WO2021142996A1

WO2021142996A1 - Point cloud denoising method, system, and device employing image segmentation, and storage medium

Info

Publication number: WO2021142996A1
Application number: PCT/CN2020/091751
Authority: WO
Inventors: 张建民; 陈富健; 龙佳乐
Original assignee: 五邑大学
Priority date: 2020-01-13
Filing date: 2020-05-22
Publication date: 2021-07-22
Also published as: CN111275633A; CN111275633B; WO2021142996A9

Abstract

Disclosed are a point cloud denoising method, system, and device employing image segmentation, and a storage medium. The method can effectively remove all noise point clouds from a point cloud, enabling the noise point clouds to be removed one by one whether they are outliers, scattered points, or a large number of noise points suspended near a main body. In addition, the method employs image processing, and thus avoids the complexity of three-dimensional point computation. The method is simple and reliable, and can be used to remove noise from a 3D reconstructed point cloud by using absolute phase in an arbitrary manner.

Description

Point cloud denoising method, system, device and storage medium based on image segmentation

Technical field

The invention relates to the technical field of three-dimensional measurement, in particular to a point cloud denoising method, system, device and storage medium based on image segmentation.

Background technique

With the rapid development of computer science and the emergence of modern precision instruments, the acquisition of point cloud models has become easier and easier, and it has been widely used in urban modeling, computer graphics, topographic surveying and mapping, cultural relics protection, reverse engineering and other fields. . As the 3D scanning process will be affected by instrument accuracy, environmental noise, and the structure of the measured object itself, noise point clouds will inevitably be generated, which affects the accuracy of 3D reconstruction. At present, most point cloud denoising methods cannot adapt to the existence of various noisy point clouds, and most of them calculate the characteristics of point clouds in three-dimensional space to remove the noise point clouds, but the coordinate accuracy of the point clouds is high and three calculations are required. The value of each coordinate has the disadvantages of complicated calculation and large amount of calculation.

Summary of the invention

In order to solve the above problems, the purpose of the present invention is to provide a point cloud denoising method, system, device and storage medium based on image segmentation, no matter what kind of noise point cloud exists, it can be filtered out, and based on the image The processing avoids the complexity of calculation in the three-dimensional space, and the method is simple and reliable, and is suitable for any point cloud noise removal using absolute phase for three-dimensional reconstruction.

The technical solutions adopted by the present invention to solve its problems are:

In the first aspect, an embodiment of the present invention proposes a point cloud denoising method based on image segmentation, including:

According to the one-to-one correspondence between the Z value of the three-dimensional point after three-dimensional reconstruction and the absolute phase map, a point cloud mapping image is obtained;

The image segmentation method of region growth is used to perform image segmentation on the point cloud mapping image until all regions are segmented, and the larger region is defined as the reference region, and the other regions are defined as the pending region, and then it is judged whether the pending region is noise area;

By calculating the relationship between the to-be-determined area and the reference area, the to-be-determined area is divided into the internal and external conditions of the reference area and processed separately;

In the internal case, calculate the distance between any point in the to-be-determined area and any nearest neighbor point in the reference area, and in the external case, calculate the distance between any point on the contour of the to-be-determined area and any nearest neighbor point on the contour of the reference area, the distance is larger The noise area is the noise area, and the noise point cloud is removed by deleting the location of the noise area.

Further, the one-to-one correspondence between the Z value of the three-dimensional point after the three-dimensional reconstruction and the absolute phase map to obtain a point cloud mapping image includes: using the absolute phase map and the internal and external parameters calibrated by the camera and the projector to perform the three-dimensional point cloud Reconstruction, the Z coordinate in the three-dimensional point (X, Y, Z) reconstructed in three dimensions is one-to-one correspondence with the two-dimensional point (u, v) of the absolute phase map to construct a point cloud mapping image, a point cloud mapping image The pixel value of each pixel point (u, v) in the plane is the value of the three-dimensional point Z.

Further, the image segmentation method using region growth cyclically performs image segmentation on the point cloud mapping image until all regions are segmented, and a larger region is defined as a reference region, and other regions are defined as pending regions, and then the determination is pending Whether the area is a noisy area includes: using the image segmentation method of area growth, randomly selecting the initial point to segment the point cloud mapping image to obtain an area, deleting an area obtained by segmentation in the point cloud mapping image, and continuing to randomly select the initial point-to-point cloud mapping image Perform segmentation and repeat the loop until all areas are segmented. The area with the smaller segmentation area is judged as a noise area, the area with the smaller segmentation area is deleted, and the larger segmentation area is judged as the reference area of the noise-free point cloud. The other divided areas are determined as pending areas.

Further, by calculating the relationship between the pending area and the reference area, dividing the pending area into the internal and external conditions of the reference area for processing respectively includes: filling the reference area with holes, and marking the pixel coordinates of the reference area after the holes are filled Point (u, v), use the k-nearest neighbor algorithm to calculate the nearest neighbor of any pixel coordinate point (u, v) of the to-be-determined area and all the pixel coordinate points (u, v) of the reference area after filling the hole, if the nearest If the distance between neighboring points is 0, it means the area to be determined is in the reference area and recorded as an internal area; if the distance between the nearest neighbors is greater than 0, it means the area to be determined is outside the reference area and recorded as an external area.

Further, in the internal case, the distance between any point in the to-be-determined area and any nearest neighbor point in the reference area is calculated, and in the external case, the distance between any point on the contour of the to-be-determined area and any nearest neighbor point on the contour of the reference area is calculated. , The larger distance is the noise area. Removing the noise point cloud by deleting the location of the noise area includes: when the area to be determined is in the reference area, that is, the internal area, using the k-nearest neighbor algorithm to calculate any pixel of the area to be determined ( u,v) and the N neighbors of all pixels (u,v) in the reference area after the hole is filled, calculate the Z value corresponding to any pixel (u,v) in the to-be-determined area and any selected pixel ( u,v) The corresponding Euclidean distance of the Z value is greater than the set threshold, indicating that it is a noise point cloud, delete this pending area; when the pending area is outside the reference area, that is, the external area, calculate the reference area and the pending area respectively The pixel coordinate point (u, v) of the contour point, using the k-nearest neighbor algorithm to calculate one of all the pixel coordinate points (u, v) on the contour of the to-be-determined area and all the pixel coordinate points (u, v) on the contour of the reference area Nearest neighbor, according to the distance of the found nearest neighbor, select the pixel on the contour of the reference area that is closest to the pixel on the contour of the to-be-determined area, and calculate the Z value corresponding to the pixel point (u, v) on the selected contour of the to-be-determined area and The Euclidean distance of the Z value corresponding to the pixel point (u, v) on the contour of the reference area is greater than the set threshold, indicating that it is a noise point cloud, delete this pending area; use the point cloud mapping image after the noise pending area is deleted to the absolute phase map Dot multiplication obtains the absolute phase map of the noise-free point cloud, and finally uses the absolute phase map and the internal and external parameters calibrated by the camera and projector to reconstruct the three-dimensional point cloud to obtain a noise-free three-dimensional point cloud.

In the second aspect, an embodiment of the present invention also proposes a point cloud denoising system based on image segmentation, including:

Constructing a mapping module to perform a one-to-one correspondence between the Z value of the three-dimensional reconstructed three-dimensional point and the absolute phase map to obtain a point cloud mapping image;

The image segmentation module is used to perform image segmentation on the point cloud mapping image by using the image segmentation method of region growth until all regions are segmented, and the larger region is defined as the reference region, and the other regions are defined as pending regions, and then Determine whether the area to be determined is a noise area;

The area division module is used to divide the area to be determined into the internal and external conditions of the reference area by calculating the relationship between the area to be determined and the reference area for processing;

The noise removal module is used to calculate the distance between any point in the to-be-determined area and any nearest neighbor point in the reference area when it is located in the interior, and to calculate any point on the contour of the to-be-determined area and any nearest neighbor point on the contour of the reference area in the external case The larger the distance is the noise area, and the noise point cloud is removed by deleting the location of the noise area.

In the third aspect, an embodiment of the present invention also proposes a point cloud denoising device based on image segmentation, including:

At least one processor; and,

A memory communicatively connected with the at least one processor; wherein,

The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor, so that the at least one processor can execute the method according to the first aspect of the present invention.

In the fourth aspect, the embodiment of the present invention also provides a computer-readable storage medium, the computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to make a computer execute the first aspect of the present invention. The method described.

One or more technical solutions provided in the embodiments of the present invention have at least the following beneficial effects: The present invention provides a point cloud denoising method, system, device, and storage medium based on image segmentation. A one-to-one correspondence between the Z value of the three-dimensional point and the absolute phase map is performed to obtain a point cloud mapping image. Then the image segmentation method of region growing is used to segment the point cloud mapping image until all regions are segmented. Define a larger area as a reference area, define other areas as a pending area, and then determine whether the pending area is a noise area. By calculating the relationship between the to-be-determined area and the reference area, the to-be-determined area is divided into the internal and external conditions of the reference area and processed separately. For the internal case, the distance between any point in the to-be-determined area and any nearest neighbor in the reference area is calculated, In the external case, the distance between any point on the contour of the to-be-determined area and any nearest neighbor on the contour of the reference area is calculated. The larger distance is the noise area, and the noise point cloud is removed by deleting the position of the noise area. This method effectively removes all the noise point clouds on the point cloud, whether it is outliers, scattered points, large noise points floating near the subject, etc., can be removed one by one, and this method is based on the image. The processing avoids the complexity of three-dimensional point calculation, and the method is simple and reliable, and is suitable for any point cloud noise removal using absolute phase for three-dimensional reconstruction.

Description of the drawings

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

Figure 1 is a schematic flow chart of a point cloud denoising method based on image segmentation in the first embodiment of the present invention;

Figure 2 is a schematic structural diagram of a point cloud denoising system based on image segmentation in a second embodiment of the present invention;

Fig. 3 is a schematic structural diagram of a point cloud denoising device based on image segmentation in a third 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.

The embodiments of the present invention will be further described below in conjunction with the accompanying drawings.

As shown in Figure 1, the first embodiment of the present invention provides a point cloud denoising method based on image segmentation, including but not limited to the following steps:

S100: Perform a one-to-one correspondence between the Z value of the three-dimensional point after three-dimensional reconstruction and the absolute phase map to obtain a point cloud mapping image;

S200: Perform image segmentation on the point cloud mapping image by using the region growing image segmentation method until all regions are segmented, and define a larger region as a reference region, define other regions as a pending region, and then determine whether the pending region is Is a noisy area;

S300: By calculating the relationship between the to-be-determined area and the reference area, the to-be-determined area is divided into the internal and external conditions of the reference area and processed separately;

S400: Calculate the distance between any point in the to-be-determined area and any nearest neighbor point in the reference area for the internal case, and calculate the distance between any point on the contour of the to-be-determined area and any nearest neighbor point on the contour of the reference area for the external case. The larger is the noise area, and the noise point cloud is removed by deleting the location of the noise area.

Specifically, the absolute phase map and the internal and external parameters calibrated by the camera and the projector are used to reconstruct the three-dimensional point cloud. Since the three-dimensional point is reconstructed according to each point of the absolute phase map in the image plane, each point of the absolute phase map in the image plane has a one-to-one correspondence with the reconstructed three-dimensional point.

One-to-one correspondence between the Z coordinates of the three-dimensional points (X, Y, Z) reconstructed in three dimensions and the two-dimensional points (u, v) of the absolute phase map to construct a point cloud mapping image, that is, the point cloud mapping image plane The pixel value of each pixel point (u, v) of is the value of the three-dimensional point Z, that is, the point cloud mapping image contains both the position information of the absolute phase map and the information of the three-dimensional point Z.

Using the region growing image segmentation method, the initial point is randomly selected to segment the point cloud mapping image to obtain a region.

Delete an area obtained by segmentation in the point cloud mapping image, and continue to randomly select the initial point to segment the point cloud mapping image. The cycle repeats until all areas are divided.

The area with the smaller segmentation area is judged as a noise area, and the area with the smaller segmentation area is deleted.

The larger segmented area is determined as the reference area of the noise-free point cloud, and the other segmented areas are determined as the pending area, that is, it may be a noise point cloud area or a noise-free point cloud area.

Fill the reference area with holes, and mark the pixel coordinates (u, v) of the reference area after the holes are filled. This is to judge whether the points in the pending area are within the range of the reference area. If there are holes, it cannot be judged. Therefore, the holes in the reference area must be filled first.

Use the k-nearest neighbor algorithm to calculate the nearest neighbor of any pixel coordinate point (u, v) in the area to be determined and all the pixel coordinate points (u, v) in the reference area after filling the hole. If the distance between the nearest neighbor points is 0, It means that the area to be determined is in the reference area and is recorded as an internal area; if the distance between the nearest neighbors is greater than 0, it means that the area to be determined is outside the reference area and recorded as an external area.

When the area to be determined is in the reference area, that is, the internal area, the k-nearest neighbor algorithm is used to calculate any pixel (u, v) of the area to be determined and all pixels (u, v) of the reference area after filling the hole. Neighbor points (Note: N is the number of pixels in the area to be determined plus one, because the area to be determined is in the reference area after the hole is filled. In order to avoid the calculated neighbor point being its own point, adding one to ensure that even the previous calculations are all for itself Point, the last point is also a point in the reference area), take any pixel point (u, v) with a distance greater than 0 (excluding the location of the pixel point of the pending area itself), and calculate any pixel point (u, v) in the pending area ) The Euclidean distance between the corresponding Z value and the Z value corresponding to any selected pixel (u, v). If it is greater than the set threshold, it is a noise point cloud. Delete this pending area.

When the area to be determined is outside the reference area, that is, in the case of an external area, the pixel coordinate points (u, v) of the contour points of the reference area and the area to be determined are calculated respectively.

Use the k-nearest neighbor algorithm to calculate a nearest neighbor of all pixel coordinate points (u, v) on the contour of the to-be-determined area and all pixel coordinate points (u, v) on the contour of the reference area. According to the distance of the nearest neighbors found, Select the pixels on the contour of the reference area that are closest to the pixels on the contour of the area to be determined.

Calculate the Euclidean distance between the Z value corresponding to the pixel point (u, v) on the contour of the selected area to be determined and the Z value corresponding to the pixel point (u, v) on the contour of the reference area. If it is greater than the set threshold, it is a noise point cloud. Delete this pending area.

Use the point cloud mapping image after removing the noise to be determined area to multiply the absolute phase map to obtain the absolute phase map of the noise-free point cloud, and finally use the absolute phase map and the internal and external parameters calibrated by the camera and projector to reconstruct the three-dimensional point cloud to obtain 3D point cloud of noise.

In summary, compared with the prior art, the point cloud denoising method based on image segmentation has the advantage that: firstly, a point cloud is obtained by one-to-one correspondence between the Z value of the three-dimensional point after three-dimensional reconstruction and the absolute phase map. Map the image. Then the image segmentation method of region growing is used to segment the point cloud mapping image until all regions are segmented. Define a larger area as a reference area, define other areas as a pending area, and then determine whether the pending area is a noise area. By calculating the relationship between the to-be-determined area and the reference area, the to-be-determined area is divided into the internal and external conditions of the reference area and processed separately. For the internal case, the distance between any point in the to-be-determined area and any nearest neighbor in the reference area is calculated, In the external case, the distance between any point on the contour of the to-be-determined area and any nearest neighbor on the contour of the reference area is calculated. The larger distance is the noise area, and the noise point cloud is removed by deleting the position of the noise area. This method effectively removes all the noise point clouds on the point cloud, whether it is outliers, scattered points, large noise points floating near the subject, etc., can be removed one by one, and this method is based on the image. The processing avoids the complexity of three-dimensional point calculation, and the method is simple and reliable, and is suitable for any point cloud noise removal using absolute phase for three-dimensional reconstruction.

In addition, as shown in FIG. 2, the second embodiment of the present invention provides a point cloud denoising system based on image segmentation, including:

Constructing a mapping module 110 for performing one-to-one correspondence between the Z value of the three-dimensional reconstructed three-dimensional point and the absolute phase map to obtain a point cloud mapping image;

The image segmentation module 120 is configured to perform image segmentation on the point cloud mapping image by using the image segmentation method of region growth until all regions are segmented, and define a larger region as a reference region, and define other regions as pending regions, Then determine whether the area to be determined is a noise area;

The area division module 130 is configured to divide the area to be determined into the internal and external conditions of the reference area by calculating the relationship between the area to be determined and the reference area for processing respectively;

The noise removal module 140 is used to calculate the distance between any point in the to-be-determined area and any nearest neighbor in the reference area when it is located in the interior, and to calculate the distance between any point on the contour of the to-be-determined area and any nearest neighbor on the contour of the reference area in the external case The distance between the points, the larger the distance is the noise area, and the noise point cloud is removed by deleting the location of the noise area.

The point cloud denoising system based on image segmentation in this embodiment is based on the same inventive concept as the point cloud denoising method based on image segmentation in the first embodiment. Therefore, the point cloud denoising system based on image segmentation in this embodiment is The noise system has the same beneficial effects: firstly, a point cloud mapping image is obtained by one-to-one correspondence between the Z value of the three-dimensional point after the three-dimensional reconstruction and the absolute phase map. Then the image segmentation method of region growing is used to segment the point cloud mapping image until all regions are segmented. Define a larger area as a reference area, define other areas as a pending area, and then determine whether the pending area is a noise area. By calculating the relationship between the to-be-determined area and the reference area, the to-be-determined area is divided into the internal and external conditions of the reference area and processed separately. For the internal case, the distance between any point in the to-be-determined area and any nearest neighbor in the reference area is calculated, In the external case, the distance between any point on the contour of the to-be-determined area and any nearest neighbor on the contour of the reference area is calculated. The larger distance is the noise area, and the noise point cloud is removed by deleting the position of the noise area. This method effectively removes all the noise point clouds on the point cloud, whether it is outliers, scattered points, large noise points floating near the subject, etc., can be removed one by one, and the system is based on the image. The processing avoids the complexity of three-dimensional point calculation, and the system is simple and reliable, and is suitable for any point cloud noise removal using absolute phase for three-dimensional reconstruction.

As shown in FIG. 3, the third embodiment of the present invention also provides a point cloud denoising device based on image segmentation, including:

At least one processor;

And a memory communicatively connected with the at least one processor;

Wherein, the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor, so that the at least one processor can execute any one of the instructions in the first embodiment. A point cloud denoising method based on image segmentation.

As a non-transitory computer-readable storage medium, the memory can be used to store non-transitory software programs, non-transitory computer-executable programs and modules, such as program instructions/modules corresponding to the virtual image control method in the embodiment of the present invention . The processor executes various functional applications and data processing of the stereo imaging processing device by running non-transitory software programs, instructions, and modules stored in the memory, that is, to realize the point cloud removal based on image segmentation in any of the above-mentioned method embodiments. Noise method.

The memory may include a storage program area and a storage data area, where the storage program area can store an operating system and an application program required by at least one function; the storage data area can store data created according to the use of the stereo imaging processing device, 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 processor, and these remote memories may be connected to the stereoscopic projection apparatus 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 one or more processors, the point cloud denoising method based on image segmentation in any of the foregoing method embodiments is executed, for example, the first embodiment The method steps S100 to S400 in.

The fourth embodiment of the present invention also provides a computer-readable storage medium that stores computer-executable instructions that are executed by one or more control processors to enable the foregoing One or more processors execute a point cloud denoising method based on image segmentation in the foregoing method embodiments, for example, the method steps S100 to S400 in the first embodiment.

The device embodiments described above are merely illustrative, and the units 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 units. Some or all of the modules can be selected according to actual needs to achieve the objectives of the solutions of the embodiments.

Through the description of the above implementation manners, those of ordinary skill in the art can clearly understand that each implementation manner can be implemented by means of software plus a general hardware platform, and of course, it can also be implemented by hardware. A person of ordinary skill in the art can understand that all or part of the processes in the methods of the foregoing embodiments can be implemented by instructing relevant hardware through a computer program. The program can be stored in a computer readable storage medium, and the program can be stored in a computer readable storage medium. When executed, it may include the procedures of the above-mentioned method embodiments. 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 point cloud denoising method based on image segmentation, which is characterized in that it includes:

According to the one-to-one correspondence between the Z value of the three-dimensional point after three-dimensional reconstruction and the absolute phase map, a point cloud mapping image is obtained;

The image segmentation method of region growth is used to perform image segmentation on the point cloud mapping image until all regions are segmented, and the larger region is defined as the reference region, and the other regions are defined as the pending region, and then it is judged whether the pending region is noise area;

By calculating the relationship between the to-be-determined area and the reference area, the to-be-determined area is divided into the internal and external conditions of the reference area and processed separately;

In the internal case, calculate the distance between any point in the to-be-determined area and any nearest neighbor point in the reference area, and in the external case, calculate the distance between any point on the contour of the to-be-determined area and any nearest neighbor point on the contour of the reference area, the distance is larger The noise area is the noise area, and the noise point cloud is removed by deleting the location of the noise area.
A point cloud denoising method based on image segmentation according to claim 1, wherein the Z value of the three-dimensional point after the three-dimensional reconstruction is in a one-to-one correspondence with the absolute phase map to obtain a point cloud mapping The image includes: using the absolute phase map and the internal and external parameters calibrated by the camera and projector to reconstruct the three-dimensional point cloud, the Z coordinate of the three-dimensional point (X, Y, Z) reconstructed from the three-dimensional and the two-dimensional of the absolute phase map The points (u, v) correspond one-to-one to construct a point cloud mapping image, and the pixel value of each pixel point (u, v) in the point cloud mapping image plane is the value of the three-dimensional point Z.
The point cloud denoising method based on image segmentation according to claim 1, characterized in that the image segmentation method using region growth cyclically performs image segmentation on the point cloud mapping image until all regions are segmented. And define a larger area as a reference area, and define other areas as a pending area, and then determine whether the pending area is a noise area includes: using the image segmentation method of area growth, randomly selecting the initial point to segment the point cloud mapping image to obtain an area, Delete an area obtained by segmentation in the point cloud mapping image, continue to randomly select the initial point to segment the point cloud mapping image, and repeat until all areas are segmented. The area with the smaller segmentation area is judged as a noise area, and the segmentation area is deleted For the smaller area, the larger segmented area is determined as the reference area of the noise-free point cloud, and the other segmented areas are determined as the pending area.
The method of point cloud denoising based on image segmentation according to claim 1, characterized in that, by calculating the relationship between the pending area and the reference area, the pending area is divided into internal and external conditions of the reference area. The processing includes: filling the reference area with holes, marking the pixel coordinate points (u, v) of the reference area after the hole filling, using the k-nearest neighbor algorithm to calculate any pixel coordinate point (u, v) of the pending area and filling the hole A nearest neighbor of all pixel coordinate points (u, v) in the reference area. If the distance of the nearest neighbor point is 0, it means that the area to be determined is in the reference area, and it is recorded as an internal area; if the distance of the nearest neighbor point is greater than 0 , Indicating that the pending area is outside the reference area and recorded as an external area.
The method of point cloud denoising based on image segmentation according to claim 1, characterized in that the distance between any point in the to-be-determined area and any nearest neighbor in the reference area is calculated for the internal situation, and for the external situation, the distance between any point in the to-be-determined region and any nearest neighbor in the reference region is calculated. Calculate the distance between any point on the contour of the to-be-determined area and any nearest neighbor on the contour of the reference area. The larger the distance is the noise area, and the removal of the noise point cloud by deleting the location of the noise area includes: when the area to be determined is in the reference area In the case of the internal area, use the k-nearest neighbor algorithm to calculate the N neighbors of any pixel (u, v) in the area to be determined and all pixels (u, v) in the reference area after filling the hole to calculate the area to be determined The Euclidean distance between the Z value corresponding to any pixel (u, v) and the Z value corresponding to any selected pixel (u, v), if it is greater than the set threshold, it is a noise point cloud, delete this pending area; When the area to be determined is outside the reference area, that is, in the case of the outer area, the pixel coordinate points (u, v) of the contour points of the reference area and the area to be determined are calculated respectively, and all the pixel coordinate points (u, v) on the contour of the area to be determined are calculated using the k-nearest neighbor algorithm ,v) A nearest neighbor of all pixel coordinate points (u,v) on the contour of the reference area. According to the distance of the found nearest neighbor, select the pixel on the contour of the reference area that is closest to the pixel on the contour of the area to be determined , Calculate the Euclidean distance between the Z value corresponding to the pixel point (u, v) on the contour of the selected area to be determined and the Z value corresponding to the pixel point (u, v) on the contour of the reference area. If it is greater than the set threshold, it is a noise point cloud , Delete this pending area; use the point cloud mapping image after deleting the noise pending area to multiply the absolute phase map to obtain the absolute phase map of the noise-free point cloud, and finally use the absolute phase map and the internal and external parameters calibrated by the camera and projector to make three-dimensional points The reconstruction of the cloud results in a noise-free three-dimensional point cloud.
A point cloud denoising system based on image segmentation, which is characterized in that it includes:

Constructing a mapping module to perform a one-to-one correspondence between the Z value of the three-dimensional reconstructed three-dimensional point and the absolute phase map to obtain a point cloud mapping image;

The image segmentation module is used to perform image segmentation on the point cloud mapping image by using the image segmentation method of region growth until all regions are segmented, and the larger region is defined as the reference region, and the other regions are defined as pending regions, and then Determine whether the area to be determined is a noise area;

The area division module is used to divide the area to be determined into the internal and external conditions of the reference area by calculating the relationship between the area to be determined and the reference area for processing;

The noise removal module is used to calculate the distance between any point in the to-be-determined area and any nearest neighbor point in the reference area when it is located in the interior, and to calculate any point on the contour of the to-be-determined area and any nearest neighbor point on the contour of the reference area in the external case The larger the distance is the noise area, and the noise point cloud is removed by deleting the location of the noise area.
A point cloud denoising device based on image segmentation, which is characterized in that it comprises:

At least one processor; and,

A memory communicatively connected with the at least one processor; wherein,

The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor, so that the at least one processor can execute any one of claims 1-5 Methods.
A computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to make a computer execute the method according to any one of claims 1-5 .