Disclosure of Invention
The technical problem to be solved by the invention is as follows: the defects of the prior art are overcome, and the track real-time extraction method suitable for the automatic steel member spraying device is provided.
The technical scheme adopted by the invention for solving the technical problem is as follows: the track real-time extraction method suitable for the automatic steel member spraying device comprises the following steps of installing and fixing laser radars on the assembly line conveyor belt through an installing support, uniformly distributing the connecting lines of the laser radars over the assembly line conveyor belt, arranging the scanning ranges of the adjacent laser radars in a cross mode, and scanning the members conveyed by the assembly line conveyor belt through the laser radars:
step 1: fusing scanning frames: acquiring scanning frames of all laser radars, and fusing the scanning frames corresponding to all the laser radars into one scanning frame;
step 2: carrying out point cloud filtering processing on the scanning frame;
and step 3: extracting edge features and face features;
and 4, step 4: parameterizing the outline;
and 5: and generating a spraying track.
Laser radar's number is 3, and a lateral opposite side of gravity flow water line conveyer belt is 1 laser radar, 2 laser radar and 3 laser radar in proper order, step 1 is as follows in the concrete step:
aligning the scanning frames of No. 1 and No. 3 laser radars in each period to the scanning frames of No. 2 laser radars in each period through space transformation to form a certain componentFor a complete scan frame at time, the scan frame formula for the i time component is as follows:
in the formula Fi 1Scan frame for time i of lidar number 1, Fi 2Scan frame for time i of lidar number 2, Fi 3Scan frame, T, for time of laser radar No. 31 2Transformation matrix, T, for No. 1 lidar to No. 2 lidar3 2Transformation matrix for No. 3 lidar to No. 2 lidar, FiIs the scan frame at component i time.
Step 2 comprises the following substeps:
2-1 scanning frame F by pass-through filter
iExtracting the extract satisfying the constraint condition
Inner point of (1), wherein z
minIs the range threshold of the highest point of the top surface of the member, z
maxRange threshold of lowest point of bottom surface of member, y
minFor the coordinate at the minimum width of the conveyer belt of the production line, y
maxThe position of the maximum width of the conveyor belt of the assembly line is a coordinate;
2-2: removing outliers from the interior points by using a RadiusOutlierRemoval filter, wherein the radius R is searched, and the minimum neighbor number K is a hyper-parameter related to point cloud data distribution;
2-3: and finally, performing down-sampling on the point cloud data processed in the step 2-2 by using a voxel grid filter, wherein the leaf node size L is a hyper-parameter related to the point cloud data distribution, and the scanning frame processed by the three filters is marked as Fi′。
Scanning frame F for component contour in step 3i' extraction of feature points Fi={Fi e,Fi pIn which Fi eAs edge points, which are the nodes of the lance action, Fi pThe surface points correspond to areas to be sprayed;
the specific method is to utilize
Calculating the curvature of each discrete point, wherein S is the neighborhood set of the points to be calculated,
is the coordinate point of the ith scanning frame under the local coordinate system L,
is the j-th coordinate point of the k-th scanning frame under the local coordinate system L.
In step 4, an edge point threshold c is set according to the type of the componenteAnd a flat point threshold cpIs greater than ceIs set as an edge point, is less than cpIs set as a plane point;
n plane points exist between every two edge points, and a linear equation y of each section of plane point is obtained by utilizing least square polynomial linear fitting
i=f
i(x) Parameterizing the profile data of the scanning frame of the component, setting a discrete step length delta j according to the spraying process, and discretizing a parameter equation
And converting the scanning frame data into uniformly distributed discrete points.
And in the step 5, according to the profile parameterized equation and the discrete points in the step 4, combining related spraying process parameters to form a spraying track G ═ f (v, T) with time and speed characteristics, and then converting and aligning the spraying track G ═ f (v, T) into a spraying mechanical arm operation space by using a formula G ═ Tg, wherein T is a conversion matrix from No. 2 laser radar to a spraying mechanical arm.
Compared with the prior art, the invention has the following beneficial effects:
(1) the flexible spraying device is suitable for assembly line type flexible spraying operation of various steel members;
(2) compared with a method for reconstructing a complete 3D model, the method has the advantages of small calculated amount, high real-time property, strong universality, high efficiency and the like, can save the arrangement space of a production line and save the production cost while realizing instant scanning and instant spraying.
Detailed Description
Embodiments of the invention are further described below with reference to the accompanying drawings:
examples
As shown in fig. 1 to 3, install fixed lidar through installing support 4 on the assembly line conveyer belt 6, the top evenly distributed that assembly line conveyer belt 6 was spanned to each lidar's line, adjacent lidar's scanning range cross arrangement, installing support 4 spanes assembly line conveyer belt 6 and passes through the lower margin fixed good, avoid with assembly line conveyer belt 6's direct contact, lidar installs on installing support 4, lidar scans component 5 that assembly line conveyer belt 6 carried, lidar connection control system, control system communication connection spraying arm, be provided with the spraying device on the spraying arm, including the following step:
step 1: fusing scanning frames: acquiring scanning frames of all laser radars, and fusing the scanning frames corresponding to all the laser radars into one scanning frame; laser radar's number is 3, 6 one side of water line conveyer belt that flows is 1 laser radar 1 in proper order, 2 laser radar 2 and 3 laser radar 3, this embodiment sets up 3 LMS4111 laser radar fixed mounting on installing support 4, one 2 laser radar 2 be located and produce the line positive centre promptly, two other 1 laser radar 1 and 3 laser radar 3 be fixed in respectively and produce the line outside promptly, form the cross scanning visual field, accomplish the complete cover to the component on the assembly line, concrete distribution is as shown in FIG. 1. The specific steps in step 1 are as follows:
laser radars No. 1 and No. 3Aligning the scanning frame of each period of 3 to the scanning frame of each period of No. 2 laser radar through space transformation to form a complete scanning frame of the component at a certain moment, wherein the scanning frame formula of the component at the moment i is as follows:
in the formula Fi 1Scanning frame at time 1i for lidar number 1, Fi 2Scanning frame at time 2i of laser radar number 2, Fi 3Scanning frame for time 3 of laser radar No. 3, T1 2Transformation matrix, T, for No. 1 lidar to No. 2 lidar3 2Transformation matrix for No. 3 lidar to No. 2 lidar, FiIs the scan frame at component i time.
Wherein the matrix T is transformed
1 2、T
3 2Needs to be calculated and calibrated in advance according to the installation pose of the laser radar,
R
1、t
1is the rotation matrix and translation vector of No. 1 laser radar to No. 2 laser radar
3、t
3The laser radar is a rotation matrix and a translation vector of a No. 3 laser radar to a No. 2 laser radar.
Calculating included angle between each frame data point and laser radar x-axis
The visual angle range of LMS4111 laser radar is [ -35 DEG, 35 DEG ]]And eliminating points close to the edge of the field angle.
Step 2: carrying out point cloud filtering processing on the scanning frame; the acquired point cloud data can generate some noise points and outliers due to measurement random errors such as equipment precision, field environment, installation errors and the like, and the data density is uneven, so that the subsequent data processing such as feature point extraction and the like is greatly influenced. The setting step 2 therefore comprises the following sub-steps:
2-1 scanning frame F by pass-through filter
iExtracting the extract satisfying the constraint condition
I.e. the complete contour scan frame at time instant of component i. Wherein z is
minIs the range threshold of the highest point of the top surface of the member, z
maxRange threshold of lowest point of bottom surface of member, y
minFor the coordinate at the minimum width of the conveyer belt of the production line, y
maxThe position of the maximum width of the conveyor belt of the assembly line is a coordinate;
2-2: removing outliers from the interior points by using a RadiusOutlierRemoval filter, wherein the radius R is searched, and the minimum neighbor number K is a hyper-parameter related to point cloud data distribution;
2-3: and finally, performing down-sampling on the point cloud data processed in the step 2-2 by using a voxel grid filter to ensure that the point cloud is uniformly and smoothly distributed in the space, and improving the operation speed and robustness of a subsequent processing algorithm, wherein the leaf node size L is a hyper-parameter related to the point cloud data distribution, and a scanning frame processed by the three filters is recorded as Fi′。
And step 3: extracting edge features and face features; scanning frame F for component contour in step 3i' extraction of feature points Fi={Fi e,Fi pIn which Fi eAs edge points, which are the nodes of the lance action, Fi pThe surface points correspond to areas to be sprayed;
the specific method is to utilize
Calculating the curvature of each discrete point, wherein S is the neighborhood set of the points to be calculated,
is the coordinate point of the ith scanning frame under the local coordinate system L,
is the j-th coordinate point of the k-th scanning frame under the local coordinate system L.
And 4, step 4: parameterizing the outline; the member has the characteristics of regular appearance surface, plane spraying surface, clear edge angle and the like. In step 4, an edge point threshold c is set according to the type of the componenteAnd a flat point threshold cpIs greater than ceIs set as an edge point, is less than cpIs set as a plane point;
n plane points exist between every two edge points, and a linear equation y of each section of plane point is obtained by utilizing least square polynomial linear fitting
i=f
i(x) Parameterizing the profile data of the scanning frame of the
component 5, setting a discrete step length delta j according to the spraying process, and discretizing a parameter equation
Thereby converting the irregularly distributed scan frame data into uniformly distributed discrete points that better characterize the profile.
And 5: and (3) generating a spraying track, combining related spraying process parameters according to the profile parameterized equation and discrete points in the step (4) in the step (5), forming the spraying track G-f (v, T) with time and speed characteristics, and performing conversion and alignment to the operation space of the spraying mechanical arm by using a formula G-Tg, wherein T is a conversion matrix from No. 2 laser radar to the spraying mechanical arm. And then data preparation is carried out for the analysis of the spraying process, so that the control system controls the spraying mechanical arm to carry out spraying operation according to the spraying track. The treatment method can be applied to steel members with the characteristics of regular appearance surface, flat plate type member, groove type member, square member and the like, such as round member, H-shaped member, flat plate type member, groove type member, square member and the like, and the sprayed surface is a plane and has clear edges and corners.