A point cloud noise filtering method, apparatus and medium

Abstract

This application discloses a point cloud noise filtering method, comprising the following steps: acquiring a first point cloud image of a three-dimensional object; acquiring a set plane, and projecting the first point cloud image onto the set plane to obtain a binary image; acquiring first position information of the points to be filtered in the binary image; acquiring second position information of the points to be filtered in the first point cloud image based on the correspondence between points on the binary image and points on the first point cloud image; and filtering out the points to be filtered in the first point cloud image based on the second position information to obtain a filtered point cloud image. This point cloud noise filtering method has higher noise filtering efficiency.

Description

Technical Field

[0001] This application generally relates to the field of three-dimensional data measurement technology, and in particular to a point cloud noise filtering method, apparatus and medium. Background Technology

[0002] To obtain data about a 3D object, measuring instruments such as LiDAR are typically used to collect point data from the object's outer surface. This collected point data is called a point cloud. After obtaining the point cloud, it needs to be filtered to remove noise and outliers before accurate information about the 3D object's surface can be obtained in subsequent data processing. Current point cloud filtering methods generally involve acquiring initial 3D point cloud data using measuring instruments and then processing this data to remove noise and outliers. However, processing systems for 3D data are relatively slow. When the 3D object surface is complex, meaning the corresponding point cloud data volume is large, the noise filtering efficiency will be low, thus affecting the real-time performance of subsequent processing. Summary of the Invention

[0003] In view of the above-mentioned defects or deficiencies in the prior art, it is desirable to provide a point cloud noise filtering method, apparatus and medium with higher noise filtering efficiency.

[0004] The specific technical solution is as follows:

[0005] First aspect

[0006] This application provides a point cloud noise filtering method, including the following steps:

[0007] Obtain the first point cloud image of the 3D object;

[0008] Obtain a set plane, and project the first point cloud image onto the set plane to obtain its binary image;

[0009] Obtain the first position information of the points to be filtered out in the binary image;

[0010] Based on the correspondence between each point on the binary image and each point on the first point cloud image, the second position information of the points to be filtered out in the first point cloud image is obtained.

[0011] Based on the second location information, the points to be filtered out in the first point cloud image are filtered out to obtain a filtered point cloud image.

[0012] As a further limitation of this application, the process of obtaining the designated plane specifically includes the following steps:

[0013] Before acquiring the first point cloud image of a 3D object, the measuring instrument is used to scan other 3D objects to acquire the first test point cloud image of the 3D object.

[0014] Observe the position distribution of the points to be filtered in the first test point cloud image, and obtain the set plane according to the position distribution. The set plane is the plane in which the projection point corresponding to the point to be filtered has the lowest overlap with the projection point of the three-dimensional object after the first test point cloud image is projected into the plane.

[0015] As a further limitation of this application, obtaining the first location information of the points to be filtered out in the binary image specifically includes the following steps:

[0016] The binary image is statistically analyzed and quantized to obtain a set of quantized values ​​for each point in the binary image.

[0017] Obtain a set range, and filter out points in the binary image whose quantized values ​​are outside the set range, and record them as filter points;

[0018] The location information of the selected points in the binary image is obtained, which is the first location information.

[0019] As a further limitation of this application, the process of obtaining the set range specifically includes the following steps:

[0020] Before acquiring the set of quantized values ​​for each point in the binary image, the measuring instrument is used to scan other three-dimensional objects to acquire a second test point cloud image of the three-dimensional objects.

[0021] The first point cloud image is projected onto a set plane to obtain its first test binary image;

[0022] The test binary image is statistically analyzed and quantized to obtain a set of quantized values ​​for each test point in the test binary image;

[0023] The set range is defined as the points in the test binary image where the quantized value falls between the first set value and the second set value.

[0024] The values ​​of the first set value and the second set value are continuously changed, and then the second test binary images are obtained after the points outside the above range are filtered out. The processing time required to obtain each second test binary image is recorded.

[0025] Based on the number of points to be filtered in each of the second test binary images and the corresponding processing time, the values ​​of the first set value and the second set value that best balance the number of points to be filtered and the corresponding processing time are obtained, and the set range is obtained using the values.

[0026] As a further limitation of this application, after statistical analysis and quantization of the binary image, the quantized values ​​of each point conform to a Gaussian distribution.

[0027] Second aspect

[0028] This application provides a point cloud noise filtering device, comprising:

[0029] The first point cloud image acquisition module is used to acquire the first point cloud image of a 3D object.

[0030] The plane acquisition module is used to acquire the specified plane.

[0031] The binary image acquisition module is used to project the first point cloud image onto a set plane to obtain its binary image;

[0032] The first location information acquisition module is used to acquire the first location information of the points to be filtered out in the binary image;

[0033] The second location information acquisition module is used to acquire the second location information of the points to be filtered in the first point cloud image based on the correspondence between each point on the binary image and each point on the first point cloud image.

[0034] The filtered point cloud image acquisition module is used to filter out the points to be filtered out in the first point cloud image based on the second location information, so as to obtain a filtered point cloud image.

[0035] As a further limitation of this application, the setting plane acquisition module includes:

[0036] The first test point cloud image acquisition unit is used to scan other three-dimensional objects with the measuring instrument before acquiring the first point cloud image of the three-dimensional object, so as to acquire the first test point cloud image of the three-dimensional object; wherein the first test point cloud image is used to display the position distribution of the points to be filtered out, so as to acquire the set plane according to the position distribution, wherein the set plane is the plane with the lowest overlap between the projection point corresponding to the point to be filtered out and the projection point of the three-dimensional object after the first test point cloud image is projected into the plane.

[0037] As a further limitation of this application, the first location information acquisition module includes:

[0038] The point quantization value set acquisition unit is used to perform statistics and quantization on the binary image to obtain the point quantization value set in the binary image;

[0039] A set range acquisition unit is used to acquire a set range;

[0040] The filtering point acquisition unit is used to filter out points in the binary image whose quantization values ​​are outside the set range, based on the correspondence between each point in the binary image and each quantization value in the set of quantization values ​​of each point, and record them as filtering points;

[0041] The first location information acquisition unit is used to acquire the location information of the filter point in the binary image, which is the first location information.

[0042] As a further limitation of this application, the set range acquisition unit includes:

[0043] The second test point cloud image acquisition mechanism is used to scan other three-dimensional objects with the measuring instrument before acquiring the set of quantized values ​​of each point in the binary image, so as to acquire the second test point cloud image of the three-dimensional object.

[0044] The first test binary image acquisition mechanism is used to project the first point cloud image onto a set plane to obtain its first test binary image.

[0045] A set of quantized values ​​for each test point is used to perform statistics and quantization on the test binary image to obtain a set of quantized values ​​for each test point in the test binary image.

[0046] A range acquisition mechanism is configured to, based on the number of points to be filtered out in each of the second test binary images and the corresponding processing time, acquire the value of the first set value and the second set value that best balance the number of points to be filtered out and the corresponding processing time, and acquire the set range using the value. The set range is defined as the point in the test binary image whose quantization value is located between the first set value and the second set value. The first set value and the second set value are continuously changed, and each of the second test binary images obtained after filtering out the points outside the range is acquired. The processing time required to acquire each of the second test binary images is recorded.

[0047] Third aspect

[0048] This application provides a point cloud noise filtering medium for implementing the point cloud noise filtering method described above.

[0049] The beneficial effects of this application are:

[0050] This solution projects 3D point cloud data onto a predetermined plane to obtain a binary image of the 3D point cloud. Then, by applying predefined rules to filter out points in the binary image, the first location information of these points can be obtained. Compared to existing methods that process 3D data using specific rules, this solution offers a faster processing speed. After obtaining the first location information, the second location information of the points to be filtered out within the first point cloud image can be obtained through the correspondence between the first point cloud image and the binary image. Using this second location information to filter out the points in the first point cloud image yields the filtered point cloud image. This solution features a simpler calculation method and higher noise reduction efficiency. Attached Figure Description

[0051] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0052] Figure 1 A flowchart illustrating the steps of the point cloud noise filtering method provided in this application embodiment;

[0053] Figure 2 This is a point cloud image before noise filtering.

[0054] Figure 3 This is a diagram of the point cloud after noise filtering. Detailed Implementation

[0055] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0056] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0057] Example 1

[0058] Please refer to Figure 1 The flowchart of a point cloud noise filtering method provided in this embodiment includes the following steps:

[0059] S1: Obtain the first point cloud image of the 3D object;

[0060] S2: Obtain a set plane, and project the first point cloud image into the set plane to obtain its binary image;

[0061] S3: Obtain the first position information of the points to be filtered out in the binary image;

[0062] S4: Based on the correspondence between each point on the binary image and each point on the first point cloud image, obtain the second position information of the points to be filtered out in the first point cloud image;

[0063] S5: Based on the second location information, filter out the points to be filtered in the first point cloud image to obtain a filtered point cloud image.

[0064] This solution projects 3D point cloud data onto a predetermined plane to obtain a binary image of the 3D point cloud. Then, by applying predefined rules to filter out points in the binary image, the first location information of these points can be obtained. Compared to existing methods that process 3D data using specific rules, this solution offers a faster processing speed. After obtaining the first location information, the second location information of the points to be filtered out within the first point cloud image can be obtained through the correspondence between the first point cloud image and the binary image. Using this second location information to filter out the points in the first point cloud image yields the filtered point cloud image. This solution features a simpler calculation method and higher noise reduction efficiency.

[0065] In a preferred embodiment that further improves noise filtering efficiency, obtaining the set plane specifically includes the following steps:

[0066] Before acquiring the first point cloud image of a 3D object, the measuring instrument is used to scan other 3D objects to acquire the first test point cloud image of the 3D object.

[0067] Observe the position distribution of the points to be filtered in the first test point cloud image, and obtain the set plane according to the position distribution. The set plane is the plane in which the projection point corresponding to the point to be filtered has the lowest overlap with the projection point of the three-dimensional object after the first test point cloud image is projected into the plane.

[0068] When using measuring instruments to collect point clouds, due to the inherent characteristics of these instruments, the collected point clouds will generate noise points and outliers with increased density in certain areas, and this property will not change even for different 3D objects. Therefore, before acquiring the binary image of the 3D object to be measured, the measuring instrument needs to be pre-tested according to the method described above to obtain the set plane that facilitates the differentiation between 3D object points and points to be filtered after projection. Projecting onto this plane will be most effective in filtering out points to be filtered from the point cloud, thus this solution further improves the noise filtering efficiency.

[0069] In a preferred embodiment of obtaining the first location information of the points to be filtered out in the binary image, the step of obtaining the first location information of the points to be filtered out in the binary image specifically includes the following steps:

[0070] The binary image is statistically analyzed and quantized to obtain a set of quantized values ​​for each point in the binary image.

[0071] Obtain a set range, and filter out points in the binary image whose quantized values ​​are outside the set range, and record them as filter points;

[0072] The location information of the selected points in the binary image is obtained, which is the first location information.

[0073] By statistically analyzing and quantifying the location information of the points to be filtered out, it can be transformed into an image with a set distribution characteristics. The characteristics of this image can then be used to filter out the points to be filtered out.

[0074] The process of obtaining the set range specifically includes the following steps:

[0075] Before acquiring the set of quantized values ​​for each point in the binary image, the measuring instrument is used to scan other three-dimensional objects to acquire a second test point cloud image of the three-dimensional objects.

[0076] The first point cloud image is projected onto a set plane to obtain its first test binary image;

[0077] The test binary image is statistically analyzed and quantized to obtain a set of quantized values ​​for each test point in the test binary image;

[0078] The set range is defined as the points in the test binary image where the quantized value falls between the first set value and the second set value.

[0079] The values ​​of the first set value and the second set value are continuously changed, and then the second test binary images are obtained after the points outside the above range are filtered out. The processing time required to obtain each second test binary image is recorded.

[0080] Based on the number of points to be filtered in each of the second test binary images and the corresponding processing time, the values ​​of the first set value and the second set value that best balance the number of points to be filtered and the corresponding processing time are obtained, and the set range is obtained using the values.

[0081] According to the above solution, it is possible to determine the approximate position of the points to be filtered in each point distribution image formed after the point cloud with points to be filtered is statistically analyzed and quantified. In order to ensure that the number of points to be filtered in the point cloud obtained after filtering is not excessive and the overall processing efficiency is not affected. Therefore, it is necessary to pre-test the above characteristics of the measuring instrument before testing to prepare for more efficiently obtaining a point cloud with fewer points to be filtered based on this distribution characteristic in the follow-up.

[0082] After statistically analyzing and quantifying the binary image, the quantization values of each point conform to the Gaussian distribution.

[0083] The specific operation process is as follows:

[0084] Calculate the maximum value max_z and the minimum value min_z of the point cloud data z;

[0085] Quantize the z value of the point cloud data uniformly, with each quantization value interval being (max_z – min_z) / 5000;

[0086] Statistically analyze the number of points corresponding to each quantization value after quantization;

[0087] Calculate the maximum quantization value min_valz corresponding to the lowest 5% of the points and the minimum quantization value max_valz corresponding to the highest 5%;

[0088] Use a band-pass filter to filter out the values less than min_valz and greater than max_valz;

[0089] Set the memory area storing the binary projection to 0;

[0090] Project the filtered point cloud in the yz plane direction, calculate the corresponding position in the image after projection, set the position in the image to 1, and record the position in the image corresponding to this point;

[0091] Perform median filtering on the projected binary image;

[0092] Filter out the points in the point cloud corresponding to the points filtered out in the binary image to complete the filtering process.

[0093] The effects of the point cloud before and after processing are as Figure 2 and Figure 3 shown.

[0094] Embodiment 2

[0095] A point cloud noise filtering device provided in this embodiment includes:

[0096] A first point cloud image acquisition module for acquiring a first point cloud image of a three-dimensional object;

[0097] A set plane acquisition module for acquiring a set plane;

[0098] The binary image acquisition module is used to project the first point cloud image onto a set plane to obtain its binary image;

[0099] The first location information acquisition module is used to acquire the first location information of the points to be filtered out in the binary image;

[0100] The second location information acquisition module is used to acquire the second location information of the points to be filtered in the first point cloud image based on the correspondence between each point on the binary image and each point on the first point cloud image.

[0101] The filtered point cloud image acquisition module is used to filter out the points to be filtered out in the first point cloud image based on the second location information, so as to obtain a filtered point cloud image.

[0102] The defined plane acquisition module includes:

[0103] The first test point cloud image acquisition unit is used to scan other three-dimensional objects with the measuring instrument before acquiring the first point cloud image of the three-dimensional object, so as to acquire the first test point cloud image of the three-dimensional object; wherein the first test point cloud image is used to display the position distribution of the points to be filtered out, so as to acquire the set plane according to the position distribution, wherein the set plane is the plane with the lowest overlap between the projection point corresponding to the point to be filtered out and the projection point of the three-dimensional object after the first test point cloud image is projected into the plane.

[0104] The first location information acquisition module includes:

[0105] The point quantization value set acquisition unit is used to perform statistics and quantization on the binary image to obtain the point quantization value set in the binary image;

[0106] A set range acquisition unit is used to acquire a set range;

[0107] The filtering point acquisition unit is used to filter out points in the binary image whose quantization values ​​are outside the set range, based on the correspondence between each point in the binary image and each quantization value in the set of quantization values ​​of each point, and record them as filtering points;

[0108] The first location information acquisition unit is used to acquire the location information of the filter point in the binary image, which is the first location information.

[0109] The set range acquisition unit includes:

[0110] The second test point cloud image acquisition mechanism is used to scan other three-dimensional objects with the measuring instrument before acquiring the set of quantized values ​​of each point in the binary image, so as to acquire the second test point cloud image of the three-dimensional object.

[0111] The first test binary image acquisition mechanism is used to project the first point cloud image onto a set plane to obtain its first test binary image.

[0112] A set of quantized values ​​for each test point is used to perform statistics and quantization on the test binary image to obtain a set of quantized values ​​for each test point in the test binary image.

[0113] A range acquisition mechanism is configured to, based on the number of points to be filtered out in each of the second test binary images and the corresponding processing time, acquire the value of the first set value and the second set value that best balance the number of points to be filtered out and the corresponding processing time, and acquire the set range using the value. The set range is defined as the point in the test binary image whose quantization value is located between the first set value and the second set value. The first set value and the second set value are continuously changed, and each of the second test binary images obtained after filtering out the points outside the range is acquired. The processing time required to acquire each of the second test binary images is recorded.

[0114] Example 3

[0115] This embodiment provides a point cloud noise filtering medium for implementing the point cloud noise filtering method described above.

[0116] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the inventive concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this application.

Claims

1. A point cloud noise filtering method, characterized in that, Includes the following steps: Obtain the first point cloud image of the 3D object; A predetermined plane is obtained, and the first point cloud image is projected onto the predetermined plane to obtain its binary image; the specific steps of obtaining the predetermined plane are as follows: Before acquiring the first point cloud image of a 3D object, other 3D objects are scanned with a measuring instrument to obtain the first test point cloud image of the 3D object. Observe the position distribution of the points to be filtered in the first test point cloud image, and obtain the set plane according to the position distribution. The set plane is the plane in which the projection point corresponding to the point to be filtered has the lowest overlap with the projection point of the three-dimensional object after the first test point cloud image is projected into the plane. Obtain the first position information of the points to be filtered out in the binary image; Based on the correspondence between each point on the binary image and each point on the first point cloud image, the second position information of the points to be filtered out in the first point cloud image is obtained. Based on the second location information, the points to be filtered out in the first point cloud image are filtered out to obtain a filtered point cloud image.

2. The point cloud noise filtering method according to claim 1, characterized in that, The step of obtaining the first location information of the points to be filtered out in the binary image specifically includes the following steps: The binary image is statistically analyzed and quantized to obtain a set of quantized values ​​for each point in the binary image. Obtain a set range, and filter out points in the binary image whose quantized values ​​are outside the set range, and record them as filter points; The location information of the selected points in the binary image is obtained, which is the first location information.

3. The point cloud noise filtering method according to claim 2, characterized in that, The process of obtaining the set range specifically includes the following steps: Before acquiring the set of quantized values ​​for each point in the binary image, the measuring instrument is used to scan other three-dimensional objects to acquire a second test point cloud image of the three-dimensional objects. The first point cloud image is projected onto a set plane to obtain its first test binary image; The test binary image is statistically analyzed and quantized to obtain a set of quantized values ​​for each test point in the test binary image; The set range is defined as the points in the test binary image where the quantized value falls between the first set value and the second set value. The values ​​of the first set value and the second set value are continuously changed, and then the second test binary images are obtained after the points outside the above range are filtered out. The processing time required to obtain each second test binary image is recorded. Based on the number of points to be filtered in each of the second test binary images and the corresponding processing time, the values ​​of the first set value and the second set value that best balance the number of points to be filtered and the corresponding processing time are obtained, and the set range is obtained using the values.

4. The point cloud noise filtering method according to claim 2, characterized in that, After statistical analysis and quantization of the binary image, the quantized values ​​of each point conform to a Gaussian distribution.

5. A point cloud noise filtering device, characterized in that, include: The first point cloud image acquisition module is used to acquire the first point cloud image of a 3D object. The plane acquisition module is used to acquire the specified plane. The binary image acquisition module is used to project the first point cloud image onto a set plane to obtain its binary image; The first location information acquisition module is used to acquire the first location information of the points to be filtered out in the binary image; The second location information acquisition module is used to acquire the second location information of the points to be filtered in the first point cloud image based on the correspondence between each point on the binary image and each point on the first point cloud image. The filtered point cloud image acquisition module is used to filter out the points to be filtered out in the first point cloud image based on the second location information, so as to obtain a filtered point cloud image; The defined plane acquisition module includes: The first test point cloud image acquisition unit is used to scan other three-dimensional objects with a measuring instrument before acquiring the first point cloud image of the three-dimensional object, so as to acquire the first test point cloud image of the three-dimensional object; wherein the first test point cloud image is used to display the position distribution of the points to be filtered out, so as to acquire the set plane according to the position distribution, wherein the set plane is the plane with the lowest overlap between the projection point corresponding to the point to be filtered out and the projection point of the three-dimensional object after the first test point cloud image is projected into the plane.

6. The point cloud noise filtering device according to claim 5, characterized in that, The first location information acquisition module includes: The point quantization value set acquisition unit is used to perform statistics and quantization on the binary image to obtain the point quantization value set in the binary image; A set range acquisition unit is used to acquire a set range; The filtering point acquisition unit is used to filter out points in the binary image whose quantization values ​​are outside the set range, based on the correspondence between each point in the binary image and each quantization value in the set of quantization values ​​of each point, and record them as filtering points; The first location information acquisition unit is used to acquire the location information of the filter point in the binary image, which is the first location information.

7. The point cloud noise filtering device according to claim 6, characterized in that, The set range acquisition unit includes: The second test point cloud image acquisition mechanism is used to scan other three-dimensional objects with the measuring instrument before acquiring the set of quantized values ​​of each point in the binary image, so as to acquire the second test point cloud image of the three-dimensional object. The first test binary image acquisition mechanism is used to project the first point cloud image onto a set plane to obtain its first test binary image. A set of quantized values ​​for each test point is used to perform statistics and quantization on the test binary image to obtain a set of quantized values ​​for each test point in the test binary image. A range acquisition mechanism is used to acquire, based on the number of points to be filtered out in each second test binary image and the corresponding processing time, a first set value and a second set value that provide the best balance between the number of points to be filtered out and the corresponding processing time, and to acquire a set range using the set value. The set range is defined as the points in the test binary image whose quantization values ​​are located between the first set value and the second set value. The first set value and the second set value are continuously changed, and each second test binary image is acquired after the points outside the range are filtered out. The processing time required to acquire each second test binary image is recorded.

8. A point cloud noise filtering medium, characterized in that, Used to implement the point cloud noise filtering method according to any one of claims 1-4.

Images