CN112102178A - Point cloud feature-preserving denoising method and device, electronic equipment and storage medium - Google Patents
Point cloud feature-preserving denoising method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN112102178A CN112102178A CN202010748731.XA CN202010748731A CN112102178A CN 112102178 A CN112102178 A CN 112102178A CN 202010748731 A CN202010748731 A CN 202010748731A CN 112102178 A CN112102178 A CN 112102178A
- Authority
- CN
- China
- Prior art keywords
- point
- neighborhood
- vector
- point set
- main data
- 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 44
- 239000013598 vector Substances 0.000 claims abstract description 139
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 22
- 230000011218 segmentation Effects 0.000 claims description 25
- 238000004364 calculation method Methods 0.000 claims description 14
- 238000004590 computer program Methods 0.000 claims description 8
- 238000010276 construction Methods 0.000 claims description 5
- 238000005259 measurement Methods 0.000 abstract description 4
- 238000010586 diagram Methods 0.000 description 12
- 230000008569 process Effects 0.000 description 7
- 238000000513 principal component analysis Methods 0.000 description 5
- 238000012545 processing Methods 0.000 description 4
- 238000012935 Averaging Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000009466 transformation Effects 0.000 description 3
- 230000002146 bilateral effect Effects 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000011426 transformation method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- 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)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to the technical field of three-dimensional measurement, and provides a point cloud feature preserving and denoising method, a point cloud feature preserving and denoising device, electronic equipment and a storage medium, wherein the method comprises the following steps: constructing a preset tree model according to the point cloud data, and searching a neighborhood point set corresponding to a main data point in the point cloud data based on the preset tree model; calculating a point normal vector of the main data point and a centroid point of the neighborhood point set according to the neighborhood point set corresponding to the main data point through a preset algorithm; calculating a vector according to the centroid point and the neighborhood point set, and a projection distance of the vector on the normal vector of the point; if the projection distance meets the threshold distance, recording the point to be denoised corresponding to the projection distance as a noise point, and updating the neighborhood point set based on the noise point until the neighborhood point set of the main data point is completely denoised after updating. The method can be used for denoising small cluster noises and strip-shaped small-section micro-bulge noises around the point cloud main body, and the denoising efficiency is improved.
Description
Technical Field
The invention relates to the technical field of three-dimensional measurement, in particular to a point cloud feature-preserving denoising method and device, electronic equipment and a storage medium.
Background
At present, in the field of three-dimensional measurement, a three-dimensional point cloud denoising technology has important application in the aspect of reverse engineering. For three-dimensional scanning, due to influences of scanning environment, operation problems of scanning personnel, scanner precision and the like, the acquired three-dimensional point cloud data often has a plurality of noise points and outliers, and the quality of point cloud feature extraction, point cloud splicing and three-dimensional surface reconstruction can be directly influenced. In the existing point cloud denoising technology, noise deviating from a point cloud main body by a certain distance and a long distance can be processed. The denoising effect is poor for small cluster noises around a point cloud main body, and a bilateral filtering algorithm is used, so that point coordinates can be changed, and a large error can be introduced into point cloud data while denoising is performed. In addition, a fringe light projection technology can be used for measurement, the obtained data are in a strip shape, and if small-section trace convex noise appears in the point cloud data, the bilateral filtering algorithm can mistakenly consider the point cloud data to be the characteristic. Therefore, in the prior art, the problem of low denoising efficiency exists for small cluster noises around a point cloud main body.
Disclosure of Invention
The embodiment of the invention provides a point cloud feature preserving denoising method which can denoise small cluster noises and strip-shaped small-segment micro-protrusion noises around a point cloud main body, improve denoising efficiency and provide more reliable point cloud data.
In a first aspect, an embodiment of the present invention provides a point cloud feature preserving and denoising method, including the following steps:
constructing a preset tree model according to the point cloud data, and searching a neighborhood point set corresponding to a main data point in the point cloud data based on the preset tree model;
calculating a point normal vector of the main data point and a centroid point of the neighborhood point set according to the neighborhood point set corresponding to the main data point through a preset algorithm;
calculating a vector according to the centroid point and the neighborhood point set and the projection distance of the vector on the normal vector of the point, wherein the vector is formed by the point to be denoised in the neighborhood point set and the centroid point;
if the projection distance meets the threshold distance, recording the point to be denoised corresponding to the projection distance as a noise point, and updating the neighborhood point set based on the noise point until the neighborhood point set of the main data point is completely denoised after updating.
In a second aspect, an embodiment of the present invention further provides a device for denoising a point cloud with preserved features, including:
the searching module is used for constructing a preset tree model according to the point cloud data and searching a neighborhood point set corresponding to the main data point in the point cloud data based on the preset tree model;
the first calculation module is used for calculating a point normal vector of the main data point and a centroid point of the neighborhood point set according to the neighborhood point set corresponding to the main data point through a preset algorithm;
the second calculation module is used for calculating a vector according to the centroid point and the neighborhood point set and the projection distance of the vector on the normal vector of the point, wherein the vector is formed by the point to be denoised in the neighborhood point set and the centroid point;
and the updating module is used for recording the point to be denoised corresponding to the projection distance as a noise point if the projection distance meets the threshold distance, and updating the neighborhood point set based on the noise point until the updated neighborhood point set of the main data point is completely denoised.
In the embodiment of the invention, a preset tree model is constructed according to point cloud data, and a neighborhood point set corresponding to a main data point in the point cloud data is searched based on the preset tree model; calculating a point normal vector of the main data point and a centroid point of the neighborhood point set according to the neighborhood point set corresponding to the main data point through a preset algorithm; calculating a vector according to the centroid point and the neighborhood point set and the projection distance of the vector on the normal vector of the point, wherein the vector is formed by the point to be denoised in the neighborhood point set and the centroid point; if the projection distance meets the threshold distance, recording the point to be denoised corresponding to the projection distance as a noise point, and updating the neighborhood point set based on the noise point until the neighborhood point set of the main data point is completely denoised after updating. According to the embodiment of the invention, by selecting the neighborhood point set of the main data points, the geometric information of the main data can be more accurately expressed by calculating the normal vector of the points through a preset algorithm; under the condition of not changing coordinates (point cloud characteristics) of point cloud data, a mass center point is calculated according to a neighborhood point set, a vector between the mass center point and the neighborhood point set is calculated by taking the mass center point as a reference point, the projection distance of the vector on a point normal vector is taken as a standard for judging noise, and noise points with the projection distance reaching a threshold distance are denoised so as to realize denoising of small-cluster noise and denoising of strip-shaped small-section micro-bulge noise, so that the denoising efficiency is improved, and reliable point cloud data are provided.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a flowchart illustrating a method for denoising a point cloud feature according to an embodiment of the present invention;
FIG. 2 is a flowchart illustrating another embodiment of a point cloud feature-preserving denoising method according to the present invention;
FIG. 3 is a flowchart illustrating another embodiment of a point cloud feature-preserving denoising method according to the present invention;
fig. 4 is a schematic structural diagram of a point cloud feature preserving denoising device according to an embodiment of the present invention;
FIG. 5 is a schematic structural diagram of another point cloud feature preserving denoising apparatus according to an embodiment of the present invention;
FIG. 6 is a schematic structural diagram of another point cloud feature preserving denoising apparatus according to an embodiment of the present invention;
FIG. 7 is a schematic structural diagram of another point cloud feature preserving denoising apparatus according to an embodiment of the present invention;
FIG. 8 is a schematic structural diagram of another point cloud feature preserving denoising apparatus according to an embodiment of the present invention;
fig. 9 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
As shown in fig. 1, fig. 1 is a flowchart of a point cloud feature preserving denoising method provided by an embodiment of the present invention, which specifically includes the following steps:
101. and constructing a preset tree model according to the point cloud data, and searching a neighborhood point set corresponding to the main data point in the point cloud data based on the preset tree model.
In the embodiment of the invention, the scenes in which the point cloud feature-preserving denoising method is applied include, but are not limited to, a three-dimensional scanning system and the like. The electronic equipment on which the point cloud feature-preserving denoising method operates can acquire point cloud data in a wired connection mode or a wireless connection mode. The Wireless connection mode may include, but is not limited to, a 3G/4G connection, a WiFi (Wireless-Fidelity) connection, a bluetooth connection, a wimax (worldwide Interoperability for Microwave access) connection, a Zigbee (low power local area network protocol) connection, a uwb (ultra wideband) connection, and other Wireless connection modes known now or developed in the future.
The point cloud data may be point cloud data corresponding to a location where a subject (including a vehicle, a building, and the like) described in a 3D point cloud space is located. In an embodiment of the present invention, the preset tree model provided may be a kd (K-dimension tree) tree, where the kd tree is a data structure that divides subject data points in a K-dimensional space (e.g., two-dimensional (x, y), three-dimensional (x, y, z), and K-dimensional (x1, y, z … K)), and is mainly applied to searching of key data in a multidimensional space, for example: range search and nearest neighbor search. Essentially, the kd-tree may be a balanced binary tree. After the point cloud data of the main body is obtained, the kd tree construction can be carried out according to the construction mode of the kd tree, the adjacent points corresponding to all main body data points in the point cloud data can be obtained according to the function that the kd tree is applied to nearest neighbor search, and the adjacent points of all main body data points can be collected to obtainA neighborhood point set corresponding to each subject data point one to one, for example: the neighborhood point set corresponding to the main body data point A is NP1＝{p1，…，pmAnd m represents the number of points in the neighborhood point set.
102. And calculating the normal vector of the point of the main data point and the centroid point of the neighborhood point set according to the neighborhood point set corresponding to the main data point by a preset algorithm.
The preset algorithm may include a Principal Component Analysis (PCA), which is a mathematical transformation method that converts a given set of related variables into another set of uncorrelated variables through linear transformation, and the new variables are arranged in an order of decreasing variance. The total variance of the variables is kept constant in the mathematical transformation, such that the first variable has the largest variance, called the first principal component, and the second variable has the second largest variance and is uncorrelated with the first variable, called the second principal component. By analogy, m variables have m principal components. In the embodiment of the present invention, the variable is m main data points in the neighborhood point set.
The point normal vector may be a target variable output after linear transformation by using the neighborhood point set as an initial variable through the principal component analysis algorithm, and the target variable may be a point normal vector of the neighborhood point set corresponding to the main data point and may be represented by N. The point normal vector has both a direction and a magnitude. The centroid point can be used as a reference point for judging denoising, the centroid point can be a point with the highest stability obtained by calculation according to a neighborhood point set, and the preset algorithm can further comprise an averaging algorithm for calculating the centroid point.
103. And calculating a vector according to the centroid point and the neighborhood point set, and calculating the projection distance of the vector on the normal vector of the point, wherein the vector is composed of the point to be denoised and the centroid point in the neighborhood point set.
Vectors formed between each point in the neighborhood point set and the centroid point correspond to each other in sequence, and if m points exist in the neighborhood point set, m vectors correspond to each other. In the embodiment of the invention, the vector can be piOPWhere i denotes the ith point in the neighborhood set of points, (i ═ i1, …, m). The point to be denoised can represent each point in the neighborhood point set, and the point to be denoised can be judged as a noise point or can be reserved as a non-noise point. When the point to be denoised is judged as the noise point, the three-dimensional scanning system can generate a deletion instruction to delete the noise point, so that the noise reduction processing of each main data point in the point cloud data is realized, and the quality of the point cloud data is ensured.
After the corresponding vector is calculated according to the centroid point and each point in the neighborhood point set, the calculation can be performed according to the vector of each main data point and the point normal vector to obtain the projection distance of the vector in the point normal vector direction. Wherein there are m vectors piOPCorresponding to m projection distances, L for projection distanceiRepresents (i ═ 1, …, m).
104. And if the projection distance meets the threshold distance, recording the point to be denoised corresponding to the projection distance as a noise point, and updating the neighborhood point set based on the noise point until the updated neighborhood point set of the main data point is completely denoised.
The threshold distance can be obtained according to the centroid point of the neighborhood point set and the pre-selected parameters. And comparing the projection distance with the threshold distance, and when the projection distance is greater than the threshold distance, indicating that the projection distance meets the threshold distance, and marking the point to be denoised corresponding to the projection distance as a noise point. Obtaining m projection distances, there are m comparisons correspondingly, and m results are obtained. At least one noisy point and at least one non-noisy point are included in the result. And forming a noise point set after all the noise points are obtained, marking a set ID on the noise point set, searching all the noise points in the set ID in the neighborhood point set, and deleting all the noise points to update the neighborhood point set.
The above-mentioned complete denoising can indicate that the denoising effect reaches 100%. In addition, the denoising degree can be preset to reach a denoising threshold, for example: the denoising degree reaches 95%, and 95% is used as a denoising threshold. Specifically, the embodiments of the present invention are not limited. It should be noted that, in the embodiment of the present invention, the main data point, the centroid point, the noise point to be removed, and the like in the described neighborhood may be two-dimensional coordinate data, three-dimensional coordinate data, …, k-dimensional coordinate data. In the embodiments of the present invention, no specific limitation is imposed.
In the embodiment of the invention, a preset tree model is constructed according to point cloud data, and a neighborhood point set corresponding to a main data point in the point cloud data is searched based on the preset tree model; calculating a point normal vector of the main data point and a centroid point of the neighborhood point set according to the neighborhood point set corresponding to the main data point through a preset algorithm; calculating a vector according to the centroid point and the neighborhood point set, and calculating the projection distance of the vector on a normal vector of the point, wherein the vector is composed of the point to be denoised and the centroid point in the neighborhood point set; and if the projection distance meets the threshold distance, recording the point to be denoised corresponding to the projection distance as a noise point, and updating the neighborhood point set based on the noise point until the updated neighborhood point set of the main data point is completely denoised. In the embodiment of the invention, by selecting the neighborhood point set, the geometric information of the main body can be more accurately expressed by calculating the normal vector of the point through a principal component analysis algorithm; under the condition of not changing coordinates (point cloud characteristics) of point cloud data, a mass center point is calculated according to a neighborhood point set, a vector between the mass center point and the neighborhood point set is calculated by taking the mass center point as a reference point, the projection distance of the vector on a point normal vector is taken as a standard for judging noise, and noise points with the projection distance reaching a threshold distance are denoised so as to realize denoising of small-cluster noise and denoising of strip-shaped small-section micro-bulge noise, so that the denoising efficiency is improved, and reliable point cloud data are provided.
As shown in fig. 2, fig. 2 is a flowchart of another point cloud feature preserving and denoising method provided in the embodiment of the present invention, which specifically includes the following steps:
201. and acquiring point cloud data, and segmenting a point cloud space corresponding to the point cloud data into multi-dimensional areas.
Point cloud data is acquired from a point cloud space, wherein the point cloud data can be point cloud data of a certain subject and comprises a plurality of subject data points, and the subject can comprise a vehicle, a house, a human body and the like. The above-mentioned division of the point cloud space into multidimensional areas may be divided into specific parts, and then related search operations are performed within the parts of the specific space. The division rule can be customized, for example: a three-dimensional space is divided into 8-dimensional spaces.
202. And selecting a segmentation point in the multi-dimensional region according to the main body data point, and respectively inserting the main body data point into the left subspace and the right subspace of the segmentation point based on the segmentation point so as to construct a preset tree model.
The segmentation points can represent root nodes, the selection of the root nodes can judge that the main data points of the point cloud data belong to the multi-dimensional data points, then the data variance of the main data points on each dimension can be counted, the number of the data variances can be obtained according to the number of the dimensions, and the dimension corresponding to the maximum value is selected from the data variances to be used as the domain value. Then, the data in the domain value may be sorted to select a median, and the data point corresponding to the median is used as the above-mentioned segmentation point (root node). Thus, the segmentation hyperplane of the segmentation point would be through the segmentation point and perpendicular to the median, for example: the data variances of the 6 two-dimensional data points { (2,3), (5,4), (9,6), (4,7), (8,1), (7,2) }, 6 data points in the x and y dimensions are 39, 28.63, respectively, so the difference is larger above the x-axis, so the threshold value is x; if the median value in the x dimension is selected to be 7, the segmentation point location (7,2) is determined.
After the segmentation point is obtained by calculation, the entire space is divided into two parts, one of which may be a part smaller than or equal to the median value, i.e., the left subspace, and the other may be a part larger than the median value, i.e., the right subspace. The data of the dimension of the median can be compared in size, and all the subject data points can be inserted into the left subspace and the right subspace respectively. The ranking of the data points in the left subspace and the right subspace can be executed according to the process of selecting the root node, and the kd tree construction is a recursive process and is executed repeatedly until the point cloud space only contains one data point, so that the kd tree construction is completed. Of course, the method for constructing the kd-tree provided above may also include other methods that can construct the kd-tree, and is not limited in the embodiment of the present invention.
203. Presetting an initial neighborhood radius, searching for neighboring points corresponding to the subject data point in the kd tree for multiple times in the initial neighborhood radius based on the subject data point, and collecting the searched neighboring points to form a neighborhood point set.
Before the initial neighborhood radius is preset, binary tree search may be performed first, the subject data point to be searched currently is searched layer by layer from the root node, the current nearest point is found from the search path, then the distance between the current nearest point and the data point to be searched may be calculated, and the distance may be calculated according to a distance formula between the two points, where the point may be a k-dimensional coordinate point.
And performing backtracking search after completing binary tree search, backtracking from the found current nearest neighbor near point to the father node of the father node, and judging whether data points smaller than the distance exist in other child node spaces of the father node. Specifically, an initial neighborhood radius may be preset, where the initial neighborhood radius may be a distance value corresponding to a current nearest neighbor; and drawing a circle by taking the main data point to be searched as the center of the circle to see whether the main data point intersects with the boundary line of the hyperplane divided into the multi-dimensional area. If there is no intersection, the area of the dimension corresponding to the boundary of the hyperplane is not entered, i.e., data points in the dimension at smaller distances are not considered. And then backtracking to the father node of the father node, and continuing searching in the same way until the root node. Therefore, the neighborhood points can be found, and then all the searched neighborhood points are gathered to obtain the neighborhood point set.
In addition, when the currently calculated distance is smaller than the previously calculated distance in the process of searching for multiple times, the current nearest point of the current nearest point is to be updated, and the initial neighborhood radius of the subsequent search also needs to be updated to the corresponding minimum distance. It should be noted that the neighboring points may also be found by other ways, for example: BBF algorithm, etc.
204. And calculating the normal vector of the point of the main data point and the centroid point of the neighborhood point set according to the neighborhood point set corresponding to the main data point by a preset algorithm.
205. And calculating a vector according to the centroid point and the neighborhood point set, and calculating the projection distance of the vector on the normal vector of the point, wherein the vector is composed of the point to be denoised and the centroid point in the neighborhood point set.
206. And if the projection distance meets the threshold distance, recording the point to be denoised corresponding to the projection distance as a noise point, and updating the neighborhood point set based on the noise point until the updated neighborhood point set of the main data point is completely denoised.
Optionally, step 204 includes:
and calculating the mean value of the neighborhood point set according to the neighborhood point set corresponding to the main data point, and determining the mean value as the centroid point corresponding to the main data point, wherein the mean value corresponds to the main data point.
Each main body data point has a corresponding neighborhood point set, a plurality of to-be-denoised points can exist in the neighborhood point set, the mean value of all to-be-denoised points can be obtained, the obtained mean value is used as the centroid point of the main body data point, and the specific calculation formula is as follows:
wherein Op is a centroid point, n is the number of points to be denoised in the neighborhood point set, and Pi represents the ith point to be denoised. The method for calculating the projection distance by taking the centroid point and taking the centroid point as the datum point can not change the coordinates of the main body data point in the point cloud data so as to remove small cluster noise and strip-shaped small-section micro-protrusion noise which are far away from the periphery of the point cloud main body.
Optionally, step 205 includes:
and calculating a vector according to the point to be denoised in the neighborhood point set and the centroid point.
Wherein, can do the difference with the centroid point respectively with every point of waiting to denoise in the neighborhood point set, the neighborhood point is concentrated to have m to wait to denoise the point, then can calculate m and wait to denoise the vector PiPo between the point and the centroid point, and specific computational formula is as follows:
piOp＝pi-Op(i＝1，...，m) (2)
wherein Pi is the ith point to be denoised, PiPo is the ith vector, and Op is the centroid point.
And calculating the projection distance of the vector in the direction of the point normal vector according to the vector and the point normal vector until the projection distance of the vector corresponding to all the points to be denoised in the neighborhood point set on the corresponding point normal vector is calculated.
Wherein, there are m points to be denoised, m projection distances are calculated, and the specific calculation formula for calculating the projection distance of the vector in the direction of the point normal vector is as follows:
wherein Ni represents the ith dot normal vector, Ni-represents the length of the ith dot normal vector, PiPo is the ith vector, PiPo-is the length of the ith vector, and Li is the ith projection distance. According to the above formula (3), m projection distances can be calculated. The calculated projection distance is used as a standard for judging noise in the embodiment of the invention, and can be compared with a threshold distance, and a noise point is selected from the points to be denoised according to the size to remove the noise point.
Optionally, after the step of calculating the projection distance of the vector corresponding to all to-be-denoised points in the neighborhood point set on the corresponding point normal vector, the method may further include:
and averaging the projection distances of the vectors corresponding to all to-be-denoised points in the calculated neighborhood point set on the corresponding point normal vector to obtain an average projection distance.
After the m projection distances are calculated, all the projection distances may be averaged to obtain an average projection distance, which is expressed by means of a mean l, and a specific calculation formula is shown below.
Where meanL is the average projection distance, Li is the ith projection distance, and m is the total number of projection distances. The calculated average projection distance can be used for calculating a threshold distance, so that the projection distance can be conveniently compared with the threshold distance, and whether the corresponding point to be denoised is a noise point or not is judged.
In the embodiment of the invention, a kd tree is established, a neighborhood point set of main data points is selected in the form of the kd tree, the distance between the main data points and the data points in each dimensional space is calculated, and then a mode of whether closer neighborhood points exist or not is searched for by the main data points and the calculated distance to form a neighborhood point set of each main data point. The accuracy of acquiring the neighborhood of the subject data point can be improved. The geometric information of the main body can be more accurately expressed by calculating the point normal vector through a principal component analysis algorithm; under the condition of not changing the coordinates of the point cloud data, a mass center point is calculated according to a neighborhood point set, a vector between the mass center point and the neighborhood point set is calculated by taking the mass center point as a reference point, the projection distance of the vector on a point normal vector is taken as a standard for judging noise, and the noise point with the projection distance reaching a threshold distance is denoised to realize denoising of small-cluster noise and denoising of strip-shaped small-section micro-bulge noise, so that the denoising efficiency is improved, and reliable point cloud data are provided.
As shown in fig. 3, fig. 3 is a flowchart of another point cloud feature preserving and denoising method provided in the embodiment of the present invention, which specifically includes the following steps:
301. and acquiring point cloud data, and segmenting a point cloud space corresponding to the point cloud data into multi-dimensional areas.
302. And selecting a segmentation point in the multi-dimensional region according to the main body data point, and respectively inserting the main body data point into the left subspace and the right subspace of the segmentation point based on the segmentation point so as to construct a preset tree model.
303. Presetting an initial neighborhood radius, searching for neighboring points corresponding to the subject data point in the kd tree for multiple times in the initial neighborhood radius based on the subject data point, and collecting the searched neighboring points to form a neighborhood point set.
304. And calculating the normal vector of the point of the main data point and the centroid point of the neighborhood point set according to the neighborhood point set corresponding to the main data point by a preset algorithm.
305. And calculating a vector according to the centroid point and the neighborhood point set, and calculating the projection distance of the vector on the normal vector of the point, wherein the vector is composed of the point to be denoised and the centroid point in the neighborhood point set.
306. And selecting a radius coefficient, and calculating the current neighborhood radius according to the initial neighborhood radius and a preset radius coefficient.
The radius coefficient may be a preset coefficient, and may be self-defined, and may be represented by f in the embodiment of the present invention, and the selection range is (0.0 < f < 1.0). The above calculating the current neighborhood radius according to the initial neighborhood radius may be multiplying the radius coefficient by the initial neighborhood radius, as shown in the following formula:
r1＝f×r (5)
wherein r is1Is the current neighborhood radius, r is the initial neighborhood radius, and f is the radius coefficient. From the value ranges of the formula (5) and f, the current neighborhood radius r1Will be smaller than the initial neighborhood radius r, i.e. the set of neighborhood points in the smaller neighborhood.
307. Selecting a neighborhood point set in the current neighborhood radius, comparing the projection distance corresponding to the to-be-denoised point in the neighborhood point set in the current neighborhood radius with the threshold distance, and calculating the threshold distance according to the average projection distance and the radius coefficient.
The threshold distance may be obtained by multiplying the average projection distance by a radius coefficient, and the threshold distance is expressed by threshold dl, and the specific calculation formula is as follows:
thresholdL＝f1*meanL (6)
where f1 represents a radius coefficient selected within the range of f. After the current neighborhood radius is calculated, a point set in the neighborhood of the current neighborhood radius r1 is selected, and the distance between a point to be denoised in the point set and a threshold value is compared.
308. And if the projection distance meets the threshold distance, recording the data point corresponding to the projection distance as a noise point, and removing the noise point in the neighborhood point set to obtain an updated neighborhood point set.
When the projection distance Li is greater than the threshold distance threshold dl, it indicates that the point to be denoised corresponding to the projection distance Li causes noise interference to the main data point, and after the point to be denoised needs to be marked as a noise point, the noise point is removed from the neighborhood point set, so that the neighborhood point set is updated. The number of times of comparing the projection distance with the threshold distance may be greater than or equal to the number of times of updating the neighborhood point set, because there may be a case where it is not a noise point when it is determined whether it is a noise point.
309. If the updated neighborhood point set does not meet the requirement of complete denoising, the initial neighborhood radius and the radius coefficient are reselected, and the steps 301 to 308 are executed in a circulating manner.
After performing multiple updates, it is necessary to determine whether the neighborhood point set is completely denoised, so as to determine whether the steps 301 and 308 need to be repeatedly performed. And when the neighborhood point set cannot be updated with the noise point, the denoising processing is finished.
If the main data point is still affected by noise after m comparisons, step 301 and step 308 are executed in a loop until complete denoising is achieved. In addition, a denoising condition may also be set, and when the updated neighborhood point set reaches the denoising condition, the execution may be stopped. According to the test condition, the test is usually performed for 2-3 times. When the updated neighborhood point set does not meet the complete denoising or denoising condition, the parameters r, f and f1 may be reset, and then the reset parameters are carried into the above steps 301 to 308 to be repeatedly executed, so as to find out noise points for denoising, and obtain point cloud data with better quality.
In the embodiment of the invention, after m projection distances Li are calculated, the radius coefficient is selected, the current neighborhood radius is calculated according to the initial neighborhood radius and the preset radius coefficient, a neighborhood point set is selected in the range of the current neighborhood radius, the m projection distances Li in the neighborhood point set are respectively compared with the threshold distance (obtained by calculating the radius coefficient and the average projection distance), the points to be denoised which are larger than the threshold distance are recorded as the denoised points, all the denoised points can be found and deleted in the neighborhood point set in a set form, and the updating of the neighborhood point set is realized. And under the condition that the complete denoising is not judged, the parameters of the initial neighborhood radius, the current neighborhood radius and the radius coefficient are reapplied and defined, and the complete denoising can be realized after repeated execution is carried out at least once. The method has the advantages that no noise removing point is generated in the neighborhood point set, so that small cluster noise removing and strip-shaped small-section micro-bulge noise removing are realized, the point cloud data corresponding to the main body data point and reserved are high in data quality, and reliable point cloud data are improved for surface reconstruction of the point cloud data Chinese main body.
As shown in fig. 4, fig. 4 is a schematic structural diagram of a point cloud feature preserving denoising device provided in an embodiment of the present invention, where the point cloud feature preserving denoising device 400 includes:
a searching module 401, configured to construct a preset tree model according to the point cloud data, and search a neighborhood point set corresponding to the main data point in the point cloud data based on the preset tree model;
a first calculating module 402, configured to calculate, according to a neighborhood point set corresponding to a main data point, a point normal vector of the main data point and a centroid point of the neighborhood point set through a preset algorithm;
a second calculating module 403, configured to calculate a vector according to the centroid point and the neighborhood point set, and a projection distance of the vector on a normal vector of the point, where the vector is formed by the point to be denoised and the centroid point in the neighborhood point set;
and an updating module 404, configured to record, as a noise point, a point to be denoised corresponding to the projection distance if the projection distance satisfies the threshold distance, and update the neighborhood point set based on the noise point until the neighborhood point set of the updated main data point is completely denoised.
Optionally, as shown in fig. 5, fig. 5 is a schematic structural diagram of another point cloud feature preserving and denoising apparatus provided in the embodiment of the present invention, and the search module 401 includes:
the acquiring unit 4011 is configured to acquire point cloud data and partition a point cloud space corresponding to the point cloud data into multidimensional areas;
the building unit 4012 is configured to select a segmentation point in the multidimensional region according to the subject data point, and insert the subject data point into a left subspace and a right subspace of the segmentation point respectively based on the segmentation point to build a preset tree model, where the preset number model is a kd tree;
the searching unit 4013 is configured to preset an initial neighborhood radius, search, for multiple times, neighboring points corresponding to the subject data point in the kd-tree within the initial neighborhood radius based on the subject data point, and form a neighborhood point set from a set of the searched neighboring points.
Optionally, the first calculating module 402 is configured to calculate a mean value of the neighborhood point set according to the neighborhood point set corresponding to the subject data point, and determine the mean value as a centroid point corresponding to the subject data point, where the mean value corresponds to the subject data point.
Optionally, as shown in fig. 6, fig. 6 is a schematic structural diagram of another point cloud feature preserving and denoising apparatus provided in the embodiment of the present invention, and the second calculating module 403 includes:
the first calculation unit 4031 is used for calculating vectors according to the points to be denoised in the neighborhood point set and the centroid points;
a second calculating unit 4032, configured to calculate, according to the vector and the point normal vector, a projection distance of the vector in the direction where the point normal vector is located until the projection distance of the vector corresponding to all the points to be denoised in the neighborhood point set on the corresponding point normal vector is calculated.
Optionally, as shown in fig. 7, fig. 7 is a schematic structural diagram of another point cloud feature preserving and denoising apparatus provided in the embodiment of the present invention, and the apparatus 400 further includes:
the third calculating module 405 is configured to average projection distances of vectors corresponding to all to-be-denoised points in the calculated neighborhood point set on the corresponding point normal vectors to obtain an average projection distance.
Optionally, as shown in fig. 8, fig. 8 is a schematic structural diagram of another point cloud feature preserving and denoising apparatus provided in the embodiment of the present invention, and the update module 404 includes:
a third calculating unit 4041, configured to select a radius coefficient, and calculate a current neighborhood radius according to the initial neighborhood radius and a preset radius coefficient;
the comparing unit 4042 is configured to select a neighborhood point set within a current neighborhood radius, compare a projection distance corresponding to a to-be-denoised point in the neighborhood point set within the current neighborhood radius with a threshold distance, and calculate the threshold distance according to an average projection distance and a radius coefficient;
a removing unit 4043, configured to record a main data point corresponding to the projection distance as a noise point if the projection distance satisfies the threshold distance, and remove the noise point in the neighborhood point set to obtain an updated neighborhood point set;
the loop unit 4044 is configured to reselect the initial neighborhood radius and the radius coefficient if the updated neighborhood point set does not meet the complete denoising requirement, and return to the module 401-404 to execute the corresponding step.
As shown in fig. 9, fig. 9 is a schematic structural diagram of an electronic device according to an embodiment of the present invention, where the electronic device 900 includes: the system comprises a memory 902, a processor 901, a network interface 903 and a computer program which is stored on the memory 902 and can run on the processor 901, wherein the processor 901 implements the steps of the point cloud feature-preserving denoising method provided by the embodiment when executing the computer program.
Specifically, the processor 901 is configured to perform the following steps:
constructing a preset tree model according to the point cloud data, and searching a neighborhood point set corresponding to the main data point in the point cloud data based on the preset tree model;
calculating a point normal vector of the main data point and a centroid point of the neighborhood point set according to the neighborhood point set corresponding to the main data point through a preset algorithm;
calculating a vector according to the centroid point and the neighborhood point set, and calculating the projection distance of the vector on a normal vector of the point, wherein the vector is composed of the point to be denoised and the centroid point in the neighborhood point set;
and if the projection distance meets the threshold distance, recording the point to be denoised corresponding to the projection distance as a noise point, and updating the neighborhood point set based on the noise point until the updated neighborhood point set of the main data point is completely denoised.
Optionally, the step of constructing a preset tree model according to the point cloud data and searching a neighborhood point set corresponding to the main data point in the point cloud data based on the preset tree model, executed by the processor 901, includes:
acquiring point cloud data, and segmenting a point cloud space corresponding to the point cloud data into multi-dimensional areas;
selecting segmentation points in the multi-dimensional area according to the main data points, and respectively inserting the main data points into the left subspace and the right subspace of the segmentation points based on the segmentation points to construct a preset tree model, wherein the preset number model is a kd tree;
presetting an initial neighborhood radius, searching for neighboring points corresponding to the subject data point in the kd tree for multiple times in the initial neighborhood radius based on the subject data point, and collecting the searched neighboring points to form a neighborhood point set.
Optionally, the step of calculating the centroid point of the neighborhood point set performed by the processor 901 includes:
and calculating the mean value of the neighborhood point set according to the neighborhood point set corresponding to the main data point, and determining the mean value as the centroid point corresponding to the main data point, wherein the mean value corresponds to the main data point.
Optionally, the step of calculating a vector according to the centroid point and the neighborhood point set, and the projection distance of the vector on the normal vector of the point, performed by the processor 901, includes:
calculating vectors according to the point to be denoised and the centroid point in the neighborhood point set;
and calculating the projection distance of the vector in the direction of the point normal vector according to the vector and the point normal vector until the projection distance of the vector corresponding to all the points to be denoised in the neighborhood point set on the corresponding point normal vector is calculated.
Optionally, after the step of calculating the projection distance of the vector corresponding to all to-be-denoised points in the neighborhood point set on the normal vector of the corresponding point, the processor 901 is further configured to execute:
and averaging the projection distances of the vectors corresponding to all to-be-denoised points in the calculated neighborhood point set on the corresponding point normal vector to obtain an average projection distance.
Optionally, if the projection distance satisfies the threshold distance, the step executed by the processor 901 of recording the main data point corresponding to the projection distance as a noise point, and updating the neighborhood point set based on the noise point until the neighborhood point set of the updated main data point is completely denoised includes:
selecting a radius coefficient, and calculating the radius of the current neighborhood according to the initial neighborhood radius and a preset radius coefficient;
selecting a neighborhood point set in the current neighborhood radius, comparing the projection distance corresponding to the to-be-denoised point in the neighborhood point set in the current neighborhood radius with a threshold distance, and calculating the threshold distance according to the average projection distance and the radius coefficient;
if the projection distance meets the threshold distance, recording a main data point corresponding to the projection distance as a noise point, and removing the noise point in the neighborhood point set to obtain an updated neighborhood point set;
and if the updated neighborhood point set does not meet the complete denoising requirement, reselecting the initial neighborhood radius and the radius coefficient, circularly executing the steps of constructing a preset tree model according to the point cloud data and searching the neighborhood point set corresponding to the main data point in the point cloud data based on the preset tree model.
The electronic device 900 provided by the embodiment of the present invention can implement each implementation manner in the point cloud feature-preserving denoising method embodiment, and has corresponding beneficial effects, and for avoiding repetition, details are not repeated here.
It is noted that only 901 and 903 having components are shown, but it is understood that not all of the shown components are required and that more or fewer components may alternatively be implemented. As will be understood by those skilled in the art, the electronic device 900 is a device capable of automatically performing numerical calculation and/or information processing according to a preset or stored instruction, and the hardware includes, but is not limited to, a microprocessor, an Application Specific Integrated Circuit (ASIC), a Programmable gate array (FPGA), a Digital Signal Processor (DSP), an embedded device, and the like.
The electronic device 900 may be a desktop computer, a notebook, a palmtop, or other computing device. The electronic device 900 may interact with a user through a keyboard, mouse, remote control, touch pad, or voice-activated device.
The memory 902 includes at least one type of readable storage medium including a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., SD or DX memory, etc.), a Random Access Memory (RAM), a Static Random Access Memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, an optical disk, etc. In some embodiments, the storage 902 may be an internal storage unit of the electronic device 900, such as a hard disk or a memory of the electronic device 900. In other embodiments, the memory 902 may also be an external storage device of the electronic device 900, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), or the like, provided on the electronic device 900. Of course, the memory 902 may also include both internal and external memory units of the electronic device 900. In this embodiment, the memory 902 is generally used for storing an operating system installed in the electronic device 900 and various application software, such as program codes of a point cloud feature-preserving denoising method. In addition, the memory 902 may also be used to temporarily store various types of data that have been output or are to be output.
The network interface 903 may comprise a wireless network interface or a wired network interface, and the network interface 903 is typically used to establish communication connections between the electronic device 900 and other electronic devices.
The embodiment of the present invention further provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when being executed by the processor 901, the computer program implements each process in the point cloud feature preserving and denoising method provided in the embodiment, and can achieve the same technical effect, and is not described here again to avoid repetition.
It will be understood by those skilled in the art that all or part of the processes for implementing the point cloud feature-preserving denoising method according to the embodiments may be implemented by a computer program instructing associated hardware, and the program may be stored in a computer-readable storage medium, and when executed, may include the processes according to the embodiments of the methods. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), or the like.
It should be understood that, although the steps in the flowcharts of the figures are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and may be performed in other orders unless explicitly stated herein. Moreover, at least a portion of the steps in the flow chart of the figure may include multiple sub-steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, which are not necessarily performed in sequence, but may be performed alternately or alternately with other steps or at least a portion of the sub-steps or stages of other steps.
The terms "including" and "having," and any variations thereof, in the description and claims of this application and the description of the above figures are intended to cover non-exclusive inclusions. The terms "first," "second," and the like in the description and claims of this application or in the above-described drawings are used for distinguishing between different objects and not for describing a particular order. Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. It is explicitly and implicitly understood by one skilled in the art that the embodiments described herein can be combined with other embodiments.
The above disclosure is only for the purpose of illustrating the preferred embodiments of the present invention, and it is therefore to be understood that the invention is not limited by the scope of the appended claims.
Claims (10)
1. A point cloud feature preserving denoising method is characterized by comprising the following steps:
constructing a preset tree model according to the point cloud data, and searching a neighborhood point set corresponding to a main data point in the point cloud data based on the preset tree model;
calculating a point normal vector of the main data point and a centroid point of the neighborhood point set according to the neighborhood point set corresponding to the main data point through a preset algorithm;
calculating a vector according to the centroid point and the neighborhood point set and the projection distance of the vector on the normal vector of the point, wherein the vector is formed by the point to be denoised in the neighborhood point set and the centroid point;
if the projection distance meets the threshold distance, recording the point to be denoised corresponding to the projection distance as a noise point, and updating the neighborhood point set based on the noise point until the neighborhood point set of the main data point is completely denoised after updating.
2. The method of claim 1, wherein the constructing a preset tree model from the point cloud data, and the searching a set of neighborhood points corresponding to the subject data points in the point cloud data based on the preset tree model comprises:
acquiring the point cloud data, and segmenting a point cloud space corresponding to the point cloud data into multi-dimensional areas;
selecting a segmentation point in the multi-dimensional region according to the main data point, and respectively inserting the main data point into a left subspace and a right subspace of the segmentation point based on the segmentation point to construct the preset tree model, wherein the preset number model is a kd tree;
presetting an initial neighborhood radius, searching adjacent points corresponding to the main data point in the kd tree for multiple times in the initial neighborhood radius based on the main data point, and forming the neighborhood point set by the searched adjacent points.
3. The method of claim 1, wherein the step of computing the centroid point of the set of neighborhood points comprises:
calculating the mean value of the neighborhood point set according to the neighborhood point set corresponding to the main data point, and determining the mean value as the centroid point corresponding to the main data point, wherein the mean value corresponds to the main data point.
4. The method of claim 2, wherein the step of computing a vector from the centroid point and the set of neighborhood points, and wherein the projection distance of the vector onto the normal vector of points comprises:
calculating the vector according to the point to be denoised in the neighborhood point set and the centroid point;
and calculating the projection distance of the vector in the direction of the point normal vector according to the vector and the point normal vector until the projection distance of the vector corresponding to all the points to be denoised in the neighborhood point set on the corresponding point normal vector is calculated.
5. The method of claim 4, wherein after the step of calculating the projection distances of the vectors corresponding to all the points to be denoised in the neighborhood point set on the corresponding point normal vectors, the method further comprises:
and calculating the average value of the projection distances of the vectors corresponding to all the to-be-denoised points in the calculated neighborhood point set on the corresponding point normal vector so as to obtain the average projection distance.
6. The method of claim 5, wherein if the projection distance satisfies a threshold distance, recording a subject data point corresponding to the projection distance as a noise point, and updating the neighborhood point set based on the noise point until the updated neighborhood point set of the subject data point is completely denoised comprises:
selecting a radius coefficient, and calculating the radius of the current neighborhood according to the initial neighborhood radius and a preset radius coefficient;
selecting the neighborhood point set in the current neighborhood radius, and comparing the projection distance corresponding to the to-be-denoised point in the neighborhood point set in the current neighborhood radius with a threshold distance, wherein the threshold distance is obtained by calculation according to the average projection distance and the radius coefficient;
if the projection distance meets the threshold distance, recording a main data point corresponding to the projection distance as a noise point, and removing the noise point in the neighborhood point set to obtain an updated neighborhood point set;
if the updated neighborhood point set does not meet the complete denoising requirement, reselecting the initial neighborhood radius and the radius coefficient, circularly executing the steps of constructing a preset tree model according to the point cloud data and searching the neighborhood point set corresponding to the main data point in the point cloud data based on the preset tree model.
7. A point cloud feature preserving denoising device is characterized by comprising:
the searching module is used for constructing a preset tree model according to the point cloud data and searching a neighborhood point set corresponding to the main data point in the point cloud data based on the preset tree model;
the first calculation module is used for calculating a point normal vector of the main data point and a centroid point of the neighborhood point set according to the neighborhood point set corresponding to the main data point through a preset algorithm;
the second calculation module is used for calculating a vector according to the centroid point and the neighborhood point set and the projection distance of the vector on the normal vector of the point, wherein the vector is formed by the point to be denoised in the neighborhood point set and the centroid point;
and the updating module is used for recording the point to be denoised corresponding to the projection distance as a noise point if the projection distance meets the threshold distance, and updating the neighborhood point set based on the noise point until the updated neighborhood point set of the main data point is completely denoised.
8. The apparatus of claim 7, the search module comprising:
the acquisition unit is used for acquiring the point cloud data and segmenting a point cloud space corresponding to the point cloud data into multi-dimensional areas;
a construction unit, configured to select a segmentation point in the multidimensional region according to the subject data point, and insert the subject data point into a left subspace and a right subspace of the segmentation point respectively based on the segmentation point to construct the preset tree model, where the preset number model is a kd tree;
and the searching unit is used for presetting an initial neighborhood radius, selecting the main data point, searching the neighboring points corresponding to the main data point in the kd tree for multiple times in the initial neighborhood radius based on the main data point, and collecting the searched neighboring points to form the neighborhood point set.
9. An electronic device, comprising: memory, processor and computer program stored on the memory and executable on the processor, the processor implementing the steps in the point cloud feature preserving denoising method of any one of claims 1-6 when executing the computer program.
10. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the point cloud feature-preserving denoising method according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010748731.XA CN112102178A (en) | 2020-07-29 | 2020-07-29 | Point cloud feature-preserving denoising method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010748731.XA CN112102178A (en) | 2020-07-29 | 2020-07-29 | Point cloud feature-preserving denoising method and device, electronic equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112102178A true CN112102178A (en) | 2020-12-18 |
Family
ID=73750511
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010748731.XA Pending CN112102178A (en) | 2020-07-29 | 2020-07-29 | Point cloud feature-preserving denoising method and device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112102178A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112733813A (en) * | 2021-03-30 | 2021-04-30 | 北京三快在线科技有限公司 | Data noise reduction method and device |
| CN114627017A (en) * | 2022-03-17 | 2022-06-14 | 南京航空航天大学 | Point cloud denoising method based on multi-level attention perception |
Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100092059A1 (en) * | 2008-10-10 | 2010-04-15 | Siemens Corporate Research, Inc. | Creation Of Motion Compensated MRI M-Mode Images Of The Myocardial Wall |
| CN103065354A (en) * | 2012-12-24 | 2013-04-24 | 中国科学院深圳先进技术研究院 | Device and method for point cloud optimization |
| CN103268631A (en) * | 2013-05-23 | 2013-08-28 | 中国科学院深圳先进技术研究院 | Method and device for extracting point cloud framework |
| CN104123746A (en) * | 2014-07-10 | 2014-10-29 | 上海大学 | Calculating method for three-dimension scanning point cloud real-time normal vectors |
| CN106530342A (en) * | 2016-10-31 | 2017-03-22 | 武汉大学 | Measurable panorama image generation method assisted by laser point cloud |
| CN106548520A (en) * | 2016-11-16 | 2017-03-29 | 湖南拓视觉信息技术有限公司 | A kind of method and system of cloud data denoising |
| CN106599129A (en) * | 2016-12-02 | 2017-04-26 | 山东科技大学 | Multi-beam point cloud data denoising method considering terrain characteristics |
| CN108364257A (en) * | 2018-02-06 | 2018-08-03 | 深圳市菲森科技有限公司 | The joining method and system of 3-D scanning point cloud data |
| CN108830931A (en) * | 2018-05-23 | 2018-11-16 | 上海电力学院 | A kind of laser point cloud compressing method based on dynamic grid k neighborhood search |
| CN108846809A (en) * | 2018-05-28 | 2018-11-20 | 河海大学 | A kind of noise eliminating method towards point off density cloud |
| CN108876744A (en) * | 2018-06-27 | 2018-11-23 | 大连理工大学 | A kind of large scale point cloud noise denoising method based on region segmentation |
| WO2019010916A1 (en) * | 2017-07-10 | 2019-01-17 | 南京埃斯顿机器人工程有限公司 | Workpiece three-dimensional point cloud data smoothing method |
| CN109740227A (en) * | 2018-12-26 | 2019-05-10 | 江南大学 | Miniature complex parts modeling method based on feature identification |
| US20200043186A1 (en) * | 2017-01-27 | 2020-02-06 | Ucl Business Plc | Apparatus, method, and system for alignment of 3d datasets |
| CN111161179A (en) * | 2019-12-26 | 2020-05-15 | 华南理工大学 | Point cloud smoothing filtering method based on normal vector |
-
2020
- 2020-07-29 CN CN202010748731.XA patent/CN112102178A/en active Pending
Patent Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100092059A1 (en) * | 2008-10-10 | 2010-04-15 | Siemens Corporate Research, Inc. | Creation Of Motion Compensated MRI M-Mode Images Of The Myocardial Wall |
| CN103065354A (en) * | 2012-12-24 | 2013-04-24 | 中国科学院深圳先进技术研究院 | Device and method for point cloud optimization |
| CN103268631A (en) * | 2013-05-23 | 2013-08-28 | 中国科学院深圳先进技术研究院 | Method and device for extracting point cloud framework |
| CN104123746A (en) * | 2014-07-10 | 2014-10-29 | 上海大学 | Calculating method for three-dimension scanning point cloud real-time normal vectors |
| CN106530342A (en) * | 2016-10-31 | 2017-03-22 | 武汉大学 | Measurable panorama image generation method assisted by laser point cloud |
| CN106548520A (en) * | 2016-11-16 | 2017-03-29 | 湖南拓视觉信息技术有限公司 | A kind of method and system of cloud data denoising |
| CN106599129A (en) * | 2016-12-02 | 2017-04-26 | 山东科技大学 | Multi-beam point cloud data denoising method considering terrain characteristics |
| US20200043186A1 (en) * | 2017-01-27 | 2020-02-06 | Ucl Business Plc | Apparatus, method, and system for alignment of 3d datasets |
| WO2019010916A1 (en) * | 2017-07-10 | 2019-01-17 | 南京埃斯顿机器人工程有限公司 | Workpiece three-dimensional point cloud data smoothing method |
| CN108364257A (en) * | 2018-02-06 | 2018-08-03 | 深圳市菲森科技有限公司 | The joining method and system of 3-D scanning point cloud data |
| CN108830931A (en) * | 2018-05-23 | 2018-11-16 | 上海电力学院 | A kind of laser point cloud compressing method based on dynamic grid k neighborhood search |
| CN108846809A (en) * | 2018-05-28 | 2018-11-20 | 河海大学 | A kind of noise eliminating method towards point off density cloud |
| CN108876744A (en) * | 2018-06-27 | 2018-11-23 | 大连理工大学 | A kind of large scale point cloud noise denoising method based on region segmentation |
| CN109740227A (en) * | 2018-12-26 | 2019-05-10 | 江南大学 | Miniature complex parts modeling method based on feature identification |
| CN111161179A (en) * | 2019-12-26 | 2020-05-15 | 华南理工大学 | Point cloud smoothing filtering method based on normal vector |
Non-Patent Citations (2)
| Title |
|---|
| 卢钰仁 等: "基于法向修正的双边滤波点云去噪处理", 《仪表技术与传感器》, no. 7, 31 July 2018 (2018-07-31), pages 111 - 115 * |
| 张佰春: "条纹投影三维测量的若干关键技术的研究", 《中国优秀硕士学位论文全文数据库信息科技辑》, no. 7, 15 July 2017 (2017-07-15) * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112733813A (en) * | 2021-03-30 | 2021-04-30 | 北京三快在线科技有限公司 | Data noise reduction method and device |
| CN114627017A (en) * | 2022-03-17 | 2022-06-14 | 南京航空航天大学 | Point cloud denoising method based on multi-level attention perception |
| CN114627017B (en) * | 2022-03-17 | 2022-12-13 | 南京航空航天大学 | Point cloud denoising method based on multi-level attention perception |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20140125663A1 (en) | 3d model shape analysis method based on perception information | |
| EP2945071B1 (en) | Index generating device and method, and search device and search method | |
| CN110599506B (en) | Point cloud segmentation method for three-dimensional measurement of complex special-shaped curved surface robot | |
| JP5754055B2 (en) | Information representation method of object or shape | |
| El‐Sayed et al. | Plane detection in 3D point cloud using octree‐balanced density down‐sampling and iterative adaptive plane extraction | |
| CN112102178A (en) | Point cloud feature-preserving denoising method and device, electronic equipment and storage medium | |
| Agathos et al. | Protrusion-oriented 3D mesh segmentation | |
| CN113920148A (en) | Building boundary extraction method and equipment based on polygon and storage medium | |
| Hu et al. | Curve skeleton extraction from 3D point clouds through hybrid feature point shifting and clustering | |
| CN111260564A (en) | Image processing method and device and computer storage medium | |
| CN110490893B (en) | Rapid Euclidean distance point cloud segmentation method | |
| KR102189811B1 (en) | Method and Apparatus for Completing Knowledge Graph Based on Convolutional Learning Using Multi-Hop Neighborhoods | |
| CN115546116A (en) | Method and system for extracting and calculating spacing of discontinuous surface of fully-covered rock mass | |
| JP5192437B2 (en) | Object region detection apparatus, object region detection method, and object region detection program | |
| CN111340145B (en) | Point cloud data classification method and device and classification equipment | |
| CN113436223A (en) | Point cloud data segmentation method and device, computer equipment and storage medium | |
| CN113159103A (en) | Image matching method, image matching device, electronic equipment and storage medium | |
| Li et al. | An improved Douglas-Peucker algorithm for fast curve approximation | |
| CN109800815B (en) | Training method, wheat recognition method and training system based on random forest model | |
| CN115375699B (en) | Point cloud segmentation method, mobile robot and computer-readable storage medium | |
| CN116452826A (en) | Coal gangue contour estimation method based on machine vision under shielding condition | |
| JP2005165969A (en) | Image processor and method | |
| JP6725310B2 (en) | Image processing device and program | |
| CN114051277A (en) | High-precision positioning method and device based on artificial intelligence | |
| CN108595686B (en) | Method, device, equipment, medium and system for mining approximate frequent graph in single graph |
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 |