A cable bending radius calculation method based on mask center line algebraic reconstruction

Abstract

The application discloses a cable bending radius calculation method based on mask center line algebraic reconstruction, which comprises the following steps: an RGB-D dual-mode semantic segmentation network is used to acquire a cable mask, and a depth image is used to generate a cable space point cloud; the mask is subjected to an image thinning algorithm to obtain a cable plane feature center line, and the cable plane feature center line track is tracked, and the cable space point cloud is indexed according to the track pixel coordinates to obtain a cable space feature center line; the cable space feature center line is non-uniformly sampled to obtain a space sampling point, and the space sampling point is subjected to algebraic reconstruction to obtain a cable space feature curve; the cable space feature curve is regarded as a particle space motion track, and the space bending radius of the cable space feature curve is solved by using a particle space motion law.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of cable laying quality detection, and particularly relates to a cable bending radius calculation method based on algebraic reconstruction of a mask center line. BACKGROUND

[0002] Cables play an important role in signal transmission in large equipment in the fields of rail transportation, aerospace, and ship transportation. Detecting the bending radius of the cable during the cable laying process is an important part of large equipment quality detection. If the bending radius of the cable is too small, stress concentration will occur, affecting the quality of signal transmission, and in a more serious case, the insulation layer will be damaged, which will have a significant impact on the normal operation of large equipment. Traditional manual measurement of the bending radius of the cable is greatly affected by the subjectivity of the measurer, has low detection efficiency, and has large measurement errors.

[0003] There are methods for measuring the bending radius of a cable, such as CN113624103A, which uses a bending radius detector to measure the bending radius of a cable, and CN111412849B and CN113935958A, which use visual measurement to measure the bending radius.

[0004] The above specifically includes:

[0005] 1) "Cable Bending Radius Detector, CN113624103A". The present application discloses a cable bending radius detector, which includes a main ruler and two lengthwise rulers. The front section of the main ruler is provided with a sliding groove along the length direction, and a sliding block is arranged in the sliding groove. The main ruler is provided with a scale to the length of the front end of the main ruler. The front ends of the two lengthwise rulers are connected by a rubber band, and the rear ends are hinged to the same point of the sliding block. The length of the lengthwise ruler is known, and after the main ruler is perpendicular to the straightened rubber band, the scale value is read to quickly calculate the size of the arc radius. The whole process is simple and fast. The cable bending radius measurement of the present application is different from the above method. The cable mask is obtained through a double-modal semantic segmentation network, the spatial feature center line is constructed based on the mask, and the spatial feature center line is algebraically reconstructed to obtain the spatial feature curve of the cable. The bending radius of the cable is solved, and the measurement of the bending radius of the cable in a narrow space is more feasible. The algebraic method based on images for calculating the bending radius of the cable has high precision.

[0006] 2) "Method and system for quickly measuring the bending radius of a spacecraft cable", patent number CN111412849B. The invention discloses a method and system for quickly measuring the bending radius of a spacecraft cable. By obtaining the image of the cable to be measured and the measurement strip, the type of the cable to be measured is determined based on the RGB values of the cable in the image and the RGB values of the preset cable. The diameter of the cable is calculated by obtaining the radial dimension and axial dimension ratio of the ring-shaped ring with the RGB value difference between the measurement strip and the cable to be measured greater than the preset threshold. Based on the edge trajectory, the bending radius of the cable to be measured is obtained. According to the ratio of the bending radius and the diameter, it is judged whether the bending radius meets the specification requirements. Through the technical path of photographing, recognition and fitting, the measurement of the bending radius of different thickness cables and the quick comparison with the standard specification are realized quickly, and the consistency and efficiency of the bending radius measurement are improved. The present invention is different from the above method, which uses a deep learning RGB-D dual-mode semantic segmentation method to obtain a cable mask image. The cable mask obtained by segmentation is more accurate and has less interference. The cable space feature curve is constructed based on the cable mask and the cable space point cloud, and the cable space bending radius is solved, which is more consistent with the actual situation of cable laying.

[0007] 3) "Cable bending radius detection method and device, CN113935958A". The invention relates to a cable bending radius detection method and device. The positioning mark image and the real cable image are collected at the preset position in the real cable space. According to the coordinate information of the positioning mark image, the projection matrix of the camera is determined, and the real cable is inversely projected to obtain the corresponding curve of the real cable in the cable bending plane, and the bending radius of the real cable is determined. The method can improve the detection accuracy of the cable bending radius. The present invention is different from the above method, which extracts the mask feature center line based on the mask image and constructs the space feature center to reduce the calculation amount. The parameter polynomial equation set is algebraic reconstruction to solve the cable space bending radius. The space curvature of any point on the cable can be calculated, which has strong practicality and generalizability. SUMMARY

[0008] To solve the above technical problems, the purpose of the present invention is to provide a cable bending radius calculation method based on mask center line algebraic reconstruction.

[0009] The purpose of the present invention is achieved by the following technical solutions:

[0010] A cable bending radius calculation method based on mask center line algebraic reconstruction, comprising:

[0011] A, the cable image mask is obtained by a dual-mode semantic segmentation network, and the cable space point cloud P is constructed by a depth image cable ;

[0012] B. Obtain the center line of the cable plane feature by applying an image thinning algorithm to the mask. 2D And track the trajectory C of the center line of the cable plane feature. 2D_order According to the trajectory pixel coordinates C 2D_order (u,v) Index cable spatial point cloud P cable Obtain the center line C of the cable spatial characteristics 3D_order ;

[0013] C. After non-uniform sampling of the center line of the cable spatial features, spatial sampling point C is obtained. 3D_sampling The spatial characteristic curve F of the cable is obtained by algebraically reconstructing the spatial sampling points. cable ;

[0014] D. Cable spatial characteristic curve F cable Treating the trajectory of a point mass in space, the spatial bending radius R of the cable's spatial characteristic curve is solved using the laws governing the spatial motion of the point mass. cable .

[0015] Compared with the prior art, one or more embodiments of the present invention may have the following advantages:

[0016] In the process of measuring cable bending radius, a cable mask image is obtained through a dual-modal semantic segmentation network. Based on the mask image, the cable feature centerline is extracted, and then the cable bending radius is solved by an algebraic method. The image-based cable bending radius measurement method can adapt to the bending radius measurement task under narrow cable laying environment conditions, effectively improving the efficiency of cable bending radius detection. The cable planar feature curve trajectory is tracked and the cable spatial feature curve is constructed according to its pixel coordinate index to solve the cable spatial bending radius, which meets the actual situation of non-coplanarity in the actual cable laying process. The algebraic reconstruction of the cable feature curve transforms the discrete point set of the cable spatial feature centerline into a continuous algebraic expression, which meets the smoothness requirements in the actual cable laying process, and can also avoid the influence of point cloud noise in the spatial feature centerline, effectively improving the accuracy of cable bending radius calculation. Attached Figure Description

[0017] Figure 1 This is a flowchart of a cable bending radius calculation method based on algebraic reconstruction of the mask centerline;

[0018] Figure 2 This is the RGB diagram of the cable under test;

[0019] Figure 3 It is a diagram of the centerline of the cable mask and planar features;

[0020] Figure 4 It is a centerline diagram of the spatial characteristics of the cable;

[0021] Figure 5 It is an algebraic reconstruction of the spatial characteristic curve of the cable;

[0022] Figure 6 is a cable space bending radius visualization result map. DETAILED DESCRIPTION

[0023] In order to make the objects, technical solutions and advantages of the present application clearer, the following will further describe the present application in detail with examples and drawings.

[0024] As shown in Figure 1 , it is a cable bending radius calculation method based on mask center line algebraic reconstruction, including the following steps:

[0025] Step 10: RGB-D dual-mode semantic segmentation network obtains a cable mask Mask, and generates a cable space point cloud P from a depth image cable ;

[0026] Input the cable RGB-D image into the dual-mode semantic segmentation network improved ESANet to obtain the cable mask Mask; generate the cable point cloud P from the depth image cable , and each depth pixel coordinate (u p ,v p ) can obtain a corresponding space point coordinate P (X, Y, Z):

[0027]

[0028] Where f x , f y are the focal lengths of the camera in the imaging plane x and y directions respectively; (u0, v0) is the coordinate of the camera projection center in the image coordinate system, that is, the offset of the camera image coordinate system in the pixel coordinate system; d is the depth value at the pixel coordinate (u p ,v p ); the to-be-measured cable RGB image is as shown in Figure 2 .

[0029] Step 20: obtain the cable plane feature center line C 2D from the mask through an image thinning algorithm, and track the cable plane feature center line trajectory C 2D_order , index the cable space point cloud P cable according to the trajectory pixel coordinate C 2D_order (u, v) to obtain the cable space feature center line C 3D_order ;

[0030] Obtain the single-pixel-width cable plane feature center line (as shown in Figure 3 and Figure 4 ) from the mask image through the rosenfeld thinning algorithm, track the cable plane feature center line trajectory, and index the cable plane feature center line trajectory pixel coordinate C 2D_order(u, v) index cable space point cloud P cable , to obtain the cable space feature centerline trajectory ordered point set C 3D_order .

[0031] The specific method for tracking the cable plane feature centerline trajectory is as follows: first, search all end points on the cable plane feature centerline, select one end point as the starting point, and traverse the neighborhood non-0 pixels in turn until the other end point is reached to complete the trajectory tracking, when a branch point is encountered, save the branch point coordinates and the trajectory vector before the branch point, after tracking the first trajectory, continue to traverse the neighborhood non-0 pixels from the branch point along the trajectory before the branch point to complete the tracking of other trajectories, and realize the conversion of the cable plane feature centerline image point C 2D into the ordered feature center point coordinate point C 2D_order . For the multi-branch cable plane feature centerline, a common starting point of the multi-branch trajectories is selected to realize the separate tracking of each branch, and the trajectory coordinate vector is obtained.

[0032] Step 30: non-uniformly sampling the cable space feature centerline to obtain the space sampling point C 3D_sampling , and algebraically reconstructing the space sampling point to obtain the cable space feature curve F cable .

[0033] After the cable space feature centerline is non-uniformly sampled by the equal chord distance sampling method, the space sampling point C 3D_sampling is obtained. The equal chord distance sampling method makes the sampling point C 3D_sampling more consistent with the curvature variation characteristics of the original space feature point set C 3D_order , that is, the sampling points are dense when the curvature is large, and the sampling points are sparse when the curvature is small.

[0034] The equal chord distance sampling method uses the Euclidean distance (chord distance) between the arc midpoint and the chord midpoint as the sampling measurement index instead of the chord height, and the specific content is as follows: the space feature point set C 3D_order is sampled by equal arc length with a large step size to obtain an initial sampling interval, the arc chord distance is calculated interval by interval, and a sampling threshold is set. If the arc chord distance is greater than the threshold, the arc length is halved to obtain a new sampling interval, and the arc chord distance is calculated again until all intervals meet the equal arc chord distance sampling condition, and the cable space feature sampling point set C 3D_sampling is obtained.

[0035] The space sampling point is algebraically reconstructed by a parameter 8th order polynomial equation group to obtain the cable space feature curve F cable (X(t), Y(t), Z(t)):

[0036]

[0037] where a i , bi , c i respectively represent polynomial fitting coefficients; t is the equation set parameter, representing the position of each coordinate point in the ordered point set, t ∈ [1, m], m is the number of spatial points in the cable spatial feature sampling point set C 3D_sampling .

[0038] Substitute the coordinates of the cable spatial feature sampling point set C 3D_sampling into equation (2) to solve the polynomial fitting coefficients a i , b i , c i by the least squares method, and obtain the cable spatial feature curve F cable represented by the parameter equation set (as shown in Figure 5 ), and the point coordinates on the feature curve are represented as (X(t), Y(t), Z(t)).

[0039] Step 40: regarding the cable spatial feature curve F cable as a particle spatial motion trajectory, the spatial bending radius R cable of the cable spatial feature curve is solved by using the particle spatial motion law.

[0040] Regarding the cable spatial feature curve F cable as a particle spatial motion trajectory, the size of the cable bending radius R cable can be solved by the particle spatial motion normal acceleration a n and the motion speed v:

[0041]

[0042] Specifically, it can be described as follows: given that the particle spatial motion trajectory is the cable spatial feature curve F cable , the particle spatial motion trajectory velocity acceleration are respectively represented as:

[0043]

[0044] The particle spatial motion acceleration can be decomposed into the tangential acceleration and the normal acceleration The tangential acceleration has the same direction as the velocity , and by vector composition, the tangential acceleration has the size:

[0045]

[0046] where θ is the angle between the tangential acceleration and the acceleration The angle between them. The normal acceleration can be obtained from the vector difference The size of which is:

[0047]

[0048] The bending radius R of each point on the characteristic curve of the cable can be obtained from equation (3) cable is:

[0049]

[0050] The spatial bending radius of the cable can be regarded as a spatial vector in the spatial coordinate system of the characteristic curve of the cable, the length of the vector is the size of the bending radius, and the direction of the vector is collinear with the normal acceleration , as shown in FIG. 6. Figure 6 The visualization result of the spatial bending radius of the cable is shown in FIG. 7.

[0051] Although the embodiments of the present application are disclosed as above, the content described is only for the purpose of facilitating understanding of the present application, and is not intended to limit the present application. Any person skilled in the art of the present application can make any modification and change in the form and details without departing from the spirit and scope of the present application, and the patent protection scope of the present application shall be subject to the scope defined by the appended claims.

Claims

1. A cable bend radius calculation method based on mask centerline algebraic reconstruction, characterized in that, Comprise: Step A, obtain a cable image mask through a dual-modal semantic segmentation network, and construct a cable spatial point cloud P from a depth image cable ; Step B, the mask is obtained by the image thinning algorithm to get the cable plane feature center line C 2D And track the cable plane feature center point coordinate point C 2D_order , according to the trajectory pixel coordinate C 2D_order (u,v) index cable space point cloud P cable Get cable space feature center line trajectory ordered point set C 3D_order ; Step C, the cable space feature center line is non-uniformly sampled to obtain a space sampling point C 3D_sampling The space sampling point is algebraically reconstructed to obtain a cable space feature curve F cable ; Step D, the cable space characteristic curve F cable The space bending radius R of the cable space characteristic curve is solved by using the particle space motion law as the particle space motion trajectory cable ; In the step B, the image thinning algorithm adopts the rosenfeld thinning algorithm, tracks the cable plane feature center line track, and obtains the cable plane feature center line track pixel coordinates C 2D_order (u,v) index cable space point cloud P cable , to obtain an ordered point set C 3D_order of the cable space feature center line track that can be used for algebraic reconstruction. The tracking cable plane feature center line trajectory specifically comprises: Search all end points on the cable plane feature center line, select one of the end points as the starting point, and traverse the non-zero pixels in the neighborhood in turn until the other end point is reached, complete the trajectory tracking, when encountering a branch point, save the branch point coordinates and the trajectory vector before the branch point, after tracking the first trajectory, continue to traverse the non-zero pixels in the neighborhood from the branch point following the trajectory before the branch point to complete the tracking of other trajectories, realize the conversion of the cable plane feature center line image point C 2D to the ordered cable plane feature center point coordinate point C 2D_order For multi-branch cable plane feature center line, the common starting point of multiple branches must be selected to realize the separation tracking of each branch and obtain the trajectory coordinate vector respectively.

2. The cable bend radius calculation method based on mask centerline algebraic reconstruction according to claim 1, wherein, The cable RGB-D image input in step A is input into the improved bimodal semantic segmentation network ESANet to obtain a cable mask Mask; Generating cable point cloud P from depth map cable Each depth pixel coordinate (u p ,v p ) can get a corresponding spatial point coordinate P(X,Y,Z): where f x , f y are the focal lengths of the camera in the x, y directions of the imaging plane, respectively; (u0, v0) are the coordinates of the camera projection center in the image coordinate system; d is the depth value at pixel coordinates (u p , v p ).

3. The cable bend radius calculation method based on mask centerline algebraic reconstruction according to claim 1, wherein, The spatial sampling point C obtained by the non-uniform sampling of the cable spatial feature center line in step C by the equiangular chord length sampling method 3D_sampling The equiangular chord length sampling method makes the spatial sampling point C 3D_sampling Conform to the curvature variation characteristics of the ordered point set C of the cable spatial feature center line trajectory 3D_order The sampling points are dense when the curvature is large, and the sampling points are sparse when the curvature is small. The equal arc chord distance sampling method uses the Euclidean distance between the arc midpoint and the chord midpoint in the sampling interval to approximate the chord height as a sampling measurement index. The specific content is as follows: for the ordered point set C of the spatial characteristic center line track of the cable 3D_order Equal arc length sampling with large step length is performed to obtain an initial sampling interval. The arc chord distance is calculated interval by interval, and a sampling threshold is set. If the arc chord distance is greater than the threshold, the arc length is halved to obtain a new sampling interval, and the arc chord distance is calculated again until all intervals meet the equal arc chord distance sampling condition, and the spatial sampling points C are obtained 3D_sampling .

Images