WO2020155159A1 - 增加点云采样密度的方法、点云扫描系统、可读存储介质 - Google Patents

增加点云采样密度的方法、点云扫描系统、可读存储介质 Download PDF

Info

Publication number
WO2020155159A1
WO2020155159A1 PCT/CN2019/074630 CN2019074630W WO2020155159A1 WO 2020155159 A1 WO2020155159 A1 WO 2020155159A1 CN 2019074630 W CN2019074630 W CN 2019074630W WO 2020155159 A1 WO2020155159 A1 WO 2020155159A1
Authority
WO
WIPO (PCT)
Prior art keywords
point cloud
blank area
pixels
point
pixel
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2019/074630
Other languages
English (en)
French (fr)
Inventor
李延召
张富
陈涵
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
SZ DJI Technology Co Ltd
Original Assignee
SZ DJI Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by SZ DJI Technology Co Ltd filed Critical SZ DJI Technology Co Ltd
Priority to PCT/CN2019/074630 priority Critical patent/WO2020155159A1/zh
Priority to CN201980005414.4A priority patent/CN111819602A/zh
Publication of WO2020155159A1 publication Critical patent/WO2020155159A1/zh
Priority to US17/308,056 priority patent/US20210256740A1/en
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T19/00—Manipulating three-dimensional [3D] models or images for computer graphics
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T12/00—Tomographic reconstruction from projections
    • G06T12/20—Inverse problem, i.e. transformations from projection space into object space
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
    • G01S17/88—Lidar systems specially adapted for specific applications
    • G01S17/89—Lidar systems specially adapted for specific applications for mapping or imaging
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2210/00—Indexing scheme for image generation or computer graphics
    • G06T2210/56—Particle system, point based geometry or rendering

Definitions

  • the embodiments of the present disclosure relate to the field of control technology, in particular to a method for increasing the sampling density of a point cloud, a point cloud scanning system, and a readable storage medium.
  • the lidar system uses point cloud technology to obtain spatial sampling patterns. Under the same sampling pattern, the more points there are, the denser the point cloud; under the same number of points, the more uniform the sampling pattern is, the more conducive to the presentation of FOV (Field of View). Therefore, it is necessary to increase the point cloud sampling density as much as possible.
  • the first is to improve the hardware solution, that is, to systematically increase the sampling frequency and improve the sampling pattern.
  • the use of multi-channel mode to achieve parallel acquisition can double the sampling frequency and sampling pattern, thereby significantly increasing the point cloud density.
  • the second is to improve the point cloud sampling density through software interpolation, such as nearest neighbor difference and linear interpolation in three-dimensional space. Because of the sparseness of the point cloud when acquiring the point cloud pattern in three-dimensional space, the effect and adaptability of direct interpolation are poor. In addition, the presence of noise points in the point cloud pattern will also cause interpolation errors, thereby further deteriorating the effect of the point cloud pattern.
  • software interpolation such as nearest neighbor difference and linear interpolation in three-dimensional space.
  • the embodiments of the present disclosure provide a method for increasing the sampling density of a point cloud, a point cloud scanning system, and a readable storage medium.
  • embodiments of the present disclosure provide a method for increasing the sampling density of a point cloud, including:
  • embodiments of the present disclosure provide a point cloud scanning system, including a memory and a processor; the memory is connected to the processor through a communication bus, and is used to store computer instructions executable by the processor; The processor is used to read computer instructions from the memory to realize:
  • an embodiment of the present disclosure provides a readable storage medium having a number of computer instructions stored on the readable storage medium, and when the computer instructions are executed, the steps of the method described in the first aspect are implemented.
  • the first point cloud is used to project the three-dimensional first point cloud based on a given plane to obtain a first plane image; then, the blank space in the first plane image A number of pixels are inserted into the area to obtain a second plane image; finally, the second plane image is back-projected and transformed to obtain a reconstructed three-dimensional second point cloud.
  • pixels are inserted into the planar image, which can reduce the difficulty of the insertion operation.
  • the density of points in the second point cloud is significantly increased, which reduces the sparsity of the point distribution, and facilitates the user to observe objects in the corresponding scene based on the second point cloud.
  • Fig. 1 is a block diagram of a point cloud scanning system provided by an embodiment of the present disclosure
  • Figure 2 is a schematic structural diagram of a distance detection device using a coaxial optical path provided by an embodiment of the present disclosure
  • FIG. 3 is a typical point cloud scanning trajectory diagram provided by an embodiment of the present disclosure.
  • FIG. 4 is a schematic flowchart of a method for increasing the sampling density of a point cloud provided by an embodiment of the present disclosure
  • FIG. 5 is a block diagram of the state of the point cloud at different stages provided by an embodiment of the present disclosure
  • Fig. 6 is a schematic diagram of a projection plane as a given plane provided by an embodiment of the present disclosure.
  • FIG. 7 is a schematic diagram of a process for acquiring a second planar image provided by an embodiment of the present disclosure.
  • FIG. 8 is a schematic diagram of acquiring an object in a first plane image provided by an embodiment of the present disclosure.
  • FIG. 9 is a schematic diagram of obtaining a blank area on an object in a first planar image according to an embodiment of the present disclosure.
  • FIG. 10 is a schematic diagram of an image acquisition area in a first plane provided by an embodiment of the present disclosure.
  • FIG. 11 is another schematic diagram of obtaining a blank area in an area provided by an embodiment of the present disclosure.
  • FIG. 12 is a schematic diagram of inserting physical parameters into a target point in a blank area provided by an embodiment of the present disclosure
  • FIG. 13 is a schematic flowchart of determining a target point provided by an embodiment of the present disclosure.
  • FIG. 14 is an effect diagram of a second point cloud provided by an embodiment of the present disclosure.
  • 15 is a schematic flowchart of another method for increasing the sampling density of a point cloud according to an embodiment of the present disclosure
  • FIG. 16 is a block diagram of the state of the point cloud at different stages provided by an embodiment of the present disclosure.
  • FIG. 17 is a schematic diagram of a process for obtaining a third point cloud provided by an embodiment of the present disclosure.
  • FIG. 18 is a schematic flowchart of another method for increasing the sampling density of a point cloud according to an embodiment of the present disclosure
  • FIG. 19 is a block diagram of the state of the point cloud at different stages provided by an embodiment of the present disclosure.
  • FIG. 20 is a schematic flowchart of another method for increasing the sampling density of a point cloud according to an embodiment of the present disclosure
  • FIG. 21 is a block diagram of the state of the point cloud at different stages provided by an embodiment of the present disclosure.
  • Fig. 22 is a block diagram of another point cloud scanning system provided by an embodiment of the present disclosure.
  • the two methods of increasing the sampling density of the point cloud in related technologies have the following problems: First, it is more difficult to build hardware to improve the hardware solution, and the power consumption of the lidar system also increases. Second, the point cloud sampling density is improved through software interpolation in the three-dimensional space. Therefore, the point cloud pattern will have the characteristics of sparseness, resulting in poor direct interpolation effect and poor adaptability. In addition, the presence of noise points in the point cloud pattern will also cause interpolation errors, thereby further deteriorating the effect of the point cloud pattern.
  • FIG. 1 is a block diagram of a point cloud scanning system provided by an embodiment of the present disclosure.
  • a point cloud scanning system includes a distance detection device 100, a processor 200 and a memory 300.
  • the processor 200 may be connected to the distance detection device 100 and the memory 300 respectively, and the distance detection device 100 and the memory 300 are connected respectively.
  • the distance detection device 10 is used to obtain a point cloud and send the point cloud to the memory 300 or the processor 200.
  • the distance detection device 100 can be integrated in the point cloud scanning system; it can also be set separately, and the point cloud can be output through the connection with the point cloud scanning system.
  • the solution of the present application will be described by taking the distance detection device 100 installed in the point cloud scanning system as an example.
  • the embodiment of the present disclosure also provides a method for increasing the sampling density of the point cloud.
  • the processor 200 may execute the method for increasing the sampling density of the point cloud when receiving the point cloud, so as to achieve the effect of increasing the sampling density of the point cloud. After that, the processor 200 may also send the reconstructed point cloud to the memory 300 for storage.
  • the distance detection device may include radar, such as lidar.
  • the detection device can detect the distance between the detection device and the detection device by measuring the time of light propagation between the detection device and the detection object, that is, the time-of-flight (TOF).
  • TOF time-of-flight
  • a coaxial optical path can be used in the distance detection device, that is, the beam emitted by the detection device and the reflected beam share at least part of the optical path in the detection device.
  • the detection device may also adopt an off-axis optical path, that is, the light beam emitted by the detection device and the reflected light beam are respectively transmitted along different optical paths in the detection device.
  • Fig. 2 shows a schematic diagram of an embodiment in which the distance detection device of the present disclosure adopts a coaxial optical path.
  • the distance detection device 100 includes an optical transceiver 110, and the optical transceiver 110 includes a light source 103, a collimating element 104, a detector 105 and an optical path changing element 106.
  • the optical transceiver 110 is used to emit light beams, receive return light, and convert the return light into electrical signals.
  • the light source 103 is used to emit a light beam. In one embodiment, the light source 103 may emit a laser beam.
  • the light source 103 may include a laser diode packaging module 106 for emitting laser pulses at a certain angle with the first surface of the substrate of the laser diode packaging module 106, wherein the angle is less than or equal to 90 degrees.
  • the laser beam emitted by the light source 103 is a narrow-bandwidth beam with a wavelength outside the visible light range.
  • the collimating element 104 is arranged on the exit light path of the light source 103, and is used to collimate the light beam emitted from the light source 103 and collimate the light beam emitted from the light source 103 into parallel light.
  • the collimating element 104 is also used to condense at least a part of the return light reflected by the probe.
  • the collimating element 104 may be a collimating lens or other elements capable of collimating light beams.
  • the distance detection device 100 further includes a scanning module 102.
  • the scanning module 102 is placed on the exit light path of the optical transceiver 110.
  • the scanning module 102 is used to change the transmission direction of the collimated beam 119 emitted by the collimating element 104 and project it to the external environment, and project the return light to the collimating element 104 .
  • the returned light is collected on the detector 105 via the collimating element 104.
  • the scanning module 102 may include one or more optical elements, for example, lenses, mirrors, prisms, gratings, optical phased arrays (Optical Phased Array) or any combination of the foregoing optical elements.
  • the multiple optical elements of the scanning module 102 can be rotated around a common rotation axis 109, and each rotating optical element is used to continuously change the propagation direction of the incident light beam.
  • the multiple optical elements of the scanning module 102 may rotate at different rotation speeds.
  • the multiple optical elements of the scanning module 102 may rotate at substantially the same rotation speed.
  • the multiple optical elements of the scanning module 102 may also rotate around different axes, or vibrate in the same direction, or vibrate in different directions, which is not limited herein.
  • the scanning module 102 includes a first optical element 114 and a driver 116 connected to the first optical element 114.
  • the driver 116 may include a motor or other driving device for driving the first optical element 114 to rotate around the rotation axis 109 so that the first optical element 114 changes the direction of the collimated light beam 119.
  • the first optical element 114 projects the collimated light beam 119 to different directions.
  • the angle between the direction of the collimated light beam 119 changed by the first optical element 114 and the rotation axis 109 changes with the rotation of the first optical element 114.
  • the first optical element 114 includes a pair of opposed non-parallel surfaces through which the collimated light beam 119 passes.
  • the first optical element 114 includes a prism whose thickness varies in at least one radial direction. In one embodiment, the first optical element 114 includes a wedge prism to collimate the beam 119 for refracting. In one embodiment, the first optical element 114 is coated with an anti-reflection coating, and the thickness of the anti-reflection coating is equal to the wavelength of the light beam emitted by the light source 103, which can increase the intensity of the transmitted light beam.
  • the scanning module 102 further includes a second optical element 115, the second optical element 115 rotates around the rotation axis 109, and the rotation speed of the second optical element 115 is different from the rotation speed of the first optical element 114.
  • the second optical element 115 is used to change the direction of the light beam projected by the first optical element 114.
  • the second optical element 115 is connected to another driver 117.
  • the driver 117 may include a motor or other driving device to drive the second optical element 115 to rotate.
  • the first optical element 114 and the second optical element 115 can be driven by different drivers, so that the rotation speed of the first optical element 114 and the second optical element 115 are different, so that the collimated beam 119 is projected to different directions of the external environment and can be scanned. Larger space range.
  • the controller 118 controls the driver 116 and the driver 117 to drive the first optical element 114 and the second optical element 115, respectively.
  • the rotational speeds of the first optical element 114 and the second optical element 115 can be determined according to the area and pattern expected to be scanned in actual applications. For example, adjusting the rotational speeds of the first optical element 114 and the second optical element 115 can obtain the one shown in FIG. 3 A typical point cloud scanning trajectory diagram.
  • the second optical element 115 includes a pair of opposite non-parallel surfaces through which the light beam passes. In one embodiment, the second optical element 115 includes a prism whose thickness varies in at least one radial direction. In one embodiment, the second optical element 115 includes a wedge prism. In one embodiment, the second optical element 115 is coated with an anti-reflection coating to increase the intensity of the transmitted light beam.
  • the rotation of the scanning module 102 can project light to different orientations, such as the orientation 111 and the orientation 113, so as to scan the space around the detection device 100.
  • the light beam projected by the scanning module 102 hits the detection object 101 along the azimuth 111, a part of the light is reflected by the detection object 101 to the detection device 100 in a direction opposite to the projected light beam.
  • the scanning module 102 receives the return light 112 reflected by the probe 101 and projects the return light 112 to the collimating element 104.
  • the collimating element 104 condenses at least a part of the return light 112 reflected by the probe 101.
  • an anti-reflection coating is plated on the collimating element 104 to increase the intensity of the transmitted light beam.
  • the detector 105 and the light source 103 are placed on the same side of the collimating element 104, and the detector 105 is used to convert at least part of the return light passing through the collimating element 104 into an electrical signal.
  • the light source 103 may include a laser diode through which nanosecond laser light is emitted.
  • the laser pulse emitted by the light source 103 lasts for 10 ns.
  • the laser pulse receiving time can be determined, for example, the laser pulse receiving time can be determined by detecting the rising edge time and/or the falling edge time of the electrical signal pulse. In this way, the distance detection device 100 can calculate the TOF using the pulse receiving time information and the pulse sending time information, so as to determine the distance from the detection object 101 to the distance detection device 100.
  • the distance and orientation detected by the distance detection device 100 can be used for remote sensing, obstacle avoidance, surveying and mapping, modeling, navigation, and the like.
  • the distance detection device 100 can calculate the physical parameters of the points on the detection object from the back light of different azimuths.
  • FIG. 4 is a method for increasing a point cloud according to an embodiment of the present disclosure. Schematic diagram of the flow of the sampling density method. Referring to Fig. 4, a method for increasing the sampling density of a point cloud includes steps 401 to 403:
  • the processor 200 first obtains a given plane. There is a mapping relationship between the given plane and the image plane of the distance detection device 100, and the mapping relationship is the mathematical relationship of translation and rotation between the given plane and the image plane.
  • the given plane is the image plane of the distance detection device 100, so that the subsequent processing can be simplified.
  • the processor 200 may send a request that characterizes the acquisition of a given plane to the distance detection device 100, and the distance detection device 100 responds to the request to send its image plane to the processor 200 as the given plane of the processor 200.
  • the image plane of the distance detection device 200 can also be stored in the memory 300 in advance, and when a given plane is needed, the processor 200 can read it from the memory 300.
  • the processor 200 performs projection transformation on the three-dimensional first point cloud based on a given plane, and can convert the three-dimensional point cloud to a two-dimensional point cloud, thereby obtaining a first plane image.
  • the projection transformation can be implemented in multiple methods.
  • the given plane is used as the projection surface
  • the position of the detection device such as lidar
  • the point cloud points of the first point cloud are respectively perspectively projected to the given plane to obtain the respective locations.
  • the projection point on the given plane is used as the projection surface
  • the position of the detection device such as lidar
  • a given plane (that is, a projection plane) is a plane perpendicular to the central axis of the light pulse sequence emitted by the detection device, where the plane can also be the image plane of the detection device, Or other suitable planes, the given plane can be used as the projection plane, the position of the detection device (such as lidar) as the center point, and the point cloud point of the first point cloud can be perspectively projected to the reference plane to obtain The projection point on the reference plane.
  • the processor 200 obtains a blank area in the first plane image. Ways to obtain the blank area can include:
  • the processor 200 divides the first plane image according to the physical parameters of each pixel in the first plane image to obtain objects contained in the first plane image and blank areas in each object (corresponding to step 701). Then, for the blank area of each object, the processor 200 inserts a number of pixels in the blank area according to the pixels around the blank area, and inserts the blank area of each object into the planar image as the second planar image ( Corresponding to step 702).
  • FIG. 8 is a schematic diagram of a first planar image provided by an embodiment of the present disclosure.
  • the processor 200 may scan the first planar image. Get the depth value of each point. Then, the processor 200 can segment object 1, object 2, object 3, and object 4 contained in the first planar image according to the depth value of each point.
  • the way for the processor 200 to segment the first plane image may include at least one of the following: semantic segmentation and instance segmentation.
  • semantic segmentation refers to segmenting and recognizing the content (ie, objects) in an image and making corresponding annotations. The same objects are labeled the same.
  • Instance segmentation refers to segmenting and identifying the content (ie objects) in the image, and each object corresponds to a label.
  • Technicians can choose a suitable segmentation method according to a specific scene, which is not limited here.
  • FIG. 9 is a schematic diagram of obtaining a blank area on an object in a first planar image provided by an embodiment of the present disclosure.
  • the processor 200 can determine the first plane according to the depth value of each point in the object 4 At least one blank area contained in the image. For example, the processor 200 takes one of the points A as a starting point, and then spreads around. If there are points around point A, it means that point A is not a boundary point, that is, point A has nothing to do with the blank area. In this case, the processor 200 updates the starting point and iterates the above steps until it finds the starting point that is the boundary of the blank area. Then continue to detect the boundary points of the blank area from the adjacent points of the starting point, until a closed blank area as shown in FIG. 9 is obtained.
  • the processor 200 divides the first plane image into multiple regions, and the multiple regions may have the same area, or may be partially the same, or may be different from one another. Then, for each area, taking an area in the small rectangular area in FIG. 11 as an example, the processor 200 obtains the blank area 1, the blank area 2, the blank area 3, and the blank area 4 in the area.
  • the processor 200 obtains the pixels around the blank area, and inserts several pixels in the blank area according to the physical parameters of the surrounding pixels, so that the second plane image can be obtained .
  • the distance detection device 100 Since the distance detection device 100 emits a light beam into the scene space, the light beam directed to the sky does not hit the detected object. At this time, the distance detection device 100 cannot receive the echo information, so the depth value of the point cannot be obtained.
  • the corresponding point below is called the sky point. In other words, there is no depth value between the sky point and the point in the first plane image that has not been scanned by the distance detection device 100 (ie, the unscanned point), so these two points can be included when segmenting the blank area.
  • the processor 200 may obtain the blank area in accordance with a preset step size. Or get the points to be inserted in the blank area in turn. Then, the processor 200 may determine whether the point to be inserted is a scanning point. This is because for sky points, there is no need to insert data, and for unscanned points, it needs to insert data. Therefore, before inserting data, it is necessary to determine whether the point to be inserted data is a sky point or a non-scanned point, that is, to be Whether the point where the data is inserted is the target point (corresponding to step 1202).
  • the processor 200 determines the value of the physical parameter of the target point according to the preset algorithm ( Corresponding to step 1203).
  • the preset algorithm may be an interpolation algorithm.
  • the interpolation algorithm may be at least one of the following: nearest neighbor interpolation, linear interpolation, Lanczos interpolation algorithm, inverse distance weighting method, spline interpolation method, discrete smooth interpolation and trend surface smooth interpolation.
  • the processor 200 determines whether the point to be inserted in the blank area is the target point, which may include the following methods
  • the processor 200 obtains the physical parameter values of the pixels around the pixel to be inserted (corresponding to step 1301).
  • the number of surrounding pixels can be selected as 4, 8 or more, which is not limited here.
  • the processor 200 compares the value of the physical parameter with the parameter threshold to obtain a comparison result (corresponding to step 1302). After that, the processor 200 can determine whether the pixel to be inserted is the target point based on the comparison result (corresponding to step 1303).
  • the processor 200 determines that the pixel to be inserted is the target point when the value of the comparison result representing the physical parameter is less than or equal to the parameter threshold; or, when the value of the comparison result representing the physical parameter is greater than the parameter threshold, it determines The pixel to be inserted is not the target point.
  • Manner 2 In the process of generating the first point cloud, the distance detection device 100 sets a flag when the sky point is detected.
  • the processor 200 can parse out the identifier in the physical parameters of the pixel to be inserted, and determine whether the pixel to be inserted is the target point based on the identifier.
  • the technical personnel can also choose other methods to determine whether the pixel to be inserted is the target point according to the specific scene.
  • the target point can be determined, the corresponding solution also falls within the protection scope of this application.
  • the processor 200 performs back-projection transformation on the second plane image, where the back-projection transformation is the inverse process of the projection transformation.
  • the projection transformation scheme can be combined to restore each The positions of the projection points can then be used to obtain a reconstructed three-dimensional second point cloud.
  • the density of points in the second point cloud is significantly higher than that of the first point cloud.
  • the processor 200 may send the second point cloud to the memory 300 for storage, or send it to a display (not shown in the figure) for display.
  • the user can directly determine that the scene space includes 4 features, for example, object 1 is a chair back, object 2 is a bonsai, object 3 is a blackboard, and object 4 is a chair handle.
  • pixel points are inserted in a planar image instead of points in a three-dimensional point cloud, which can reduce the difficulty of inserting data.
  • the density of the points in the second point cloud in this embodiment is significantly increased, which reduces the sparseness of the point distribution in the first point cloud, making it convenient for users to observe the corresponding scene in the second point cloud Objects.
  • FIG. 15 is a schematic flowchart of another method for increasing the sampling density of a point cloud provided by an embodiment of the present disclosure
  • FIG. 16 is a block diagram of the state of the point cloud provided by an embodiment of the present disclosure at different stages.
  • a method for increasing the sampling density of a point cloud includes steps 1501 to 1504:
  • step 1501 and step 401 are the same.
  • FIG. 4 and related content of step 401 which will not be repeated here.
  • step 1502 and step 402 are the same. For detailed description, please refer to FIG. 4 and related content of step 402, which will not be repeated here.
  • step 1503 and step 403 are the same. For detailed description, please refer to FIG. 4 and related content of step 403, which will not be repeated here.
  • the third point cloud includes the second point cloud, and points located in the first point cloud but not located in the second point cloud.
  • the processor 200 further uses the first point cloud to correct the second point cloud.
  • the processor 200 compares the physical parameters of each point in the first point cloud and the second point cloud (corresponding to step 1701). If there are different points in the second point cloud in the first point cloud, then these different points are corrected to the second point cloud to obtain the third point cloud. That is, the third point cloud includes all the points in the second point cloud and the points located in the first point cloud but not in the second point cloud (corresponding to step 1702).
  • a number of points can be inserted into the first point cloud with a lower difficulty in data insertion, to obtain a second point cloud that shows an increase in the density of the points in the first point cloud.
  • the density of the points in the third point cloud can be further increased, and the sparsity of the point distribution in the first point cloud can be further reduced. It is more convenient for the user to observe objects in the corresponding scene according to the second point cloud.
  • FIG. 18 is a schematic flowchart of another method for increasing the sampling density of a point cloud provided by an embodiment of the present disclosure
  • FIG. 19 is a block diagram of a state of a point cloud provided by an embodiment of the present disclosure at different stages.
  • a method for increasing the sampling density of a point cloud includes steps 1801 to 1805:
  • step 1801 and step 401 are the same.
  • FIG. 4 and related content of step 401 which will not be repeated here.
  • the processor 200 discretizes the first plane image according to a preset algorithm. Specifically, the processor 200 divides the first plane image into a plurality of regions, and the plurality of regions includes a partial blank region. The blank area includes an area that does not include a point in the first plane image.
  • the preset discrete algorithm includes at least one of the following: quadtree segmentation algorithm, uniform segmentation algorithm, ordinary segmentation algorithm, semantic segmentation algorithm, and instance segmentation algorithm.
  • Technicians can also choose other discrete algorithms, and if the first plane image can be discretized, it also falls within the protection scope of the present application.
  • the number of points contained in any two of the multiple regions may be different. In an embodiment, at least some of the multiple regions have different areas. In an embodiment, the processor 200 may continue to discretize each area. For example, if the number of points contained in a region is large, the number of discretizations can be increased until the resolution requirement is reached. For another example, if the number of points contained in a region is small, the discretization can be eliminated or the number of discretizations can be reduced. The number of discretizations can be set according to specific scenarios and is not limited here.
  • step 1803 and step 402 The specific methods and principles of determining the blank area in step 1803 and step 402 are the same. For detailed description, please refer to FIG. 4 and the related content of step 402, which are not repeated here.
  • the processor 200 determines the physical parameters of the pixels inserted in the blank area according to the physical parameters of the pixels in at least one adjacent area of the blank area.
  • the processor 200 determines the physical parameters of the pixels inserted in the blank area according to the physical parameters of the point closest to the detection device in the at least one adjacent area. For example, the depth value of the closest point is used as the depth value of the inserted pixel.
  • the processor 200 may also determine the physical parameters of the pixels inserted in the blank area according to the average value of the physical parameters of the points in the at least one adjacent area. For example, the average value of the physical parameters is used as the value of the physical parameter inserted into the pixel.
  • the processor 200 may also determine the physical parameter of the pixel to be inserted in the blank area according to the physical parameter of the point closest to the pixel to be inserted in the at least one adjacent area. For example, the value of the physical parameter of the closest point is used as the value of the physical parameter of the inserted pixel.
  • step 1804 For other content of step 1804, please refer to FIG. 4 and related content of step 402, which will not be repeated here.
  • step 1805 and step 403 are the same. For detailed description, please refer to FIG. 4 and related content of step 403, which will not be repeated here.
  • the difficulty of inserting data can be further reduced.
  • pixel points are inserted in the planar image instead of the points in the three-dimensional point cloud, which can reduce the difficulty of inserting data.
  • the density of the points in the second point cloud in this embodiment is significantly increased, which reduces the sparseness of the point distribution in the first point cloud, making it convenient for users to observe the corresponding scene in the second point cloud Objects.
  • FIG. 20 is a schematic flowchart of another method for increasing the sampling density of a point cloud provided by an embodiment of the present disclosure
  • FIG. 21 is a block diagram of a state of a point cloud provided by an embodiment of the present disclosure at different stages.
  • a method for increasing the sampling density of a point cloud includes steps 2001 to 2004:
  • the three-dimensional first point cloud was projected and transformed based on a given plane to obtain a first plane image.
  • step 2001 and step 401 are the same.
  • FIG. 4 and related content of step 401 which will not be repeated here.
  • step 2002 and step 402 are the same. For detailed description, please refer to FIG. 4 and related content of step 402, which will not be repeated here.
  • the second plane image is filtered according to a preset filtering algorithm.
  • the processor 200 invokes a preset filtering algorithm set in advance to filter the second plane image.
  • the purpose of filtering is to improve the smoothness of the newly inserted point and the surrounding points in the second plane, so that the newly inserted point is more matched with the surrounding points, that is, filtering is beneficial to improve the accuracy of the inserted data and is beneficial to the subsequent three-dimensional Point cloud reconstruction operation.
  • the preset filtering algorithm includes at least one of the following: Gaussian filtering, mean filtering, limit filtering, median filtering, recursive average filtering, median average filtering, limit average filtering, one First-order lag filtering method, weighted recursive average filtering method, debounce filtering method and limit debounce filtering method.
  • Gaussian filtering mean filtering, limit filtering, median filtering, recursive average filtering, median average filtering, limit average filtering, one First-order lag filtering method, weighted recursive average filtering method, debounce filtering method and limit debounce filtering method.
  • back-projection transformation was performed based on the filtered second plane image to obtain a reconstructed three-dimensional second point cloud.
  • step 2004 and step 403 are the same. For detailed description, please refer to FIG. 4 and related content of step 403, which will not be repeated here.
  • pixels are inserted into the planar image, which can reduce the difficulty of the insertion operation.
  • the newly inserted points can be more matched with the surrounding points, which is beneficial to improve the accuracy of the reconstructed second point cloud, thereby facilitating the user to accurately observe the second point. Cloud objects.
  • solutions of the embodiments shown in FIGS. 4-21 include different technical features. If the technical features do not conflict, multiple technical features can be combined to obtain different solutions, for example, discretization and first One point cloud is combined with the second point cloud, discretization and filtering are combined, and the first point cloud is corrected with the second point cloud combined with filtering, etc.
  • the corresponding solutions also fall within the protection scope of this application.
  • FIG. 22 is a block diagram of another point cloud scanning system provided by the embodiment of the present disclosure.
  • a point cloud scanning system 2200 includes at least a processor 2201 and a memory 2202; the memory 2202 is connected to the processor 2201 through a communication bus 2203, and is used to store computer instructions executable by the processor 2201; the processor 2201 is used for Read computer instructions from the memory 2202 to realize:
  • the first point cloud is acquired by the distance detection device; a given plane has a mapping relationship with the image plane of the distance detection device.
  • the given plane is the image plane of the distance detection device.
  • that the processor 2201 is configured to insert several pixels in the blank area based on the pixels around the blank area in the first plane image includes:
  • the value of the physical parameter of the target point is determined according to a preset algorithm.
  • the physical parameter is at least one of the following including: depth value, reflectivity, angle value, and color information.
  • the pixel points in the blank area may be non-scanning points; the non-scanning points refer to pixels in the image corresponding to directions in the scene space that are not scanned.
  • the pixels in the blank area may be sky points; the sky points refer to pixels in the scene space that are scanned in a direction but have not received echo information.
  • the processor 2201 configured to determine whether the pixel to be inserted in the blank area is a target point includes:
  • the pixel to be inserted is a non-scanning point, it is determined that the pixel is the target point; and/or,
  • the pixel to be inserted is a sky point, it is determined that the pixel is not a target point.
  • the processor 2201 configured to determine whether the pixel to be inserted in the blank area is a target point includes:
  • the processor 2201 configured to determine whether the pixel to be inserted is a target point based on the comparison result includes:
  • the comparison result indicates that the value of the physical parameter is less than or equal to the parameter threshold, it is determined that the pixel to be inserted is a target point
  • the comparison result indicates that the value of the physical parameter is greater than the parameter threshold, it is determined that the pixel to be inserted is not a target point.
  • the preset algorithm is an interpolation algorithm.
  • the interpolation algorithm is at least one of the following: nearest neighbor interpolation, linear interpolation, Lanczos interpolation algorithm, inverse distance weighting method, spline interpolation method, discrete smooth interpolation, and trend surface smooth interpolation.
  • the processor configured to insert several pixels in the blank area based on the pixels around the blank area in the first planar image to obtain the second planar image includes:
  • a number of pixels are inserted into the blank area according to the pixels around the blank area, and the planar image after the blank area of each object is inserted into the pixel is used as the second planar image.
  • the segmentation method includes at least one of the following: semantic segmentation and instance segmentation.
  • the processor is configured to perform back-projection transformation on the second planar image to obtain a reconstructed three-dimensional second point cloud, and then is further configured to:
  • the second point cloud is corrected based on the first point cloud to obtain a third point cloud.
  • the third point cloud includes the second point cloud and those that are located in the first point cloud and are not located at all Describe the points in the second point cloud.
  • that the processor 2201 is configured to correct the second point cloud based on the first point cloud includes:
  • the points with different physical parameters are added to the second point cloud to obtain the third point cloud.
  • the processor 2201 is configured to: based on the pixels surrounding the blank area in the first plane image, before inserting a number of pixels in the blank area, it is further configured to:
  • the blank area is determined according to the discretized first plane image.
  • that the processor 2201 is configured to discretize the planar image obtained after the projection transformation according to a preset discretization algorithm includes:
  • the first plane image is divided into a plurality of regions, the plurality of regions include a partial blank region, and the blank region includes a region that does not contain dots.
  • the number of points contained in any two regions may be different.
  • each area may continue to be discretized at least once.
  • At least some of the multiple regions have different areas.
  • that the processor 2201 is configured to insert several pixels in the blank area based on the pixels around the blank area in the first plane image includes:
  • the physical parameters of the pixels inserted in the blank area are determined according to the physical parameters of the pixels in at least one adjacent area of the blank area.
  • that the processor 2201 is configured to insert several pixels in the blank area based on the pixels around the blank area in the first plane image includes:
  • the physical parameters of the pixels inserted in the blank area are determined according to the physical parameters of the at least one adjacent area, wherein the physical parameters of the area are based on the points in the area that are closest to the detection device The physical parameters are determined.
  • the preset discrete algorithm includes at least one of the following: quadtree segmentation algorithm, uniform segmentation algorithm, ordinary segmentation algorithm, semantic segmentation algorithm, and instance segmentation algorithm.
  • the processor 2201 is configured to perform back-projection transformation on the second plane image to obtain a reconstructed three-dimensional second point cloud, and is further configured to:
  • the preset filtering algorithm includes at least one of the following: Gaussian filtering, average filtering, limiting filtering, median filtering, recursive average filtering, median average filtering, and limited average filtering Method, first-order lag filter method, weighted recursive average filter method, de-shake filter method and limit de-shake filter method.
  • the embodiment of the present disclosure also provides a readable storage medium having a number of computer instructions stored on the readable storage medium, and when the computer instructions are executed, the steps of the methods described in FIGS. 4-22 are implemented.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Hardware Design (AREA)
  • Computer Graphics (AREA)
  • Software Systems (AREA)
  • Optical Radar Systems And Details Thereof (AREA)
  • Algebra (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Mathematical Physics (AREA)
  • Pure & Applied Mathematics (AREA)

Abstract

提供了一种增加点云采样密度的方法、点云扫描系统和可读存储介质。增加点云采样密度的方法包括：基于给定的平面对三维的第一点云进行投影变换，得到第一平面图像（401）；基于第一平面图像中空白区域周围的像素点在空白区域内插入若干个像素点，得到第二平面图像（402）；对第二平面图像进行反投影变换，得到重建后三维的第二点云（403）。增加点云采样密度的方法以在平面图像中插入像素点替代在三维点云中插入点，这样可以降低插入数据操作的难度。并且，与第一点云相比，第二点云中点的密度明显增加，降低了第一点云中点分布的稀疏性，方便用户根据第二点云观察所对应场景内的物体。

Description

增加点云采样密度的方法、点云扫描系统、可读存储介质 技术领域
本公开实施例涉及控制技术领域，尤其涉及增加点云采样密度的方法、点云扫描系统、可读存储介质。
背景技术
激光雷达系统在空间扫描的过程中，采用点云技术获取空间的采样图案。在同样的采样图案下，点数越多点云越密；在同样的点数下，采样图案越均匀越利于FOV(Field of View，视野)的呈现，因此需要尽可能的增加点云采样密度。
为增加点云采样密度，相关技术中有两种方式：
第一种为改善硬件方案，即以系统地提升采样频率、改善采样图案。例如采用多通道的模式来实现并行采集，可以成倍改善采样频率和采样图案，从而显著提升点云密度。然而本模式需要搭建硬件的难度较大，且激光雷达系统的功耗也随之提升。
第二种为通过软件插值来改善点云采样密度，如在三维空间进行最近邻差值、线性插值等。由于在三维空间中获取点云图案时会存在点云稀疏性的特点，导致直接插值的效果及适应性较差。另外，点云图案中存在的噪声点还会引起插值误差，从而使点云图案的效果进一步恶化。
发明内容
本公开实施例提供一种增加点云采样密度的方法、点云扫描系统、可读存储介质。
第一方面，本公开实施例提供一种增加点云采样密度的方法，包括：
基于给定的平面对三维的第一点云进行投影变换，得到第一平面图像；
基于所述第一平面图像中空白区域周围的像素点在所述空白区域内插 入若干个像素点，得到第二平面图像；
对所述第二平面图像进行反投影变换，得到重建后三维的第二点云。
第二方面，本公开实施例提供一种点云扫描系统，包括存储器和处理器；所述存储器通过通信总线和所述处理器连接，用于存储所述处理器可执行的计算机指令；所述处理器用于从所述存储器读取计算机指令以实现：
基于给定的平面对三维的第一点云进行投影变换，得到第一平面图像；
基于所述第一平面图像中空白区域周围的像素点在所述空白区域内插入若干个像素点，得到第二平面图像；
对所述第二平面图像进行反投影变换，得到重建后三维的第二点云。
第三方面，本公开实施例提供一种可读存储介质，所述可读存储介质上存储有若干计算机指令，所述计算机指令被执行时实现第一方面所述方法的步骤。
由上述的技术方案可见，本实施例中通过第一点云基于给定的平面对三维的第一点云进行投影变换，得到第一平面图像；然后，在所述第一平面图像中的空白区域插入若干个像素点，得到第二平面图像；最后，对所述第二平面图像进行反投影变换，得到重建后三维的第二点云。这样，本实施例中在平面图像中插入像素点，可以降低插入操作的难度。并且，本实施例中与第一点云相比，第二点云中点的密度明显增加，降低了点分布的稀疏性，方便用户根据第二点云观察所对应场景内的物体。
附图说明
为了更清楚地说明本公开实施例中的技术方案，下面将对实施例描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本公开的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动性的前提下，还可以根据这些附图获得其他的附图。
图1是本公开实施例提供的一种点云扫描系统的框图；
图2是本公开实施例提供的采用同轴光路的距离探测装置的结构示意 图；
图3是本公开实施例提供的一种典型的点云扫描轨迹图；
图4是本公开实施例提供的一种增加点云采样密度的方法的流程示意图；
图5是本公开实施例提供的点云在不同阶段所处状态的框图；
图6是本公开实施例提供的给定的平面为投影面的示意图；
图7是本公开实施例提供的获取第二平面图像的流程示意图；
图8是本公开实施例提供的在第一平面图像获取物体的示意图；
图9是本公开实施例提供的一种获取第一平面图像中物体上空白区域的示意图；
图10是本公开实施例提供的在第一平面图像获取区域的示意图；
图11是本公开实施例提供的另一种获取区域中空白区域的示意图；
图12是本公开实施例提供的一种向空白区域中目标点插入物理参数的示意图；
图13是本公开实施例提供的确定目标点的流程示意图；
图14是本公开实施例提供的第二点云的效果图；
图15是本公开实施例提供的另一种增加点云采样密度的方法的流程示意图；
图16是本公开实施例提供的点云在不同阶段所处状态的框图；
图17是本公开实施例提供的获取第三点云的流程示意图；
图18是本公开实施例提供的又一种增加点云采样密度的方法的流程示意图；
图19是本公开实施例提供的点云在不同阶段所处状态的框图；
图20是本公开实施例提供的又一种增加点云采样密度的方法的流程示意图；
图21是本公开实施例提供的点云在不同阶段所处状态的框图；
图22是本公开实施例提供的另一种点云扫描系统的框图。
具体实施方式
下面将结合本公开实施例中的附图，对本公开实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本公开一部分实施例，而不是全部的实施例。基于本公开中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本公开保护的范围。
由于相关技术中增加点云采样密度的两种方式存在如下问题：一，改善硬件方案需要搭建硬件的难度较大，且激光雷达系统的功耗也随之提升。二，通过软件在三维空间中插值来改善点云采样密度，因此点云图案时会存在稀疏性的特点，导致直接插值的效果及适应性较差。另外，点云图案中存在的噪声点还会引起插值误差，从而使点云图案的效果进一步恶化。
为解决上述问题，本公开实施例提供了一种点云扫描系统，图1是本公开实施例提供的一种点云扫描系统的框图。参见图1，一种点云扫描系统包括距离探测装置100、处理器200和存储器300。其中，处理器200可以分别与距离探测装置100和存储器300连接，距离探测装置100和存储器300连接。距离探测装置10用于获取点云，并将点云发送给存储器300，或者处理器200。可理解的是，距离探测装置100可以集成在点云扫描系统内；还可以单独设置，通过与点云扫描系统之间的连接输出点云。后续以距离探测装置100设置在点云扫描系统内为例描述本申请的方案。
本公开实施例中还提供了一种增加点云采样密度的方法，处理器200可以在接收到点云时，执行增加点云采样密度的方法，达到增加点云采样密度的效果。之后，处理器200还可以将重建后的点云发送给存储器300保存。
在一实施例中，距离探测装置可以包括雷达，例如激光雷达。探测装置可以通过测量探测装置和探测物之间光传播的时间，即光飞行时间(Time-of-Flight，TOF)，来探测探测物到探测装置的距离。
距离探测装置中可以采用同轴光路，也即探测装置出射的光束和经反射回来的光束在探测装置内共用至少部分光路。或者，探测装置也可以采用异轴光路，也即探测装置出射的光束和经反射回来的光束在探测装置内分别沿不同的光路传输。图2示出了本公开的距离探测装置采用同轴光路的一种实施例的示意图。
参见图2，距离探测装置100包括光收发装置110，光收发装置110包括光源103、准直元件104、探测器105和光路改变元件106。光收发装置110用于发射光束，且接收回光，将回光转换为电信号。光源103用于发射光束。在一个实施例中，光源103可发射激光束。其中，光源103可以包括激光二极管封装模块106，用于以与激光二极管封装模块106的基板的第一表面呈一定夹角的方向出射激光脉冲，其中夹角小于或等于90度。可选的，光源103发射出的激光束为波长在可见光范围之外的窄带宽光束。准直元件104设置于光源103的出射光路上，用于准直从光源103发出的光束，将光源103发出的光束准直为平行光。准直元件104还用于会聚经探测物反射的回光的至少一部分。该准直元件104可以是准直透镜或者是其他能够准直光束的元件。
距离探测装置100还包括扫描模块102。扫描模块102放置于光收发装置110的出射光路上，扫描模块102用于改变经准直元件104出射的准直光束119的传输方向并投射至外界环境，并将回光投射至准直元件104。回光经准直元件104汇聚到探测器105上。
在一个实施例中，扫描模块102可以包括一个或多个光学元件，例如，透镜、反射镜、棱镜、光栅、光学相控阵(Optical Phased Array)或上述光学元件的任意组合。在一些实施例中，扫描模块102的多个光学元件可以绕共同的转动轴109旋转，每个旋转的光学元件用于不断改变入射光束的传播方向。在一个实施例中，扫描模块102的多个光学元件可以以不同的转速旋转。在另一个实施例中，扫描模块102的多个光学元件可以以基本相同的转速旋转。
在一些实施例中，扫描模块102的多个光学元件也可以是绕不同的轴旋转，或者沿相同的方向振动，或者沿不同的方向振动，在此不作限制。
在一个实施例中，扫描模块102包括第一光学元件114和与第一光学元件114连接的驱动器116。驱动器116可以包括电机或其他驱动装置，用于驱动第一光学元件114绕转动轴109转动，使第一光学元件114改变准直光束119的方向。第一光学元件114将准直光束119投射至不同的方向。在一个实施例中，准直光束119经第一光学元件114改变后的方向与转动轴109的夹角随着第一光学元件114的转动而变化。在一个实施例中，第一光学元件114包括相对的非平行的一对表面，准直光束119穿过该对表面。在一个实施例中，第一光学元件114包括厚度沿至少一个径向变化的棱镜。在一个实施例中，第一光学元件114包括楔角棱镜，对准直光束119进行折射。在一个实施例中，第一光学元件114上镀有增透膜，增透膜的厚度与光源103发射出的光束的波长相等，能够增加透射光束的强度。
在一个实施例中，扫描模块102还包括第二光学元件115，第二光学元件115绕转动轴109转动，第二光学元件115的转动速度与第一光学元件114的转动速度不同。第二光学元件115用于改变第一光学元件114投射的光束的方向。在一个实施例中，第二光学元件115与另一驱动器117连接。驱动器117可以包括电机或其他驱动装置，驱动第二光学元件115转动。第一光学元件114和第二光学元件115可以由不同的驱动器驱动，使第一光学元件114和第二光学元件115的转速不同，从而将准直光束119投射至外界环境不同的方向，可以扫描较大的空间范围。在一个实施例中，控制器118控制驱动器116和驱动器117，分别驱动第一光学元件114和第二光学元件115。第一光学元件114和第二光学元件115的转速可以根据实际应用中预期扫描的区域和样式确定，例如，调整第一光学元件114和第二光学元件115的转速可以得到图3所示的一种典型的点云扫描轨迹图。
在一个实施例中，第二光学元件115包括相对的非平行的一对表面， 光束穿过该对表面。在一个实施例中，第二光学元件115包括厚度沿至少一个径向变化的棱镜。在一个实施例中，第二光学元件115包括楔角棱镜。在一个实施例中，第二光学元件115上镀有增透膜，能够增加透射光束的强度。
扫描模块102旋转可以将光投射至不同的方位，例如方位111和方位113，如此对探测装置100周围的空间进行扫描。当扫描模块102投射出的光束沿着方位111打到探测物101时，一部分光被探测物101沿与投射的光束相反的方向反射至探测装置100。扫描模块102接收探测物101反射的回光112，将回光112投射至准直元件104。
准直元件104会聚探测物101反射的回光112的至少一部分。在一个实施例中，准直元件104上镀有增透膜，能够增加透射光束的强度。探测器105与光源103放置于准直元件104的同一侧，探测器105用于将穿过准直元件104的至少部分回光转换为电信号。
在一些实施例中，光源103可以包括激光二极管，通过激光二极管发射纳秒级别的激光。例如，光源103发射的激光脉冲持续10ns。进一步地，可以确定激光脉冲接收时间，例如，通过探测电信号脉冲的上升沿时间和/或下降沿时间确定激光脉冲接收时间。如此，距离探测装置100可以利用脉冲接收时间信息和脉冲发出时间信息计算TOF，从而确定探测物101到距离探测装置100的距离。
距离探测装置100探测到的距离和方位可以用于遥感、避障、测绘、建模、导航等。
在一实施例中，距离探测装置100可以将不同方位的回光计算出探测物上点的物理参数。其中物理参数可以包括如下至少一种：距离即深度值、反射率、方位即角度值和颜色信息，以及不同方位的回光相对于距离探测装置的像平面的坐标数据，可以得到一个三维空间中一个点，即Pi＝{Xi，Yi，Zi，……}，其中省略号表示物理参数，后续实施例中也称之为扫描点。然后，不同方位的点构成点云Point Cloud＝{P1，P2，P3，…，Pn}。可理 解的是，该点云是一个三维结构的点的集合，后续称其为第一点云。
在一实施例中，处理器200在接收到距离探测装置100发送的点云数据时，执行一种增加点云采样密度的方法的步骤，图4是本公开实施例提供的一种增加点云采样密度的方法的流程示意图。参见图4，一种增加点云采样密度的方法，包括步骤401～步骤403：
401，基于给定的平面对三维的第一点云进行投影变换，得到第一平面图像。
本实施例中，处理器200先获取给定的平面。其中给定的平面与距离探测装置100的像平面存在映射关系，映射关系即是给定平面与像平面之间的平移、旋转的数学关系。在一实施例中，给定的平面为距离探测装置100的像平面，从而可以简化后续的处理过程。具体地，处理器200可以向距离探测装置100发送表征获取给定的平面的请求，距离探测装置100响应该请求将其像平面发送给处理器200，作为处理器200的给定的平面。当然，还可以预先将距离探测装置200的像平面保存在存储器300中，在需要给定的平面时，处理器200从存储器300中读取即可。
本实施例中，参见图5，处理器200基于给定的平面，对三维的第一点云进行投影变换，可以将三维点云转换到二维点云，从而得到第一平面图像。
本实施例中，投影变换可以采用多种方法实现。在一个示例中，以该给定平面作为投影面，以探测装置(例如激光雷达)所在的位置作为中心点，将第一点云的点云点分别向该给定平面进行透视投影得到各自位于该给定平面上的投影点。
例如，如图6所示，给定平面(也即投影面)为与所述探测装置所发射的光脉冲序列的中心轴相垂直的面，其中，该平面也可以为探测装置的像平面，或者其他适合的平面，可以以该给定平面作为投影面，以探测装置(例如激光雷达)所在的位置作为中心点，将所述第一点云的点云点向该参考面进行透视投影得到位于该参考面上的投影点。
402，基于所述第一平面图像中空白区域周围的像素点在所述空白区域内插入若干个像素点，得到第二平面图像。
本实施例中，处理器200获取第一平面图像中的空白区域。获取空白区域的方式可以包括：
方式一，参见图7，处理器200根据第一平面图像内各像素点的物理参数分割第一平面图像，得到第一平面图像包含的物体以及各物体内的空白区域(对应步骤701)。然后，针对各物体的空白区域，处理器200根据空白区域周围的像素点在所述空白区域内插入若干个像素点，将各物体的空白区域插入像素点后的平面图像作为第二平面图像(对应步骤702)。
图8是本公开实施例提供的一种第一平面图像的示意图，参见图8，以第一平面图像中各点的物理参数中包含深度值为例，处理器200可以扫描第一平面图像中的各点，获取各点的深度值。然后，处理器200根据各点的深度值可以分割出第一平面图像中包含的物体1、物体2、物体3和物体4。
在一实施例中，处理器200分割第一平面图像的方式可以包括以下至少一种：语义分割和实例分割。其中，语义分割是指，分割并识别出图像中的内容(即物体)，并作相应的标注。其中相同的物体标注相同。实例分割是指，分割并识别出图像中的内容(即物体)，每个物体对应一个标注。技术人员可以根据具体场景选择合适的分割方式，在此不作限定。
图9是本公开实施例提供的获取第一平面图像中物体上空白区域的示意图，参见图9，以物体4为例，处理器200根据物体4中各点的深度值可以确定出第一平面图像中包含的至少一个空白区域。例如，处理器200以其中一个点A为起点，然后向周围扩散。若点A的周围存在点，则说明点A不是边界点，即点A与空白区域无关。此情况下，处理器200更新起点，迭代上述步骤，直至找到是空白区域边界的起点。之后从起点的相邻点中继续检测空白区域的边界点，直至得到如图9所示的一个封闭的空白区域。
方式二，参见图10，处理器200将第一平面图像分割为多个区域，多个区域可以面积相同，也可以部分相同，还可以各不相同。然后，针对各区域，以图11中小矩形区域内的一个区域为例，处理器200获取该区域内的空白区域1、空白区域2、空白区域3和空白区域4。
需要说明的是，技术人员还可以根据具体场景选择其他分割方式来分割第一平面图像，在能够确定空白区域的情况下，相应方案同样落入本申请的保护范围。
本实施例中，针对每个空白区域，处理器200获取该空白区域周围的像素点，并根据周围的像素点的物理参数在该空白区域内插入若干个像素点，这样可以得到第二平面图像。
由于距离探测装置100向场景空间中出射光束，射向天空的光束因未打到探测物上，此时距离探测装置100接收不到回波信息，从而无法获取到该点的深度值，此情况下对应的点称之为天空点。换言之，天空点与第一平面图像中未被距离探测装置100扫描过的点(即未扫描点)，都不存在深度值，因此在分割出空白区域时可以包括这两种点。
在一实施例中，参见图12，处理器根据第一平面图像中的点确定出第一平面图像中的空白区域后(对应步骤1201)，处理器200可以按照预设步长获取空白区域中的待插入的点或者依次获取空白区域中的待插入的点。然后，处理器200可以确定待插入的点是否为扫描点。这是因为，对于天空点而言，其无需插入数据，而对于未扫描点而言，其需要插入数据，因此在插入数据之前需要判断待插入数据的点是天空点还是非扫描点，即待插入数据的点是不是目标点(对应步骤1202)。
在一实施例中，若待插入的点为天空点，则该待插入的点不是目标点，无需插入数据。在另一实施例中，若待插入的点为非扫描点，则该待插入的点为是目标点，需要插入数据，此时处理器200根据预设算法确定目标点的物理参数的数值(对应步骤1203)。本实施例中，预设算法可以插值算法。其中插值算法可以为以下至少一种：最近邻插值、线性插值、Lanczos 插值算法、反距离加权法、样条插值法、离散平滑插值和趋势面光滑插值。
在一实施例中，处理器200确定空白区域中待插入点是否为目标点，可以包括以下方式
方式一，参见图13，处理器200获取待插入的像素点周围像素点的物理参数的数值(对应步骤1301)。周围像素点的数量可以选择为4个、8点，或者更多，在此不作限定。然后，处理器200对比物理参数的数值和参数阈值，可以得到比对结果(对应步骤1302)。之后，处理器200基于对比结果可以确定出待插入的像素点是否为目标点(对应步骤1303)。
在一实施例中，处理器200在对比结果表征物理参数的数值小于或等于参数阈值时，确定待插入的像素点是目标点；或者，在对比结果表征物理参数的数值大于参数阈值时，确定待插入的像素点不是目标点。
方式二，距离探测装置100在生成第一点云的过程中，在检测到天空点时设置一个标识。处理器200可以解析出待插入像素点的物理参数中的标识，基于该标识判断该待插入的像素点是否为目标点。
需要说明的是，技术人员还可以根据具体场景选择其他方式来判断待插入的像素点是否为目标点，在能够确定目标点的情况下，相应方案同样落入本申请的保护范围。
403，对所述第二平面图像进行反投影变换，得到重建后三维的第二点云。
本实施例中，处理器200对第二平面图像进行反投影变换，其中，反投影变换为投影变换的逆过程，在确定投影面和深度值的情况下，可以结合投影变换的方案还原出各投影点的位置，进而可以得到重建后的三维的第二点云，第二点云中点的密度较第一点云有明显的提升。之后，处理器200可以将第二点云发送给存储器300保存，或者发送给显示器(图中未示出)显示。
参见图14，当第二点云由显示设备显示后，用户可以直接确定出场景空间中包括4个特征，例如，物体1为椅背，物体2为盆景，物体3为黑 板，物体4为椅子把手。
至此，本实施例中以在平面图像中插入像素点替代在三维点云中插入点，这样可以降低插入数据操作的难度。并且，与第一点云相比，本实施例中第二点云中点的密度明显增加，降低了第一点云中点分布的稀疏性，方便用户根据第二点云观察所对应场景内的物体。
图15是本公开实施例提供的另一种增加点云采样密度的方法的流程示意图，图16是本公开实施例提供的点云在不同阶段所处状态的框图。参见图15和图16，一种增加点云采样密度的方法，包括步骤1501～步骤1504：
1501，基于给定的平面对三维的第一点云进行投影变换，得到第一平面图像。
步骤1501和步骤401的具体方法和原理一致，详细描述请参考图4及步骤401的相关内容，此处不再赘述。
1502，基于所述第一平面图像中空白区域周围的像素点在所述空白区域内插入若干个像素点，得到第二平面图像。
步骤1502和步骤402的具体方法和原理一致，详细描述请参考图4及步骤402的相关内容，此处不再赘述。
1503，对所述第二平面图像进行反投影变换，得到重建后三维的第二点云。
步骤1503和步骤403的具体方法和原理一致，详细描述请参考图4及步骤403的相关内容，此处不再赘述。
1504，基于所述第一点云修正所述第二点云，得到第三点云。所述第三点云包括所述第二点云，以及位于所述第一点云中的、且不位于所述第二点云中的点。
在一实施例中，处理器200还利用第一点云对第二点云进行修正。参见图17，处理器200对比第一点云和第二点云中各点的物理参数(对应步骤1701)。若第一点云中存在第二点云中不同的点，则将这些不同的点修正到第二点云中，得到第三点云。即第三点云中包括第二点云中所有的点， 以及位于第一点云中且不位于第二点云中的点(对应步骤1702)。
至此，本实施例中可以在较低的数据插入难度在第一点云中插入若干点，得到比第一点云中点的密度显示增加的第二点云。同时，本实施例中，由于利用第一点云修正第二点云得到第三点云，可以使第三点云中点的密度进一步增加，进一步降低第一点云中点分布的稀疏性，更方便用户根据第二点云观察所对应场景内的物体。
图18是本公开实施例提供的又一种增加点云采样密度的方法的流程示意图，图19是本公开实施例提供的点云在不同阶段所处状态的框图。参见图18和图19，一种增加点云采样密度的方法，包括步骤1801～步骤1805：
1801，基于给定的平面对三维的第一点云进行投影变换，得到第一平面图像。
步骤1801和步骤401的具体方法和原理一致，详细描述请参考图4及步骤401的相关内容，此处不再赘述。
1802，根据预设离散算法对所述第一平面图像进行离散化。
本实施例中，处理器200根据预设算法对第一平面图像进行离散化。具体地，处理器200将第一平面图像分割成多个区域，多个区域中包括部分空白区域。空白区域包括未包含第一平面图像中点的区域。
其中，预设离散算法包括以下至少一种：四叉树分割算法、均匀分割算法、普通分割算法、语义分割算法和实例分割算法。技术人员还可以选择其他离散算法，在能够对第一平面图像进行离散化的情况下，同样落入要本申请的保护范围。
在一实施例中，多个区域中任意两个区域内所包含的点的数量可以不同。在一实施例中，多个区域中至少部分区域的面积不同。在一实施例中，处理器200可以对各区域进行继续离散化。例如，若一个区域内包含点的数量较多，则离散化次数可以增加，直至达到分辨率要求停止。又如，若一个区域内包含点的数量较少，则可以不再离散化或者离散化的数量减少。离散化次数可以根据具体场景进行设置，在此不作限定。
1803，根据离散化后的第一平面图像确定所述空白区域。
步骤1803和步骤402中确定空白区域的具体方法和原理一致，详细描述请参考图4及步骤402的相关内容，此处不再赘述。
1804，基于所述第一平面图像中空白区域周围的像素点在所述空白区域内插入若干个像素点，得到第二平面图像。
本实施例中，处理器200根据空白区域的至少一个相邻的区域内像素点的物理参数，确定插入空白区域内的像素点的物理参数。
在一实施例中，处理器200根据所述至少一个相邻区域中与距离探测装置之间距离最近的点的物理参数确定插入所述空白区域内的像素点的物理参数。例如，将距离最近的点的深度值作为插入像素点的深度值。
在另一实施例中，处理器200还可以根据至少一个相邻区域中点的物理参数的平均值确定插入所述空白区域内的像素点的物理参数。例如，将物理参数的平均值作为插入像素点的物理参数的数值。
在另一实施例中，处理器200还可以根据至少一个相邻区域中与待插入的像素点距离最近的点的物理参数确定插入所述空白区域内的像素点的物理参数。例如，将距离最近的点的物理参数的数值作为插入像素点的物理参数的数值。
步骤1804的其他内容还可以参考参考图4及步骤402的相关内容，此处不再赘述。
1805，对所述第二平面图像进行反投影变换，得到重建后三维的第二点云。
步骤1805和步骤403的具体方法和原理一致，详细描述请参考图4及步骤403的相关内容，此处不再赘述。
至此，本实施例中通过对第一平面图像离散化，可以进一步降低插入数据的难度。并且，本实施例中以在平面图像中插入像素点替代在三维点云中插入点，这样可以降低插入数据操作的难度。并且，与第一点云相比，本实施例中第二点云中点的密度明显增加，降低了第一点云中点分布的稀 疏性，方便用户根据第二点云观察所对应场景内的物体。
图20是本公开实施例提供的又一种增加点云采样密度的方法的流程示意图，图21是本公开实施例提供的点云在不同阶段所处状态的框图。参见图20和图21，一种增加点云采样密度的方法，包括步骤2001～步骤2004：
2001，基于给定的平面对三维的第一点云进行投影变换，得到第一平面图像。
步骤2001和步骤401的具体方法和原理一致，详细描述请参考图4及步骤401的相关内容，此处不再赘述。
2002，基于所述第一平面图像中空白区域周围的像素点在所述空白区域内插入若干个像素点，得到第二平面图像。
步骤2002和步骤402的具体方法和原理一致，详细描述请参考图4及步骤402的相关内容，此处不再赘述。
2003，根据预设滤波算法对所述第二平面图像进行滤波。
本实施例中，处理器200调用预先设置的预设滤波算法对第二平面图像进行滤波。其中，滤波的目的在于，改善第二平面中新插入点与周围点的平滑程度，使新插入的点与周围点更匹配，即通过滤波有利于提升插入数据的准确度，且有利于后续三维点云的重建操作。
具体地，预设滤波算法包括以下至少一种：高斯滤波、均值滤波、限幅滤波法、中位值滤波法、递推平均滤波法、中位值平均滤波法、限幅平均滤波法、一阶滞后滤波法、加权递推平均滤波法、消抖滤波法和限幅消抖滤波法。当然，技术人员还可以选择其他滤波算法，在能够平滑新插入点与周围点的情况下，相应的算法同样落入本申请的保护范围。
2004，基于滤波后的第二平面图像进行反投影变换，得到重建后三维的第二点云。
步骤2004和步骤403的具体方法和原理一致，详细描述请参考图4及步骤403的相关内容，此处不再赘述。
至此，本实施例中在平面图像中插入像素点，可以降低插入操作的难 度。并且，本实施例中通过对第二平面图像进行滤波，可以使新插入的点与周围的点更匹配，有利于提升重建后第二点云的准确度，进而方便用户准确观察出第二点云的物体。
需要说明的是，图4～图21所示实施例的方案中包括不同的技术特征，在技术特征不冲突的情况下，多个技术特征可以相互组合得到不同的方案，例如，离散化和第一点云修正第二点云相结合，离散化和滤波相结合，第一点云修正第二点云和滤波相结合等等，相应的方案同样落入本申请的保护范围。
本公开实施例还提供了一种点云扫描系统，图22是本公开实施例提供的另一种点云扫描系统的框图。参见图22，一种点云扫描系统2200，至少包括处理器2201和存储器2202；存储器2202通过通信总线2203和处理器2201连接，用于存储处理器2201可执行的计算机指令；处理器2201用于从存储器2202读取计算机指令以实现：
基于给定的平面对三维的第一点云进行投影变换，得到第一平面图像；
基于第一平面图像中空白区域周围的像素点在空白区域内插入若干个像素点，得到第二平面图像；
对第二平面图像进行反投影变换，得到重建后三维的第二点云。
在一实施例中，第一点云由距离探测装置获取；给定的平面与距离探测装置的像平面存在映射关系。
在一实施例中，给定的平面为距离探测装置的像平面。
在一实施例中，处理器2201用于基于所述第一平面图像中空白区域周围的像素点在所述空白区域内插入若干个像素点包括：
根据所述第一平面图像中的像素点确定所述第一平面图像中的空白区域；
确定所述空白区域中待插入的像素点是否为目标点；
若是目标点，则根据预设算法确定所述目标点的物理参数的数值。
在一实施例中，物理参数为以下至少一种包括：深度值、反射率、角 度值和颜色信息。
在一实施例中，空白区域中的像素点可以为非扫描点；所述非扫描点是指场景空间中未被扫描到的方向在图像中对应的像素点。
在一实施例中，空白区域中的像素点可以为天空点；所述天空点是指场景空间中被扫描到的方向但未接收到回波信息的像素点。
在一实施例中，处理器2201用于确定所述空白区域中待插入的像素点是否为目标点包括：
若待插入的像素点为非扫描点，则确定所述像素点是目标点；和/或，
若待插入的像素点为天空点，则确定所述像素点不是目标点。
在一实施例中，处理器2201用于确定所述空白区域中待插入的像素点是否为目标点包括：
获取待插入的像素点周围像素点的物理参数的数值；
比对所述物理参数的数值和参数阈值，得到比对结果；
基于所述比对结果确定所述待插入的像素点是否为目标点。
在一实施例中，处理器2201用于基于所述比对结果确定所述待插入的像素点是否为目标点包括：
若所述对比结果表征所述物理参数的数值小于或等于所述参数阈值，则确定所述待插入的像素点是目标点；
若所述对比结果表征所述物理参数的数值大于所述参数阈值，则确定所述待插入的像素点不是目标点。
在一实施例中，预设算法为插值算法。
在一实施例中，插值算法为以下至少一种：最近邻插值、线性插值、Lanczos插值算法、反距离加权法、样条插值法、离散平滑插值和趋势面光滑插值。
在一实施例中，处理器用于基于所述第一平面图像中空白区域周围的像素点在所述空白区域内插入若干个像素点，得到第二平面图像包括：
根据所述第一平面图像内各像素点的物理参数分割所述第一平面图 像，得到所述第一平面图像包含的物体以及各物体内的空白区域；
针对各物体的空白区域，根据所述空白区域周围的像素点在所述空白区域内插入若干个像素点，将各物体的空白区域插入像素点后的平面图像作为第二平面图像。
在一实施例中，分割方式包括以下至少一种：语义分割和实例分割。
在一实施例中，处理器用于对所述第二平面图像进行反投影变换，得到重建后三维的第二点云之后，还用于：
基于所述第一点云修正所述第二点云，得到第三点云，所述第三点云包括所述第二点云，以及位于所述第一点云中的、且不位于所述第二点云中的点。
在一实施例中，处理器2201用于基于所述第一点云修正所述第二点云包括：
对比第一点云中各点的物理参数和第二点云中各点的物理参数；
将物理参数不同的点增加到第二点云中，得到所述第三点云。
在一实施例中，处理器2201用于基于所述第一平面图像中空白区域周围的像素点在所述空白区域内插入若干个像素点之前，还用于：
根据预设离散算法对所述第一平面图像进行离散化；
根据离散化后的第一平面图像确定所述空白区域。
在一实施例中，处理器2201用于根据预设离散算法对投影变换后得到的平面图像进行离散化包括：
将所述第一平面图像分割成多个区域，所述多个区域包括部分空白区域,所述空白区域包括未包含点的区域。
在一实施例中，任意两个区域内所包含的点的数量可以不同。
在一实施例中，各区域可以继续至少一次离散化。
在一实施例中，多个区域中至少部分区域的面积不同。
在一实施例中，处理器2201用于基于所述第一平面图像中空白区域周围的像素点在所述空白区域内插入若干个像素点包括：
根据所述空白区域的至少一个相邻的区域内像素点的物理参数，确定插入所述空白区域内的像素点的物理参数。
在一实施例中，处理器2201用于基于所述第一平面图像中空白区域周围的像素点在所述空白区域内插入若干个像素点包括：
根据所述至少一个相邻的区域的物理参数确定插入所述空白区域内的像素点的物理参数，其中，所述区域的物理参数是基于所述区域内与距离探测装置之间距离最近的点的物理参数确定的。
在一实施例中，预设离散算法包括以下至少一种：四叉树分割算法、均匀分割算法、普通分割算法、语义分割算法和实例分割算法。
在一实施例中，处理器2201用于对所述第二平面图像进行反投影变换，得到重建后三维的第二点云之前，还用于：
根据预设滤波算法对所述第二平面图像进行滤波；
基于滤波后的第二平面图像进行反投影变换，得到重建后三维的第二点云。
在一实施例中，预设滤波算法包括以下至少一种：高斯滤波、均值滤波、限幅滤波法、中位值滤波法、递推平均滤波法、中位值平均滤波法、限幅平均滤波法、一阶滞后滤波法、加权递推平均滤波法、消抖滤波法和限幅消抖滤波法。
本公开实施例还提供了一种可读存储介质，所述可读存储介质上存储有若干计算机指令，所述计算机指令被执行时实现图4～图22所述方法的步骤。
需要说明的是，在本文中，诸如第一和第二等之类的关系术语仅仅用来将一个实体或者操作与另一个实体或操作区分开来，而不一定要求或者暗示这些实体或操作之间存在任何这种实际的关系或者顺序。术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、物品或者设备不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、物品或者设备 所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括所述要素的过程、方法、物品或者设备中还存在另外的相同要素。
以上对本公开实施例所提供的检测装置和方法进行了详细介绍，本公开中应用了具体个例对本公开的原理及实施方式进行了阐述，以上实施例的说明只是用于帮助理解本公开的方法及其核心思想；对于本领域的一般技术人员，依据本公开的思想，在具体实施方式及应用范围上均会有改变之处，综上所述，本说明书内容不应理解为对本公开的限制。

Claims (53)

  1. 一种增加点云采样密度的方法，其特征在于，包括：
    基于给定的平面对三维的第一点云进行投影变换，得到第一平面图像；
    基于所述第一平面图像中空白区域周围的像素点在所述空白区域内插入若干个像素点，得到第二平面图像；
    对所述第二平面图像进行反投影变换，得到重建后三维的第二点云。
  2. 根据权利要求1所述的方法，其特征在于，所述第一点云由距离探测装置获取；所述给定的平面与所述距离探测装置的像平面存在映射关系。
  3. 根据权利要求2所述的方法，其特征在于，所述给定的平面为所述距离探测装置的像平面。
  4. 根据权利要求1所述的方法，其特征在于，基于所述第一平面图像中空白区域周围的像素点在所述空白区域内插入若干个像素点包括：
    根据所述第一平面图像中的像素点确定所述第一平面图像中的空白区域；
    确定所述空白区域中待插入的像素点是否为目标点；
    若是目标点，则根据预设算法确定所述目标点的物理参数的数值。
  5. 根据权利要求4所述的方法，其特征在于，所述物理参数为以下至少一种包括：深度值、反射率、角度值和颜色信息。
  6. 根据权利要求4所述的方法，其特征在于，所述空白区域中的像素点可以为非扫描点；所述非扫描点是指场景空间中未被扫描到的方向在图像中对应的像素点。
  7. 根据权利要求4所述的方法，其特征在于，所述空白区域中的像素点可以为天空点；所述天空点是指场景空间中被扫描到的方向但未接收到回波信息的像素点。
  8. 根据权利要求6或7所述的方法，其特征在于，确定所述空白区域中待插入的像素点是否为目标点包括：
    若待插入的像素点为非扫描点，则确定所述像素点是目标点；和/或，
    若待插入的像素点为天空点，则确定所述像素点不是目标点。
  9. 根据权利要求4所述的方法，其特征在于，确定所述空白区域中待插入的像素点是否为目标点包括：
    获取待插入的像素点周围像素点的物理参数的数值；
    比对所述物理参数的数值和参数阈值，得到比对结果；
    基于所述比对结果确定所述待插入的像素点是否为目标点。
  10. 根据权利要求9所述的方法，其特征在于，基于所述比对结果确定所述待插入的像素点是否为目标点包括：
    若所述对比结果表征所述物理参数的数值小于或等于所述参数阈值，则确定所述待插入的像素点是目标点；
    若所述对比结果表征所述物理参数的数值大于所述参数阈值，则确定所述待插入的像素点不是目标点。
  11. 根据权利要求4所述的方法，其特征在于，所述预设算法为插值算法。
  12. 根据权利要求11所述的方法，其特征在于，所述插值算法为以下至少一种：最近邻插值、线性插值、Lanczos插值算法、反距离加权法、样条插值法、离散平滑插值和趋势面光滑插值。
  13. 根据权利要求1所述的方法，其特征在于，基于所述第一平面图像中空白区域周围的像素点在所述空白区域内插入若干个像素点，得到第二平面图像包括：
    根据所述第一平面图像内各像素点的物理参数分割所述第一平面图像，得到所述第一平面图像包含的物体以及各物体内的空白区域；
    针对各物体的空白区域，根据所述空白区域周围的像素点在所述空白区域内插入若干个像素点，将各物体的空白区域插入像素点后的平面图像作为第二平面图像。
  14. 根据权利要求13所述的方法，其特征在于，分割方式包括以下至 少一种：语义分割和实例分割。
  15. 根据权利要求1所述的方法，其特征在于，对所述第二平面图像进行反投影变换，得到重建后三维的第二点云之后，所述方法还包括：
    基于所述第一点云修正所述第二点云，得到第三点云，所述第三点云包括所述第二点云，以及位于所述第一点云中的、且不位于所述第二点云中的点。
  16. 根据权利要求15所述的方法，其特征在于，基于所述第一点云修正所述第二点云包括：
    对比第一点云中各点的物理参数和第二点云中各点的物理参数；
    将物理参数不同的点增加到第二点云中，得到所述第三点云。
  17. 根据权利要求1所述的方法，其特征在于，基于所述第一平面图像中空白区域周围的像素点在所述空白区域内插入若干个像素点之前，所述方法还包括：
    根据预设离散算法对所述第一平面图像进行离散化；
    根据离散化后的第一平面图像确定所述空白区域。
  18. 根据权利要求17所述的方法，其特征在于，根据预设离散算法对投影变换后得到的平面图像进行离散化包括：
    将所述第一平面图像分割成多个区域，所述多个区域包括部分空白区域,所述空白区域包括未包含点的区域。
  19. 根据权利要求18所述的方法，其特征在于，所述任意两个区域内所包含的点的数量可以不同。
  20. 根据权利要求18所述的方法，其特征在于，各区域可以继续至少一次离散化。
  21. 根据权利要求18所述的方法，其特征在于，所述多个区域中至少部分区域的面积不同。
  22. 根据权利要求18所述的方法，其特征在于，基于所述第一平面图像中空白区域周围的像素点在所述空白区域内插入若干个像素点包括：
    根据所述空白区域的至少一个相邻的区域内像素点的物理参数，确定插入所述空白区域内的像素点的物理参数。
  23. 根据权利要求22所述的方法，其特征在于，基于所述第一平面图像中空白区域周围的像素点在所述空白区域内插入若干个像素点包括：
    根据所述至少一个相邻的区域的物理参数确定插入所述空白区域内的像素点的物理参数，其中，所述区域的物理参数是基于所述区域内与距离探测装置之间距离最近的点的物理参数确定的。
  24. 根据权利要求17所述的方法，其特征在于，所述预设离散算法包括以下至少一种：四叉树分割算法、均匀分割算法、普通分割算法、语义分割算法和实例分割算法。
  25. 根据权利要求1所述的方法，其特征在于，对所述第二平面图像进行反投影变换，得到重建后三维的第二点云之前，所述方法还包括：
    根据预设滤波算法对所述第二平面图像进行滤波；
    基于滤波后的第二平面图像进行反投影变换，得到重建后三维的第二点云。
  26. 根据权利要求1所述的方法，其特征在于，所述预设滤波算法包括以下至少一种：高斯滤波、均值滤波、限幅滤波法、中位值滤波法、递推平均滤波法、中位值平均滤波法、限幅平均滤波法、一阶滞后滤波法、加权递推平均滤波法、消抖滤波法和限幅消抖滤波法。
  27. 一种点云扫描系统，其特征在于，至少包括存储器和处理器；所述存储器通过通信总线和所述处理器连接，用于存储所述处理器可执行的计算机指令；所述处理器用于从所述存储器读取计算机指令以实现：
    基于给定的平面对三维的第一点云进行投影变换，得到第一平面图像；
    基于所述第一平面图像中空白区域周围的像素点在所述空白区域内插入若干个像素点，得到第二平面图像；
    对所述第二平面图像进行反投影变换，得到重建后三维的第二点云。
  28. 根据权利要求27所述的点云扫描系统，其特征在于，所述第一点 云由距离探测装置获取；所述给定的平面与所述距离探测装置的像平面存在映射关系。
  29. 根据权利要求28所述的点云扫描系统，其特征在于，所述给定的平面为所述距离探测装置的像平面。
  30. 根据权利要求27所述的点云扫描系统，其特征在于，所述处理器用于基于所述第一平面图像中空白区域周围的像素点在所述空白区域内插入若干个像素点包括：
    根据所述第一平面图像中的像素点确定所述第一平面图像中的空白区域；
    确定所述空白区域中待插入的像素点是否为目标点；
    若是目标点，则根据预设算法确定所述目标点的物理参数的数值。
  31. 根据权利要求30所述的点云扫描系统，其特征在于，所述物理参数为以下至少一种包括：深度值、反射率、角度值和颜色信息。
  32. 根据权利要求30所述的点云扫描系统，其特征在于，所述空白区域中的像素点可以为非扫描点；所述非扫描点是指场景空间中未被扫描到的方向在图像中对应的像素点。
  33. 根据权利要求30所述的点云扫描系统，其特征在于，所述空白区域中的像素点可以为天空点；所述天空点是指场景空间中被扫描到的方向但未接收到回波信息的像素点。
  34. 根据权利要求32或33所述的点云扫描系统，其特征在于，所述处理器用于确定所述空白区域中待插入的像素点是否为目标点包括：
    若待插入的像素点为非扫描点，则确定所述像素点是目标点；和/或，
    若待插入的像素点为天空点，则确定所述像素点不是目标点。
  35. 根据权利要求30所述的点云扫描系统，其特征在于，所述处理器用于确定所述空白区域中待插入的像素点是否为目标点包括：
    获取待插入的像素点周围像素点的物理参数的数值；
    比对所述物理参数的数值和参数阈值，得到比对结果；
    基于所述比对结果确定所述待插入的像素点是否为目标点。
  36. 根据权利要求35所述的点云扫描系统，其特征在于，所述处理器用于基于所述比对结果确定所述待插入的像素点是否为目标点包括：
    若所述对比结果表征所述物理参数的数值小于或等于所述参数阈值，则确定所述待插入的像素点是目标点；
    若所述对比结果表征所述物理参数的数值大于所述参数阈值，则确定所述待插入的像素点不是目标点。
  37. 根据权利要求30所述的点云扫描系统，其特征在于，所述预设算法为插值算法。
  38. 根据权利要求37所述的点云扫描系统，其特征在于，所述插值算法为以下至少一种：最近邻插值、线性插值、Lanczos插值算法、反距离加权法、样条插值法、离散平滑插值和趋势面光滑插值。
  39. 根据权利要求27所述的点云扫描系统，其特征在于，所述处理器用于基于所述第一平面图像中空白区域周围的像素点在所述空白区域内插入若干个像素点，得到第二平面图像包括：
    根据所述第一平面图像内各像素点的物理参数分割所述第一平面图像，得到所述第一平面图像包含的物体以及各物体内的空白区域；
    针对各物体的空白区域，根据所述空白区域周围的像素点在所述空白区域内插入若干个像素点，将各物体的空白区域插入像素点后的平面图像作为第二平面图像。
  40. 根据权利要求39所述的点云扫描系统，其特征在于，分割方式包括以下至少一种：语义分割和实例分割。
  41. 根据权利要求27所述的点云扫描系统，其特征在于，所述处理器用于对所述第二平面图像进行反投影变换，得到重建后三维的第二点云之后，还用于：
    基于所述第一点云修正所述第二点云，得到第三点云，所述第三点云包括所述第二点云，以及位于所述第一点云中的、且不位于所述第二点云 中的点。
  42. 根据权利要求41所述的点云扫描系统，其特征在于，所述处理器用于基于所述第一点云修正所述第二点云包括：
    对比第一点云中各点的物理参数和第二点云中各点的物理参数；
    将物理参数不同的点增加到第二点云中，得到所述第三点云。
  43. 根据权利要求27所述的点云扫描系统，其特征在于，所述处理器用于基于所述第一平面图像中空白区域周围的像素点在所述空白区域内插入若干个像素点之前，还用于：
    根据预设离散算法对所述第一平面图像进行离散化；
    根据离散化后的第一平面图像确定所述空白区域。
  44. 根据权利要求43所述的点云扫描系统，其特征在于，所述处理器用于根据预设离散算法对投影变换后得到的平面图像进行离散化包括：
    将所述第一平面图像分割成多个区域，所述多个区域包括部分空白区域,所述空白区域包括未包含点的区域。
  45. 根据权利要求44所述的点云扫描系统，其特征在于，所述任意两个区域内所包含的点的数量可以不同。
  46. 根据权利要求44所述的点云扫描系统，其特征在于，各区域可以继续至少一次离散化。
  47. 根据权利要求44所述的点云扫描系统，其特征在于，所述多个区域中至少部分区域的面积不同。
  48. 根据权利要求44所述的点云扫描系统，其特征在于，所述处理器用于基于所述第一平面图像中空白区域周围的像素点在所述空白区域内插入若干个像素点包括：
    根据所述空白区域的至少一个相邻的区域内像素点的物理参数，确定插入所述空白区域内的像素点的物理参数。
  49. 根据权利要求48所述的点云扫描系统，其特征在于，所述处理器用于基于所述第一平面图像中空白区域周围的像素点在所述空白区域内插 入若干个像素点包括：
    根据所述至少一个相邻的区域的物理参数确定插入所述空白区域内的像素点的物理参数，其中，所述区域的物理参数是基于所述区域内与距离探测装置之间距离最近的点的物理参数确定的。
  50. 根据权利要求43所述的点云扫描系统，其特征在于，所述预设离散算法包括以下至少一种：四叉树分割算法、均匀分割算法、普通分割算法、语义分割算法和实例分割算法。
  51. 根据权利要求27所述的点云扫描系统，其特征在于，所述处理器用于对所述第二平面图像进行反投影变换，得到重建后三维的第二点云之前，还用于：
    根据预设滤波算法对所述第二平面图像进行滤波；
    基于滤波后的第二平面图像进行反投影变换，得到重建后三维的第二点云。
  52. 根据权利要求27所述的点云扫描系统，其特征在于，所述预设滤波算法包括以下至少一种：高斯滤波、均值滤波、限幅滤波法、中位值滤波法、递推平均滤波法、中位值平均滤波法、限幅平均滤波法、一阶滞后滤波法、加权递推平均滤波法、消抖滤波法和限幅消抖滤波法。
  53. 一种可读存储介质，其特征在于，所述可读存储介质上存储有若干计算机指令，所述计算机指令被执行时实现权利要求1～26任一项所述方法的步骤。
PCT/CN2019/074630 2019-02-02 2019-02-02 增加点云采样密度的方法、点云扫描系统、可读存储介质 Ceased WO2020155159A1 (zh)

Priority Applications (3)

Application Number Priority Date Filing Date Title
PCT/CN2019/074630 WO2020155159A1 (zh) 2019-02-02 2019-02-02 增加点云采样密度的方法、点云扫描系统、可读存储介质
CN201980005414.4A CN111819602A (zh) 2019-02-02 2019-02-02 增加点云采样密度的方法、点云扫描系统、可读存储介质
US17/308,056 US20210256740A1 (en) 2019-02-02 2021-05-05 Method for increasing point cloud sampling density, point cloud processing system, and readable storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2019/074630 WO2020155159A1 (zh) 2019-02-02 2019-02-02 增加点云采样密度的方法、点云扫描系统、可读存储介质

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US17/308,056 Continuation US20210256740A1 (en) 2019-02-02 2021-05-05 Method for increasing point cloud sampling density, point cloud processing system, and readable storage medium

Publications (1)

Publication Number Publication Date
WO2020155159A1 true WO2020155159A1 (zh) 2020-08-06

Family

ID=71840688

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/074630 Ceased WO2020155159A1 (zh) 2019-02-02 2019-02-02 增加点云采样密度的方法、点云扫描系统、可读存储介质

Country Status (3)

Country Link
US (1) US20210256740A1 (zh)
CN (1) CN111819602A (zh)
WO (1) WO2020155159A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115760588A (zh) * 2022-09-09 2023-03-07 北京百度网讯科技有限公司 点云的校正方法和三维模型的生成方法、装置、设备
CN116452403A (zh) * 2023-06-16 2023-07-18 瀚博半导体(上海)有限公司 点云数据处理方法、装置、计算机设备及存储介质
CN119445298A (zh) * 2024-09-20 2025-02-14 同济大学 基于深度一致性多模态数据联合增强方法及系统

Families Citing this family (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113427168A (zh) * 2021-05-12 2021-09-24 广州中国科学院先进技术研究所 一种焊接机器人实时焊缝跟踪装置及方法
CN113838125B (zh) * 2021-09-17 2024-10-22 中国第一汽车股份有限公司 目标位置确定方法、装置、电子设备以及存储介质
CN114119692B (zh) * 2021-10-08 2025-04-22 江汉大学 一种刚性物体几何信息恢复方法、装置及存储介质
CN115601275A (zh) * 2022-09-07 2023-01-13 北京紫光展锐通信技术有限公司（Cn） 点云增广方法及装置、计算机可读存储介质、终端设备
CN121860860A (zh) * 2023-09-04 2026-04-14 深圳市速腾聚创科技有限公司 补点处理方法、设备及存储介质
CN117423102A (zh) * 2023-10-26 2024-01-19 腾讯科技（深圳）有限公司 点云数据处理方法以及相关设备
CN117830087B (zh) * 2024-01-02 2024-08-02 常州市新创智能科技有限公司 超声波检测中c扫图像填充方法、装置、设备及介质
CN118258356B (zh) * 2024-04-02 2024-09-06 慈溪市诚正建设工程检测有限公司 一种基于计算机视觉的建筑地基沉降检测方法
CN118657936B (zh) * 2024-08-21 2024-10-22 北京海天瑞声科技股份有限公司 目标检测方法、装置、电子设备及存储介质
CN119251422B (zh) * 2024-12-04 2025-05-02 山东省国土空间生态修复中心(山东省地质灾害防治技术指导中心、山东省土地储备中心) 一种地热资源勘查区三维地质模型构建方法及系统

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105825543A (zh) * 2016-03-31 2016-08-03 武汉大学 基于低空遥感影像的多视密集点云生成方法及系统
CN106023303A (zh) * 2016-05-06 2016-10-12 西安电子科技大学 一种基于轮廓有效性提高三维重建点云稠密程度的方法
CN106683173A (zh) * 2016-12-22 2017-05-17 西安电子科技大学 一种基于邻域块匹配提高三维重建点云稠密程度的方法
CN107194983A (zh) * 2017-05-16 2017-09-22 华中科技大学 一种基于点云与影像数据的三维可视化方法和系统
CN107830800A (zh) * 2017-10-26 2018-03-23 首都师范大学 一种基于车载扫描系统生成精细立面图的方法
EP3382645A2 (en) * 2017-03-27 2018-10-03 3Dflow srl Method for generation of a 3d model based on structure from motion and photometric stereo of 2d sparse images
CN109300190A (zh) * 2018-09-06 2019-02-01 百度在线网络技术（北京）有限公司 三维数据的处理方法、装置、设备和存储介质

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9811880B2 (en) * 2012-11-09 2017-11-07 The Boeing Company Backfilling points in a point cloud
CN107316325B (zh) * 2017-06-07 2020-09-22 华南理工大学 一种基于图像配准的机载激光点云与影像配准融合方法
GB2564642A (en) * 2017-07-10 2019-01-23 Nokia Technologies Oy Methods and apparatuses for panoramic image processing

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105825543A (zh) * 2016-03-31 2016-08-03 武汉大学 基于低空遥感影像的多视密集点云生成方法及系统
CN106023303A (zh) * 2016-05-06 2016-10-12 西安电子科技大学 一种基于轮廓有效性提高三维重建点云稠密程度的方法
CN106683173A (zh) * 2016-12-22 2017-05-17 西安电子科技大学 一种基于邻域块匹配提高三维重建点云稠密程度的方法
EP3382645A2 (en) * 2017-03-27 2018-10-03 3Dflow srl Method for generation of a 3d model based on structure from motion and photometric stereo of 2d sparse images
CN107194983A (zh) * 2017-05-16 2017-09-22 华中科技大学 一种基于点云与影像数据的三维可视化方法和系统
CN107830800A (zh) * 2017-10-26 2018-03-23 首都师范大学 一种基于车载扫描系统生成精细立面图的方法
CN109300190A (zh) * 2018-09-06 2019-02-01 百度在线网络技术（北京）有限公司 三维数据的处理方法、装置、设备和存储介质

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115760588A (zh) * 2022-09-09 2023-03-07 北京百度网讯科技有限公司 点云的校正方法和三维模型的生成方法、装置、设备
CN115760588B (zh) * 2022-09-09 2026-02-17 北京百度网讯科技有限公司 点云的校正方法和三维模型的生成方法、装置、设备
CN116452403A (zh) * 2023-06-16 2023-07-18 瀚博半导体(上海)有限公司 点云数据处理方法、装置、计算机设备及存储介质
CN116452403B (zh) * 2023-06-16 2023-09-01 瀚博半导体(上海)有限公司 点云数据处理方法、装置、计算机设备及存储介质
CN119445298A (zh) * 2024-09-20 2025-02-14 同济大学 基于深度一致性多模态数据联合增强方法及系统

Also Published As

Publication number Publication date
US20210256740A1 (en) 2021-08-19
CN111819602A (zh) 2020-10-23

Similar Documents

Publication Publication Date Title
US20210256740A1 (en) Method for increasing point cloud sampling density, point cloud processing system, and readable storage medium
US10983218B2 (en) Multiple pixel scanning LIDAR
US12625269B2 (en) Multiple resolution, simultaneous localization and mapping based on 3-D LIDAR measurements
WO2022126427A1 (zh) 点云处理方法、点云处理装置、可移动平台和计算机存储介质
Joe et al. Sensor fusion of two sonar devices for underwater 3D mapping with an AUV
WO2020243962A1 (zh) 物体检测方法、电子设备和可移动平台
US11328446B2 (en) Combining light-field data with active depth data for depth map generation
CN112020660A (zh) 具有分层功率控制的基于lidar的距离测量
EP2221639A1 (en) Systems and methods for detecting and analyzing objects
US20240036210A1 (en) Laser radar system, and spatial measurement device and method
US20210255289A1 (en) Light detection method, light detection device, and mobile platform
WO2020215252A1 (zh) 测距装置点云滤噪的方法、测距装置和移动平台
WO2021239054A1 (zh) 空间测量装置、方法、设备以及计算机可读存储介质
WO2022217520A1 (zh) 探测方法、装置、可移动平台及存储介质
EP3465249A1 (en) Multiple pixel scanning lidar
WO2020155142A1 (zh) 一种点云重采样的方法、装置和系统
WO2020177076A1 (zh) 一种探测装置初始状态标定方法及装置
CN206019574U (zh) 带有倾斜式窗口镜的三维激光扫描仪
WO2021253429A1 (zh) 数据处理方法、装置、激光雷达和存储介质
CN119126062A (zh) 实现roi的激光雷达架构、激光雷达架构的控制方法及装置
WO2020237663A1 (zh) 一种多通道激光雷达点云插值的方法和测距装置
WO2020107379A1 (zh) 应用于测距装置的反射率校正方法、测距装置
WO2022226984A1 (zh) 扫描视场的控制方法、测距装置和可移动平台
CN222014444U (zh) 一种透镜单元、激光雷达及机器人
KR101840328B1 (ko) 3d 레이저 스캐너

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19913391

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19913391

Country of ref document: EP

Kind code of ref document: A1