WO2021142995A1

WO2021142995A1 - Image processing-based k-nearest neighbor point cloud filtering method, apparatus, and storage medium

Info

Publication number: WO2021142995A1
Application number: PCT/CN2020/091749
Authority: WO
Inventors: 张建民; 陈富健; 龙佳乐
Original assignee: 五邑大学
Priority date: 2020-01-17
Filing date: 2020-05-22
Publication date: 2021-07-22
Also published as: CN111275810B; CN111275810A

Abstract

Disclosed in the present invention are an image processing-based K-nearest neighbor point cloud filtering method, an apparatus, and a storage medium. Said method comprise: after acquiring a three-dimensional point cloud, constructing a two-dimensional image, selecting several points in the two-dimensional image by means of a K-nearest neighbor filtering template, calculating the Euclidean distance between the center point and each K-nearest neighbor point, determining whether the center point is a noise point on the basis of the Euclidean distance, a preset distance threshold and a preset ratio threshold, if the center point satisfies the distance threshold and the ratio threshold, counting the number of K-nearest neighbor points in the template on the basis of a number threshold, and if the number of K-nearest neighbor points is less than the number threshold, setting the center point of the template as an outlier noise point. By setting a number threshold, the present invention realizes space limitation, effectively removing outlier, scattered and elongated noise point clouds, and improving the accuracy and scope of application of filtering.

Description

K-nearest neighbor point cloud filtering method, device and storage medium based on image processing

Technical field

The invention relates to the technical field of three-dimensional shape measurement, in particular to a k-nearest neighbor point cloud filtering method, device and storage medium based on image processing.

Background technique

At present, with the development of 3D scanning technology, it is becoming easier to obtain 3D point cloud data on the surface of an object, which promotes the wide application of 3D scanning technology in urban modeling, reverse engineering, survey engineering and other fields. However, due to the manual operation of 3D scanning equipment, coupled with the structure of the camera and projector and the influence of external factors, there are noisy point clouds in the acquired 3D point cloud. The existing denoising methods are mainly based on K nearest neighbors. Point cloud filtering method, but the existing method only removes individual noise points, and cannot remove the long and narrow noise point cloud, and the filtering effect is not good.

Summary of the invention

In order to overcome the deficiencies of the prior art, the purpose of the present invention is to provide a K-nearest neighbor point cloud filtering method, device and storage medium based on image processing, which can remove long and narrow noise point clouds and improve the application range of the K-nearest neighbor point cloud filtering method. .

The technical solution adopted by the present invention to solve its problem is: In the first aspect, the present invention provides a K-nearest neighbor point cloud filtering method based on image processing, which includes the following steps:

Acquiring a three-dimensional point cloud, and constructing a two-dimensional image, where the two-dimensional points in the two-dimensional image uniquely correspond to the three-dimensional points in the three-dimensional point cloud;

Obtain a preset K neighboring filter template, perform convolution calculation on the two-dimensional image according to the K neighboring filter template, and obtain the Euclidean distance between the center point of the K neighboring filter template and all K neighboring points , If the Euclidean distance is less than the preset distance threshold, the K neighboring point is set as the first K neighboring point, and if the Euclidean distance is greater than the preset distance threshold, the K neighboring point is set Is the second K neighboring point;

Obtain a preset ratio threshold and number threshold, if the ratio of the number of the first K neighboring points to the second K neighboring points is greater than the ratio threshold, or the first K neighboring points and the second K neighboring points The ratio of the number of neighboring points is less than the ratio threshold and the sum of the numbers of the first K neighboring points and the second K neighboring points is less than the number threshold, and the center point is set as a noise point.

Further, the three-dimensional point cloud is obtained by reconstructing the absolute phase map and internal and external parameters of an image acquisition device, and the image acquisition device includes a camera and a projector.

Further, the performing convolution calculation on the two-dimensional image according to the K neighboring filter template specifically includes the following steps:

Acquiring a first point coordinate corresponding to the center point and a second point coordinate corresponding to the K neighboring point in the three-dimensional point cloud;

Euclidean distance is calculated according to the coordinates of the first point and the coordinates of the second point.

Further, the shape of the K neighboring filter template includes but is not limited to a rectangle and a circle, and the K neighboring filter template includes at least a center point and several K neighboring points.

Further, if the ratio of the number of the first K neighboring points to the second K neighboring points is less than the ratio threshold and the sum of the number of the first K neighboring points and the second K neighboring points is less than the number threshold , Also includes: setting the center point as an outlier noise point.

Further, after setting the center point as a noise point, it further includes: moving the K neighboring filter template to an unfiltered area in the two-dimensional image, and repeating the filtering operation until the traversal of the two-dimensional image is completed .

In a second aspect, the present invention provides a device for performing a K-nearest neighbor point cloud filtering method based on image processing, including a CPU unit configured to perform the following steps:

Acquiring a three-dimensional point cloud and constructing a two-dimensional image, where the two-dimensional points in the two-dimensional image uniquely correspond to the three-dimensional points in the three-dimensional point cloud;

Further, the CPU unit is also used to perform the following steps:

The K neighboring filter template is moved to an unfiltered area in the two-dimensional image, and the filtering operation is repeated until the traversal of the two-dimensional image is completed.

In a third aspect, the present invention provides a device for performing a K-nearest neighbor point cloud filtering method based on image processing, including at least one control processor and a memory for communicating with the at least one control processor; the memory stores the The instruction executed by the at least one control processor is executed by the at least one control processor, so that the at least one control processor can execute the K-nearest neighbor point cloud filtering method based on image processing as described above.

In a fourth aspect, the present invention provides a computer-readable storage medium, the computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to make the computer execute the K-nearest neighbor point cloud filtering based on image processing as described above method.

In a fifth aspect, the present invention also provides a computer program product, the computer program product includes a computer program stored on a computer-readable storage medium, the computer program includes program instructions, when the program instructions are executed by a computer , Let the computer execute the K-nearest neighbor point cloud filtering method based on image processing as described above.

The one or more technical solutions provided in the embodiments of the present invention have at least the following beneficial effects: the present invention constructs a two-dimensional image after acquiring a three-dimensional point cloud, and selects several points in the two-dimensional image through the K neighbor filter template, Calculate the Euclidean distance between the center point and K neighboring points, and determine whether the center point is a noise point based on the Euclidean distance and the preset distance threshold and ratio threshold. For the center point that meets the distance threshold and the ratio threshold, the number threshold is compared The number of K neighboring points in the template is counted, and the center point of the template whose number is less than the number threshold is set as outlier noise points. By setting the number threshold, the space level limitation is realized, and outliers, scattered and narrow noise points are effectively removed Cloud, improve the accuracy and scope of filtering.

Description of the drawings

The present invention will be further explained below with reference to the drawings and examples.

Fig. 1 is a flowchart of a K-nearest neighbor point cloud filtering method based on image processing provided by an embodiment of the present invention;

2 is a flowchart of performing convolution calculation on the two-dimensional image according to the K neighbor filter template in a K neighbor point cloud filtering method based on image processing according to an embodiment of the present invention;

Fig. 3 is a complete flow chart of a K-nearest neighbor point cloud filtering method based on image processing provided by an embodiment of the present invention;

Fig. 4 is a schematic diagram of an apparatus for performing a K-nearest neighbor point cloud filtering method based on image processing according to another embodiment of the present invention.

Detailed ways

In order to make the objectives, technical solutions, and advantages of the present invention clearer, the following further describes the present invention in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present invention, but not used to limit the present invention.

It should be noted that if there is no conflict, the various features in the embodiments of the present invention can be combined with each other, and all fall within the protection scope of the present invention. In addition, although functional modules are divided in the schematic diagram of the device, and the logical sequence is shown in the flowchart, in some cases, the module division in the device may be different from the module division in the device, or the sequence shown in the flowchart may be executed. Or the steps described.

Referring to Fig. 1, the first embodiment of the present invention provides a K-nearest neighbor point cloud filtering method based on image processing, including the following steps:

Step S100, acquiring a three-dimensional point cloud, and constructing a two-dimensional image, where the two-dimensional points in the two-dimensional image uniquely correspond to the three-dimensional points in the three-dimensional point cloud;

Step S200: Obtain a preset K neighboring filter template, and perform convolution calculation on the two-dimensional image according to the K neighboring filter template, and obtain the distance between the center point of the K neighboring filter template and all K neighboring points If the Euclidean distance is less than the preset distance threshold, set the K neighboring point as the first K neighboring point; if the Euclidean distance is greater than the preset distance threshold, set the K The neighboring point is set as the second K neighboring point;

Step S300: Obtain a preset ratio threshold and number threshold. If the ratio of the number of the first K neighboring points to the second K neighboring points is greater than the ratio threshold, or the first K neighboring points and the number The number ratio of the second K neighboring points is less than the ratio threshold and the sum of the number of the first K neighboring points and the second K neighboring points is less than the number threshold, and the center point is set as a noise point.

Among them, it should be noted that the three-dimensional points in this embodiment are reconstructed according to each point of the absolute phase map in the image plane, so each point of the absolute phase map in the image plane is the same as the reconstructed three-dimensional point. The points have a one-to-one correspondence, that is, the constructed two-dimensional image contains not only the position information of the absolute phase map but also the three-dimensional point information, which is beneficial to the filtering calculation. Among them, it should be noted that the reconstruction method can use any three-dimensional to two-dimensional graphics reconstruction method in the prior art, and the present invention does not involve the improvement of specific reconstruction methods, and will not be repeated here.

Among them, it should be noted that this embodiment preferably uses the K neighbor filter template to define a certain spatial range from the two-dimensional image. Therefore, the number of K neighbor points is limited. If the number is too small, the representative area is outlying or scattered noise. Area, so as to increase the spatial information to filter the noise of the point cloud. It can be understood that the Euclidean distance can better reflect the spatial distance between two points, which is the preferred embodiment of this embodiment, and other types of parameters can also be used to achieve similar effects. Those skilled in the art can understand that, due to different filtering accuracy requirements, the distance threshold, the ratio threshold, and the quantity threshold can be adjusted according to actual needs, and this embodiment is not limited due to specific numerical values.

Wherein, in step S300, if the ratio of the number of the first K neighboring points to the second K neighboring points is less than the ratio threshold, it means that the distance between the center point and the three-dimensional point corresponding to the neighboring points is relatively short. Therefore, the center point is based on image processing. The level of is not a noise point, but if there are fewer points in the K neighboring filter template, the corresponding is a large area of outlier noise. In this case, this embodiment preferably judges by the number threshold. If the first K neighboring point If the sum of the number of neighboring points with the second K is less than the number threshold, the center point is set as outlier noise, so as to remove the long and narrow patchy noise. In addition, if the ratio of the number of the first K neighboring points to the second K neighboring points is greater than the ratio threshold, it means that the center point is far away from other neighboring points, which is probably a noise point in the point cloud. Set the center point directly to For noise points, there is no need to judge the number threshold, which can simplify the calculation process.

Among them, it should be noted that the three-dimensional point cloud can be acquired by any device in the prior art. In this embodiment, a camera and a projector are preferably used. The specific three-dimensional point cloud acquisition method is not an improvement of the present invention. Go into details again.

Referring to FIG. 2, further, the performing convolution calculation on the two-dimensional image according to the K neighboring filter template specifically includes the following steps:

Step S210, acquiring the first point coordinates corresponding to the center point and the second point coordinates corresponding to the K neighboring points in the three-dimensional point cloud;

Step S220: Calculate the Euclidean distance according to the coordinates of the first point and the coordinates of the second point.

Among them, it should be noted that since the K neighboring filter template includes at least several K neighboring points, this embodiment preferably compares the center point with other K neighboring points one by one, and repeating step S210 and step S220 until all the K neighboring points are completed. Point calculation.

Among them, it should be noted that the specific shape of the K neighboring filter template can be determined according to actual needs, which does not limit the embodiment in this city, and the circle and rectangle are only preferred in this embodiment. It can be understood that, for example, if the K neighboring filter template is a rectangle with a size of M×N, then K=M×N-1, avoiding repeated calculation of the center point.

Among them, it should be noted that the ratio of the number of the first K neighboring points to the second K neighboring points is less than the ratio threshold and the sum of the number of the first K neighboring points and the second K neighboring points is less than the number threshold is set as outlier noise points The preferred embodiment of this embodiment can also only be set as noise points. In this embodiment, outlier noise points can be distinguished from ordinary noise points, which is convenient for subsequent processing.

Further, after setting the center point as a noise point, the method further includes: moving the K neighboring filter template to an unfiltered area in the two-dimensional image, and repeating the filtering operation until the traversal of the two-dimensional image is completed .

Among them, it should be noted that since the K neighboring filter template is smaller than the two-dimensional image, that is, it does not include all points in the two-dimensional image, after performing one filtering, the K neighboring filter template is moved so that the K neighboring filter template is in two Slide in the two-dimensional image until the traversal is completed, and filter all points to ensure the integrity and accuracy of the filtering.

Referring to FIG. 3, in addition, another embodiment of the present invention also provides a K-nearest neighbor point cloud filtering method based on image processing, including the following steps:

Step S3100: Obtain the coordinates of the first point corresponding to the center point and the coordinates of the second point corresponding to the K neighboring points in the three-dimensional point cloud;

Step S3200: Obtain a preset K neighbor filter template, and acquire the first point coordinates corresponding to the center point of the K neighbor filter template and the second point coordinates corresponding to the K neighbor points in the three-dimensional point cloud;

Step S3300, it is judged whether the Euclidean distance is greater than the distance threshold, if yes, go to step S3310, otherwise go to step S3320;

Step S3310, the K neighboring point is set as the first K neighboring point, and step S3330 is executed;

Step S3320, the K neighboring point is set as the second K neighboring point, and step S3330 is executed;

Step S3330: Calculate the ratio of the number of the first K neighboring points to the second K neighboring points;

Step S3400, it is judged whether the ratio obtained in step S3330 is greater than the ratio threshold value, if yes, go to step S3410, otherwise go to step S3420;

Step S3410, set the center point as a noise point, move the K-nearest neighbor filter template, if the two-bit image is not completely traversed, perform step S3100;

Step S3420: Calculate the sum of the numbers of the first K neighboring points and the second K neighboring points. Compared with the number threshold, if it is greater than the number threshold, perform step S3422, and if it is less than the number threshold, perform step S3423;

Step S3422, set the center point as an outlier noise point, move the K-nearest neighbor filter template, if the two-bit image is not completely traversed, perform step S3100;

In step S3423, the center point is retained, and the K-nearest neighbor filter template is moved. If the two-bit image is not completely traversed, step S3100 is executed.

It should be noted that steps S3100 to S3330 of this embodiment implement point cloud filtering based on image processing, and steps S3400 to S3423 increase the spatial information of k-nearest neighbors to implement a large and narrow noise point cloud.

4, the second embodiment of the present invention also provides a device for performing K-nearest neighbor point cloud filtering method based on image processing. The device is a smart device, such as a smart phone, a computer, and a tablet. Take the computer as an example to illustrate.

The computer 4000 for performing the K-nearest neighbor point cloud filtering method based on image processing includes a CPU unit 4100, and the CPU unit 4100 is configured to perform the following steps:

Further, the CPU unit is further configured to perform the following steps: further, in another embodiment of the present invention, the CPU unit 4100 is further configured to perform the following steps:

Further, in another embodiment of the present invention, the CPU unit 4100 is further configured to perform the following steps:

The computer 4000 and the CPU unit 4100 can be connected by a bus or other means. The computer 4000 also includes a memory. As a non-transitory computer-readable storage medium, the memory can be used to store non-transitory software programs and non-transitory Computer-executable programs and modules, such as program instructions/modules corresponding to the device for executing the K-nearest neighbor point cloud filtering method based on image processing in the embodiment of the present invention. The computer 4000 runs the non-transitory software programs, instructions, and modules stored in the memory to control the CPU unit 4100 to execute various functional applications and data processing for executing the K-nearest point cloud filtering method based on image processing, that is, to achieve the above The K-nearest neighbor point cloud filtering method based on image processing of the method embodiment.

The memory may include a storage program area and a storage data area, where the storage program area may store an operating system and an application program required by at least one function; the storage data area may store data created according to the use of the CPU unit 4100 and the like. In addition, the memory may include a high-speed random access memory, and may also include a non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some embodiments, the memory may optionally include a memory remotely provided with respect to the CPU unit 4100, and these remote memories may be connected to the computer 4000 via a network. Examples of the aforementioned networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

The one or more modules are stored in the memory, and when executed by the CPU unit 4100, the K-nearest neighbor point cloud filtering method based on image processing in the foregoing method embodiment is executed.

The embodiment of the present invention also provides a computer-readable storage medium that stores computer-executable instructions that are executed by the CPU unit 4100 to implement the above-mentioned K based on image processing. Neighbor point cloud filtering method.

The device embodiments described above are merely illustrative, and the devices described as separate components may or may not be physically separated, that is, they may be located in one place, or they may be distributed on multiple network devices. Some or all of the modules can be selected according to actual needs to achieve the objectives of the solutions of the embodiments.

It should be noted that, since the apparatus for performing the K-nearest neighbor point cloud filtering method based on image processing in this embodiment is based on the same inventive concept as the above-mentioned K-nearest neighbor point cloud filtering method based on image processing, the method embodiment The corresponding content in is also applicable to this device embodiment, and will not be described in detail here.

Through the description of the above implementation manners, those skilled in the art can clearly understand that each implementation manner can be implemented by means of software plus a general hardware platform. Those skilled in the art can understand that all or part of the processes in the methods of the foregoing embodiments can be implemented by computer programs instructing relevant hardware. The programs can be stored in a computer-readable storage medium. When the program is executed, , May include the flow of the embodiment of the above-mentioned method. Wherein, the storage medium may be a magnetic disk, an optical disc, a read-only memory (Read Only Memory, ROM), or a random access memory (Random Access Memory, RAM), etc.

The above is a detailed description of the preferred implementation of the present invention, but the present invention is not limited to the above-mentioned embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention. Equivalent modifications or replacements are all included in the scope defined by the claims of this application.

Claims

A K-nearest neighbor point cloud filtering method based on image processing is characterized in that it comprises the following steps:

Acquiring a three-dimensional point cloud and constructing a two-dimensional image, where the two-dimensional points in the two-dimensional image uniquely correspond to the three-dimensional points in the three-dimensional point cloud;

Obtain a preset K neighboring filter template, perform convolution calculation on the two-dimensional image according to the K neighboring filter template, and obtain the Euclidean distance between the center point of the K neighboring filter template and all K neighboring points , If the Euclidean distance is less than the preset distance threshold, the K neighboring point is set as the first K neighboring point, and if the Euclidean distance is greater than the preset distance threshold, the K neighboring point is set Is the second K neighboring point;

Obtain a preset ratio threshold and number threshold, if the ratio of the number of the first K neighboring points to the second K neighboring points is greater than the ratio threshold, or the first K neighboring points and the second K neighboring points The ratio of the number of neighboring points is less than the ratio threshold and the sum of the numbers of the first K neighboring points and the second K neighboring points is less than the number threshold, and the center point is set as a noise point.
The K-nearest neighbor point cloud filtering method based on image processing according to claim 1, wherein the three-dimensional point cloud is reconstructed from the absolute phase map and the internal and external parameters of the image acquisition device, and the image acquisition The equipment includes a camera and a projector.
The K-nearest neighbor point cloud filtering method based on image processing according to claim 1, wherein the performing convolution calculation on the two-dimensional image according to the K-nearest filter template specifically comprises the following steps:

Acquiring a first point coordinate corresponding to the center point and a second point coordinate corresponding to the K neighboring point in the three-dimensional point cloud;

Euclidean distance is calculated according to the coordinates of the first point and the coordinates of the second point.
The K-nearest neighbor point cloud filtering method based on image processing according to claim 1, wherein the shape of the K-nearest filter template includes but is not limited to a rectangle and a circle, and the K-nearest filter template includes at least The central point and several K neighboring points.
The K-nearest neighbor point cloud filtering method based on image processing according to claim 1, wherein if the ratio of the number of the first K neighboring points to the second K neighboring points is less than the ratio threshold and the ratio The sum of the number of the first K neighboring points and the second K neighboring points is less than the number threshold, and the method further includes: setting the center point as an outlier noise point.
The K-nearest neighbor point cloud filtering method based on image processing according to claim 1, wherein after setting the center point as a noise point, the method further comprises: moving the K-nearest filter template to the two-dimensional For the unfiltered area in the image, the filtering operation is repeated until the traversal of the two-dimensional image is completed.
A device for performing the K-nearest neighbor point cloud filtering method based on image processing, characterized in that it comprises a CPU unit, and the CPU unit is configured to perform the following steps:

Acquiring a three-dimensional point cloud and constructing a two-dimensional image, where the two-dimensional points in the two-dimensional image uniquely correspond to the three-dimensional points in the three-dimensional point cloud;

Obtain a preset K neighboring filter template, perform convolution calculation on the two-dimensional image according to the K neighboring filter template, and obtain the Euclidean distance between the center point of the K neighboring filter template and all K neighboring points If the Euclidean distance is less than the preset distance threshold, then the K neighboring point is set as the first K neighboring point, and if the Euclidean distance is greater than the preset distance threshold, then the K neighboring point is set Is the second K neighboring point;

Obtain a preset ratio threshold and number threshold, if the ratio of the number of the first K neighboring points to the second K neighboring points is greater than the ratio threshold, or the first K neighboring points and the second K neighboring points The ratio of the number of neighboring points is less than the ratio threshold and the sum of the numbers of the first K neighboring points and the second K neighboring points is less than the number threshold, and the center point is set as a noise point.
The apparatus for performing K-nearest neighbor point cloud filtering method based on image processing according to claim 7, wherein the CPU unit is further configured to perform the following steps: acquiring the three-dimensional point cloud and the The coordinates of the first point corresponding to the center point and the coordinates of the second point corresponding to the K neighboring points;

Euclidean distance is calculated according to the coordinates of the first point and the coordinates of the second point.
The apparatus for performing the K-nearest neighbor point cloud filtering method based on image processing according to claim 7, wherein the CPU unit is further configured to perform the following step: move the K-nearest filter template to all For the unfiltered area in the two-dimensional image, the filtering operation is repeated until the traversal of the two-dimensional image is completed.
A computer-readable storage medium, characterized in that: the computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to make a computer execute any one of claims 1-6. A K-nearest neighbor point cloud filtering method based on image processing.