CN116242277A - Automatic measurement method for size of power supply cabinet structural member based on full-field three-dimensional vision - Google Patents
Automatic measurement method for size of power supply cabinet structural member based on full-field three-dimensional vision Download PDFInfo
- Publication number
- CN116242277A CN116242277A CN202310131114.9A CN202310131114A CN116242277A CN 116242277 A CN116242277 A CN 116242277A CN 202310131114 A CN202310131114 A CN 202310131114A CN 116242277 A CN116242277 A CN 116242277A
- Authority
- CN
- China
- Prior art keywords
- point cloud
- measurement
- structural member
- power supply
- shape feature
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
- G01B11/00—Measuring arrangements characterised by the use of optical techniques
- G01B11/24—Measuring arrangements characterised by the use of optical techniques for measuring contours or curvatures
- G01B11/25—Measuring arrangements characterised by the use of optical techniques for measuring contours or curvatures by projecting a pattern, e.g. one or more lines, moiré fringes on the object
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
- G01B11/00—Measuring arrangements characterised by the use of optical techniques
- G01B11/24—Measuring arrangements characterised by the use of optical techniques for measuring contours or curvatures
- G01B11/25—Measuring arrangements characterised by the use of optical techniques for measuring contours or curvatures by projecting a pattern, e.g. one or more lines, moiré fringes on the object
- G01B11/254—Projection of a pattern, viewing through a pattern, e.g. moiré
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T19/00—Manipulating 3D models or images for computer graphics
- G06T19/20—Editing of 3D images, e.g. changing shapes or colours, aligning objects or positioning parts
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10028—Range image; Depth image; 3D point clouds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- Computational Mathematics (AREA)
- Pure & Applied Mathematics (AREA)
- Evolutionary Computation (AREA)
- Mathematical Optimization (AREA)
- Mathematical Analysis (AREA)
- Architecture (AREA)
- Computer Graphics (AREA)
- Software Systems (AREA)
- Length Measuring Devices By Optical Means (AREA)
Abstract
The method is characterized in that the method is based on a full-field three-dimensional vision-based automatic measurement method for the size of a power supply cabinet structural member, and based on a phase-shift fringe projection measurement method, single measurement viewpoint point clouds under a measurement viewpoint of the power supply cabinet structural member are obtained through multi-brightness fringe projection synthesis; acquiring all measurement viewpoint point clouds according to the method, splicing point clouds by using mark points with random positions on a measurement turntable to obtain complete point clouds of a power supply case structural member, and denoising the complete point cloud data to obtain measurement point clouds of the power supply case structural member; and obtaining a model point cloud according to the CAD model of the power supply case structural member, registering the measured point cloud and the model point cloud to form a unified coordinate system, extracting a local measured point cloud corresponding to the shape feature on the measured point cloud based on the shape feature extracted from the power supply case structural member model, fitting the local measured point cloud to obtain geometric parameters, and calculating the actually measured size information of the shape feature based on the geometric parameters. The invention has strong adaptability, high precision and high speed, and can realize the automatic detection of structural members.
Description
Technical Field
The invention belongs to the field of three-dimensional measurement, and relates to an automatic measurement method for the size of a power supply cabinet structural member based on full-field three-dimensional vision.
Background
The satellite-borne equipment structural parts such as a power supply case and the like are used as important satellite parts and are mainly used for highly integrated spacecraft electronic systems, and the detection efficiency and the precision requirements of related structural parts are high. The power cabinet structural part is usually a high-precision machined part, and has a plurality of single parts and pieces. At present, manual caliper measurement is mainly relied on, and when the size to be measured is more, the time consumption and the precision of a manual measurement mode are low, so that the detection requirement cannot be met; the contact type measuring method such as a three-coordinate machine is high in precision, but when the contact type measuring method is used, a measuring path needs to be programmed, high offline programming cost is achieved, and actual production requirements cannot be met. .
Disclosure of Invention
Aiming at the defects or improvement demands of the prior art, the invention provides an automatic measurement method for the size of a power cabinet structural member based on full-field three-dimensional vision, which solves the technical problems that the existing manual measurement and contact measurement are low in efficiency and precision and cannot meet the actual detection demands of the power cabinet structural member.
The invention provides a full-field three-dimensional vision-based automatic measurement method for the size of a power supply cabinet structural member, which comprises the following steps:
the three-dimensional vision measurement system obtains a single measurement viewpoint point cloud under one measurement viewpoint of the power supply case structural member through multi-brightness stripe projection and synthesis;
acquiring all measured point clouds of the power cabinet structural member according to the method of the previous step, splicing all measured point clouds by using the marking points with random positions on the measuring turntable to obtain the complete point cloud containing the power cabinet structural member, and denoising the complete point cloud data containing the power cabinet structural member to obtain the measured point cloud of the power cabinet structural member;
and obtaining a model point cloud according to the CAD model of the power supply case structural member, registering the measuring point cloud and the model point cloud to unify the measuring point cloud and the model point cloud under the same coordinate system, extracting a local measuring point cloud corresponding to the shape feature on the measuring point cloud based on the shape feature extracted from the CAD model of the power supply case structural member, fitting the local measuring point cloud to obtain geometric parameters, and calculating the measured size information of the shape feature based on the geometric parameters.
Further, the three-dimensional vision measurement system obtains a single measurement viewpoint point cloud under one measurement viewpoint of the power cabinet structural member through multi-brightness stripe projection and synthesis, and specifically comprises:
forming a plurality of groups of phase shift fringe patterns by adopting a plurality of three-frequency four-step phase shift fringe patterns under different fringe projection brightness, projecting the plurality of groups of phase shift fringe patterns from dark to light onto the surface of a power supply cabinet structural member, and synchronously acquiring fringe patterns modulated by the power supply cabinet structural member by a double camera of a three-dimensional vision measurement system so as to generate a plurality of groups of mask images;
counting the number of unsaturated pixels of the same pixel point of the mask image in the plurality of groups of phase-shift fringe patterns under each brightness level, and enabling the corresponding mask image to be effective when the number exceeds a saturation threshold value, or else disabling the mask image;
for the pixel points of the effective mask image, only the mask image pixel points corresponding to the brightness level with the maximum modulation degree are reserved as a final mask image, and the final mask image is utilized to synthesize a phase shift stripe image;
and processing the phase-shift fringe image to obtain a single-measurement viewpoint point cloud.
Further, the method according to the previous step obtains all measured viewpoint point clouds of the power cabinet structural member, specifically including:
the measurement view point of the three-dimensional vision measurement system is set, so that the working distance of the three-dimensional vision measurement system is within the depth of field range of the double cameras, at least 3 pairs of mark points in the double camera view field are simultaneously shot, the circle centers of the 3 pairs of mark points are not collinear, the double cameras measure the point cloud data on the top surface of the power cabinet structural member from the upper part of the power cabinet structural member, and then the point cloud data on the side surface of the power cabinet structural member is supplemented and measured from four directions, namely front, back, left and right.
Furthermore, the splicing of all the measurement viewpoint point clouds by using the random marker points on the measurement turntable to obtain a complete point cloud containing the power cabinet structural member specifically comprises the following steps:
step 1, extracting circle centers of mark points in an acquired image, carrying out polar line matching, calculating the spatial positions of the circle centers of the mark points, solving a rigid transformation matrix between two groups of point clouds by using a least square method for the two groups of point clouds acquired by two adjacent measurement viewpoints, and obtaining a splicing result of two groups of adjacent point cloud data;
step 2, continuously splicing point clouds of two adjacent viewpoints based on mark points according to the method of the step 1, and further registering by using an ICP algorithm until splicing of point clouds of all viewpoints is completed; and Gaussian filtering is carried out on the point clouds in the spliced overlapping area, so that the complete point cloud containing the power supply case structural member is obtained.
Further, the performing gaussian filtering on the point cloud of the spliced overlapping area specifically includes:
setting the side length of a voxel grid, dividing point cloud data into a plurality of voxel grids with the same size for all single measurement viewpoint point clouds, calculating the gravity center of the point clouds in each voxel grid, and replacing points in the voxel grid by the gravity center;
traversing all measured viewpoint point clouds, and calculating the distance d between each point and k adjacent points of the point i I=1, 2, …, k, the subscript i is the number of the k nearest neighbor, and the average distance is calculated
And d i Standard deviation std, knockout->
Is formed by the following steps ofWith adjacent points, t is an adjustable parameter.
Further, the denoising processing is performed on the complete point cloud data including the power cabinet structural member to obtain a measurement point cloud of the power cabinet structural member, specifically including:
three main directions of a complete point cloud containing a power supply case structural member are obtained by utilizing a principal component analysis method, and a normal vector perpendicular to a turntable plane is determined so as to determine the turntable plane; and removing the turntable Ping Miandian cloud according to the size of the power supply cabinet structure, and finally obtaining the measuring point cloud of the power supply cabinet structure.
Further, the obtaining the model point cloud according to the CAD model of the power cabinet structural member specifically includes:
and converting the CAD model into an STL format, obtaining the information of the triangular patches on the surface of the CAD model, randomly up-sampling each triangular patch to fill point cloud data into the triangular patches, and finally discretizing the CAD model into model point clouds.
Further, registering the measurement point cloud with the model point cloud to unify the measurement point cloud and the model point cloud under the same coordinate system, extracting a local measurement point cloud corresponding to the shape feature on the measurement point cloud based on the shape feature extracted from the CAD model of the power machine case structural member, fitting the local measurement point cloud to obtain geometric parameters, and calculating actual measurement size information of the shape feature based on the geometric parameters, wherein the method specifically comprises the following steps:
based on an FRG algorithm, carrying out first registration on the measurement point cloud and the model point cloud of the power supply case structural member, so that the measurement point cloud and the model point cloud are unified into the same coordinate system;
performing second registration on the measurement point cloud and the model point cloud based on an ICP algorithm;
dividing the point cloud of the shape feature according to the shape feature type of the CAD model to obtain a local point cloud corresponding to the shape feature, fitting the local point cloud, and further obtaining the geometric dimension parameter of the shape feature;
positioning the position of the corresponding shape feature on the measurement point cloud according to the spatial position of the shape feature extracted from the CAD model, and removing the point cloud except the shape feature by using a straight-through filter to segment out the local measurement point cloud of the corresponding shape feature;
and firstly denoising the local measurement point cloud of the shape feature, then adopting fitting based on a random sampling consistency algorithm and a whole least squares algorithm to obtain geometric parameters of the shape feature, and calculating actual measurement size information of the shape feature based on the geometric parameters.
Further, the method further comprises: comparing the actually measured size information of the shape features with the design size extracted from the design model, judging whether the processing is qualified or not, and giving a report.
Further, the power supply chassis may be replaced with other on-board devices.
The method for automatically measuring the size of the power supply case structural member based on full-field three-dimensional vision is used for detecting the three-dimensional size of the power supply case structural member according to the input three-dimensional point cloud model and the marked size of the power supply case structural member and generating a detection report, so that the detection efficiency and the detection accuracy are improved, powerful guarantee is provided for the processing quality of the power supply case structural member of the spacecraft electronic system, and the method can be further applied to measuring the size of the payload of the on-board equipment such as an on-board computer and on-board communication equipment, and has good universality.
Drawings
FIG. 1 is a schematic flow chart of a three-dimensional vision-based automatic measurement method for scanning a power supply case structural member;
FIG. 2 is a schematic diagram of a phase shift fringe projection measurement system based on binocular stereo vision;
FIG. 3 is a schematic diagram of a three-dimensional vision-based automatic power supply case structural member scanning measurement system provided by the invention;
FIG. 4 is a flow chart of the registration of the real point cloud and the model point cloud;
fig. 5 shows the result of the point cloud feature segmentation, (a) is the complete part point cloud, (b) is the feature associated with the distance dimension 1 between two planes, and (c) is the feature associated with the distance dimension 2 between two planes.
Fig. 6 shows the results of the feature segmentation for the industrial part dimensions, (a) a CAD model of the dimension, and (b) a point cloud of segmented features.
Fig. 7 shows the fitting results of the point cloud features, (a) the fitting results of the distance dimension 1 correlation features between two planes, and (b) the fitting results of the distance dimension 2 correlation features between two planes.
Detailed Description
In order to make the objects and technical solutions of the present invention clearer, the present invention will be further described in detail with reference to the accompanying drawings and examples.
The invention provides a full-field three-dimensional vision-based automatic measurement method for the size of a power supply cabinet structural member. Payloads of on-board computers, on-board communication devices, etc. typically have a regular shape resembling a cube. The method of the invention can also be applied to the measurement of the size of the effective load of the on-board equipment such as an on-board computer, an on-board communication equipment and the like, and has good universality. The invention can realize automatic scanning and size measurement of the structural parts of the on-board equipment. The following takes a structural component of the on-board power supply case as an example to describe the real-time mode of the invention in detail.
The embodiment of the invention provides a full-field three-dimensional vision-based automatic measurement method for the size of a power supply cabinet structural member, which is shown in fig. 1 and comprises the following steps:
the automatic measuring method of the invention is realized by utilizing a three-dimensional vision measuring system, and referring to fig. 2-3, the three-dimensional vision measuring system consists of a turntable, a mechanical arm, a phase matching full-field measuring sensor, a computer and the like.
Specifically, the upper surface of the turntable is provided with a special flexible clamp, and round mark points which are randomly distributed are adhered at the same time. The phase matching full-field measuring sensor is driven by a mechanical arm and is matched with the turntable to rotate, so that the three-dimensional point cloud of the power supply case structural member is obtained. The computer is used to store data and perform control and computing functions.
The phase matching full-field measuring sensor consists of a double camera, a stripe projector, a cooling fan, a data interface and a laser projector. The camera and the stripe projector are selected according to factors such as the size design camera field of view, the volume and the like of the measured object, and the working distance is determined by the measuring range and the projector field angle. The maximum size of the measured piece can be 500mm multiplied by 500mm, and because the single-view field measurement has higher requirements on the projector and the measuring head base line, the increase of the view field can increase the point distance and influence the measurement precision, the multi-view field measurement is adopted, and each view field measurement range is recommended to be 260mm multiplied by 220mm; in order to cover the view field, the projector adopts various medium-sized short Jiao Toushe projectors with larger view field angles; and calculating according to the view angle and the view field transverse range to obtain the working distance of the sensor, wherein the calculated result is 500mm. Then, the resolution of the camera is selected according to the measurement requirement, and when the curvature of the measured object is changed greatly and the step edges are more, the accuracy of detail measurement can be ensured by the high-resolution camera, and the 2448 multiplied by 2048 resolution camera is selected. Secondly, according to camera parameters, the lens is selected, the focal length of the lens can be calculated and obtained according to the size of the target surface of the camera, the working distance of the sensor and the measuring range, and the lens with the focal length of 16mm is preferable. Meanwhile, the optical resolution of the lens is larger than the resolution of the camera. And finally, simulating according to system parameters, calculating a random error of measurement, and judging whether the measurement precision index is met. The cooling fan is used for cooling the camera and the stripe projector and inhibiting temperature drift; the data interface is used for data transmission between the sensor and the computer; the laser projector is used for manually assisting in judging whether the sensor is located at the designed working distance, and when the laser points on the surface of the measured object are overlapped, the surface position is the designed working distance.
The implementation process of the method for realizing automatic measurement of the size of the power supply cabinet structural part based on full-field three-dimensional vision by utilizing the three-dimensional vision measurement system is as follows:
s1, acquiring local point clouds of a power cabinet structural member based on phase matching three-dimensional vision measurement, and acquiring monoscopic point clouds of a strong reflection surface of the power cabinet structural member in a measurement angle view field of a sensor by utilizing a phase shift fringe projection measurement principle and through a multi-brightness fringe projection and synthesis technology.
Specifically, the phase matching full-field measuring sensor utilizes a stripe projector to project phase shifting sinusoidal stripes to an object to be measured, two cameras are used for shooting the stripes, the phase of the stripes is unfolded, and then the three-dimensional shape of the object is calculated by utilizing a structured light triangulation principle or a binocular stereoscopic vision principle. The sensor utilizes the phase shift fringe projection measurement principle and combines the multi-brightness synthesis technology to realize the monoscopic accurate measurement of the strong reflection surface of the power supply case structural member, and the specific flow is as follows:
s11, adopting a three-frequency four-step phase shift fringe pattern, namely, fringe periods are respectively 15, 16 and 17 pixels, and throwing 4 phase shift fringe patterns taking pi/2 as steps under each period, wherein the fringe patterns can be expressed as:
wherein I is i (x, y) represents the ith phase-shifted fringe pattern, (x, y) represents the coordinates of the camera image plane coordinate system, I '(x, y) represents the average brightness of the background light, I' (x, y) represents the modulation degree of the fringe pattern,
representing the wrapping phase.
S12, projecting the phase shift fringe patterns onto the surface of a tested object, designing 7 groups of three-frequency four-step phase shift fringe patterns with different projection brightness for the surface of a power cabinet structural member with different reflectivities, and projecting 7 groups of phase shift fringe patterns onto the surface of the power cabinet structural member from dark to light by a fringe projector, wherein the two cameras synchronously acquire the fringe patterns modulated by the surface of the tested object.
S13, generating a group of mask images M k (x, y) counting the number of unsaturated pixels in the N phase-shift fringe patterns of the same pixel point (x, y) under each brightness level, and marking as m k (x, y), assigning a value to the mask image:
wherein, when the mask value is 1, the mask value is effective, and when the mask value is 0, the mask value is ineffective.
The modulation degree of the stripe pattern was calculated:
wherein k is the number of brightness levels, I' k (x, y) is the modulation degree under k brightness level, i is the phase shift number, the value is 0 to 3,
for the brightness of the fringe pattern acquired under the k brightness level, (x, y) represents the coordinates of the camera image plane coordinate system;
for the pixel points with the effective number of the mask values being larger than 1, only the corresponding mask value is effective under the brightness level with the maximum modulation degree, and the mask values under other brightness are set as 0:
mask image pixel points with effective mask values are used as final mask images, and finally, a group of high dynamic phase shift stripe images are synthesized by using the mask images:
s14, according to the synthesized high-dynamic phase shift fringe image, carrying out phase principal value solving, multi-frequency heterodyne phase unfolding, polar line correction, phase matching and three-dimensional reconstruction to obtain a monoscopic point cloud.
S2, according to the principle of obtaining single-viewpoint point clouds, a measuring head is driven by a rotary turntable in cooperation with mechanical arm path planning, measurement of different angles of a power supply case structural member is achieved, all surfaces of the power supply case structural member are covered, splicing is achieved based on mark points on the turntable, complete dense point clouds are obtained, as shown in fig. 3, point cloud data preprocessing is conducted aiming at noise and invalid points, and the specific steps are as follows:
s21, setting a measurement viewpoint of a camera by combining the size of a power cabinet structural member, and requiring the working distance of the camera to be within the depth of field of the camera, so that a clear mark point image is shot, and the precision of the extracted center coordinates is higher. Because the power cabinet structural member shields the marking points on the turntable, at least 3 pairs of marking points in the double camera view field are required to be shot at the same time, and the center coordinates of the 3 pairs of marking points are not collinear. In cooperation with the movement of the turntable and the mechanical arm, the two camera respectively measures the point cloud data of the top surface of the power cabinet structural member from the upper part of the power cabinet structural member, and then supplements and measures the point cloud data of the side surface of the power cabinet structural member from four directions, namely front, back, left and right.
S22, the marker points are fixedly adhered to the turntable, so that the space relative positions are unchanged all the time, according to the characteristics, firstly, the circle centers of the marker points in the double-camera image are extracted, epipolar matching is carried out, the space positions of the circle centers of the marker points are calculated, two groups of point clouds acquired by two adjacent measurement viewpoints are subjected to central coordinates of the extracted marker points, a rigid transformation matrix between the two groups of point clouds is solved by using a least square method, and a rough splicing result of two groups of adjacent point cloud data is obtained, so that the two groups of point clouds are approximately unified under the same coordinate system.
S23, continuously performing rough stitching on two adjacent viewpoint point clouds based on the mark points, and performing accurate registration by using an ICP algorithm until stitching of all view field point clouds is completed.
S24, aiming at the problems that the point clouds in the overlapped area are too dense and the details are disordered, the effect of eliminating the overlapped points is achieved while the geometrical characteristics are maintained by utilizing a voxel grid method, and aiming at noise in original point cloud data, point cloud filtering is carried out by adopting Gaussian filtering.
Specifically, the voxel grid method sets the side length of a voxel grid, divides point cloud data into a plurality of voxel grids with the same size, calculates the gravity center of the point cloud data in the grids, reserves the gravity center, and replaces the data points in the grids by the gravity center so as to achieve the effect of simplifying the point cloud data. Specifically, a K neighborhood method is utilized to traverse all measured viewpoint point clouds, and the distance d between each point and K adjacent points in the K neighborhood of the point is calculated i (i=1, 2, …, k, subscript i is the number of k neighbors), average distance
And d i Standard deviation std, knockout->
T is an adjustable parameter.
S25, aiming at irrelevant point cloud noise such as a turntable, three main directions and centroids of the point cloud are obtained by using a principal component analysis method, in the three-dimensional point cloud, a covariance matrix eigenvalue represents the amplitude with the largest variance of data change, an eigenvector represents the direction with the largest variance, and the direction perpendicular to the plane of the turntable, namely the eigenvector direction corresponding to the largest eigenvalue. And a single-point random sampling consistency algorithm is adopted, a plane is determined by randomly selected points in the actual measurement point cloud and the normal vector of the plane of the turntable to serve as the plane of the turntable, and the turntable Ping Miandian cloud can be removed to obtain the point cloud of the measured object, so that the measurement point cloud of the power supply case structural member is obtained. In order to further remove irrelevant point cloud noise in the whole point cloud data, a plane is determined for randomly selected points and normal vectors in all measurement point clouds based on a single-point random sampling consistency algorithm, a plurality of planes are calculated for a plurality of randomly selected points, and a distance judgment threshold value theta of each plane is obtained h The points in the plane are taken as the turntable Ping Miandian cloud, and since the number of turntable plane points is the largest and the turntable size is known, the turntable Ping Miandian cloud can be determined according to the number of the extracted points in the plane and the plane size. And removing the turntable Ping Miandian cloud to obtain the point cloud of the measured object so as to obtain the measurement point cloud of the power supply case structural member. Wherein the threshold value theta is determined h The adaptability adjustment can be carried out according to actual application scenes and experience.
S3, registering the measured point cloud and the model point cloud based on the FRG algorithm and the ICP algorithm, performing point cloud segmentation, feature fitting and size calculation based on the size-related features, and performing comparison and identification with the size information extracted from the chassis CAD model to realize size measurement, wherein the flow is shown in FIG. 4.
S31, converting the CAD model into an STL format, obtaining information of triangular patches on the surface of the model, randomly up-sampling the triangular patches according to each triangular patch, filling point cloud data into the triangular patches, and finally discretizing the STL model into point clouds.
S32, performing primary rough registration on the measured point cloud of the piece to be detected and the CAD model point cloud based on the FRG algorithm, enabling the model point cloud and the real point cloud under different coordinate systems to be approximately unified to the same coordinate system, providing a good initial position for subsequent ICP accurate registration, and ensuring that the accurate registration algorithm is effective.
S33, accurately registering the actually measured point cloud with the CAD model for the second time based on an ICP algorithm, and improving registration accuracy to enable the point cloud to achieve best fitting.
S34, dividing the point cloud near the feature according to the size-related feature types (such as a plane, a cylinder and the like) provided on the CAD model to obtain a local point cloud corresponding to the feature, and fitting the local point cloud to obtain the geometric size parameter of the feature.
The size types extracted from the CAD model include linear size, radius size, and diameter size, wherein the linear size can be calculated from the point-to-plane distance and the point-to-line distance, and the radius size and the diameter size can be directly read from the fitting equation.
The method comprises the steps of based on the point cloud segmentation of the dimension association features, rapidly positioning the position of the dimension association features on the real-time point cloud according to the space position information in the dimension association feature information extracted from the three-dimensional CAD model, and rapidly cutting out the parts except the features by using a straight-pass filter to achieve the purpose of separating the features, as shown in fig. 5. After the complete real-time point cloud is segmented, the segmentation result is shown in fig. 6.
Fitting of point cloud surface characteristic parameters, firstly denoising, reducing errors, and then adopting fitting characteristics based on random sampling consistency (RANSAC) and integral least squares (TLS). The random sampling consistency algorithm gives up some invalid points, uses data which meets the conditions as much as possible, searches a model to fit, and accordingly denoising the point cloud and reserving effective data. The integral least square method is based on the improvement of the least square method, the interference factors in the coefficient matrix during fitting are taken into consideration, errors on three coordinate axes of x, y and z are taken into consideration, and compared with the least square method, the fitting result obtained is more accurate, and the fitting result is shown in figure 7.
S35, calculating the corresponding actually measured size information in the actual measurement point cloud based on the geometric parameters, comparing the actually measured size information with the design size information extracted from the CAD design model, judging whether the machining is qualified or not, and outputting a report.
Claims (10)
1. The automatic measurement method for the size of the power cabinet structural member based on full-field three-dimensional vision is characterized by comprising the following steps of:
the three-dimensional vision measurement system obtains a single measurement viewpoint point cloud under one measurement viewpoint of the power supply case structural member through multi-brightness stripe projection and synthesis;
acquiring all measured point clouds of the power cabinet structural member according to the method of the previous step, splicing all measured point clouds by using the marking points with random positions on the measuring turntable to obtain the complete point cloud containing the power cabinet structural member, and denoising the complete point cloud data containing the power cabinet structural member to obtain the measured point cloud of the power cabinet structural member;
and obtaining a model point cloud according to the CAD model of the power supply case structural member, registering the measuring point cloud and the model point cloud to unify the measuring point cloud and the model point cloud under the same coordinate system, extracting a local measuring point cloud corresponding to the shape feature on the measuring point cloud based on the shape feature extracted from the CAD model of the power supply case structural member, fitting the local measuring point cloud to obtain geometric parameters, and calculating the measured size information of the shape feature based on the geometric parameters.
2. The method according to claim 1, wherein the three-dimensional vision measurement system obtains a single measurement viewpoint point cloud under one measurement viewpoint of the power box structure through multi-brightness stripe projection and synthesis, and specifically comprises:
forming a plurality of groups of phase shift fringe patterns by adopting a plurality of three-frequency four-step phase shift fringe patterns under different fringe projection brightness, projecting the plurality of groups of phase shift fringe patterns from dark to light onto the surface of a power supply cabinet structural member, and synchronously acquiring fringe patterns modulated by the power supply cabinet structural member by a double camera of a three-dimensional vision measurement system so as to generate a plurality of groups of mask images;
counting the number of unsaturated pixels of the same pixel point of the mask image in the plurality of groups of phase-shift fringe patterns under each brightness level, and enabling the corresponding mask image to be effective when the number exceeds a saturation threshold value, or else disabling the mask image;
for the pixel points of the effective mask image, only the mask image pixel points corresponding to the brightness level with the maximum modulation degree are reserved as a final mask image, and the final mask image is utilized to synthesize a phase shift stripe image;
and processing the phase-shift fringe image to obtain a single-measurement viewpoint point cloud.
3. The method according to claim 2, wherein the obtaining all measured point-of-view clouds of the power box structural member according to the method of the previous step specifically includes:
the measurement view point of the three-dimensional vision measurement system is set, so that the working distance of the three-dimensional vision measurement system is within the depth of field range of the double cameras, at least 3 pairs of mark points in the double camera view field are simultaneously shot, the circle centers of the 3 pairs of mark points are not collinear, the double cameras measure the point cloud data on the top surface of the power cabinet structural member from the upper part of the power cabinet structural member, and then the point cloud data on the side surface of the power cabinet structural member is supplemented and measured from four directions, namely front, back, left and right.
4. The method of claim 3, wherein the splicing all the measurement viewpoint point clouds by using the random marker points on the measurement turntable to obtain a complete point cloud comprising the power supply chassis structural member specifically comprises:
step 1, extracting circle centers of mark points in an acquired image, carrying out polar line matching, calculating the spatial positions of the circle centers of the mark points, solving a rigid transformation matrix between two groups of point clouds by using a least square method for the two groups of point clouds acquired by two adjacent measurement viewpoints, and obtaining a splicing result of two groups of adjacent point cloud data;
step 2, continuously splicing point clouds of two adjacent viewpoints based on mark points according to the method of the step 1, and further registering by using an ICP algorithm until splicing of point clouds of all viewpoints is completed; and Gaussian filtering is carried out on the point clouds in the spliced overlapping area, so that the complete point cloud containing the power supply case structural member is obtained.
5. The method according to claim 4, wherein the gaussian filtering of the spliced overlapping area point cloud specifically comprises:
setting the side length of a voxel grid, dividing point cloud data into a plurality of voxel grids with the same size for all single measurement viewpoint point clouds, calculating the gravity center of the point clouds in each voxel grid, and replacing points in the voxel grid by the gravity center;
traversing all measured viewpoint point clouds, and calculating the distance d between each point and k adjacent points of the point i I=1, 2, …, k, the subscript i is the number of the k nearest neighbor, and the average distance is calculated
And d i Standard deviation std, knockout->
T is an adjustable parameter.
6. The method of claim 4, wherein denoising the complete point cloud data including the power box structure to obtain a measurement point cloud of the power box structure, specifically comprises:
three main directions of a complete point cloud containing a power supply case structural member are obtained by utilizing a principal component analysis method, and a normal vector perpendicular to a turntable plane is determined so as to determine the turntable plane; and removing the turntable Ping Miandian cloud according to the size of the power supply cabinet structure, and finally obtaining the measuring point cloud of the power supply cabinet structure.
7. The method according to claim 6, wherein the obtaining a model point cloud according to the CAD model of the power chassis structure specifically includes:
and converting the CAD model into an STL format, obtaining the information of the triangular patches on the surface of the CAD model, randomly up-sampling each triangular patch to fill point cloud data into the triangular patches, and finally discretizing the CAD model into model point clouds.
8. The method of claim 7, wherein registering the measurement point cloud with the model point cloud to unify the measurement point cloud and the model point cloud under a same coordinate system, extracting a local measurement point cloud corresponding to the shape feature on the measurement point cloud based on the shape feature extracted from the CAD model of the power chassis structural member, fitting the local measurement point cloud to obtain the geometric parameter, and calculating the measured dimension information of the shape feature based on the geometric parameter, and specifically comprising:
based on an FRG algorithm, carrying out first registration on the measurement point cloud and the model point cloud of the power supply case structural member, so that the measurement point cloud and the model point cloud are unified into the same coordinate system;
performing second registration on the measurement point cloud and the model point cloud based on an ICP algorithm;
dividing the point cloud of the shape feature according to the shape feature type of the CAD model to obtain a local point cloud corresponding to the shape feature, fitting the local point cloud, and further obtaining the geometric dimension parameter of the shape feature;
positioning the position of the corresponding shape feature on the measurement point cloud according to the spatial position of the shape feature extracted from the CAD model, and removing the point cloud except the shape feature by using a straight-through filter to segment out the local measurement point cloud of the corresponding shape feature;
and firstly denoising the local measurement point cloud of the shape feature, then adopting fitting based on a random sampling consistency algorithm and a whole least squares algorithm to obtain geometric parameters of the shape feature, and calculating actual measurement size information of the shape feature based on the geometric parameters.
9. The method according to claim 1, wherein the method further comprises: comparing the actually measured size information of the shape features with the design size extracted from the design model, judging whether the processing is qualified or not, and giving a report.
10. The method of any of claims 1-9, wherein the power chassis may be replaced with other on-board devices.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310131114.9A CN116242277A (en) | 2023-02-17 | 2023-02-17 | Automatic measurement method for size of power supply cabinet structural member based on full-field three-dimensional vision |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310131114.9A CN116242277A (en) | 2023-02-17 | 2023-02-17 | Automatic measurement method for size of power supply cabinet structural member based on full-field three-dimensional vision |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116242277A true CN116242277A (en) | 2023-06-09 |
Family
ID=86630815
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310131114.9A Pending CN116242277A (en) | 2023-02-17 | 2023-02-17 | Automatic measurement method for size of power supply cabinet structural member based on full-field three-dimensional vision |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116242277A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116740060A (en) * | 2023-08-11 | 2023-09-12 | 安徽大学绿色产业创新研究院 | Method for detecting size of prefabricated part based on point cloud geometric feature extraction |
-
2023
- 2023-02-17 CN CN202310131114.9A patent/CN116242277A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116740060A (en) * | 2023-08-11 | 2023-09-12 | 安徽大学绿色产业创新研究院 | Method for detecting size of prefabricated part based on point cloud geometric feature extraction |
| CN116740060B (en) * | 2023-08-11 | 2023-10-20 | 安徽大学绿色产业创新研究院 | Method for detecting size of prefabricated part based on point cloud geometric feature extraction |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110288642B (en) | Three-dimensional object rapid reconstruction method based on camera array | |
| CN110672039B (en) | Object omnibearing three-dimensional measurement method based on plane reflector | |
| CN110514143B (en) | Stripe projection system calibration method based on reflector | |
| CN106959078B (en) | A kind of contour measuring method for measuring three-dimensional profile | |
| CA2731680C (en) | System for adaptive three-dimensional scanning of surface characteristics | |
| CN111563921B (en) | Underwater point cloud acquisition method based on binocular camera | |
| CN105066909B (en) | A kind of many laser stripe quick three-dimensional measuring methods of hand-held | |
| CN107063129A (en) | A kind of array parallel laser projection three-dimensional scan method | |
| US11544860B2 (en) | Combined point cloud generation using a stationary laser scanner and a mobile scanner | |
| CN113205592B (en) | Light field three-dimensional reconstruction method and system based on phase similarity | |
| CN112308963B (en) | Non-inductive three-dimensional face reconstruction method and acquisition reconstruction system | |
| JP2016075637A (en) | Information processing apparatus and method for the same | |
| CN102376089A (en) | Target correction method and system | |
| CN110702025B (en) | Grating type binocular stereoscopic vision three-dimensional measurement system and method | |
| Ahmed et al. | Pothole 3D reconstruction with a novel imaging system and structure from motion techniques | |
| CN112489109B (en) | Three-dimensional imaging system method and device and three-dimensional imaging system | |
| CA3233222A1 (en) | Method, apparatus and device for photogrammetry, and storage medium | |
| CN106500625A (en) | A kind of telecentricity stereo vision measuring apparatus and its method for being applied to the measurement of object dimensional pattern micron accuracies | |
| CN116242277A (en) | Automatic measurement method for size of power supply cabinet structural member based on full-field three-dimensional vision | |
| CN107367245B (en) | Invalid point detection and elimination method in optical three-dimensional profile measurement | |
| Bergström et al. | Automatic in-line inspection of shape based on photogrammetry | |
| CN113012238B (en) | Method for quick calibration and data fusion of multi-depth camera | |
| CN112525106B (en) | Three-phase machine cooperative laser-based 3D detection method and device | |
| CN113393413B (en) | Water area measuring method and system based on monocular and binocular vision cooperation | |
| JP2003065737A (en) | System and method for measuring surface shape and system for illustrating surface condition |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |