CN115564811A - Point cloud registration method for transformer wiring terminal - Google Patents
Point cloud registration method for transformer wiring terminal Download PDFInfo
- Publication number
- CN115564811A CN115564811A CN202211191262.1A CN202211191262A CN115564811A CN 115564811 A CN115564811 A CN 115564811A CN 202211191262 A CN202211191262 A CN 202211191262A CN 115564811 A CN115564811 A CN 115564811A
- Authority
- CN
- China
- Prior art keywords
- point cloud
- point
- registration
- target
- source
- 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
- 238000000034 method Methods 0.000 title claims abstract description 24
- 239000013598 vector Substances 0.000 claims abstract description 33
- 239000011159 matrix material Substances 0.000 claims abstract description 25
- 238000009616 inductively coupled plasma Methods 0.000 claims abstract description 24
- 238000013519 translation Methods 0.000 claims abstract description 23
- 238000005070 sampling Methods 0.000 claims abstract description 17
- 238000007781 pre-processing Methods 0.000 claims abstract description 8
- 230000009466 transformation Effects 0.000 claims description 25
- 238000001914 filtration Methods 0.000 claims description 4
- 230000011218 segmentation Effects 0.000 claims description 4
- 230000008859 change Effects 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 7
- 230000006872 improvement Effects 0.000 abstract description 4
- 238000011156 evaluation Methods 0.000 abstract description 3
- 238000010586 diagram Methods 0.000 description 11
- 238000005516 engineering process Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000012847 principal component analysis method Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- 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
- G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/40—Image enhancement or restoration using histogram techniques
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a point cloud registration method for a transformer wiring terminal, which comprises the following specific steps: s1: collecting source point cloud data and target point cloud data, and preprocessing; s2: respectively selecting feature points of the source point cloud and the target point cloud, and finding n random sampling points in the source point cloud and corresponding points of the n random sampling points in the target point cloud according to a feature vector obtained by each feature point; s3: carrying out coarse registration on the initial matching point pairs by using a random sampling consistency algorithm, and calculating an initial rotation translation matrix of the source point cloud; s4: and finishing point cloud fine registration by adopting a symmetrical ICP (inductively coupled plasma) algorithm based on a target function. The invention is based on the target function symmetrical ICP algorithm, when the point pair is positioned on the second-order (constant curvature) surface of the curved surface, the symmetrical target is minimized, but not minimized when the point pair is positioned on the plane, the effective improvement of the registration precision is realized on the premise of ensuring the registration speed, the evaluation of the registration effect is optimized, and the problem of non-ideal registration effect is avoided.
Description
Technical Field
The invention belongs to the field of image three-dimensional point cloud registration, and particularly relates to a point cloud registration method for a transformer wiring terminal.
Background
The high-precision three-dimensional vision positioning technology of the target is used as a key technology of the wiring robot, and as the vision sensor can only scan and acquire data in a limited visual field range, a point cloud registration algorithm with excellent performance is required to generate a complete three-dimensional scene. Therefore, the accuracy, the speed and the robustness of the optimized point cloud registration algorithm become the key points of the research of the three-dimensional reconstruction technology.
At present, most point cloud registration algorithms are obtained by improving feature point extraction and iterative optimization algorithms on the basis of the traditional ICP algorithm, and the method comprises the following specific steps:
song Cheng navigation et al propose to improve ICP algorithm by using characteristic point sampling consistency, extract characteristic points through normal vector neighborhood included angle characteristics and establish fast point characteristic histogram for characteristic description, then use sampling consistency algorithm to register, improve registration accuracy, but when the point cloud number is great, registration efficiency is low.
Li Hui et al propose an improved ICP laser point cloud accurate registration method, wherein a principal component analysis method is used in coarse registration, a probability value is calculated and distributed for each point according to a two-way distance to improve an ICP algorithm, and although registration accuracy is guaranteed, registration efficiency is sacrificed.
Disclosure of Invention
In order to solve the defects in the prior art, the invention provides a point cloud registration method for a wiring terminal of a transformer, which is based on a target function symmetric ICP algorithm, wherein when a point pair is positioned on a second-order (constant curvature) surface of a curved surface, a symmetric target is minimized, rather than minimized only when the point pair is positioned on a plane, on the premise of ensuring the registration speed, the registration precision is effectively improved, meanwhile, the evaluation of the registration effect is optimized, and the problem of unsatisfactory registration effect is avoided.
The specific technical scheme of the invention is as follows:
a point cloud registration method for a transformer wiring terminal comprises the following specific steps:
s1: acquiring source point cloud data and target point cloud data by using a three-dimensional point cloud camera, and respectively preprocessing the source point cloud and the target point cloud;
s2: respectively inputting a source point cloud and a target point cloud, respectively selecting characteristic points of the source point cloud and the target point cloud according to the change of a local normal vector of each point, respectively calculating a characteristic descriptor FPFH (fast point feature histogram) of the source point cloud and the target point cloud by using an FPFH (fast point feature histogram) algorithm, and finding n random sampling points in the source point cloud and corresponding points in the target point cloud according to the characteristic vector obtained by each characteristic point, namely forming an initial matching point pair, wherein n is more than or equal to 3; s3: carrying out coarse registration on the initial matching point pairs by using a random sampling consistency algorithm, and calculating an initial rotation translation matrix of the source point cloud to enable the source point cloud to obtain an initial position; s4: and finishing point cloud fine registration by adopting a symmetrical ICP (inductively coupled plasma) algorithm based on a target function.
Preferably, in S1, the point cloud data is subjected to point cloud segmentation, filtering, and downsampling in sequence, so as to ensure that the point cloud quality meets the registration requirement, and improve the accuracy and speed of point cloud registration.
Preferably, in S2, the specific steps of comparing the feature vectors of the feature points of the source point cloud and the target point cloud to obtain the initial matching point pair are as follows:
s2-1: judging the similarity of the feature points in the source point cloud and the target point cloud by taking the Euclidean distance between the feature vectors of the matching point pairs as a comparison criterion;
s2-2: setting a threshold value epsilon, and calculating Euclidean distances among the characteristic vectors of the matching point pairs;
s2-3: and if the Euclidean distance between the characteristic vectors of the matching point pair is larger than the threshold epsilon, removing the characteristic vectors, and if the Euclidean distance between the characteristic vectors of the matching point pair is smaller than or equal to the threshold epsilon, reserving the characteristic vectors as the initial matching point pair.
Preferably, in S3, the specific steps of the coarse registration are as follows:
s3-1: estimating hypothesis transformation according to the corresponding relation of the n random sampling points found in the step S2;
s3-2: applying a hypothetical transformation to the source point cloud;
s3-3: searching an inner point corresponding point pair between the transformed source point cloud and the transformed target point cloud by adopting space nearest neighbor search according to a preset Euclidean distance threshold, and returning to S3-1 if the number of the inner point corresponding point pairs is less than that of preset point pairs;
s3-4: re-estimating the hypothesis transformation according to the found corresponding relation of the interior points;
s3-5: and calculating the distance between the corresponding point pairs of the inner points, and if the distance reaches a preset minimum distance threshold or a preset iteration number, taking the hypothesis transformation obtained by re-estimation in the step S3-4 as a final initial rotation and translation matrix.
Preferably, in S4, the specific steps of point cloud fine registration are as follows:
s4-1: source point of pairEach point P in the cloud P i Applying an initial rotation-translation matrix to obtain p' i Forming a new point set P';
s4-2: finding distance point p 'from target point cloud Q' i Nearest point q i Form corresponding point pairs (p' i ，qi)；
S4-3: according to the corresponding points obtained in the step S4-2, solving the optimal transformation based on the symmetrical objective function:
wherein, p' i Is the ith point p in the source point cloud i Transformed point, q i Is of point p' i The corresponding point in the target point cloud, then n p′，i Is a cloud midpoint p 'of a source point' i Surface normal of (2), n q,i Is the midpoint q of the target point cloud i Surface normal to (b), wherein (p ') for any corresponding point pair' i ，q i ) Vector p 'between them' i -q i Sum of normal lines perpendicular to the surface
R is a rotation matrix, t is a translation matrix,. Epsilon symm The distance values of a pair of point pairs in the source point cloud and the target point cloud after the initial transformation are obtained;
s4-4: the step S4-3 is executed by an iteration loop, and each iteration is carried out on the point P in the source point cloud P i And carrying out rotation and translation transformation by using the rotation matrix R and the translation matrix t obtained last time to obtain a new point set P' until epsilon symm If the distance value of (2) is less than the preset minimum distance value or reaches the preset maximum iteration number, outputting the final rotation and translation matrix.
Has the advantages that: the invention discloses a point cloud registration method for a transformer wiring terminal, which has the following advantages compared with the existing registration algorithm:
(1) The method adopts the target function symmetrical ICP algorithm, when the point pair is positioned on a second-order (constant curvature) surface of a curved surface, the symmetrical target is minimized, but not minimized when the point pair is positioned on a plane, the registration precision and the registration speed can be effectively optimized simultaneously, compared with the traditional registration method, the method can better integrate the point clouds at different visual angles into a specified coordinate system through rigid transformation such as rotation and translation, thereby realizing the acquisition of the actual three-dimensional coordinate information of the point clouds, and providing a good basic environment for the work of mechanical arm grabbing and the like at the later stage;
(2) The invention adopts the square sum of the nearest point distances of the point clouds after registration as the evaluation of the registration effect, thereby avoiding the problem of the registration effect.
(3) The method does not need to explicitly calculate the attribute of a second-order curved surface and a volume data structure to store the approximate value of the Euclidean distance function, and solves the problem of low registration efficiency when the number of point clouds in the existing point cloud registration method is large. Experiments prove that in the point cloud containing multiple noises and partial overlapping, the function improvement speed and the convergence range of the symmetrical target are effectively improved.
Drawings
FIG. 1 is a block diagram of a specific algorithm architecture flow proposed by the present invention;
FIG. 2 is a graph of a symmetry function on a 2D surface;
FIG. 3 is an original point cloud image (source point cloud) acquired at angle A;
FIG. 4 is an original point cloud (target point cloud) acquired at angle B;
FIG. 5 is a point cloud preprocessing diagram of an angle A, which is a source point cloud input in the registration link;
FIG. 6 is a point cloud preprocessing diagram at an angle B, which is a target point cloud input in the registration link;
fig. 7 is a rough registration result diagram of the point cloud registration algorithm proposed in this embodiment;
fig. 8 is a fine registration result diagram of the point cloud registration algorithm proposed in this embodiment;
fig. 9 is a comparison diagram (front view) between the point cloud registration diagram of the transformer terminal and the real point cloud diagram of the point cloud registration algorithm provided in this embodiment;
fig. 10 is a comparison graph (left view) between the point cloud registration graph of the transformer terminal of the point cloud registration algorithm proposed in this embodiment and the real point cloud graph;
fig. 11 is a comparison graph (back graph) between a point cloud registration graph and a real point cloud graph of a transformer terminal of the point cloud registration algorithm proposed in the present embodiment;
fig. 12 is a comparison graph (right view) between the point cloud registration graph of the transformer terminal of the point cloud registration algorithm proposed in this embodiment and the real point cloud graph.
Detailed Description
The invention is described in the following with reference to the accompanying drawings, which are intended to cover several modifications and embodiments of the invention.
Example 1
A point cloud registration method for transformer wiring terminals, as shown in fig. 1, the specific registration method is as follows:
s1: acquiring source point cloud data and target point cloud data by using a three-dimensional point cloud camera, as shown in fig. 3 and 4, respectively preprocessing the source point cloud and the target point cloud, and sequentially performing preprocessing such as point cloud segmentation, filtering, down-sampling and the like on the point cloud data, as shown in fig. 5 and 6, so that the point cloud quality is ensured to meet the registration requirement, and the precision and the speed of point cloud registration are improved;
s2: respectively inputting source point cloud and target point cloud, respectively selecting the feature points of the source point cloud and the target point cloud according to the change of the local normal vector of each point, and using a FPFH fast point feature histogram algorithm to respectively calculate the feature descriptor FPFH of the source point cloud and the target point cloud, according to the feature vector obtained by each feature point, finding n random sampling points in the source point cloud and the corresponding points in the target point cloud, namely forming an initial matching point pair, wherein n is more than or equal to 3, wherein, the specific steps of comparing the feature vectors of the feature points of the source point cloud and the target point cloud to obtain the initial matching point pair are as follows:
s2-1: judging the similarity of the feature points in the source point cloud and the target point cloud by taking the Euclidean distance between the feature vectors of the matching point pairs as a comparison criterion;
s2-2: setting a threshold value epsilon, and calculating Euclidean distances among the characteristic vectors of the matching point pairs;
s2-3: and if the Euclidean distance between the characteristic vectors of the matching point pair is larger than the threshold epsilon, removing the characteristic vectors, and if the Euclidean distance between the characteristic vectors of the matching point pair is smaller than or equal to the threshold epsilon, reserving the characteristic vectors as the initial matching point pair.
S3: carrying out coarse registration on the initial matching point pairs by using a random sampling consistency algorithm, wherein the coarse registration result is shown in FIG. 7, and calculating an initial rotation translation matrix of the source point cloud to enable the source point cloud to obtain an initial position; the rough registration comprises the following specific steps:
s3-1: estimating hypothesis transformation according to the corresponding relation of the n random sampling points found in the step S2;
s3-2: applying a hypothetical transformation to the source point cloud;
s3-3: searching an inner point corresponding point pair between the source point cloud and the target point cloud after transformation by adopting space nearest neighbor search according to a preset Euclidean distance threshold, and returning to S3-1 if the number of the inner point corresponding point pairs is less than that of preset point pairs;
s3-4: re-estimating the hypothesis transformation according to the found corresponding relation of the interior points;
s3-5: and calculating the distance between the corresponding point pairs of the inner points, and if the distance reaches a preset minimum distance threshold or a preset iteration number, taking the hypothesis transformation obtained by re-estimation in the step S3-4 as a final initial rotation and translation matrix.
S4: and finishing point cloud precise registration by adopting a symmetrical ICP (inductively coupled plasma) algorithm based on a target function, wherein the precise registration result is shown in FIG. 8, and the specific steps of the point cloud precise registration are as follows:
s4-1 pairs of each point P in source point cloud P i Applying an initial rotation-translation matrix to obtain p' i Forming a new point set P';
s4-2, finding a distance point p 'from the target point cloud Q' i Nearest point q i Form corresponding point pairs (p' i ，q i )；
S4-3: according to the corresponding points obtained in the step S4-2, solving the optimal transformation based on the symmetrical objective function:
wherein, p' i Is the ith point p in the source point cloud i Transformed point, q i Is dot p' i The corresponding point in the target point cloud is then n p′，i Is the cloud midpoint p 'of the source point' i Surface normal of (2), n q,i Is the midpoint q of the target point cloud i Wherein for any corresponding point pair (p' i ，q i ) Vector p 'between them' i -q i Perpendicular to the sum of surface normals
R is a rotation matrix, t is a translation matrix,. Epsilon symm The distance values of a pair of point pairs in the source point cloud and the target point cloud after the initial transformation are obtained;
s4-4: the iteration loop executes the step S4-3, and each iteration is carried out on the point P in the source point cloud P i Using the rotation matrix R and the translation matrix t obtained last time to carry out rotation and translation transformation to obtain a new point set P' until epsilon symm If the distance value of (2) is less than the preset minimum distance value or reaches the preset maximum iteration number, outputting the final rotation and translation matrix.
As shown in fig. 9 to 12, they are comparison diagrams (front view, left view, back view, and right view) of the point cloud registration diagram of the transformer terminal of the point cloud registration algorithm proposed in this embodiment and the real point cloud diagram.
The specific principle of the target function symmetric ICP algorithm adopted in the invention is as follows:
since point-to-plane ICP is only 0 when the local surface is planar, the distance residual at the best alignment. The invention therefore proposes a symmetrical version for the point-to-plane ICP. First, the plane in which the error is minimized is based on the surface normals of the two points in the corresponding point pair. Secondly, the optimization of the invention is carried out in a stable coordinate system, and the two grids move towards opposite directions at the same time, so that the calculated amount is hardly increased in each iteration, and the convergence of the ICP algorithm is improved. The symmetric objective function is minimized whenever a pair of point pairs lie on a second order surface, rather than only when the points lie on a plane. Therefore, the same effect as the second order distance function minimizing method can be obtained without explicitly calculating the second order surface property and the euclidean distance. Meanwhile, a rotational linearization alternative method is introduced into the algorithm of the invention to simplify optimization into a linear least square problem, and the precise transformation can still be solved when the corresponding relation is precise, so that the error reduction of each iteration is larger, and the convergence domain is increased.
As shown in fig. 2, since two surfaces should be identical under noise, if a pair of nearby points (p, q) is sampled on the surface, the error of the objective function in the point-to-plane case is:
(p-q)·n q (2)；
in the formula, n q Is the normal of the point q in the target point cloud, and p and q are respectively the corresponding points of the source point cloud and the target point cloud.
If a pair of nearby points (p, q) is sampled anywhere within a small area of a surface, then this value is zero only if this surface is a perfect plane, and therefore a more symmetric objective function is considered, as follows:
(p-q)·(n p +n q ) (3)；
for any point (p, q) sampled from the arc, the sum n of the vectors p-q between them, which is perpendicular to the normal p +n q . It can be seen that when (p, q) is sampled from one circle, the symmetric function is 0. When a rigid body transformation is applied to the point set P of the source point cloud, the value of the individual expression remains 0 as long as the relative positions of P and q are consistent.
Similarly, the same holds true in 3D, and the calculation result of equation (3) is 0 as long as p and q and their normals coincide with a certain cylinder. (p, n) on arbitrary cylinders p ) The set having a fixed correspondence (q, n) q ) Equation (2) holds as long as p and q match the local quadric surface therebetween.
Based on the principle, the invention provides an objective function symmetric ICP algorithm.
Verification experiment
In this embodiment, a three-dimensional point cloud camera is used to collect the source point cloud data and the target point cloud data of the transformer wiring terminal, as shown in fig. 3 and 4. And then respectively carrying out point cloud segmentation, down sampling, filtering and other preprocessing on the source point cloud and the target point cloud in sequence, and respectively reducing the quantity of the source point cloud and the target point cloud to 20145 and 19366, wherein the preprocessed point cloud pictures are shown in figures 5 and 6.
The processed point cloud data are registered by a plurality of algorithms (including the algorithm of the invention) respectively, so that the root mean square error of the registration result, the time and the square sum of the nearest point distance of the point cloud after registration are compared, and the comparison result is shown in table 1.
TABLE 1 comparison table of accuracy and speed of different algorithms
The result shows that the algorithm registration accuracy and the registration rate are effectively optimized, wherein compared with an ICP algorithm (SIFT + FPFH + SAC-IA + ICP) based on scale-invariant feature transform, the improved algorithm registration accuracy is respectively improved by 21.0%, and the speed is improved by 10.6%; compared with the characteristic point sampling consistency improved ICP algorithm (SAC-IA + ICP), the improved algorithm has the advantages that the registration accuracy is improved by 43.3%, the speed is improved by 30.2%, and therefore the effectiveness of the method is verified.
The above description is merely illustrative of the present invention and is a preferred embodiment of the present invention. It should be noted that, without departing from the scope of the present invention, a person skilled in the art may make several modifications and improvements, and such modifications and improvements should also be considered as the protection scope of the present invention.
Claims (5)
1. A point cloud registration method for a transformer wiring terminal is characterized by comprising the following steps:
s1: acquiring source point cloud data and target point cloud data by using a three-dimensional point cloud camera, and respectively preprocessing the source point cloud and the target point cloud;
s2: respectively inputting a source point cloud and a target point cloud, respectively selecting characteristic points of the source point cloud and the target point cloud according to the change of a local normal vector of each point, respectively calculating a characteristic descriptor FPFH (fast point histogram) of the source point cloud and the target point cloud by using an FPFH (fast point histogram) algorithm, and finding n random sampling points in the source point cloud and corresponding points in the target point cloud according to the characteristic vector obtained by each characteristic point, namely forming an initial matching point pair, wherein n is more than or equal to 3;
s3: carrying out coarse registration on the initial matching point pairs by using a random sampling consistency algorithm, and calculating an initial rotation translation matrix of the source point cloud to enable the source point cloud to obtain an initial position;
s4: and finishing point cloud fine registration by adopting a symmetrical ICP (inductively coupled plasma) algorithm based on a target function.
2. The point cloud registration method of the transformer wiring terminal according to claim 1, wherein in the step S1, the point cloud data is subjected to point cloud segmentation, filtering and downsampling pretreatment in sequence, so that the point cloud quality is ensured to meet the registration requirement, and the precision and speed of point cloud registration are improved.
3. The point cloud registration method for the transformer wiring terminals according to claim 1, wherein in S2, the specific steps of comparing the feature vectors of the feature points of the source point cloud and the target point cloud to obtain the initial matching point pair are as follows:
s2-1: judging the similarity of the feature points in the source point cloud and the target point cloud by taking the Euclidean distance between the feature vectors of the matching point pairs as a comparison criterion;
s2-2: setting a threshold value epsilon, and calculating Euclidean distances among the characteristic vectors of the matching point pairs;
s2-3: and if the Euclidean distance between the characteristic vectors of the matched point pair is greater than the threshold epsilon, removing the characteristic vectors, and if the Euclidean distance between the characteristic vectors of the matched point pair is less than or equal to the threshold epsilon, reserving the characteristic vectors as the initial matched point pair.
4. The point cloud registration method for the transformer wiring terminals according to claim 1, wherein in the step S3, the rough registration specifically comprises the following steps:
s3-1: estimating hypothesis transformation according to the corresponding relation of the n random sampling points found in the step S2;
s3-2: applying a hypothetical transformation to the source point cloud;
s3-3: searching an inner point corresponding point pair between the source point cloud and the target point cloud after transformation by adopting space nearest neighbor search according to a preset Euclidean distance threshold, and returning to S3-1 if the number of the inner point corresponding point pairs is less than that of preset point pairs;
s3-4: re-estimating the hypothesis transformation according to the searched corresponding relation of the interior points;
s3-5: and (4) calculating the distance between the corresponding point pairs of the inner points, and if the distance reaches a preset minimum distance threshold or reaches a preset iteration number, taking the hypothesis transform obtained by re-estimation in the step (S3-4) as a final initial rotation and translation matrix.
5. The point cloud registration method for the transformer wiring terminals according to claim 4, wherein in the step S4, the point cloud fine registration comprises the following specific steps:
s4-1: for each point P in the source point cloud P i Applying an initial rotation-translation matrix to obtain p' i Forming a new point set P';
s4-2: finding a distance point p 'from the target point cloud Q' i Nearest point q i Form corresponding point pairs (p' i ，q i )；
S4-3: solving the optimal transformation based on the symmetrical objective function according to the corresponding points obtained in the step S4-2:
wherein, p' i Is the ith point p in the source point cloud i Transformed point, q i Is of point p' i The corresponding point in the target point cloud, then n p′，i Is a cloud midpoint p 'of a source point' i Surface normal of (2), n q,i Is the midpoint q of the target point cloud i Wherein for any corresponding point pair (p' i ，q i ) Vector p 'between them' i -q i Sum n perpendicular to surface normal p′i +n q,i (ii) a R is a rotation matrix, t is a translation matrix,. Epsilon symm The distance value of a pair of point pairs in the source point cloud and the target point cloud after the initial transformation is obtained;
s4-4: the iteration loop executes the step S4-3, and each iteration is carried out on the point P in the source point cloud P i And carrying out rotation and translation transformation by using the rotation matrix R and the translation matrix t obtained last time to obtain a new point set P' until epsilon symm If the distance value of (2) is less than the preset minimum distance value or reaches the preset maximum iteration number, outputting the final rotation and translation matrix.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211191262.1A CN115564811A (en) | 2022-09-28 | 2022-09-28 | Point cloud registration method for transformer wiring terminal |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211191262.1A CN115564811A (en) | 2022-09-28 | 2022-09-28 | Point cloud registration method for transformer wiring terminal |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115564811A true CN115564811A (en) | 2023-01-03 |
Family
ID=84743666
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211191262.1A Pending CN115564811A (en) | 2022-09-28 | 2022-09-28 | Point cloud registration method for transformer wiring terminal |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115564811A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117495932A (en) * | 2023-12-25 | 2024-02-02 | 国网山东省电力公司滨州供电公司 | Power equipment heterologous point cloud registration method and system |
| CN117572454A (en) * | 2023-11-15 | 2024-02-20 | 武汉万曦智能科技有限公司 | Method and system for measuring safety clearance of field vehicle storage battery |
-
2022
- 2022-09-28 CN CN202211191262.1A patent/CN115564811A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117572454A (en) * | 2023-11-15 | 2024-02-20 | 武汉万曦智能科技有限公司 | Method and system for measuring safety clearance of field vehicle storage battery |
| CN117572454B (en) * | 2023-11-15 | 2024-05-10 | 武汉万曦智能科技有限公司 | Method and system for measuring safety clearance of field vehicle storage battery |
| CN117495932A (en) * | 2023-12-25 | 2024-02-02 | 国网山东省电力公司滨州供电公司 | Power equipment heterologous point cloud registration method and system |
| CN117495932B (en) * | 2023-12-25 | 2024-04-16 | 国网山东省电力公司滨州供电公司 | Power equipment heterologous point cloud registration method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115564811A (en) | Point cloud registration method for transformer wiring terminal | |
| CN110889243B (en) | Aircraft fuel tank three-dimensional reconstruction method and detection method based on depth camera | |
| JP6216508B2 (en) | Method for recognition and pose determination of 3D objects in 3D scenes | |
| Uchida et al. | 3D face recognition using passive stereo vision | |
| CA2326816C (en) | Face recognition from video images | |
| Boyer et al. | 3D surface reconstruction using occluding contours | |
| CN114972459B (en) | Point cloud registration method based on low-dimensional point cloud local feature descriptor | |
| CN111145232A (en) | Three-dimensional point cloud automatic registration method based on characteristic information change degree | |
| CN109903313A (en) | Real-time pose tracking method based on target three-dimensional model | |
| CN112017220A (en) | Point cloud accurate registration method based on robust constraint least square algorithm | |
| CN114677418A (en) | Registration method based on point cloud feature point extraction | |
| CN115861397A (en) | Point cloud registration method based on improved FPFH-ICP | |
| CN113706381A (en) | Three-dimensional point cloud data splicing method and device | |
| Hong et al. | An improved model-based matching method using footprints | |
| CN113313200B (en) | Point cloud precision matching method based on normal constraint | |
| CN109584347B (en) | Augmented reality virtual and real occlusion processing method based on active appearance model | |
| Cui et al. | Dense registration and mosaicking of fingerprints by training an end-to-end network | |
| Ma et al. | A subpixel matching method for stereovision of narrow baseline remotely sensed imagery | |
| Wang et al. | 6D pose estimation from point cloud using an improved point pair features method | |
| Wang et al. | An improved ICCP gravity matching algorithm based on Mahalanobis distance | |
| CN117541614B (en) | Space non-cooperative target close-range relative pose tracking method based on improved ICP algorithm | |
| Caglioti et al. | Methods for space line localization from single catadioptric images: new proposals and comparisons | |
| Liu et al. | Pose estimation for 3D workpiece grasping in industrial environment based on evolutionary algorithm | |
| CN115393432B (en) | Robot positioning method, storage medium and equipment for eliminating static characteristic mismatching | |
| Hu et al. | Automatic interpretation and precision detecting of target's axis in image sequences |
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 |