A stud detection method based on point cloud information

Abstract

The present invention provides a stud detection method based on point cloud information, which uses a three-dimensional visual sensor to obtain point clouds of the bottom surface and the side surface of the stud, pre-processes the point cloud of the side surface of the stud, removes clutter, and obtains the normal vector of each point in the processed point cloud; uses the normal vector to calculate cosine similarity, and removes thread points on non-inner and outer diameters; uses all internal thread points and external thread points to fit a cylinder, and obtains the central axis of the stud; calculates the angle between the central axis and the bottom surface and the coordinates of the intersection as the stud posture information. The method of the present invention directly uses the internal thread points and the external thread points to fit the central axis, and the calculation is accurate and time-saving. In addition, by pre-processing the point cloud, the noise and clutter in the point cloud are removed, which further ensures the effectiveness of the fitting result, and has the characteristics of high automation and strong anti-interference.

Description

Technical Field

[0001] The invention relates to the field of point cloud processing, and in particular to a stud detection method based on point cloud information. Background Art

[0002] Studs or connectors with threaded features are widely used in the manufacturing and construction fields. However, due to factors such as installation errors and manufacturing errors, the actual installation position (coordinates of the intersection with the bottom surface) and posture (angle with the bottom surface) of the studs deviate from the preset requirements. Therefore, in the field of precision machining and manufacturing, it is necessary to detect the actual position and posture of the studs after fixation to determine whether the position and posture between the studs and the bottom surface meet the requirements.

[0003] The most common detection methods currently include:

[0004] Manual inspection: Use a sleeve to cover the stud, and use a dial indicator, wedge-shaped feeler gauge and other tools to measure the axis of the sleeve as the axis of the stud. This method requires manual installation and removal of the sleeve and then measurement with tools, which undoubtedly increases labor costs and time costs.

[0005] Structured light measurement: Use 3D scanning equipment to obtain the stud point cloud, use the point cloud to fit the axis, and then calculate the stud's position and posture. However, this method has two problems in practical application:

[0006] One is that the metal studs themselves have strong reflections, and a lot of noise is inevitable at the reflections. There are many noise points in the point cloud, which affects the calculation accuracy.

[0007] The second is to directly use the point cloud to fit the stud, and the obtained axis is not accurate. The point cloud scattered between the internal thread and the external thread actually reduces the fitting accuracy of the stud. Patent document: CN110411341A provides a method for calculating the posture of a threaded target, which uses the internal thread points and the external thread points of the stud to fit the cylinder and obtain the central axis, and then records the central axis fitted by the external thread points as the final axis. This method eliminates the point cloud scattered between the internal thread and the external thread when fitting the stud, and improves the accuracy of obtaining the central axis; however, there are the following problems: it obtains plane points by setting sections and projecting the point cloud plane, and then interpolates and fits to obtain the reference curve, and then performs least squares movement to calculate the point cloud posture to calculate the outer diameter and inner diameter points. When many sections are set, the calculation amount is large and the speed is slow. When fewer sections are set, the accuracy decreases. The point cloud segmentation accuracy of the inner and outer diameters depends on the accuracy of the initial fitting cylinder central axis and the fitting accuracy of the reference curve. There are many interference factors and the calculation amount is large. Summary of the invention

[0008] In order to solve the above technical problems, the present invention provides a stud detection method based on point cloud information. The method of the present invention directly uses internal thread points and external thread points to fit the central axis, and the calculation is accurate and time-saving. By preprocessing the point cloud, the noise and miscellaneous points in the point cloud are removed, which further ensures the effectiveness of the fitting result. The method has the characteristics of high degree of automation and strong anti-interference.

[0009] The technical solution is as follows:

[0010] A stud detection method based on point cloud information, wherein the plane where the stud intersects with other objects when the stud is fixed is taken as the bottom surface, and the bottom surface and the stud side surface point cloud are obtained by using a three-dimensional visual sensor, and the stud side surface point cloud is processed by the following steps:

[0011] S1. Preprocess the point cloud of the side surface of the stud, remove the clutter, and obtain the normal vector of each point in the processed point cloud;

[0012] S2. Perform the following processing on each point in the processed point cloud:

[0013] Taking a single point as the center, define the area around the point as the local area, calculate the cosine similarity between the point and each point in the local area, and take the mean as the mean A corresponding to the point;

[0014] Determine whether the mean A corresponding to the point is less than the threshold. If not, remove the point; if yes, retain the point.

[0015] After traversing each point in the point cloud processed by step S1, proceed to step S3;

[0016] S3, respectively determine the relationship between the normal vector of each point and the Z-axis direction. If they are in the same direction, the point is recorded as an internal thread point; if they are in the opposite direction, the point is recorded as an external thread point;

[0017] The Z axis is the optical axis direction of the camera in the three-dimensional vision sensor;

[0018] S4. Fitting a cylinder using all internal thread points / external thread points to obtain the center axis of the stud;

[0019] Alternatively, all the internal thread points and external thread points are used to fit the cylinder respectively, and then the central axis corresponding to the two cylinders is used to synthesize the final central axis;

[0020] The angle between the central axis and the bottom surface and the coordinates of the intersection are calculated as the stud posture information, and the bottom surface is obtained by point cloud fitting.

[0021] Furthermore, the threshold in step S2 is set by an empirical value, or: first obtain the mean value A corresponding to each point, and take 1.2 to 1.5 times the minimum mean value A as the threshold.

[0022] Preferably, in step S2, the local area is an eight-neighborhood area or a four-neighborhood area around a single point.

[0023] Furthermore, in step S1, the point cloud is preprocessed and the method for removing the noise is as follows: according to the center point on the theoretical central axis of the stud and the theoretical height, the area where the middle part of the stud is located is set as the central area, and the point cloud in the central area is extracted and retained, and the point clouds in other positions are removed.

[0024] Alternatively, in step S1, the point cloud is preprocessed to remove the noise points by: at the same position, the three-dimensional visual sensor also collects a two-dimensional image of the stud;

[0025] The center point on the theoretical central axis of the stud is projected onto the two-dimensional image. A region of interest is set with this point as the center and 0.4 to 0.7 times the theoretical height of the stud as the radius. The pixel coordinates with grayscale values ​​within the preset range are screened out in this region and then projected into the point cloud respectively. The corresponding three-dimensional point coordinates are retained and the other three-dimensional point coordinates are discarded.

[0026] Furthermore, the three-dimensional visual sensor includes a projector and a camera, and records the spatial position coordinates of the camera and the projector in the global coordinate system. C , O p ;

[0027] In step S1, the point cloud of the side surface of the stud is preprocessed to remove the clutter in the following way:

[0028] 1) fitting a surface using the point cloud and recording it as a measured surface; calculating the central axis of the measured surface;

[0029] Select any point in the middle area of ​​the central axis, record it as point A, and obtain the spatial position coordinates of point A;

[0030] 2) Get the passing point O C , the straight line from point A to point O p The intersection points of the straight line from point A and the measured surface are B and C respectively;

[0031] Find the midpoint M between the intersection points B and C;

[0032] 3) Draw a vector through point A A straight line is drawn, and the intersection point K between the straight line and the measured surface is calculated; a characteristic straight line L is drawn through the intersection point K and is parallel to the central axis;

[0033] The distance from each point in the point cloud to the characteristic line L is calculated respectively, and the points with distance values ​​less than the preset value are recorded as reflective noise points and removed, while other point clouds are retained to complete the denoising of the point cloud.

[0034] Further, in step 2), the calculation method of the intersection point B is: calculate the intersection point OC , the angle θ between the straight line at point A and the central axis; use the angle θ and the theoretical radius of the cylinder to calculate the distance △d between point A and point B; move point A along the vector Move △d to obtain the position coordinates of point B;

[0035] The calculation method of the intersection point C is: calculate the intersection point O P , the angle θ1 between the straight line at point A and the central axis; using the angle θ1 and the theoretical radius of the cylinder to calculate the distance △d1 between point A and point C; moving point A along the vector Move △d1 to obtain the position coordinates of point C.

[0036] Further, in step 3), the intersection point K is calculated by obtaining the intersection point A and the intersection point K along the vector The angle θ2 between the straight line and the central axis; use the angle θ2 and the theoretical radius of the cylinder to calculate the distance △d2 between point A and point K; move point A along the vector Move △d2 to obtain the position coordinates of point K;

[0037] The points whose distance from the characteristic straight line L is less than the preset value are recorded as reflective noise points, specifically:

[0038] Note that any point p in the point cloud h The direction vector to point K is Find the vector The area of ​​the parallelogram with the characteristic line L as the side, and then divide the area value by the length of the side where the characteristic line L is located, to get the point p h The distance from the characteristic line L;

[0039] Default value R represents the theoretical radius of the stud.

[0040] In order to collect a relatively complete thread point cloud, preferably, the angle β between the camera optical axis in the three-dimensional vision sensor and the theoretical central axis of the stud satisfies the following relationship:

[0041] Thread angle < β < (180°-thread angle).

[0042] In order to improve the fitting accuracy, preferably, in step S4, before using the internal thread points and the external thread points to fit the cylinder, the RANSAC method is first used to remove the noise points in the internal thread points and the external thread points, and then the removed points are used to fit the cylinder.

[0043] Beneficial effects of the present invention:

[0044] The stud detection method of the present invention is based on the principle that the smaller the cosine similarity value, the greater the difference between the two vectors. This principle selects the points on the internal and external threads, and then classifies them according to their relationship with the Z-axis direction to distinguish between the internal thread points and the external thread points; the internal and external thread points are used to fit the central axis, the calculation is accurate, the time consumption is short, and it has the characteristics of high degree of automation and strong anti-interference.

[0045] In addition, this method also removes reflective noise in the point cloud by preprocessing the point cloud, which further ensures the effectiveness of the fitting results and is faster and more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 is the grayscale image of the stud;

[0047] Figure 2 This is a schematic diagram of the point cloud of the side surface of the stud;

[0048] Figure 3 This is the normal distribution front view of the point cloud of the inner and outer diameters of the stud;

[0049] Figure 4 It is the top view of the normal distribution of the point cloud of the inner and outer diameters of the stud;

[0050] Figure 5 It is a schematic diagram of using a grayscale image to remove noise in the second embodiment of the specific implementation method;

[0051] Figure 6 It is a structural schematic diagram of removing noise points in the third method in the specific implementation method. DETAILED DESCRIPTION

[0052] The technical solution of the present invention is described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0053] A stud detection method based on point cloud information, in which the plane where the stud intersects with other objects when the stud is fixed is taken as the bottom surface, and a three-dimensional visual sensor is used to obtain the point cloud of the bottom surface and the side surface of the stud (such as Figure 2 ), process the point cloud of the stud side surface through the following steps:

[0054] S1. Preprocess the point cloud of the side surface of the stud, remove the noise points, and obtain the normal vector of each point in the processed point cloud (such as Figure 3 , Figure 4 );

[0055] S2. Perform the following processing on each point in the processed point cloud:

[0056] Taking a single point as the center, define the area around the point as the local area, calculate the cosine similarity between the point and each point in the local area, and take the mean as the mean A corresponding to the point;

[0057] Determine whether the mean A corresponding to the point is less than the threshold. If not, remove the point; if yes, retain the point.

[0058] After traversing each point in the point cloud processed by step S1, proceed to step S3;

[0059] S3. Determine the relationship between the normal vector of each point and the Z-axis direction. If they are in the same direction, the point is recorded as an internal thread point. If they are in the opposite direction, the point is recorded as an external thread point (such as Figure 3 , Figure 4 );

[0060] The Z axis is the optical axis direction of the camera in the 3D vision sensor;

[0061] S4. Fitting a cylinder using all internal thread points / external thread points to obtain the center axis of the stud;

[0062] Alternatively, all the internal thread points and external thread points are used to fit the cylinder respectively, and then the central axis corresponding to the two cylinders is used to synthesize the final central axis;

[0063] The angle between the central axis and the bottom surface and the coordinates of the intersection are calculated as the stud posture information, where the bottom surface is obtained by point cloud fitting.

[0064] Specifically, in step S2, the cosine similarity cosβ between a single point A in the processed point cloud and a single point B in the local area is calculated as follows:

[0065] and They represent the normal vectors corresponding to point A and point B respectively. The smaller the cosβ is, the greater the difference between the two vectors is.

[0066] The normal direction of the point cloud on the inner diameter of the stud is as follows Figure 3 , Figure 4 As shown, they all point to the stud axis, while the normals of the point clouds on the outer diameter all point from the stud axis to the outside. Since the Z axis points to the stud axis, the internal thread points and external thread points can be separated by the relationship between the normals of the inner and outer diameter points and the Z axis direction.

[0067] Step S4: Calculate the angle between the central axis and the bottom surface in is the normal vector of the bottom surface, is the direction vector of the central axis.

[0068] The threshold in step S2 is set by an empirical value (such as 30°, 40°, 50°), or: first obtain the mean value A corresponding to each point, and take 1.2 to 1.5 times the minimum mean value A as the threshold;

[0069] The local area is the eight-neighborhood or four-neighborhood area around a single point.

[0070] More specifically, in step S1, the point cloud is preprocessed and the method of removing noise points can adopt the following three methods:

[0071] Method 1: According to the center point on the theoretical central axis of the stud and the theoretical height, the area where the middle part of the stud is located is set as the central area, and the point cloud in the central area is extracted and retained, and the point clouds in other positions are eliminated.

[0072] Since the top and bottom ends of the stud are usually processed at a certain cutting angle and are not regular cylinders, the point cloud at this position is not conducive to fitting the central axis. Therefore, this method aims to eliminate the point cloud at the top and bottom ends of the stud and only retain the point cloud in the cylindrical area.

[0073] Method 2: At the same position, the 3D vision sensor also collects a 2D image of the stud (such as Figure 1 );

[0074] Project the center point on the theoretical central axis of the stud onto the two-dimensional image, and set the region of interest with this point as the center and 0.4 to 0.7 times the theoretical height of the stud as the radius, such as Figure 5 As shown, pixel coordinates whose grayscale values ​​are within a preset range are screened out in the area, and then projected into the point cloud respectively, the corresponding three-dimensional point coordinates are retained, and other three-dimensional point coordinates are discarded.

[0075] Since the points between the two-dimensional image and the three-dimensional point cloud show a one-to-one correspondence, this method not only removes the point clouds at the top and bottom of the stud by setting the region of interest, but also maps the points whose grayscale values ​​meet the requirements in the region of interest to the point cloud and retains them, while removing other overexposed points or points on the background, effectively removing noise points;

[0076] During specific implementation, the preset interval can be set according to actual conditions. In this embodiment, Figure 5 As shown, it is set to (10, 80).

[0077] Method 3: 3D vision sensors include projectors and cameras, such as Figure 6 As shown, the spatial position coordinates O of the recording camera and projector in the global coordinate system C , O p ;

[0078] In step S1, the point cloud of the side surface of the stud is preprocessed to remove the clutter in the following way:

[0079] 1) Use the point cloud to fit the surface and record it as the measured surface; calculate the central axis of the measured surface;

[0080] Select any point in the middle area of ​​the central axis, record it as point A, and obtain the spatial position coordinates of point A;

[0081] 2) Get the passing point O C , the straight line from point A to point O p The intersection points of the straight line from point A and the measured surface are B and C respectively;

[0082] Find the midpoint M between the intersection points B and C;

[0083] 3) Draw a vector through point A The straight line is drawn, and the intersection point K between the straight line and the measured surface is calculated; a characteristic straight line L is drawn through the intersection point K and is parallel to the central axis;

[0084] The distance from each point in the point cloud to the characteristic line L is calculated respectively, and the points with distance values ​​less than the preset value are recorded as reflective noise points and removed, while other point clouds are retained to complete the denoising of the point cloud.

[0085] The specific explanation of method 3 is as follows: Since the side surface of the stud is a cylindrical surface, such as Figure 1 As shown, the reflective area is concentrated on the side facing the lighting component. Method three uses the geometric relationship between the camera and the lighting component and the center axis of the curved workpiece to directly eliminate all points in the cylindrical reflective area, which is simple and effective.

[0086] Among them, in step 2), the calculation method of intersection B is: calculate the intersection point O C , the angle θ between the straight line at point A and the central axis; use the angle θ and the theoretical radius of the cylinder to calculate the distance △d between point A and point B; move point A along the vector Move △d to obtain the position coordinates of point B;

[0087] Intersection point B(x B ,y B ,z B ) The specific solution is as follows:

[0088]

[0089]

[0090]

[0091] Where R is the theoretical radius of the cylinder. The direction vector of the central axis; Indicates point O C The direction vector to point A; Representation vector The model, Respectively Components in the x, y, and z coordinate directions; for ease of understanding, the vector Written as a vector Let point A(xA ,y A ,z A ) along the vector Move △d to obtain the position coordinates of point B;

[0092] The calculation method of the intersection point C is: calculate the intersection point O P , the angle θ1 between the straight line at point A and the central axis; using the angle θ1 and the theoretical radius of the cylinder to calculate the distance △d1 between point A and point C; moving point A along the vector Move △d1 to obtain the position coordinates of point C.

[0093] Intersection point C(x C ,y C ,z C ) is calculated as follows:

[0094]

[0095]

[0096]

[0097] in, Indicates point O P The direction vector to point A, Representation vector The model, Respectively Components in the x, y, and z coordinate directions; for ease of understanding, the vector Written as a vector Let point A(x A ,y A ,z A ) along the vector Move △d1 to obtain the position coordinates of point C.

[0098] Further, in step 3), the intersection point K is calculated by obtaining the intersection point A and the intersection point K along the vector The angle θ2 between the straight line and the central axis; use the angle θ2 and the theoretical radius of the cylinder to calculate the distance △d2 between point A and point K; move point A along the vector Move △d2 to obtain the position coordinates of point K;

[0099] Intersection point K(x K ,y K ,z K ) The specific solution is as follows:

[0100]

[0101]

[0102]

[0103] Where R is the theoretical radius of the cylinder. Representation vector The model, Represents vectors Components in the x, y, and z coordinate directions.

[0104] The points whose distance from the characteristic straight line L is less than the preset value are recorded as reflective noise points, specifically:

[0105] Note that any point p in the point cloud h The direction vector to point K is Find the vector The area of ​​the parallelogram with characteristic line L as side Then the area value is divided by the length of the side where the characteristic line L is located. Get some p h The distance from the characteristic line L;

[0106] The specific solution process is: calculate any point p in the point cloud h The distance value Q from the characteristic line L h :

[0107]

[0108] in, Represents point p h Direction vector to point K.

[0109] Default value R represents the theoretical radius of the stud.

[0110] As a preferred implementation, in order to collect a relatively complete thread point cloud, the three-dimensional visual sensor collects the stud point cloud at a certain angle laterally to the stud, specifically: the angle β between the camera optical axis and the theoretical central axis of the stud satisfies the following relationship:

[0111] Thread angle < β < (180°-thread angle).

[0112] In order to improve the fitting accuracy, in step S4, before using the internal thread points and the external thread points to fit the cylinder, the RANSAC method is first used to remove the noise points in the internal thread points and the external thread points, and then the removed points are used to fit the cylinder.

[0113] The studs mentioned in the method of the present invention include bolts, screws or other parts with external thread structures; the method of the present invention is suitable for the posture detection of parts with triangular threads, rectangular threads, trapezoidal threads, and serrated threads; and is suitable for stud posture detection in fields such as automobile manufacturing, construction, electronics, electricity, rail transportation, aerospace, shipbuilding and shipping, etc. where there is a need for stud posture detection.

[0114] The foregoing description of the specific exemplary embodiments of the present invention is presented for the purpose of illustration and description. The foregoing description is not intended to be exhaustive, nor is it intended to limit the present invention to the precise form disclosed, and it is apparent that many changes and variations are possible in light of the above teachings. The exemplary embodiments are selected and described to explain the specific principles of the present invention and its practical application, so that other technicians in the field can realize and utilize the various exemplary embodiments of the present invention and its different selected forms and modified forms. The scope of the present invention is intended to be limited by the appended claims and their equivalents.

Claims

1. A stud detection method based on point cloud information, characterized in that: The plane where the stud intersects with other objects when it is fixed is the bottom surface. The 3D vision sensor is used to obtain the point cloud of the bottom surface and the side surface of the stud. The point cloud of the side surface of the stud is processed by the following steps: S1. Preprocess the point cloud of the side surface of the stud, remove the clutter, and obtain the normal vector of each point in the processed point cloud; S2. Perform the following processing on each point in the processed point cloud: Taking a single point as the center, define the area around the point as the local area, calculate the cosine similarity between the point and each point in the local area, and take the mean as the mean A corresponding to the point; Determine whether the mean A corresponding to the point is less than the threshold. If not, remove the point; if yes, retain the point. After traversing each point in the point cloud processed by step S1, proceed to step S3; S3, respectively determine the relationship between the normal vector of each point and the Z-axis direction. If they are in the same direction, the point is recorded as an internal thread point; if they are in the opposite direction, the point is recorded as an external thread point; The Z axis is the optical axis direction of the camera in the three-dimensional vision sensor; S4. Fitting a cylinder using all internal thread points / external thread points to obtain the center axis of the stud; Alternatively, all the internal thread points and external thread points are used to fit the cylinder respectively, and then the central axis corresponding to the two cylinders is used to synthesize the final central axis; The angle between the central axis and the bottom surface and the coordinates of the intersection are calculated as the stud posture information, and the bottom surface is obtained by point cloud fitting.

2. The stud detection method based on point cloud information as claimed in claim 1, characterized in that: The threshold in step S2 is set by empirical value, or: first obtain the mean value A corresponding to each point, and take 1.2 to 1.5 times of the minimum mean value A as the threshold.

3. The stud detection method based on point cloud information as claimed in claim 1, characterized in that: In step S2, the local area is an eight-neighborhood or a four-neighborhood around a single point.

4. The stud detection method based on point cloud information as claimed in claim 1, characterized in that: In step S1, the point cloud is preprocessed and the method for removing the noise is as follows: according to the center point on the theoretical central axis of the stud and the theoretical height, the area where the middle part of the stud is located is set as the central area, and the point cloud in the central area is extracted and retained, and the point clouds in other positions are removed.

5. The stud detection method based on point cloud information as claimed in claim 1, characterized in that: In step S1, the point cloud is preprocessed to remove the noise points by: at the same position, the three-dimensional visual sensor also collects a two-dimensional image of the stud; The center point on the theoretical central axis of the stud is projected onto the two-dimensional image. A region of interest is set with this point as the center and 0.4 to 0.7 times the theoretical height of the stud as the radius. The pixel coordinates with grayscale values ​​within the preset range are screened out in this region and then projected into the point cloud respectively. The corresponding three-dimensional point coordinates are retained and the other three-dimensional point coordinates are discarded.

6. The stud detection method based on point cloud information as claimed in claim 1, characterized in that: The three-dimensional visual sensor includes a projector and a camera, and records the spatial position coordinates of the camera and the projector in the global coordinate system. C , O p ; In step S1, the point cloud of the side surface of the stud is preprocessed to remove the clutter in the following way: 1) fitting a surface using the point cloud and recording it as a measured surface; calculating the central axis of the measured surface; Select any point in the middle area of ​​the central axis, record it as point A, and obtain the spatial position coordinates of point A; 2) Get the passing point O C , the straight line from point A to point O p The intersection points of the straight line from point A and the measured surface are B and C respectively; Find the midpoint M between the intersection points B and C; 3) Draw a vector through point A A straight line is drawn, and the intersection point K between the straight line and the measured surface is calculated; a characteristic straight line L is drawn through the intersection point K and is parallel to the central axis; The distance from each point in the point cloud to the characteristic line L is calculated respectively, and the points with distance values ​​less than the preset value are recorded as reflective noise points and removed, while other point clouds are retained to complete the denoising of the point cloud.

7. The stud detection method based on point cloud information as claimed in claim 6, characterized in that: In step 2), the intersection point B is calculated by: calculating the intersection point O C , the angle θ between the straight line at point A and the central axis; use the angle θ and the theoretical radius of the cylinder to calculate the distance △d between point A and point B; move point A along the vector Move △d to obtain the position coordinates of point B; The calculation method of the intersection point C is: calculate the intersection point O P , the angle θ1 between the straight line at point A and the central axis; using the angle θ1 and the theoretical radius of the cylinder to calculate the distance △d1 between point A and point C; moving point A along the vector Move △d1 to obtain the position coordinates of point C.

8. The stud detection method based on point cloud information as claimed in claim 6, characterized in that: In step 3), the intersection point K is calculated by obtaining the intersection point A and the vector The angle θ2 between the straight line and the central axis; use the angle θ2 and the theoretical radius of the cylinder to calculate the distance △d2 between point A and point K; move point A along the vector Move △d2 to obtain the position coordinates of point K; The points whose distance from the characteristic straight line L is less than the preset value are recorded as reflective noise points, specifically: Note that any point p in the point cloud h The direction vector to point K is Find the vector The area of ​​the parallelogram with the characteristic line L as the side, and then divide the area value by the length of the side where the characteristic line L is located, to get the point p h The distance from the characteristic line L; Default value R represents the theoretical radius of the stud.

9. The stud detection method based on point cloud information as claimed in claim 1, characterized in that: The angle β between the camera optical axis in the three-dimensional vision sensor and the theoretical central axis of the stud satisfies the following relationship: Thread angle < β < (180°-thread angle).

10. The stud detection method based on point cloud information as claimed in claim 1, characterized in that: In step S4, before using the internal thread points and the external thread points to fit the cylinder, the RANSAC method is first used to remove the noise points in the internal thread points and the external thread points, and then the removed points are used to fit the cylinder.

Images