CN115453563A - Three-dimensional space dynamic object identification method, system and storage medium - Google Patents
Three-dimensional space dynamic object identification method, system and storage medium Download PDFInfo
- Publication number
- CN115453563A CN115453563A CN202211136420.3A CN202211136420A CN115453563A CN 115453563 A CN115453563 A CN 115453563A CN 202211136420 A CN202211136420 A CN 202211136420A CN 115453563 A CN115453563 A CN 115453563A
- Authority
- CN
- China
- Prior art keywords
- voxel
- data
- point cloud
- space
- dimensional space
- 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.)
- Pending
Links
Images
Classifications
-
- 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
- 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
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/48—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S17/00
- G01S7/4808—Evaluating distance, position or velocity data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/10—Image acquisition
- G06V10/12—Details of acquisition arrangements; Constructional details thereof
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Multimedia (AREA)
- Software Systems (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Databases & Information Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Remote Sensing (AREA)
- Radar, Positioning & Navigation (AREA)
- Medical Informatics (AREA)
- Computer Networks & Wireless Communication (AREA)
- Life Sciences & Earth Sciences (AREA)
- Electromagnetism (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a method, a system and a storage medium for identifying a dynamic object in a three-dimensional space, which are applied to the technical field of object identification, can accurately identify the dynamic object in the three-dimensional space and effectively improve the stability of the system. The method comprises the following steps: acquiring point cloud data of a target three-dimensional space; carrying out voxelization on the target three-dimensional space to obtain a plurality of space voxels; calculating a voxel value of each space voxel in a current point cloud frame according to the point cloud data; wherein the voxel values are determined from the point cloud data in the spatial voxels; updating the state accumulated data of each space voxel according to the voxel value; converting the state accumulated data into a voxel motion coefficient through a preset motion indication function; wherein the preset motion indication function comprises a non-linear function; and identifying and obtaining the dynamic object in the target three-dimensional space according to the voxel motion coefficient.
Description
Technical Field
The invention relates to the technical field of object recognition, in particular to a three-dimensional space dynamic object recognition method, a three-dimensional space dynamic object recognition system and a storage medium.
Background
In computer vision, the motion of an object is seen as a change in the environment, and the nature of the detection of moving objects in space is such that a corresponding change is recognized. The identification of the state of the spatial object has very wide practical value as an infrastructure for the upper layer. In the related art, the moving object is mainly identified by a background image proofreading or target identification mode, but the moving object identification by the background image proofreading mode is misjudged due to untimely updating of the background image, and the real-time performance and the stability are insufficient. The training target can only be identified in a target identification mode, the judgment range is limited greatly, the weather influence is easy to occur, and all-weather identification is difficult to achieve.
Disclosure of Invention
In order to solve at least one of the above technical problems, the present invention provides a method, a system and a storage medium for identifying a dynamic object in a three-dimensional space, which can identify the dynamic object in the three-dimensional space more accurately and effectively improve the stability of the system.
In one aspect, an embodiment of the present invention provides a method for identifying a three-dimensional dynamic object, including the following steps:
acquiring point cloud data of a target three-dimensional space;
voxelizing the target three-dimensional space to obtain a plurality of space voxels;
calculating a voxel value of each space voxel in a current point cloud frame according to the point cloud data; wherein the voxel values are determined from the point cloud data in the spatial voxels;
updating the state accumulated data of each space voxel according to the voxel value;
converting the state accumulated data into a voxel motion coefficient through a preset motion indication function; wherein the preset motion indication function comprises a non-linear function;
and identifying and obtaining the dynamic object in the target three-dimensional space according to the voxel motion coefficient.
The method for identifying the three-dimensional space dynamic object according to the embodiment of the invention at least has the following beneficial effects: in this embodiment, first, point cloud data of a target three-dimensional space is obtained, and a plurality of spatial voxels are obtained by performing voxelization on the target three-dimensional space. Then, the embodiment calculates the voxel value of each spatial voxel in the current point cloud frame according to the point cloud data, and updates the state accumulated data of each spatial voxel according to the voxel value. And then, converting the state accumulated data of each space voxel into a voxel motion coefficient through a preset motion indication function, so that a dynamic object in a target three-dimensional space is identified according to the voxel motion coefficient, and the dynamic object in the three-dimensional space is accurately identified. Meanwhile, the dynamic object in the three-dimensional space is identified in a manner of acquiring point cloud data in the target three-dimensional space, so that the problem that the image is greatly influenced by the environment can be effectively solved, the identification is carried out without depending on a background image, the problems of background image maintenance and instantaneity are solved, and the stability of the system is effectively improved.
According to some embodiments of the invention, the acquiring point cloud data of a target three-dimensional space comprises:
a preset point cloud acquisition device is arranged in the target three-dimensional space; the preset point cloud acquisition device comprises a laser radar, a structured light scanning device and a binocular stereoscopic vision device;
and acquiring the point cloud data of the target three-dimensional space in real time through the preset point cloud acquisition device.
According to some embodiments of the invention, the voxelizing the target three-dimensional space into a plurality of spatial voxels includes:
determining voxel size data of the space voxels according to preset voxel conditions; wherein the preset voxel condition comprises a preset error tolerance, a system computing capability and a resolution of the point cloud data;
and carrying out voxelization on the target three-dimensional space according to the voxel size data to obtain a plurality of space voxels.
According to some embodiments of the invention, the calculating a voxel value of each of the spatial voxels in the current point cloud frame from the point cloud data comprises:
determining the distribution data of the point cloud in each space voxel in the current point cloud frame according to the point cloud data;
determining cloud points exist in a first space voxel according to the distribution data, and setting the voxel value of the first space voxel to be in an activated state;
and determining that the cloud point does not exist in a second space voxel according to the distribution data, and setting the voxel value of the second space voxel to be in a static state.
According to some embodiments of the invention, the updating the state accumulation data for each of the spatial voxels according to the voxel values comprises:
determining the voxel value of the spatial voxel as the activation state, and accumulating the state accumulated data of the spatial voxel;
or determining the voxel value of the space voxel as the static state, and attenuating the state accumulated data of the space voxel.
According to some embodiments of the invention, after performing the step of identifying the dynamic object in the target three-dimensional space according to the voxel motion coefficient, the method further comprises:
constructing a four-dimensional point cloud according to the voxel motion coefficient and the point cloud data;
inputting the four-dimensional point cloud into a preset clustering algorithm for clustering analysis to obtain three-dimensional data of the dynamic object; wherein the three-dimensional data comprises length data, width data, and height data of the dynamic object.
According to some embodiments of the invention, the predetermined clustering algorithm comprises a density clustering algorithm;
inputting the four-dimensional point cloud into a preset clustering algorithm for clustering analysis to obtain three-dimensional data of the dynamic object, wherein the three-dimensional data comprises the following steps:
and carrying out clustering analysis on the four-dimensional point cloud through the density clustering algorithm to obtain the three-dimensional data.
On the other hand, an embodiment of the present invention further provides a three-dimensional space dynamic object recognition system, including:
the acquisition module is used for acquiring point cloud data of a target three-dimensional space;
the voxelization module is used for voxelizing the target three-dimensional space to obtain a plurality of space voxels;
the calculation module is used for calculating the voxel value of each space voxel in the current point cloud frame according to the point cloud data; wherein the voxel values are determined from the point cloud data in the spatial voxels;
the updating module is used for updating the state accumulated data of each space voxel according to the voxel value;
the conversion module is used for converting the state accumulated data into a voxel motion coefficient through a preset motion indication function; wherein the preset motion indication function comprises a non-linear function;
and the identification module is used for identifying and obtaining the dynamic object in the target three-dimensional space according to the voxel motion coefficient.
On the other hand, an embodiment of the present invention further provides a three-dimensional space dynamic object recognition system, including:
at least one processor;
at least one memory for storing at least one program;
when the at least one program is executed by the at least one processor, the at least one processor may implement the three-dimensional space dynamic object recognition method according to the above-described embodiment.
In another aspect, the present invention further provides a computer storage medium, in which a program executable by a processor is stored, and when the program executable by the processor is executed by the processor, the program is used to implement the three-dimensional space dynamic object identification method according to the above embodiments.
Drawings
FIG. 1 is a flow chart of a method for identifying a dynamic object in a three-dimensional space according to an embodiment of the present invention;
fig. 2 is a schematic block diagram of a three-dimensional space dynamic object recognition system according to an embodiment of the present invention.
Detailed Description
The embodiments described in the embodiments of the present application should not be construed as limiting the present application, and all other embodiments that can be obtained by a person skilled in the art without making any inventive step shall fall within the scope of protection of the present application.
In the following description, reference is made to "some embodiments" which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of the present application only and is not intended to be limiting of the application.
In computer vision, motion is generally considered to be a change in environment, and the nature of the detection of moving objects in space is such that changes in environment are recognized. The identification of the state of the space object has very wide practical value as an infrastructure for the upper layer. In the related art, for the identification of a moving object, a background image is usually saved, and the moving object is screened out by performing a calibration with the background image. However, for the background map-based collation scheme, the cost is increased due to the need of regular maintenance on the background map, and the background map update frequency is also considered, and is related to downstream services, so that various variable factors are introduced. Meanwhile, when the detection environment changes, if the background image is not updated in time according to the corresponding change, a misjudgment situation may be caused, and the accuracy and stability of the identification result may be greatly affected. In addition, for dynamic object recognition depending on a target recognition mode, only a pre-trained target can be recognized, but an untrained target cannot be recognized, and motion judgment cannot be performed, so that the application scenarios of the method are greatly limited. In addition, the target identification method based on the image is greatly influenced by weather, for example, in poor illumination conditions or rain and fog weather, all-weather identification is difficult to achieve, and the accuracy of object motion state identification is greatly influenced.
Based on this, an embodiment of the present invention provides a method, a system, and a storage medium for identifying a dynamic object in a three-dimensional space, which can identify the dynamic object in the three-dimensional space more accurately and effectively improve the stability of the system. Referring to fig. 1, the method of the embodiment of the present invention includes, but is not limited to, step S110, step S120, step S130, step S140, step S150, and step S160.
Specifically, the method application process of the embodiment of the invention includes, but is not limited to, the following steps:
s110: and acquiring point cloud data of the target three-dimensional space.
S120: and carrying out voxelization on the target three-dimensional space to obtain a plurality of space voxels.
S130: and calculating the voxel value of each space voxel in the current point cloud frame according to the point cloud data. The voxel value is determined according to point cloud data in the space voxel.
S140: and updating the state accumulated data of each space voxel according to the voxel value.
S150: and converting the state accumulated data into a voxel motion coefficient through a preset motion indication function. Wherein the preset motion indication function comprises a non-linear function.
S160: and identifying and obtaining the dynamic object in the target three-dimensional space according to the voxel motion coefficient.
In the working process of the embodiment, the embodiment first obtains point cloud data of a target three-dimensional space. In the embodiment, the point cloud data of the target three-dimensional space is obtained in real time by scanning the target three-dimensional space in real time. Then, this embodiment voxelizes the target three-dimensional space to obtain a plurality of spatial voxels. Specifically, a voxel (voxel) is a combined word of a pixel (pixel), a volume (volume), and an element (element), and is equivalent to a pixel in a three-dimensional space, and the pixel is a solid, and the whole target three-dimensional space is regarded as being composed of individual spatial voxels, and each block area in the target three-dimensional space corresponds to a corresponding voxel. Further, the embodiment calculates the voxel value of each spatial voxel in the current point cloud frame according to the cloud point data. Specifically, the voxel value of each spatial voxel is determined from the point cloud data in the spatial voxel. For example, in the embodiment, the current point cloud frame is obtained by acquiring point cloud data of a target three-dimensional space in real time. And then determining the voxel value of each space voxel in the current point cloud frame according to the point cloud data of the current point cloud frame. Then, the present embodiment updates the state accumulated data of each spatial voxel according to the calculated voxel value of each spatial voxel in the current point cloud frame. In this embodiment, each spatial voxel is correspondingly provided with state accumulation data, so that statistics is performed on the change situation of the voxel value of each spatial voxel through the state accumulation data. Further, the present embodiment converts the state accumulation data into voxel motion coefficients by presetting a motion indication function. Specifically, the preset motion indication function in the present embodiment includes a non-linear function. In the embodiment, the motion state of an object in a space is expressed through a voxel motion coefficient, and the expression capability of a detection scene is increased in a mode of converting state accumulated data into the voxel motion coefficient through a preset motion indication function. Since the state accumulation data is based on statistics, that is, the state accumulation data is linear data, the present embodiment converts the state accumulation data into voxel motion coefficients through some non-linear motion state indication functions to enhance the non-linear expression of the target three-dimensional space. The choice of the non-linear function in this embodiment is set according to the downstream required traffic. For example, when the voxel motion coefficient is needed to determine the probability that an object in the region corresponding to the spatial voxel is stationary, a function, such as a sigmod function, that maps the state accumulation data to the [0,1] interval may be selected. When objects which are absolutely static or static with a high probability in the target three-dimensional space scene need to be screened out, the preset motion indication function can be set as a selection function, as shown in the following formula (1):
wherein, x is the state accumulated data, and y is the voxel motion coefficient.
Further, the embodiment identifies and obtains the dynamic object in the target three-dimensional space according to the voxel motion coefficient. In this embodiment, after the state accumulated data is converted into the voxel motion coefficient by the preset motion indication function, the voxel motion coefficient of each spatial voxel is analyzed, so as to screen out the dynamic object in the target three-dimensional space. Illustratively, when the motion indicating function is preset, the function is selected as shown in equation (1) above. At this time, when x is greater than 90, the object corresponding to the spatial voxel is a stationary object with a high probability, and the corresponding motion coefficient is 1, otherwise, 0, so as to distinguish the dynamic object from the static object in the target three-dimensional space. According to the method, the dynamic object in the three-dimensional space is identified in a manner of acquiring the point cloud data of the target three-dimensional space, so that the problem that the image is greatly influenced by the environment is effectively solved, the background image does not need to be identified depending on the background image, the problems of background image maintenance and real-time performance are effectively solved, and the stability and reliability of the system are improved.
In some embodiments of the invention, point cloud data of a target three-dimensional space is obtained, including but not limited to:
and arranging a preset point cloud acquisition device in the target three-dimensional space. The preset point cloud acquisition device comprises a laser radar, a structured light scanning device and a binocular stereoscopic vision device.
And acquiring point cloud data of the target three-dimensional space in real time through a preset point cloud acquisition device.
In this embodiment, the preset point cloud collection device is disposed in the target three-dimensional space, so as to obtain the point cloud data of the target three-dimensional space in real time through the preset point cloud collection device. Specifically, the preset point cloud collection device in the embodiment comprises a laser radar, a structured light scanning device and a binocular stereoscopic vision device. In the embodiment, different preset point cloud acquisition devices are selected according to different target three-dimensional space scenes. For example, lidar generally performs well, but is relatively costly, and thus when not cost sensitive, one may choose to acquire point cloud data via lidar. And structured light scanning device receives illumination influence great, and the performance is relatively poor under the highlight, and detection distance is shorter, but it is good under low light environment, consequently can obtain point cloud data through structured light scanning device under the application scene that needs adaptation low light environment and not high to the distance requirement. Meanwhile, the binocular stereo vision device is based on a common camera, the cost is low, but the performance is poor in a dark scene and a weak texture scene, so that point cloud data can be acquired through the binocular stereo vision device when the illumination condition of the measuring environment is good and the texture of the scene is rich. In addition, in the process of setting the preset point cloud collection device in the target three-dimensional space, the performance of hardware, namely the performance of the preset point cloud collection device, needs to be combined. For example, when the preset point cloud acquisition device is a laser radar, the point cloud density is increased as much as possible under the condition of completely covering the detection area according to the relation between the area to be detected and the resolution, so as to improve the subsequent identification precision and accuracy. Meanwhile, the stability of fixing the device needs to be guaranteed in the process of deploying the preset point cloud acquisition device. It is easy to understand that when the preset point cloud acquisition device is unstable, the recognition result is prone to have a large error, which affects the recognition accuracy. And the stability of the device and the industrial personal computer network needs to be ensured, so that the real-time point cloud data acquired by the preset point cloud acquisition device can be timely sent out.
In some embodiments of the present invention, the target three-dimensional space is voxelized to obtain a number of spatial voxels, including but not limited to:
and determining voxel size data of the space voxel according to a preset voxel condition. The preset voxel condition comprises preset error tolerance, system computing capacity and resolution of point cloud data.
And carrying out voxelization on the target three-dimensional space according to the voxel size data to obtain a plurality of space voxels.
In this embodiment, firstly, voxel size data of a spatial voxel is determined according to a preset voxel condition, and then a target three-dimensional space is voxelized according to the voxel size data to obtain a plurality of spatial voxels. Specifically, the preset voxel condition includes a preset error tolerance, a system calculation capability, and a resolution of the point cloud data, and in this embodiment, the voxel size data of the spatial voxels is determined by combining the preset tolerance, the system calculation capability, and the resolution of the point cloud data, so that the target three-dimensional space is voxelized according to the voxel size data to obtain a plurality of spatial voxels. The voxels may be understood as a three-dimensional grid map with a fixed resolution, i.e. the target three-dimensional space is considered to be composed of several spatial voxels. Wherein, each block area in the target three-dimensional space corresponds to a corresponding voxel value. In this embodiment, the process of voxelizing the target three-dimensional space is equivalent to the process of modeling the target three-dimensional space. After the preset point cloud collection device is set, the position of the preset point cloud collection device is used as the origin of the voxel space in the embodiment, so that the calculation amount in the process of calculating the relationship between the point cloud and the space voxel is reduced. Meanwhile, the origin of the voxel space can be set by self according to downstream application requirements, at the moment, the rotation and translation matrix of converting the position of the preset point cloud acquisition device to the set origin of the voxel is obtained through measurement and calculation, and then the corresponding relation between the point cloud and the space voxel is obtained through conversion calculation of the matrix. Further, in this embodiment, the voxel size of the spatial voxel is determined according to a preset error tolerance, a system calculation capability, and a resolution of the point cloud data. Specifically, in the calculation capability and resolution bearable range of the system, the voxel size data value is made as small as possible, so as to improve the accuracy and precision of the identification. In this embodiment, when the resolution of the point cloud data is too low, if the number of spatial voxels is small, it is easy to cause that the spatial voxels corresponding to the stationary object are difficult to accumulate enough state accumulated data, and it is difficult to distinguish a moving object from the stationary object. In the voxel size data, the number of voxels in a unit space is set to be less than half of the number of point clouds in the unit space.
In addition, the present embodiment needs to be balanced in real-time and accuracy. For example, when the real-time performance of the system is required to be high, the frame rate needs to be increased, and due to the limitation of corresponding hardware, the voxel size data of the spatial voxels is increased. Conversely, when the real-time requirement is low and the recognition accuracy requirement is high, the frame rate can be lowered and the voxel size data can be reduced. When the downstream service has low requirement on the real-time performance of identification and has high requirement on the accuracy, the embodiment performs frame extraction on the point cloud so as to use the computational power saved by frame extraction for reducing the size of the spatial voxel. For example, when the frame rate of the preset point cloud collection device is 10 frames per second, 10 frames of point cloud data need to be processed per second without performing frame extraction. When the hardware performance is hard to meet the requirement or the requirement on the frame rate is low on the application level, that is, it is not necessary to reach 10 frames per second, the present embodiment reduces the amount of calculation of the system by means of frame extraction processing. In the embodiment, the acquired 10 frames of point cloud data are processed once every other preset frame, and the point cloud frames which are not processed are discarded, so that the frame extraction processing of the point cloud data is realized.
In some embodiments of the present invention, the voxel values of the respective spatial voxels in the current point cloud frame are calculated from the point cloud data, including but not limited to:
and determining the distribution data of the point cloud in each space voxel in the current point cloud frame according to the point cloud data.
And determining the existence of cloud points in the first space voxel according to the distribution data, and setting the voxel value of the first space voxel to be in an activated state.
And determining that cloud points do not exist in the second space voxel according to the distribution data, and setting the voxel value of the second space voxel to be in a static state.
In this embodiment, first, the distribution data of the point cloud in each spatial voxel in the current point cloud frame is determined according to the acquired point cloud data, so as to determine whether cloud points exist in each spatial voxel according to the respective data, and set a voxel value of the corresponding spatial voxel. Specifically, when a certain spatial voxel, namely a first spatial voxel, is determined according to distribution data of point clouds in each spatial voxel in a current point cloud frame, and the point clouds are distributed, the voxel value of the first spatial voxel is set in an activated state. It is easy to understand that, in this embodiment, if there is a cloud point in the current point cloud frame that falls within a certain spatial voxel, the spatial voxel is activated, that is, the voxel value of the spatial voxel is set to be in an activated state. Accordingly, when it is determined from the distribution data that there is no cloud point in a certain spatial voxel, i.e., a second spatial voxel, the voxel value of the second spatial voxel is set to a stationary state. In this embodiment, each spatial voxel is divided into a first spatial voxel and a second spatial voxel, and the first spatial voxel and the second spatial voxel are classified according to whether cloud points exist in the spatial voxels. In addition, since the number and resolution of the point clouds falling on the spatial voxels are related, and the distribution at each position in the same frame of point clouds is not uniform, the resolution and distribution of the point clouds have no correlation with whether the object is in a motion state or not. Therefore, the present embodiment sets the voxel value of each spatial voxel by determining whether a point in the current point cloud frame falls within the spatial voxel. For example, if there is a cloud point in the current point cloud frame that falls within a certain spatial voxel, the voxel value of the spatial voxel is set to 1, i.e., activated. Conversely, when no cloud point falls within a certain spatial voxel, the voxel value of the spatial voxel is set to 0, i.e., a stationary state.
In some embodiments of the present invention, the state accumulation data for each spatial voxel is updated according to voxel value, including but not limited to:
and determining the voxel value of the space voxel as an activated state, and accumulating the state accumulated data of the space voxel.
Or determining the voxel value of the space voxel to be in a static state, and attenuating the state accumulated data of the space voxel.
In this embodiment, the present embodiment updates the state accumulation data of each spatial voxel through an accumulation and attenuation mechanism. Specifically, in this embodiment, a voxel value of a spatial voxel is determined, and when the voxel value of the spatial voxel is in an active state, the state accumulated data of the spatial voxel is accumulated, and when the voxel value of the spatial voxel is determined to be in a static state, the state accumulated data of the spatial voxel is attenuated. For example, in this embodiment, when the voxel value of a spatial voxel is determined to be in the activated state, 1 is added to the state accumulation data of the spatial voxel. And when the voxel value of the spatial voxel is determined to be in a static state, subtracting 1 from the state accumulated data of the spatial voxel. The embodiment introduces an accumulation and attenuation mechanism, so that after statistics of continuous multiple frames, the motion state of an object in a target three-dimensional space can be reflected through state accumulation data. For example, when there is no object in a space voxel for a long time, the state accumulated data of the region represented by the space voxel will be attenuated to a lower value, and when a space voxel is always activated, that is, there is an object in the space voxel, the state accumulated data of the space voxel will be accumulated to a larger value, so as to distinguish a dynamic object from a static object in the space according to the state accumulated data of the space voxel. It is easy to understand that after statistics of accumulation and attenuation mechanisms of continuous multiple frames, state accumulated data of fixed objects and moving objects in a target three-dimensional space can have differences. For example, after statistics of accumulation and attenuation mechanisms of consecutive multiple frames, state accumulation data of a static object in a target three-dimensional space is accumulated to a larger value, while a dynamic object has accumulation and attenuation processes, and state accumulation data of the dynamic object has a certain difference from the static object. It should be noted that, in this embodiment, the value range of the state accumulated data is [0,100], that is, the minimum value of the state accumulated data after being attenuated for multiple times is 0, and the maximum value of the state accumulated data after being accumulated for multiple times is 100.
It should be noted that in some embodiments of the present invention, the accumulation coefficient and the attenuation coefficient of the accumulation and attenuation mechanism may be customized according to the downstream application. For example, when the motion data in the detected scene is dense and frequent, the attenuation coefficient and the accumulation coefficient are increased to highlight the characteristics of the moving object and the static object, so that the capability of indicating the motion state of the object by the state accumulation data is effectively improved.
In some embodiments of the present invention, after performing the step of identifying a dynamic object in the target three-dimensional space according to the voxel motion coefficient, the three-dimensional space dynamic identification method provided by this embodiment further includes, but is not limited to:
and constructing a four-dimensional point cloud according to the voxel motion coefficient and the point cloud data.
And inputting the four-dimensional point cloud into a preset clustering algorithm for clustering analysis to obtain three-dimensional data of the dynamic object. Wherein the three-dimensional data comprises length data, width data and height data of the dynamic object.
In this embodiment, after the dynamic object in the target three-dimensional space is identified and obtained according to the voxel motion coefficient, the dynamic object is screened from the target three-dimensional space by using a preset clustering algorithm. Specifically, in this embodiment, the voxel motion coefficient is first combined with the point cloud data to construct a four-dimensional point cloud. In the embodiment, the voxel motion coefficients of the spatial voxels are used as fourth-dimensional data and are fused with the three-dimensional point cloud data to obtain four-dimensional point cloud data, namely four-dimensional point cloud, so that the clustering effect is effectively improved. Further, the four-dimensional point cloud is input into a preset clustering algorithm for clustering analysis, and three-dimensional data of the dynamic object is obtained. In particular, clustering algorithms can divide a collection of physical or abstract objects into classes composed of similar objects. Wherein the cluster generated by the clustering algorithm is a set of data objects that are similar to objects in the same cluster and distinct from objects in other clusters. In the embodiment, the constructed four-dimensional point cloud is used as the input of a preset clustering algorithm, so that the dynamic objects and the static objects in the target three-dimensional space are subjected to clustering analysis. According to the difference of the voxel motion coefficients of the dynamic object and the static object, the dynamic object can be effectively screened from the target three-dimensional space. Meanwhile, the three-dimensional data of the dynamic object, such as length data, width data and height data of the dynamic object, can be obtained by calculation by combining the point cloud data corresponding to the spatial voxel, that is, the three-dimensional data of each cloud point corresponding to the spatial voxel.
It is to be understood that, in some embodiments of the present invention, after the dynamic object is screened out by the preset clustering algorithm and corresponding three-dimensional data is obtained, the present embodiment may further combine some multi-target tracking algorithms to obtain data related to the time sequence of each moving object, for example, the moving speed, the moving distance, etc. of the dynamic object in the detection of consecutive multiple frames.
In some embodiments of the invention, the predetermined clustering algorithm comprises a density clustering algorithm. Correspondingly, inputting the four-dimensional point cloud into a preset clustering algorithm for clustering analysis to obtain three-dimensional data of the dynamic object, including but not limited to:
and carrying out clustering analysis on the four-dimensional point cloud through a density clustering algorithm to obtain three-dimensional data.
In this embodiment, the four-dimensional point cloud in the target three-dimensional space is subjected to cluster analysis by a density clustering algorithm. Specifically, the density clustering algorithm defines clusters as the maximum set of points connected in density, can divide areas with high enough density into clusters, and can perform clustering of various shapes in a noisy spatial database, thereby effectively improving the accuracy and reliability of clustering analysis. The present embodiment inputs the four-dimensional point cloud into a Density Clustering algorithm, such as DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm, to perform cluster analysis on the point cloud in the target three-dimensional space. The cloud point sets of the dynamic objects are obtained after the DBSCAN algorithm analysis, and the three-dimensional data of the dynamic objects are obtained through calculation according to the three-dimensional information contained in the cloud points in the cloud point sets of the dynamic objects.
An embodiment of the present invention further provides a three-dimensional space dynamic object recognition system, including:
and the acquisition module is used for acquiring point cloud data of the target three-dimensional space.
And the voxelization module is used for voxelizing the target three-dimensional space to obtain a plurality of space voxels.
And the calculation module is used for calculating the voxel value of each space voxel in the current point cloud frame according to the point cloud data. Wherein the voxel value is determined from the point cloud data in the spatial voxel.
And the updating module is used for updating the state accumulated data of each space voxel according to the voxel value.
And the conversion module is used for converting the state accumulated data into a voxel motion coefficient through a preset motion indication function. Wherein the preset motion indication function comprises a non-linear function.
And the identification module is used for identifying and obtaining the dynamic object in the target three-dimensional space according to the voxel motion coefficient.
Referring to fig. 2, an embodiment of the present invention further provides a three-dimensional space dynamic object recognition system, including:
at least one processor 210.
At least one memory 220 for storing at least one program.
When the at least one program is executed by the at least one processor 210, the at least one processor may implement the three-dimensional space dynamic object recognition method as described in the above embodiments.
An embodiment of the present invention also provides a computer-readable storage medium storing computer-executable instructions for execution by one or more control processors, e.g., to perform the steps described in the above embodiments.
One of ordinary skill in the art will appreciate that all or some of the steps, systems, and methods disclosed above may be implemented as software, firmware, hardware, and suitable combinations thereof. Some or all of the physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application specific integrated circuit. Such software may be distributed on computer readable media, which may include computer storage media (or non-transitory media) and communication media (or transitory media). The term computer storage media includes volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data, as is well known to those skilled in the art. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital Versatile Disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium which can be used to store the desired information and which can accessed by a computer. In addition, communication media typically embodies computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism and includes any information delivery media as known to those skilled in the art.
While the preferred embodiments of the present invention have been described, the present invention is not limited to the above embodiments, and those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention, and such equivalent modifications or substitutions are to be included within the scope of the present invention defined by the appended claims.
Claims (10)
1. A three-dimensional space dynamic object identification method is characterized by comprising the following steps:
acquiring point cloud data of a target three-dimensional space;
voxelizing the target three-dimensional space to obtain a plurality of space voxels;
calculating the voxel value of each space voxel in the current point cloud frame according to the point cloud data; wherein the voxel values are determined from the point cloud data in the spatial voxels;
updating the state accumulated data of each space voxel according to the voxel value;
converting the state accumulated data into a voxel motion coefficient through a preset motion indication function; wherein the preset motion indication function comprises a non-linear function;
and identifying and obtaining the dynamic object in the target three-dimensional space according to the voxel motion coefficient.
2. The method for identifying the dynamic object in the three-dimensional space according to claim 1, wherein the acquiring point cloud data of the target three-dimensional space comprises:
a preset point cloud acquisition device is arranged in the target three-dimensional space; the preset point cloud acquisition device comprises a laser radar, a structured light scanning device and a binocular stereoscopic vision device;
and acquiring the point cloud data of the target three-dimensional space in real time through the preset point cloud acquisition device.
3. The method according to claim 1, wherein the voxelizing the target three-dimensional space to obtain a plurality of spatial voxels comprises:
determining voxel size data of the space voxels according to preset voxel conditions; wherein the preset voxel condition comprises a preset error tolerance, a system computing capability and a resolution of the point cloud data;
and carrying out voxelization on the target three-dimensional space according to the voxel size data to obtain a plurality of space voxels.
4. The method according to claim 1, wherein the calculating voxel values of the spatial voxels in the current point cloud frame according to the point cloud data comprises:
determining distribution data of point clouds in the current point cloud frame in each space voxel according to the point cloud data;
determining cloud points exist in a first space voxel according to the distribution data, and setting the voxel value of the first space voxel to be in an activated state;
and determining that the cloud point does not exist in a second space voxel according to the distribution data, and setting the voxel value of the second space voxel to be in a static state.
5. The method according to claim 4, wherein the updating the state accumulation data of each of the spatial voxels according to the voxel value comprises:
determining the voxel value of the spatial voxel as the activation state, and accumulating the state accumulated data of the spatial voxel;
or determining the voxel value of the space voxel as the static state, and attenuating the state accumulated data of the space voxel.
6. The method for identifying a dynamic object in three-dimensional space according to claim 1, wherein after the step of identifying a dynamic object in the target three-dimensional space according to the voxel motion coefficients is performed, the method further comprises:
constructing a four-dimensional point cloud according to the voxel motion coefficient and the point cloud data;
inputting the four-dimensional point cloud into a preset clustering algorithm for clustering analysis to obtain three-dimensional data of the dynamic object; wherein the three-dimensional data comprises length data, width data, and height data of the dynamic object.
7. The method according to claim 6, wherein the predetermined clustering algorithm comprises a density clustering algorithm;
inputting the four-dimensional point cloud into a preset clustering algorithm for clustering analysis to obtain three-dimensional data of the dynamic object, wherein the three-dimensional data comprises the following steps:
and carrying out clustering analysis on the four-dimensional point cloud through the density clustering algorithm to obtain the three-dimensional data.
8. A three-dimensional space dynamic object recognition system, comprising:
the acquisition module is used for acquiring point cloud data of a target three-dimensional space;
the voxelization module is used for voxelizing the target three-dimensional space to obtain a plurality of space voxels;
the calculation module is used for calculating the voxel value of each space voxel in the current point cloud frame according to the point cloud data; wherein the voxel values are determined from the point cloud data in the spatial voxels;
the updating module is used for updating the state accumulated data of each space voxel according to the voxel value;
the conversion module is used for converting the state accumulated data into a voxel motion coefficient through a preset motion indication function; wherein the preset motion indication function comprises a non-linear function;
and the identification module is used for identifying and obtaining the dynamic object in the target three-dimensional space according to the voxel motion coefficient.
9. A three-dimensional space dynamic object recognition system, comprising:
at least one processor;
at least one memory for storing at least one program;
when executed by the at least one processor, cause the at least one processor to implement the three-dimensional space dynamic object recognition method of any one of claims 1 to 7.
10. A computer storage medium in which a processor-executable program is stored, wherein the processor-executable program, when executed by the processor, is configured to implement the three-dimensional space dynamic object recognition method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211136420.3A CN115453563A (en) | 2022-09-19 | 2022-09-19 | Three-dimensional space dynamic object identification method, system and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211136420.3A CN115453563A (en) | 2022-09-19 | 2022-09-19 | Three-dimensional space dynamic object identification method, system and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115453563A true CN115453563A (en) | 2022-12-09 |
Family
ID=84305345
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211136420.3A Pending CN115453563A (en) | 2022-09-19 | 2022-09-19 | Three-dimensional space dynamic object identification method, system and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115453563A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116486283A (en) * | 2023-01-09 | 2023-07-25 | 深圳优立全息科技有限公司 | Real-time point cloud target detection method and device based on voxel division |
-
2022
- 2022-09-19 CN CN202211136420.3A patent/CN115453563A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116486283A (en) * | 2023-01-09 | 2023-07-25 | 深圳优立全息科技有限公司 | Real-time point cloud target detection method and device based on voxel division |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112526513B (en) | Millimeter wave radar environment map construction method and device based on clustering algorithm | |
| CN110472553B (en) | Target tracking method, computing device and medium for fusion of image and laser point cloud | |
| JP2023523243A (en) | Obstacle detection method and apparatus, computer device, and computer program | |
| CN111602138B (en) | Object detection system and method based on artificial neural network | |
| CN110187334B (en) | Target monitoring method and device and computer readable storage medium | |
| CN111091592B (en) | Image processing method, image processing apparatus, electronic device, and readable storage medium | |
| CN111582054B (en) | Point cloud data processing method and device and obstacle detection method and device | |
| CN111144213B (en) | Object detection method and related equipment | |
| CN110390706B (en) | Object detection method and device | |
| CN111553946B (en) | Method and device for removing ground point cloud and method and device for detecting obstacle | |
| CN112508803B (en) | Denoising method and device for three-dimensional point cloud data and storage medium | |
| CN112149476B (en) | Target detection method, device, equipment and storage medium | |
| CN115453563A (en) | Three-dimensional space dynamic object identification method, system and storage medium | |
| CN114528941A (en) | Sensor data fusion method and device, electronic equipment and storage medium | |
| CN117372995A (en) | Vehicle drivable region detection method, related device and storage medium | |
| CN114863108B (en) | Method, system, electronic equipment and computer readable storage medium for point cloud processing | |
| CN116243273A (en) | Photon counting laser radar data filtering method and device | |
| CN113256574B (en) | Three-dimensional target detection method | |
| CN113221709B (en) | Method and device for identifying user motion and water heater | |
| CN114519400A (en) | Segmentation result evaluation method and device, electronic equipment and readable storage medium | |
| CN114359891A (en) | Three-dimensional vehicle detection method, system, device and medium | |
| CN112508970A (en) | Point cloud data segmentation method and device | |
| CN115994934B (en) | Data time alignment method and device and domain controller | |
| CN117437498A (en) | Model training and data calibration method, terminal and computer readable storage medium | |
| WO2023073987A1 (en) | Electronic control device and object identification method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |