. If there is an over-limit, the over-limit part is framed in the picture according to the rectangle, so that the detection result is intuitive and easy to understand. This method can greatly improve the working efficiency, reduce labor costs, and at the same time improve the transportation safety, bringing significant economic benefits to the logistics industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is the original point cloud view of the present invention;

[0038] Figure 2 It is a schematic diagram of the concave point (contour) of the original point cloud and the alternative fitting midline obtained by the slicing method of the present invention;

[0039] Figure 3 It is a schematic diagram of the point cloud after the end-view point cloud of the present invention is folded to the left;

[0040] Figure 4 It is a schematic diagram of the concave point of the point cloud after folding in half according to the present invention;

[0041] Figure 5 This is a schematic diagram of the convex hull point cloud of the present invention;

[0042] Figure 6 This is a schematic diagram of the point cloud clustering result after removing the convex hull points near the midline in the present invention;

[0043] Figure 7 This is a schematic diagram of points after each cluster is screened according to angles in the present invention;

[0044] Figure 8 This is a schematic diagram of the leftmost cluster candidate points of the present invention;

[0045] Fig. 9 A schematic diagram of determining the leftmost protrusion according to neighboring points extending upward / downward from the leftmost cluster according to the present invention;

[0046] Fig.10 This is a schematic diagram of clustering below the leftmost protrusion of the present invention;

[0047] Fig.11 This is a schematic diagram of the height measurement points calculated by the algorithm of the present invention;

[0048] Fig.12 This is a schematic diagram of the height measurement point output result of the present invention. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] See also Figure 1-12 The present invention provides a technical solution: a method for automatically generating a cargo loading end view measurement high point based on a laser radar. The algorithm for automatically generating a cargo loading end view measurement high point based on a laser radar comprises the following steps:

[0051] Step 1. Slice the end view point cloud in the XOY plane along the Y axis;

[0052] Step 2. Find the maximum and minimum X coordinates of the point cloud in each slice, find the average X value, and generate a point based on the Y value of the current slice;

[0053] Step 3. Perform RANSAC straight line fitting on all generated points, and the fitted straight line is the center line of the object;

[0054] Step 4. Fold the object point cloud to the left along the center line of the object, and then process the folded point cloud F;

[0055] Step 5. Find the convex hull point cloud and concave hull point cloud of point cloud F;

[0056] Step 6. Remove the convex hull points whose distance from the center line is less than 100 mm from the convex hull point cloud;

[0057] Step 7. Cluster the convex hull point cloud and sort it from high to low according to the Y value;

[0058] Step 8. Sort the point clouds in each result class of the clustering from high to low according to the Y value;

[0059] Step 9. Determine whether it is an inflection point;

[0060] Step 10. Determine whether the first and last points of each class are inflection points;

[0061] Step 11. Generate a focus for each class based on the x value of the object's centerline and the average of the maximum and minimum Y values ​​of each class retained in step 10;

[0062] Step 12. In each class, find the point with the largest distance from the focus of that class as the inflection point;

[0063] Step 13. Add the inflection points in step 10 and step 12 to the key point cloud set as candidate points for height measurement

[0064] Step 14. Segment the point cloud L of the leftmost 1 / 6 of the object;

[0065] Step 15. Filter out the first 40% points with the smallest X value in the point cloud L as the new point cloud TL, and sort the points in TL from high to low according to the Y value;

[0066] Step 16. Cluster the sorted TLs and sort each class and the points in the class from high to low according to the Y value;

[0067] Step 17. Find the class with the smallest point X value, which is the most prominent class LM;

[0068] Step 18. Determine whether the class is located at the bottom of point cloud F according to the minimum Y value of the class and the minimum Y value of point cloud F. If yes, jump to step 20. If not, determine whether the class is the class with the smallest Y value. If yes, find the minimum enclosing rectangle AABB of the point cloud below the class. If not, find the class with the smallest X value below the class and execute step 18 again.

[0069] Step 19. If there is no class above LM, jump to step 20, otherwise find the minimum enclosing rectangle AABB of all classes and the middle point clouds above LM;

[0070] Step 20. Calculate the minimum enclosing rectangle AABB for all the classes found in step 16 and add it to the rectangle set as a candidate rectangle together with the minimum enclosing rectangle AABB calculated in steps 18 and 19;

[0071] Step 21. Determine whether to merge two adjacent rectangles into one rectangle based on whether the X difference between adjacent candidate rectangles is greater than 20 mm, and obtain the final rectangle set;

[0072] Step 22. Delete the candidate height measurement points whose Y values ​​are lower than the maximum Y value in the rectangle set;

[0073] Step 23. If the number of candidate points is greater than 2, determine whether to merge the adjacent points into one point based on whether the X and Y differences of the adjacent candidate points are greater than 10 mm;

[0074] Step 24. Add the highest point of point cloud F to the candidate points, and sort the candidate points from high to low by Y value;

[0075] Step 25. Merge the point with the highest Y value and the second highest Y value among the candidate points according to the merging rule in step 23 to obtain the final set of height measurement points.

[0076] Further improved, the concave point cloud of step 5 is the point cloud contour. By defining the concave point cloud as the point cloud contour, the boundary of the goods can be more accurately identified, especially when processing goods with complex shapes, it can more accurately reflect the actual shape of the goods. The recognition of the contour helps to reduce the misjudgment of non-critical areas in subsequent processing steps and improve the overall efficiency and accuracy of the algorithm.

[0077] Further improved, the rule for judging whether it is an inflection point in step 9 is: whether the angle between the line connecting the point and the point after it and the line connecting the point and the point before it is greater than 14 degrees, if it is greater, it is regarded as an inflection point, otherwise it is not an inflection point. By setting an angle threshold (such as 14 degrees), the inflection points in the point cloud can be more accurately identified, which is crucial for subsequent clustering, sorting and height point determination. In goods of different shapes and sizes, the inflection points may have different characteristics. By setting the angle threshold, the algorithm can respond to various situations more flexibly and improve the overall robustness.

[0078] Further improved, the judgment rule in step 10 is: if it is, retain the class; if not, determine whether there is an inflection point inside the class. The judgment rule is the same as step 9. If there is an inflection point inside, retain the class, otherwise do not retain it. By determining whether the first and last points of the class are inflection points, and whether the class contains inflection points, further processing of meaningless classes can be avoided, thereby reducing the amount of calculation. This judgment rule helps to quickly screen out important classes for subsequent processing and improve the overall efficiency of the algorithm.

[0079] Further improved, in step 23, only all points after the candidate point with the highest Y value are processed. If merged, they are merged according to the following rules: adjacent candidate points a, b (Y values ​​from high to low), the previous point c of a, the next point d of b, find the line L1 connecting a and c and the line L2 connecting b and d, and find the intersection of L1 and L2, which is the point after a and b are merged. By merging adjacent candidate points, the number of final height measurement points can be reduced, making the measurement results more concise and clear. Merging adjacent points can reduce measurement deviations caused by point cloud noise or errors and improve measurement accuracy.

[0080] Working principle: First, slice the end view point cloud in the XOY plane along the Y axis and divide the point cloud into multiple slices for subsequent processing. For each point cloud in each slice, the algorithm calculates the average value of the maximum and minimum X coordinates and generates a point according to the Y value of the current slice. Then, perform RANSAC straight line fitting on all generated points. The fitted straight line is the midline of the object. Along the fitted midline of the object, fold the object point cloud to the left to form a new point cloud F for subsequent processing. Calculate the convex hull point cloud and concave hull point cloud of the folded point cloud F. The convex hull point cloud represents the outer contour of the object, while the concave hull point cloud represents the concave part inside the object. Remove points less than 100 mm from the midline from the convex hull point cloud to reduce the influence of noise and irrelevant points. Cluster the processed convex hull point cloud and sort it from high to low according to the Y value. Then, the point cloud in each cluster is sorted in the same way. The angle between the point and its adjacent points is checked to see if it is an inflection point. The inflection point plays a key role in subsequent processing. A focus point is generated for each cluster based on the X value of the center line of the object and the average of the maximum and minimum Y values ​​of each cluster. Then, the point farthest from the focus point is found in each cluster as an inflection point, and these inflection points are added to the key point cloud set as candidate points for height measurement. The leftmost 1 / 6 point cloud of the object is segmented, and the first 40% of the points with the smallest X value are filtered out as the new point cloud TL. TL is sorted and clustered to find the most prominent class LM. According to the minimum Y value of class LM and the minimum Y value of point cloud F, it is determined whether LM is at the bottom of point cloud F. Then, by extending upward or downward, the neighbor points above or below LM are determined, and the minimum enclosing rectangle AABB is calculated. Whether the rectangles are merged is determined based on whether the X difference between adjacent candidate rectangles is greater than 20mm. Delete the candidate points for height measurement with a Y value lower than the maximum Y value in the rectangle set. If the number of candidate points is greater than 2, the points are merged based on whether the X and Y differences of adjacent candidate points are greater than 10 mm. Finally, the highest point of point cloud F is added to the candidate points, and the final set of height measurement points is obtained according to the merging rules.

[0081] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and the present invention can be implemented in other specific forms without departing from the spirit or basic features of the present invention. Therefore, no matter from which point of view, the embodiments should be regarded as exemplary and non-restrictive. The scope of the present invention is defined by the attached claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any figure mark in the claims should not be regarded as limiting the claims involved.

[0082] Finally, a few points should be explained: First, in the description of this application, it should be noted that, unless otherwise specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, and can be mechanical or electrical connections, or internal connectivity between two components, or direct connections. "Up", "down", "left", "right", etc. are only used to indicate relative position relationships. When the absolute position of the object being described changes, the relative position relationship may change.

[0083] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for automatically generating a cargo loading end view measurement high point based on laser radar, characterized in that: The method for automatically generating a cargo loading end view measurement high point based on laser radar comprises the following steps: Step 1. Slice the end view point cloud in the XOY plane along the Y axis; Step 2. Find the maximum and minimum X coordinates of the point cloud in each slice, find the average X value, and generate a point based on the Y value of the current slice; Step 3. Perform RANSAC straight line fitting on all generated points, and the fitted straight line is the center line of the object; Step 4. Fold the object point cloud to the left along the center line of the object, and then process the folded point cloud F; Step 5. Calculate the convex hull point cloud and concave hull point cloud of point cloud F; Step 6. Remove the convex hull points whose distance from the center line is less than 100 mm from the convex hull point cloud; Step 7. Cluster the convex hull point cloud and sort it from high to low according to the Y value; Step 8. Sort the point clouds in each result class of the clustering from high to low according to the Y value; Step 9. Determine whether it is an inflection point; Step 10. Determine whether the first and last points of each class are inflection points; Step 11. Generate a focus for each class based on the x value of the object's centerline and the average of the maximum and minimum Y values ​​of each class retained in step 10; Step 12. In each class, find the point with the largest distance from the focus of that class as the inflection point; Step 13. Add the inflection points in step 10 and step 12 to the key point cloud set as candidate points for height measurement Step 14. Segment the point cloud L of the leftmost 1 / 6 of the object; Step 15. Filter out the first 40% points with the smallest X value in the point cloud L as the new point cloud TL, and sort the points in TL from high to low according to the Y value; Step 16. Cluster the sorted TLs and sort each class and the points in the class from high to low according to the Y value; Step 17. Find the class with the smallest point X value, which is the most prominent class LM; Step 18. Determine whether the class is located at the bottom of point cloud F according to the minimum Y value of the class and the minimum Y value of point cloud F. If yes, jump to step 20. If not, determine whether the class is the class with the smallest Y value. If yes, find the minimum enclosing rectangle AABB of the point cloud below the class. If not, find the class with the smallest X value below the class and execute step 18 again. Step 19. If there is no class above LM, jump to step 20, otherwise find the minimum enclosing rectangle AABB of all classes and the middle point clouds above LM; Step 20. Calculate the minimum enclosing rectangle AABB for all the classes found in step 16 and add it to the rectangle set as a candidate rectangle together with the minimum enclosing rectangle AABB calculated in steps 18 and 19; Step 21. Determine whether to merge two adjacent rectangles into one rectangle based on whether the X difference between adjacent candidate rectangles is greater than 20 mm, and obtain the final rectangle set; Step 22. Delete the candidate height measurement points whose Y values ​​are lower than the maximum Y value in the rectangle set; Step 23. If the number of candidate points is greater than 2, determine whether to merge the adjacent points into one point based on whether the X and Y differences of the adjacent candidate points are greater than 10 mm; Step 24. Add the highest point of point cloud F to the candidate points, and sort the candidate points from high to low by Y value; Step 25. Merge the point with the highest Y value and the second highest Y value among the candidate points according to the merging rule in step 23 to obtain the final set of height measurement points.

2. According to claim 1, a method for automatically generating a cargo loading end view height measurement point based on laser radar, characterized in that: The concave hull point cloud of step 5 is the point cloud outline.

3. The method for automatically generating a cargo loading end view height measurement point based on laser radar according to claim 1, characterized in that: The rule for judging whether it is an inflection point in step 9 is: whether the angle between the line connecting the point and the point after it and the line connecting the point and the point before it is greater than 14 degrees, if so, it is considered an inflection point, otherwise it is not an inflection point.

4. The method for automatically generating a cargo loading end view height measurement point based on laser radar according to claim 1, characterized in that: The judgment rule in step 10 is: if yes, keep the class; if not, judge whether there is an inflection point inside the class. The judgment rule is the same as step 9. If there is an inflection point inside, keep the class; otherwise, do not keep it.

5. The method for automatically generating a cargo loading end view height point based on laser radar according to claim 1, characterized in that: In step 23, only all points after the candidate point with the highest Y value are processed. If they are merged, they are merged according to the following rules: adjacent candidate points a, b (Y values ​​from high to low), the previous point c of a, the next point d of b, find the line L1 connecting a and c and the line L2 connecting b and d, and find the intersection of L1 and L2, which is the point after a and b are merged.

Images