WO2022198637A1

WO2022198637A1 - Point cloud noise filtering method and system, and movable platform

Info

Publication number: WO2022198637A1
Application number: PCT/CN2021/083286
Authority: WO
Inventors: 朱晏辰; 李延召
Original assignee: 深圳市大疆创新科技有限公司
Priority date: 2021-03-26
Filing date: 2021-03-26
Publication date: 2022-09-29
Also published as: CN116547562A

Abstract

A point cloud noise filtering method and system, and a movable platform, the point cloud noise filtering method comprising: acquiring point cloud data to be processed; performing feature extraction on the point cloud data, and marking point cloud points which satisfy preset features as first target points; marking second target points in the point cloud data according to the degree of aggregation of the point cloud data; and determining overlapping portions between the first target points and the second target points as actual measurement points, and determining at least some of the point cloud data other than the actual measurement points as noisy points. By means of the method, noisy points are screened by means of combining feature extraction and the degree of aggregation, such that information of an actual object which has been subjected to measurement can be effectively retained, and noisy points can also be identified.

Description

Point cloud noise filtering method, system and movable platform

manual

technical field

Embodiments of the present invention relate to the technical field of ranging, and more particularly, to a point cloud noise filtering method, system, and movable platform.

Background technique

Laser ranging devices such as three-dimensional point cloud detection systems such as lidar and laser rangefinders can measure the time of light travel between the ranging device and the measured object, that is, the time-of-flight (TOF) of light. ) to detect the distance from the measured object to the distance measuring device.

Most of the existing laser ranging devices have the problem of decreased measurement accuracy in the near range, especially for objects with low reflectivity, the measurement results usually have large uncertainty in depth, resulting in a lot of near noise; In the specific application of laser ranging devices, there are a lot of situations that are very concerned about near measurement, such as indoor robots, unmanned delivery vehicles, etc. The noise in the vicinity greatly affects the operation of the system, which may cause robots or vehicles to stop suddenly, freeze, freeze, etc. Unable to get out of trouble, etc. In view of this situation, the method of completely filtering out the close-range point cloud cannot be used, otherwise obstacles will be easily missed, which will cause greater security risks.

SUMMARY OF THE INVENTION

A series of concepts in simplified form have been introduced in the Summary section, which are described in further detail in the Detailed Description section. The Summary of the Invention section of the present invention is not intended to attempt to limit the key features and essential technical features of the claimed technical solution, nor is it intended to attempt to determine the protection scope of the claimed technical solution.

In view of the deficiencies of the prior art, the first aspect of the embodiments of the present invention provides a point cloud noise filtering method, including:

Get the point cloud data to be processed;

Perform feature extraction on the point cloud data, and mark the point cloud point that meets the preset characteristics as the first target point;

Mark the second target point in the point cloud data according to the aggregation degree of the point cloud data;

The overlapping part of the first target point and the second target point is determined as a real measurement point, and at least part of the point cloud data other than the real measurement point is determined as a noise point.

A second aspect of the embodiments of the present invention provides a point cloud noise filtering system, including a memory and a processor, where the memory is used for storing program instructions; the processor is used for executing the program instructions stored in the memory, when the When program instructions are executed, the processor is used to:

Get the point cloud data to be processed;

Perform feature extraction on the point cloud data, and determine the point cloud point that meets the preset characteristics as the first target point;

Determine the second target point in the point cloud data according to the aggregation degree of the point cloud data;

A third aspect of the embodiments of the present invention provides a movable platform, including:

Movable platform body;

And, in the above point cloud noise filtering system, the point cloud noise filtering system is mounted on the movable platform body.

A fourth aspect of the embodiments of the present invention provides a computer storage medium on which a computer program is stored, and when the program is executed by a processor, the above-mentioned point cloud noise filtering method is implemented.

The point cloud noise filtering method, system, and movable platform of the embodiments of the present invention use a combination of feature extraction and aggregation to filter noise, which can not only effectively retain the information of the real measured object, but also identify the noise.

Description of drawings

In order to illustrate the technical solutions of the embodiments of the present invention more clearly, the following briefly introduces the accompanying drawings that need to be used in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some of the present invention. In the embodiments, for those of ordinary skill in the art, other drawings can also be obtained according to these drawings without creative labor.

1 is a schematic flowchart of a point cloud noise filtering method according to an embodiment of the present invention;

2 is a schematic diagram of a noise point in a point cloud noise filtering method according to an embodiment of the present invention;

3 is a comparison diagram of a noise point and a non-noise point in a point cloud noise filtering method according to an embodiment of the present invention;

4 is a schematic diagram of a plane point and an edge point in a point cloud noise filtering method according to an embodiment of the present invention;

5 is a schematic diagram of a second target point and a non-second target point according to an embodiment of the present invention;

6 is a schematic diagram of a second target point and a non-second target point according to another embodiment of the present invention;

7 is a schematic diagram of a real measurement point and a noise point according to an embodiment of the present invention;

FIG. 8 is a schematic block diagram of a point cloud noise filtering system according to an embodiment of the present invention.

Detailed ways

In order to make the objects, technical solutions and advantages of the present invention more apparent, exemplary embodiments according to the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some of the embodiments of the present invention, not all of the embodiments of the present invention, and it should be understood that the present invention is not limited by the example embodiments described herein. Based on the embodiments of the present invention described in the present invention, all other embodiments obtained by those skilled in the art without creative efforts shall fall within the protection scope of the present invention.

In the following description, numerous specific details are set forth in order to provide a more thorough understanding of the present invention. It will be apparent, however, to one skilled in the art that the present invention may be practiced without one or more of these details. In other instances, some technical features known in the art have not been described in order to avoid obscuring the present invention.

It should be understood that the present invention may be embodied in different forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used herein, the singular forms "a," "an," and "the/the" are intended to include the plural forms as well, unless the context clearly dictates otherwise. It should also be understood that the terms "compose" and/or "include", when used in this specification, identify the presence of stated features, integers, steps, operations, elements and/or components, but do not exclude one or more other The presence or addition of features, integers, steps, operations, elements, parts and/or groups. As used herein, the term "and/or" includes any and all combinations of the associated listed items.

For a thorough understanding of the present invention, detailed structures will be presented in the following description in order to explain the technical solutions proposed by the present invention. Alternative embodiments of the present invention are described in detail below, however, the invention is capable of other embodiments in addition to these detailed descriptions.

The laser ranging device is an active detection instrument. When it detects the surrounding environment, the laser emission module emits a laser pulse. When the laser pulse encounters the measured object on the propagation path, a laser echo signal will be generated. The echo signal is detected by the receiving module of the laser ranging device, and the three-dimensional space position of the measured object is calculated from the laser echo signal. Since the laser pulse is also reflected inside the laser ranging device, such as reflection through the window glass, etc., the internal reflected light T0 is generated; and for the measured object that is very close to the laser ranging device, the target generated by the measured object The reflected light T1 will be difficult to distinguish between the two because the distance T0 is too close, especially when the measured object is a low reflectivity material or a smooth mirror surface, the problem becomes more serious, resulting in inaccurate depth measurement. , which in turn produces noise in the point cloud data.

To solve this problem, the laser ranging device is usually optimized from the perspective of the underlying signal, and it is expected that T0 and T1 can be effectively distinguished. However, due to the limited underlying information, if the waveforms of T0 and T1 echoes are completely fused, they cannot be effectively distinguished.

Furthermore, since the laser echo signal generated by the distant measured object is likely to be a small signal, the filtering strategy based on information such as the size or pulse width of the laser echo signal is very easy to detect the real measured object detected at a distance. The point cloud is misidentified as noise. Since there is too little information available at the signal level, it is very easy to filter out the laser echo signal generated by the small measured object nearby while filtering out the noise.

In view of the above problems, an embodiment of the present invention proposes a point cloud noise filtering method, which performs point cloud noise filtering based on upper layer information. FIG. 1 shows a schematic flowchart of a point cloud noise filtering method 100 according to an embodiment of the present application. As shown in FIG. 1 , the point cloud noise filtering method 100 includes the following steps:

In step S110, obtain point cloud data to be processed;

In step S120, feature extraction is performed on the point cloud data, and the point cloud point that meets the preset characteristics is marked as the first target point;

In step S130, mark the second target point in the point cloud data according to the aggregation degree of the point cloud data;

In step S140, the overlapping part of the first target point and the second target point is determined as a real measurement point, and at least part of the point cloud data other than the real measurement point is determined as a noise point.

The point cloud noise filtering method 100 according to the embodiment of the present invention is based on the upper layer information of the point cloud, and uses a combination of feature extraction and aggregation degree comparison to filter noise points, which can not only effectively retain the information of the real measured object, but also identify the noise points. .

Specifically, in step S110, the point cloud data to be processed is the three-dimensional point cloud data obtained by the laser ranging device detecting its surrounding environment. The point cloud data includes coordinate information of the point cloud points, and may also include reflectivity information of the point cloud points.

The laser ranging device may be an electronic device such as a laser radar or a laser ranging device. In one embodiment, the laser ranging device is used to sense external environmental information, for example, distance information, orientation information, reflection intensity information, speed information, and the like of environmental objects. The laser ranging device can detect the distance from the measured object to the laser ranging device by measuring the time of light propagation between the laser ranging device and the measured object, that is, time-of-flight (TOF).

The laser ranging device may include a transmitting circuit, a receiving circuit, a sampling circuit and an arithmetic circuit. The transmit circuit can transmit a sequence of laser pulses. The receiving circuit can receive the optical pulse sequence reflected by the measured object, and perform photoelectric conversion on the optical pulse sequence to obtain an electrical signal, which can be output to the sampling circuit after processing the electrical signal. The sampling circuit can sample the electrical signal to obtain the sampling result. The arithmetic circuit may determine the distance between the laser ranging device and the measured object based on the sampling result of the sampling circuit. Optionally, the laser ranging device may further include a control circuit, which can control other circuits, for example, can control the working time of each circuit and/or set parameters for each circuit. The laser ranging device may further include a scanning module for changing the propagation direction of at least one laser pulse sequence emitted from the transmitting circuit to emit.

The laser ranging device can adopt a coaxial optical path, that is, the light beam emitted by the laser ranging device and the reflected light beam share at least part of the optical path in the laser ranging device. For example, after at least one laser pulse sequence emitted by the transmitting circuit changes its propagation direction through the scanning module, the laser pulse sequence reflected by the measured object passes through the scanning module and then enters the receiving circuit.

Exemplarily, the laser ranging device includes a ranging module, and the ranging module includes a transmitter (which may include the above-mentioned transmitting circuit), a collimating element, a detector (which may include the above-mentioned receiving circuit, sampling circuit, and arithmetic circuit), and an optical circuit. Change the element. The ranging module is used to emit light beams, receive back light, and convert the returned light into electrical signals. Therein, the transmitter can be used to transmit a sequence of optical pulses. In one embodiment, the transmitter may emit a sequence of laser pulses. Optionally, the laser beam emitted by the transmitter is a narrow bandwidth beam with a wavelength outside the visible light range. The collimating element is arranged on the outgoing light path of the transmitter, and is used for collimating the beam emitted from the transmitter, and collimating the beam emitted by the transmitter into parallel light and outputting to the scanning module. The collimating element also serves to converge at least a portion of the return light reflected by the test object. The collimating element may be a collimating lens or other elements capable of collimating light beams.

Exemplarily, the transmitting optical path and the receiving optical path in the ranging device can be combined before the collimating element through the optical path changing element, so that the transmitting optical path and the receiving optical path can share the same collimating element, making the optical path more compact. The emitter and detector may also use respective collimating elements, and the optical path changing element may be arranged on the optical path after the collimating element. The optical path changing element can be deviated from the optical axis of the collimating element, or can be located on the optical axis of the collimating element.

The laser ranging device also includes a scanning module, which is placed on the outgoing optical path of the ranging module for changing the transmission direction of the collimated beam emitted by the collimating element and projecting it to the external environment, and projecting the return light to the collimating element. The returned light is focused on the detector through the collimating element. The scanning module can include at least one optical element for changing the propagation path of the light beam, wherein the optical element can change the propagation path of the light beam by means of reflection, refraction, diffraction, etc. of the light beam. The rotation of each optical element in the scanning module can project light in different directions, thus scanning the space around the laser ranging device. When the speed of the optical elements within the scanning module changes, the scanning pattern changes accordingly.

When the light projected by the scanning module hits the measured object, a part of the light is reflected by the measured object to the laser ranging device in the opposite direction to the projected light. The returned light reflected by the measured object passes through the scanning module and then enters the collimating element. A detector is placed on the same side of the collimating element as the emitter, and the detector is used to convert at least part of the return light passing through the collimating element into an electrical signal.

Further, the laser pulse receiving time can be determined, for example, by detecting the rising edge time and/or the falling edge time of the electrical signal pulse to determine the laser pulse receiving time. In this way, the laser ranging device can calculate the time-of-flight (TOF) by using the pulse receiving time information and the pulse sending time information, so as to determine the distance from the measured object to the laser ranging device, and generate point cloud data.

Due to the use of a coaxial optical path, in addition to the return light pulse signal reflected by the measured object, the optical components of the laser ranging device (including lenses, mirrors, prisms, window glass, etc.) will also reflect the laser pulse signal, which can also be determined by received by the receiving circuit. When the measured object is close to the laser ranging device, the internal reflected light and the target reflected light will appear pulse fusion due to being too close. noise. The embodiment of the present invention can solve the problem of distinguishing the noise point from the point cloud of the real near-measured object, and while ensuring filtering out the noise point, the point cloud of the real near-measured object is retained, so as to avoid the missed detection of obstacles and lead to greater security issues.

In some embodiments, after the initial point cloud data collected by the laser ranging device is acquired, point cloud data within a preset depth range is screened from the initial point cloud data to obtain the point cloud data to be processed, that is, the point cloud data to be processed. The point cloud data is the point cloud data within a preset depth range in the original point cloud data. Since the embodiments of the present invention are mainly aimed at distinguishing the point cloud and the noise of the near-measured object, for example, when the noise is generated due to the phenomenon of signal fusion, since the signal fusion only occurs within the preset depth range, first from the The point cloud data within the preset depth range is extracted from the point cloud data for subsequent noise filtering processing, and the point cloud data outside the preset depth range may be subjected to noise filtering processing by applying other noise filtering methods. make restrictions.

In some embodiments, other preprocessing may also be performed on the initial point cloud data, for example, removing point cloud points outside the measurement range, and performing preliminary noise filtering on the initial point cloud data.

In practical application scenarios, the black car material is a typical material that is prone to noise. Figure 2 shows the noise performance of the black car. As shown in Figure 2, the noise 201 mainly has the following characteristics:

The noise is still distributed in a cone shape, and its projection pattern is still designed, that is, the main reason for the noise is that the depth estimation is inaccurate, but the angle estimation is normal; secondly, the return light pulse signal returned from the surface of the measured object is presented in the time domain. Continuous gradient characteristics, that is, on a continuous line in time domain, the distance between every two adjacent point cloud points is equal; while noise points do not show continuous gradient characteristics on a continuous line in time domain, but almost irregular changes.

In order to distinguish the noise point from the point cloud generated by the nearby measured object, the point cloud generated by various types of near measured object is analyzed, and it is concluded that it has the following characteristics: First, although the point cloud shape of the near measured object may exist Distortion, but still retains certain morphological characteristics; secondly, since the near object to be measured shields the laser pulse signal emitted by the laser ranging device in the near distance, the point cloud points in the point cloud corresponding to the measured object are very dense .

Therefore, the embodiments of the present invention perform noise filtering processing on the point cloud data to be processed based on the above features. The noise targeted by the point cloud noise filtering method 100 according to the embodiment of the present invention is not limited to the black vehicle noise mentioned above, but also includes sunlight noise or other forms of noise.

Specifically, firstly in step S120, feature extraction is performed on the point cloud data, and the point cloud points satisfying the preset characteristics are marked as the first target points. The feature extraction of point cloud data is to extract the feature points with certain particularity from the point cloud data. Since the point cloud generated by the measured object has certain morphological characteristics, and the noise has irregularity, certain feature points can be obtained by extracting the feature of the point cloud of the measured object, but the feature point cannot be obtained by the feature extraction of the noise. Therefore, the feature extraction of the point cloud data can obtain the point cloud points that reflect the shape of the measured object, that is, the first target point.

In one embodiment, considering that the measured object may have a variety of different shapes, the point cloud points that meet the preset characteristics may include point cloud points that meet the characteristics of plane points and point cloud points that meet the characteristics of edge points, that is, the plane points and edge point. For objects of different shapes, plane points and edge points can usually be extracted from their point clouds, and the plane points and edge points can reflect the shape of the object to be measured, and this part of the point cloud points is used as the first target. The points are reserved to facilitate the subsequent identification of the analyte.

For plane points and edge points, for example, the extraction of plane points can be implemented by using a sliding window method. Specifically, first obtain a predetermined number of point cloud points from the point cloud data according to the time series with a certain sliding window size, and judge whether the obtained group of point cloud points satisfies the following conditions: the spatial distribution of the group of point cloud points Approximate as a straight line, and the set of point cloud points is approximately centrosymmetric when centered on the middle point. Exemplarily, a principal component analysis method may be used to determine whether the acquired set of point cloud points satisfies the above conditions.

If a group of point cloud points in the current sliding window satisfies the above conditions, the group of point cloud points is determined as a plane point candidate point. After that, the sliding window is moved backward, so as to obtain the next set of point cloud points of the same number for judgment, and the next set of point cloud points includes at least one point cloud point in the previous set of point cloud points. After the sliding window traverses all the point cloud points and extracts all the plane point candidate points that satisfy the first preset condition, the final plane point extraction result can be determined among the plane point candidate points.

The extraction of edge points can be performed based on the results of extraction of plane points. For example, a point on the boundary line of two planes can be determined as an edge point, a point on the edge of an isolated plane can be determined as an edge point, and a point on the edge of other small objects can be determined as an edge point, etc.

It should be noted that the above feature point types and feature point extraction methods are only examples, and the first target point extracted in the embodiment of the present invention is not limited to edge points and plane points in the point cloud data, and may also include other features. The feature extraction method is not limited to the above-mentioned feature extraction method, for example, any suitable feature point extraction method such as a clustering method, a method based on model matching, and a method based on neural network can be used.

Referring to FIG. 3 and FIG. 4 , wherein, FIG. 3 shows point cloud data collected for a black car. In FIG. 3, the left part is the point cloud 301 corresponding to the black car, and the right part is the noise point 302. A large number of feature points can be extracted from the point cloud 301 corresponding to the black car, but almost no features can be extracted from the noise point 302. point. Figure 4 shows the point cloud data collected for small obstacles in the vicinity, and a large number of feature points can also be extracted from the point cloud data in Figure 4 . It can be seen that the feature points extracted by the feature extraction method basically belong to the point cloud points generated by the measured object. When extracting feature points, a relatively strict feature extraction method can be used, that is, in step S120, it is not required to extract all the point cloud points generated by the measured object, but it is necessary to make the extracted feature points not contain noise.

In step S130, the second target point in the point cloud data is marked according to the aggregation degree of the point cloud data. Considering the spatial continuity of the point cloud of the object to be measured, the point cloud corresponding to the object to be measured has a high degree of aggregation, while the degree of aggregation of noise points is low. On the surface of the object, since the reflected light comes from the same plane and the angle changes very little, the depth change of the obtained point cloud is small. Since the noise is irregular, the depth changes very sharply between different noises. Therefore, based on this characteristic, the point cloud and noise of the measured object can be distinguished according to the aggregation degree of the point cloud data. The aggregation degree represents the degree of spatial change of the point cloud points within a certain range, and further, the aggregation degree may be the depth aggregation degree, which represents the degree of depth change of the point cloud points in a certain range.

In one embodiment, when marking the second target point according to the aggregation degree of the point cloud data, the point cloud data can be projected on the target plane, the aggregation degree of the point cloud data can be calculated according to the projection of the point cloud data, and then the aggregation degree can be calculated according to the aggregation degree. A second target point is determined. Wherein, the target plane may include the imaging plane of the laser ranging device, that is, the plane perpendicular to the depth direction. When the target plane is perpendicular to the depth direction, the aggregation degree of the point cloud data can be calculated based on the depth information of the point cloud data.

Specifically, calculating the aggregation degree of the point cloud data according to the projection of the point cloud data includes: dividing the target plane into a plurality of grids, and separately calculating the aggregation degree of the point cloud points in different grids. If the aggregation degree of the point cloud points in a certain grid is small, it indicates that the spatial variation range is small, so the point cloud points whose aggregation degree is not higher than the first preset threshold can be marked as the second target point. When the target plane is the imaging plane of the laser ranging device, a grid in the target plane corresponds to an exit cone in the angle domain, and the aggregation degree of the point cloud points in the grid can be determined according to the depth of the point cloud points in the grid. calculate.

In one embodiment, the aggregation degree of point cloud points in each grid includes, but is not limited to, the variance of the depths of multiple point cloud points in the grid; The parameter of the depth difference calculates the degree of aggregation.

Since the laser ranging device such as lidar has the characteristics of continuous high-frequency scanning according to a specific scanning mode, and continuously scans the surrounding scene according to a specific trajectory, each point cloud point has obvious timing and position information. The points in the window are mapped to the scanning track to form continuous line segments. As mentioned above, on a continuous line in time series, the depth between the point cloud points generated by the object to be measured presents a continuous gradient change, while the noise point presents an irregular change. Therefore, the variance of the depth of the point cloud points generated by the measured object is small, while the variance of the noise points is large. Therefore, if the variance of the depths of all point cloud points in the grid is not higher than the first preset threshold, the point cloud points in the grid can be marked as the second target point; otherwise, if all point clouds in the grid If the variance of the depth of the point is higher than the first preset threshold, all the point cloud points in the grid can be marked as noise points.

Exemplarily, the first preset threshold may be related to the scanning angular velocity of the laser ranging device, and the higher the scanning angular velocity, the greater the change in the depth value between two adjacent point cloud points in time series, so a larger value can be selected. the first preset threshold.

In addition, since the types of the measured objects are different in different application scenarios, the first preset threshold threshold value can also be selected according to the application scenarios. The first preset threshold threshold value may also be selected according to the type of noise point targeted. For example, since the depth of noise generated by the fusion signal is relatively small, the depth of filtering out sunlight noise and natural scene noise such as rain, fog and dust is relatively large. Therefore, a smaller first preset threshold can be used when filtering out the noise generated by the fusion signal. , when used for natural scene noise, a larger first preset threshold can be used.

Exemplarily, the first preset threshold may also be adjusted according to a preset depth range of the point cloud data. Since the number of point cloud points in each grid projected to the target plane is related to the depth range of the point cloud data to be processed, the larger the pre-selected depth range, the more point cloud data in the range, so the projection to the target plane The more the number of point cloud points in each grid of , the more the first preset threshold can be appropriately increased. In some embodiments, the preset depth range of the point cloud data can also be divided into multiple depth intervals, a first preset threshold is set for each depth interval, and the point cloud data in each depth interval is respectively set Extract the second target point.

In some embodiments, the first preset threshold may be a fixed value obtained based on theoretical calculations or experimental tests. In other embodiments, the first preset threshold can also be a floating value. For example, the first preset threshold may be determined in combination with variances within a plurality of adjacent grids. In some embodiments, since the density of point clouds in different regions of the field of view is different, the grids corresponding to different regions of the laser ranging field of view may adopt different first preset thresholds.

In some embodiments, the division density of the grid is determined according to the angular resolution of the laser ranging device. Exemplarily, the division of the grid is related to the angular resolution of the laser ranging device, for example, each grid can be made to occupy one pixel, so that noise filtering can be performed in units of pixels. In some embodiments, the grids corresponding to different regions of the laser ranging field of view may have different sizes.

In addition to projecting onto a two-dimensional plane to calculate the aggregation degree, in other embodiments, grids can also be divided in the three-dimensional space, and the aggregation degree is calculated in units of grids in the three-dimensional space. The aggregation degree in the three-dimensional space can be calculated by using Calculations are performed in a manner similar to the above-mentioned two-dimensional plane.

Referring to FIG. 5 and FIG. 6 , FIG. 5 shows the second target point 501 and the non-second target point 502 in the black car scene, and FIG. 6 shows the second target point 601 and the non-second target point 601 of a small obstacle nearby Target point 602 . Through step S130, noise points and real measurement points can also be distinguished to a certain degree.

In step S140, the overlapping part of the first target point and the second target point is determined as a real measurement point, and at least part of the point cloud data other than the real measurement point is determined as a noise point. Exemplarily, the flag flag1 can be set for the first target point extracted in step S120, the flag flag2 is set for the second target point extracted in step S130, and the point cloud point marked with flag1 and flag2 at the same time is determined as a real measurement point, That is, the point cloud points corresponding to the real object to be measured, and at least some of the other point cloud points are determined as noise points. Referring to FIG. 7 , the final extracted real measurement points 701 and noise points 702 are shown.

Through the identification method based on feature extraction in step S120 and the identification method based on aggregation degree in step S130, noise points can be identified to a certain extent, and point clouds of nearby objects can also be preserved to a certain extent. If the method of step S120 or step S130 is used for noise filtering alone, in order to filter out all the noise points, the point cloud of the measured object will also become extremely rare, which may lead to missed detection of the measured object. Therefore, in the embodiment of the present invention, the two identification methods are combined with each other, and only the point cloud points that belong to both the first target point and the second target point are determined as the real measurement points, the noise filtering effect is more stable, and the measured points can be retained. point cloud of objects.

In some embodiments, the point cloud noise filtering method 100 further includes: determining isolated points in the real measurement points as noise points. Wherein, the isolated point is a point cloud point whose number of adjacent points in the neighborhood is less than the second preset threshold. Due to the spatial continuity of the real measurement points, the isolated points are usually noise points. Determining the isolated points in the real measurement points as noise points can further filter out the unrecognized noise points in the above noise filtering process and improve the noise reduction effect.

Among them, the process of identifying outliers can be carried out in a 3D point cloud. Specifically, the number of adjacent point cloud points in the neighborhood of the point cloud point determined to be the real measurement point in the point cloud space can be counted, and the number of adjacent point cloud points is less than the second preset threshold. Points are judged as noise, and a noise mark is added or noise is removed.

The neighborhood of the original point cloud point in the three-dimensional space may be a spherical three-dimensional space area with a predetermined radius centered on the current point cloud point, and the size of the radius may be set according to actual needs. In one embodiment, all point cloud points may use the same neighborhood radius and threshold for adjacent points. In another embodiment, since the point cloud collected for objects with a closer distance is denser, and the point cloud collected for objects with a longer distance is sparser, it is easier to be judged as noise points. Distant point clouds can be applied with larger radii or smaller thresholds.

The point cloud point determined to be a noise point by the above method may be processed according to the following method: in one embodiment, when the current point cloud point is determined to be a noise point, the current point cloud point is filtered out. In another example, a noise mark is added to a point cloud point determined to be a noise point, and the upper-layer algorithm can differentiate the point cloud point with the noise mark from the real measurement point.

To sum up, in view of the noise generated by the laser ranging device in practical applications, when facing the noise that the underlying signal cannot distinguish, the point cloud noise filtering method of the embodiment of the present invention combines two methods of feature extraction and aggregation degree comparison to filter. Noise, which can effectively retain the real measured object information, and can distinguish noise points, which greatly improves the applicable scope of the laser ranging device. The point cloud noise filtering method can avoid the vehicle system from being stuck and stuck due to false detection of noise as an obstacle, and at the same time, it can effectively identify the obstacle point when encountering a real obstacle, and avoid the failure due to missed detection. A security risk arises.

Next, a point cloud noise filtering system 800 according to an embodiment of the present invention will be described with reference to FIG. 8 , wherein the features of the aforementioned point cloud noise filtering method 100 may be combined into this embodiment. The point cloud noise filtering system 800 can be implemented as an electronic device such as a computer, a server, or a vehicle-mounted terminal. The point cloud noise filtering system 800 can be applied to a movable platform. The movable platform with the point cloud noise filtering system 800 can measure the external environment, for example, measure the distance between the mobile platform and obstacles for obstacle avoidance and other purposes, and perform two-dimensional or three-dimensional mapping of the external environment. In certain embodiments, the movable platform includes an unmanned vehicle or a vehicle equipped with an Advanced Driving Assistance System (ADAS). The movable platform may also include at least one of an unmanned aerial vehicle, a robot, a ship, and a camera.

The point cloud noise filtering system 800 includes one or more processors 810, and one or more memories 820, the one or more processors 810 working together or individually. Optionally, the point cloud noise filtering system 800 may further include at least one of an input device (not shown), an output device (not shown), and an image sensor (not shown), and these components are connected through a bus system and/or other A connection mechanism (not shown) in the form of interconnection.

The memory 820 is used for storing program instructions executable by the processor, for example, for storing corresponding steps and program instructions for implementing the point cloud noise filtering method according to the embodiment of the present invention. One or more computer program products may be included, which may include various forms of computer-readable storage media, such as volatile memory and/or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and/or cache memory (cache), among others. The non-volatile memory may include, for example, read only memory (ROM), hard disk, flash memory, and the like.

The input device may be a device used by a user to input instructions, and may include one or more of a keyboard, a mouse, a microphone, a touch screen, and the like. The output device can output various information (such as images or sounds) to the outside (such as a user), and can include one or more of a display, a speaker, etc., and the output device can be used to output the filtered point cloud as image or video.

A communication interface (not shown) is used to communicate with other devices, including wired or wireless communication. The laser ranging device can access wireless networks based on communication standards, such as WiFi, 2G, 3G, 8G, 5G, or a combination thereof. In one exemplary embodiment, the communication interface receives broadcast signals or broadcast related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication interface further includes a Near Field Communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

Processor 810 may be a central processing unit (CPU), graphics processing unit (GPU), application specific integrated circuit (ASIC), field programmable gate array (FPGA), or other form of processing with data processing capabilities and/or instruction execution capabilities unit, and can control other components in the point cloud noise filtering system to perform the desired function. The processor 810 can execute the instructions stored in the memory to perform the point cloud noise filtering method of the embodiments of the present invention described herein. For example, a processor can include one or more embedded processors, processor cores, microprocessors, logic circuits, hardware finite state machines (FSMs), digital signal processors (DSPs), or combinations thereof. In this embodiment, the processor includes a Field Programmable Gate Array (FPGA), wherein the arithmetic circuit of the ranging device may be a part of the Field Programmable Gate Array (FPGA).

Specifically, when the program instructions stored in the memory 820 are executed by the processor 810, the processor 810 is used to:

Get the point cloud data to be processed;

In one embodiment, before acquiring the point cloud data to be processed, the method further includes: filtering out point cloud data within a preset depth range from the initial point cloud data.

In one embodiment, the point cloud points satisfying preset characteristics include point cloud points satisfying plane point characteristics and point cloud points satisfying edge point characteristics.

In one embodiment, the marking the second target point in the point cloud data according to the aggregation degree of the point cloud data includes: projecting the point cloud data onto a target plane; according to the point cloud data The projection calculates the aggregation degree of the point cloud data, and determines the second target point according to the aggregation degree.

In one embodiment, the target plane includes an imaging plane of a ranging device used to acquire the point cloud data.

In one embodiment, calculating the aggregation degree of the point cloud data according to the projection of the point cloud data, and determining the second target point according to the aggregation degree includes: dividing the target plane into multiple each grid, respectively calculate the aggregation degree of point cloud points in different grids; mark the point cloud point whose aggregation degree is not higher than the first preset threshold as the second target point.

In one embodiment, the degree of aggregation includes a variance of depths of a plurality of point cloud points within the grid.

In one embodiment, the processor is further configured to adjust the first preset threshold according to the depth range of the point cloud data.

In one embodiment, the division density of the grid is determined according to the angular resolution of the laser ranging device.

In one embodiment, when the program instructions are executed, the processor is further configured to: filter out isolated points in the real measurement points, and the isolated points are the number of adjacent points in the neighborhood less than the second Point cloud points with preset thresholds.

In one embodiment, the processor is further configured to: filter out the noise, or add a noise mark to the noise.

The point cloud noise filtering system of the embodiment of the present invention uses the combination of feature extraction and aggregation degree to filter noise points, which can not only effectively retain the information of the real measured object, but also identify the noise points.

An embodiment of the present invention further provides a movable platform, including a movable platform body and the point cloud noise filtering system 800 as described above, and the point cloud noise filtering system 800 is mounted on the movable platform body. The movable platform further includes a laser ranging device, such as a laser radar, which is connected in communication with the point cloud noise filtering system 800, and the point cloud noise filtering system 800 is used to filter the point cloud data collected by the laser ranging device.

In some embodiments, the movable platform includes at least one of an unmanned aerial vehicle, a car, a remote control car, a robot, a camera, and a gimbal. When the movable platform is an unmanned aerial vehicle, the body of the movable platform is the fuselage of the unmanned aerial vehicle. When the movable platform is an automobile, the movable platform body is the body of the automobile. The vehicle may be an autonomous driving vehicle or a semi-autonomous driving vehicle, which is not limited herein. When the movable platform is a remote control car, the movable platform body is the body of the remote control car. When the movable platform is a robot, the movable platform body is a robot. When the movable platform is a camera, the movable platform body is the camera itself. When the movable platform is a gimbal, the movable platform body is a gimbal body.

Since the movable platform adopts the point cloud noise filtering system according to the embodiment of the present invention, it also has the advantages mentioned above.

In addition, an embodiment of the present invention further provides a computer storage medium, on which a computer program is stored. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor may execute the program instructions stored in the memory to implement the functions (implemented by the processor) in the embodiments of the present invention described herein and/or other desired functions, for example, to perform corresponding steps of the point cloud noise filtering method according to the embodiment of the present invention, various application programs and various data may also be stored in the computer-readable storage medium, such as the Various data used and/or generated by the application, etc.

For example, the computer storage medium may include, for example, a memory card for a smartphone, a storage unit for a tablet computer, a hard disk for a personal computer, read only memory (ROM), erasable programmable read only memory (EPROM), portable compact disk Read only memory (CD-ROM), USB memory, or any combination of the above storage media. The computer-readable storage medium can be any combination of one or more computer-readable storage media. For example, a computer-readable storage medium contains program code or the like for adjusting the reflectivity of the point cloud.

Since the computer program instructions stored in the computer storage medium are used to implement the point cloud noise filtering method according to the embodiment of the present invention, the above-mentioned advantages are also provided.

In the above embodiments, it may be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented in software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer may be a general purpose computer, special purpose computer, computer network, or other programmable device. The computer instructions may be stored in or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be downloaded from a website site, computer, server, or data center Transmission to another website site, computer, server, or data center by wire (eg, coaxial cable, optical fiber, digital subscriber line, DSL) or wireless (eg, infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, a data center, or the like that includes an integration of one or more available media. The usable media may be magnetic media (eg, floppy disk, hard disk, magnetic tape), optical media (eg, digital video disc (DVD)), or semiconductor media (eg, solid state disk (SSD)), etc. .

Although example embodiments have been described herein with reference to the accompanying drawings, it should be understood that the above-described example embodiments are exemplary only, and are not intended to limit the scope of the invention thereto. Various changes and modifications can be made therein by those of ordinary skill in the art without departing from the scope and spirit of the invention. All such changes and modifications are intended to be included within the scope of the invention as claimed in the appended claims.

Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Skilled artisans may implement the described functionality using different methods for each particular application, but such implementations should not be considered beyond the scope of the present invention.

In the several embodiments provided in this application, it should be understood that the disclosed apparatus and method may be implemented in other manners. For example, the device embodiments described above are only illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components may be combined or May be integrated into another device, or some features may be omitted, or not implemented.

In the description provided herein, numerous specific details are set forth. It will be understood, however, that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures and techniques have not been shown in detail in order not to obscure an understanding of this description.

Similarly, it is to be understood that in the description of the exemplary embodiments of the invention, various features of the invention are sometimes grouped together , or in its description. However, this method of the invention should not be interpreted as reflecting the intention that the invention as claimed requires more features than are expressly recited in each claim. Rather, as the corresponding claims reflect, the invention lies in the fact that the corresponding technical problem may be solved with less than all features of a single disclosed embodiment. Thus, the claims following the Detailed Description are hereby expressly incorporated into this Detailed Description, with each claim standing on its own as a separate embodiment of this invention.

It will be understood by those skilled in the art that all features disclosed in this specification (including the accompanying claims, abstract and drawings) and any method or apparatus so disclosed may be used in any combination, except that the features are mutually exclusive. Processes or units are combined. Each feature disclosed in this specification (including accompanying claims, abstract and drawings) may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise.

Furthermore, those skilled in the art will appreciate that although some of the embodiments described herein include certain features, but not others, included in other embodiments, that combinations of features of different embodiments are intended to be within the scope of the invention within and form different embodiments. For example, in the claims, any of the claimed embodiments may be used in any combination.

Various component embodiments of the present invention may be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. Those skilled in the art should understand that a microprocessor or a digital signal processor (DSP) may be used in practice to implement some or all of the functions of some modules according to the embodiments of the present invention. The present invention may also be implemented as apparatus programs (eg, computer programs and computer program products) for performing part or all of the methods described herein. Such a program implementing the present invention may be stored on a computer-readable medium, or may be in the form of one or more signals. Such signals may be downloaded from Internet sites, or provided on carrier signals, or in any other form.

It should be noted that the above-described embodiments illustrate rather than limit the invention, and that alternative embodiments may be devised by those skilled in the art without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The invention can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In a unit claim enumerating several means, several of these means may be embodied by one and the same item of hardware. The use of the words first, second, and third, etc. do not denote any order. These words can be interpreted as names.

Claims

A point cloud noise filtering method, characterized in that the method comprises:

Get the point cloud data to be processed;

Perform feature extraction on the point cloud data, and mark the point cloud point that meets the preset characteristics as the first target point;

Mark the second target point in the point cloud data according to the aggregation degree of the point cloud data;

The overlapping part of the first target point and the second target point is determined as a real measurement point, and at least part of the point cloud data other than the real measurement point is determined as a noise point.
The method according to claim 1, wherein before the acquiring the point cloud data to be processed further comprises:

Filter out the point cloud data within the preset depth range from the initial point cloud data.
The method according to claim 1 or 2, wherein the point cloud points satisfying preset characteristics include point cloud points satisfying plane point characteristics and point cloud points satisfying edge point characteristics.
The method according to claim 1 or 2, wherein the marking the second target point in the point cloud data according to the aggregation degree of the point cloud data comprises:

projecting the point cloud data onto the target plane;

The aggregation degree of the point cloud data is calculated according to the projection of the point cloud data, and the second target point is determined according to the aggregation degree.
The method according to claim 4, wherein the target plane comprises an imaging plane of a laser ranging device that collects the point cloud data.
The method according to claim 4 or 5, characterized in that, calculating the aggregation degree of the point cloud data according to the projection of the point cloud data, and determining the second target point according to the aggregation degree, comprising: :

dividing the target plane into a plurality of grids, and calculating the aggregation degree of point cloud points in different grids respectively;

Point cloud points whose aggregation degree is not higher than the first preset threshold are marked as the second target points.
The method according to claim 6, wherein the aggregation degree comprises a variance of depths of a plurality of point cloud points in the grid.
The method according to claim 6, further comprising: adjusting the first preset threshold according to the depth range of the point cloud data.
The method according to claim 6, wherein the division density of the grid is determined according to the angular resolution of the laser ranging device.
The method according to any one of claims 1-9, further comprising: determining an isolated point in the real measurement points as a noise point, and the isolated point is the number of adjacent points in the neighborhood is less than the number of the Two point cloud points with preset thresholds.
The method according to any one of claims 1-10, further comprising: filtering out the noise, or adding a noise mark to the noise.
A point cloud noise filtering system, characterized in that it comprises a memory and a processor, the memory is used to store program instructions; the processor is used to execute the program instructions stored in the memory, when the program instructions are executed , the processor is used to:

Get the point cloud data to be processed;

Perform feature extraction on the point cloud data, and determine the point cloud point that meets the preset characteristics as the first target point;

Determine the second target point in the point cloud data according to the aggregation degree of the point cloud data;

The overlapping part of the first target point and the second target point is determined as a real measurement point, and at least part of the point cloud data other than the real measurement point is determined as a noise point.
The point cloud noise filtering system according to claim 12, wherein before the acquiring the point cloud data to be processed further comprises:

Filter out the point cloud data within the preset depth range from the initial point cloud data.
The point cloud noise filtering system according to claim 12 or 13, wherein the point cloud points satisfying preset characteristics include point cloud points satisfying plane point characteristics and point cloud points satisfying edge point characteristics.
The point cloud noise filtering system according to claim 12 or 13, wherein the marking the second target point in the point cloud data according to the aggregation degree of the point cloud data comprises:

projecting the point cloud data onto the target plane;

The aggregation degree of the point cloud data is calculated according to the projection of the point cloud data, and the second target point is determined according to the aggregation degree.
The point cloud noise filtering system according to claim 15, wherein the target plane comprises an imaging plane of a laser ranging device that collects the point cloud data.
The point cloud noise filtering system according to claim 15 or 16, characterized in that the aggregation degree of the point cloud data is calculated according to the projection of the point cloud data, and the second aggregation degree is determined according to the aggregation degree. Target points, including:

dividing the target plane into a plurality of grids, and calculating the aggregation degree of point cloud points in different grids respectively;

Point cloud points whose aggregation degree is not higher than the first preset threshold are marked as the second target points.
The point cloud noise filtering system according to claim 17, wherein the aggregation degree comprises a variance of depths of a plurality of point cloud points in the grid.
The point cloud noise filtering system according to claim 17, wherein the processor is further configured to adjust the first preset threshold according to the depth range of the point cloud data.
The point cloud noise filtering system according to claim 17, wherein the division density of the grid is determined according to the angular resolution of the laser ranging device.
The point cloud noise filtering system according to any one of claims 12-20, wherein when the program instructions are executed, the processor is further configured to: filter out isolated isolated points in the real measurement points The isolated point is a point cloud point whose number of adjacent points in the neighborhood is less than the second preset threshold.
The point cloud noise filtering system according to any one of claims 12-21, wherein the processor is further configured to: filter out the noise, or add a noise mark to the noise.
A movable platform, characterized in that, comprising:

Movable platform body;

And, the point cloud noise filtering system according to any one of claims 12 to 22, the point cloud noise filtering system is mounted on the movable platform body.
A computer storage medium on which a computer program is stored, characterized in that, when the program is executed by a processor, the point cloud noise filtering method according to any one of claims 1 to 11 is implemented.