CN114200477A - Laser three-dimensional imaging radar ground target point cloud data processing method - Google Patents

Laser three-dimensional imaging radar ground target point cloud data processing method Download PDF

Info

Publication number
CN114200477A
CN114200477A CN202111520246.8A CN202111520246A CN114200477A CN 114200477 A CN114200477 A CN 114200477A CN 202111520246 A CN202111520246 A CN 202111520246A CN 114200477 A CN114200477 A CN 114200477A
Authority
CN
China
Prior art keywords
point
ground
data
target
point cloud
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
Application number
CN202111520246.8A
Other languages
Chinese (zh)
Inventor
孟宏峰
奚银
强晶晶
陈蕙心
王力
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.)
Shanghai Radio Equipment Research Institute
Original Assignee
Shanghai Radio Equipment Research Institute
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 Shanghai Radio Equipment Research Institute filed Critical Shanghai Radio Equipment Research Institute
Priority to CN202111520246.8A priority Critical patent/CN114200477A/en
Publication of CN114200477A publication Critical patent/CN114200477A/en
Pending legal-status Critical Current

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
    • G01S7/4802—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S17/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • General Physics & Mathematics (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Electromagnetism (AREA)
  • Optical Radar Systems And Details Thereof (AREA)

Abstract

The invention provides a laser three-dimensional imaging radar ground target point cloud data processing method, which is used for carrying out ground target detection and identification on laser three-dimensional imaging radar point cloud data and outputting a suspected interesting target and comprises the following steps: s1, acquiring a point cloud data set of time and space registration; s2, filtering non-ground point data to obtain a ground point data set; s3, extracting, dividing and summarizing data of each ground target, and clustering and dividing the data into respective data subsets to obtain a ground target clustering set; s4, identifying the interested target based on the ground target clustering set, and outputting the suspected interested target. The method is used for solving the problems of detection and identification of the ground target in a complex scene, more three-dimensional information of the target is kept, the operation speed is improved, and the method is more suitable for real-time processing of point cloud data and more suitable for engineering application.

Description

Laser three-dimensional imaging radar ground target point cloud data processing method
Technical Field
The invention relates to the field of laser imaging, in particular to a laser three-dimensional imaging radar ground target point cloud data processing method.
Background
The laser radar is one of the most important technical means for solving three-dimensional imaging, and is also a technology which is vigorously developed in developed countries such as the United states and the Japan in long-term planning in a plurality of fields such as topographic mapping and ocean monitoring. The laser radar works in a manner basically consistent with that of the microwave radar, but the laser radar has a series of specific advantages: the high-angle resolution, the high-distance resolution and the high-speed resolution are high in anti-interference capability, high in target stealth resistance, wide in speed measurement range and stable in imaging performance, position information and speed information of a target can be obtained, and the high-angle resolution, the high-distance resolution and the high-speed resolution have the advantages of being small in size and light in weight compared with a microwave radar, and lay a foundation for application of the high-angle resolution, the high-distance resolution and the high-speed resolution in various aspects.
The laser three-dimensional imaging radar acquires the distance information of a target, the distance image is the distribution of target distances formed by the discrete ranging of space targets, the point cloud data of the distance image describes the surface geometric shape of the target, and compared with visible light and infrared images, the laser three-dimensional imaging radar is not influenced by the target temperature, illumination shadow and the surface texture of the target, so that more reliable geometric characteristics of the target can be obtained.
In the process of detecting a target scene by the laser three-dimensional imaging radar, imaging data is seriously interfered by noise such as distance abnormality, loss information, distance measurement errors and the like due to the influence of various factors such as a carrying platform, a sensor, a natural environment, the target scene and the like. Therefore, the research on the noise suppression algorithm of the laser radar data becomes an important step in the processing of the laser radar target point cloud data. Range anomalies are typically evenly distributed over the range gate; the dropout information corresponds to a special distance anomaly, which is usually distributed to a certain set distance value; the distance measurement error belongs to Gaussian noise which takes an actual distance value as a mean value and takes laser radar distance measurement precision as variance. The distribution of distance measurement errors is relatively simple, the correlation between an error distance value and an actual distance value and the distance values of surrounding observation points is strong, and the error distance value can be filtered by using a traditional filtering algorithm (such as weighted mean filtering and the like); distance anomalies (dropping information is also regarded as a distance anomaly) can occur in various situations, the distribution of the distance anomalies is more complex, and the distance anomalies are more difficult to filter, so that the distance anomaly noise suppression becomes an important point in the research of a laser radar data noise suppression algorithm. Most of the existing algorithms judge whether the current pixel is abnormal in distance or not or directly uses the statistical value estimation (median filtering and mean filtering) of the neighborhood pixels based on the similarity of the distance values of the current pixel and most of the pixels in the neighborhood, although the effect is still good when the common edge target is processed, the denoising effect is not ideal enough when the special condition of a linear target is met; and most algorithms are difficult to consider noise removal performance and execution efficiency. The laser three-dimensional imaging radar point cloud data processing is similar to a target identification algorithm based on a traditional two-dimensional image, and mainly comprises the stages of data preprocessing, target detection, segmentation, feature extraction and classification identification. Related scholars and research institutions at home and abroad have conducted extensive research on a target detection method based on laser radar imaging data, and various target detection methods are developed according to different imaging data and target scene characteristics. According to different algorithm principles, the method can be roughly divided into four categories: the system comprises an image processing-based target detection algorithm, a correlation filtering-based target detection algorithm, a feature matching-based target detection algorithm and a ground digital elevation model estimation-based target detection algorithm.
The image processing-based target detection algorithm has the premise that the ground in an imaging target scene is relatively flat, the number of shelters is small, the targets, the ground and various ground objects are obviously different, and a good effect is achieved in the laser radar image application under the conditions of large scene, multiple targets and less shelters.
The imaging data generally processed by the target detection algorithm based on the correlation filtering is a two-dimensional image, such as a range profile or a height profile, and the three-dimensional point cloud data is difficult to directly process, which is similar to the target detection algorithm based on the image processing.
The two methods work in a two-dimensional image space, the operation processing speed is high, the occupied space is relatively small, but the space geometric characteristics of the target, such as a space normal vector, a space curvature and a space connection relation, cannot be effectively described, and the unique advantages of the laser radar sensor are lost.
In the algorithm of three-dimensional image processing, a target detection algorithm based on feature matching needs prior information such as a target model and a corresponding rotating image feature database thereof, and generally needs to be obtained by means of laser radar imaging data simulation, and when feature matching is performed on target scene data points and the target model, the calculation amount is large.
The target detection algorithm based on the ground digital elevation model estimation mainly has the idea that point cloud filtering removes non-ground points in point cloud data and only retains the process of the ground points.
The invention provides a ground target point cloud data processing method based on ground estimation by combining the four data processing algorithms and combining engineering application practice, considering not only the operation amount of the algorithms but also the robustness of the algorithms applied to different scenes.
Disclosure of Invention
The invention aims to provide a laser three-dimensional imaging radar ground target point cloud data processing method, which aims to solve the problems of ground target detection and identification in a complex scene, more retains three-dimensional information of a target, improves the operation speed, is more suitable for real-time processing of point cloud data and is more suitable for engineering application.
In order to achieve the purpose, the invention is realized by the following technical scheme:
a laser three-dimensional imaging radar ground target point cloud data processing method is used for carrying out ground target detection and identification on laser three-dimensional imaging radar point cloud data and outputting a suspected interesting target, and comprises the following steps:
s1, acquiring a time and space registration point cloud data set based on the laser three-dimensional imaging radar point cloud data;
s2, filtering non-ground point data based on the point cloud data set of the time and space registration to obtain a ground point data set;
s3, extracting, dividing and summarizing data of each ground target based on the ground point data set, and clustering and dividing the data into respective data subsets to obtain a ground target clustering set;
s4, identifying the interested target based on the ground target clustering set, and outputting the suspected interested target.
Preferably, step S2 includes:
s21, based on the point cloud data set of the time and space registration, filtering noise points and abnormal points to obtain a normal point data set;
and S22, filtering non-ground point data based on the normal point data set to obtain the ground point data set.
Preferably, step S21 includes:
s211, constructing a k-d tree point cloud data set by adopting a k-d tree algorithm based on the time and space registration point cloud data set;
s212, based on the k-d tree point cloud data set, noise points and abnormal points are filtered out, and the normal point data set is obtained.
Preferably, step S212 is implemented by a k-nearest neighbor query algorithm.
Preferably, step S22 is implemented by a modified extended window elevation threshold filtering algorithm, comprising the steps of:
s221, initialization:
setting an iteration ordinal number: i is 1;
setting the first square cell C in the normal point data seti: selecting the data point with the minimum height in the normal data set, and taking the data point as CiAnd then the height of the vehicle object of interest is set to CiLength of side of
Figure BDA0003407042520000041
Set up Ci；
Set up CiS of the initial slope valuei；
S222, at CiSelecting a point with the minimum elevation as a seed point;
s223, based on CiLength of side of
Figure BDA0003407042520000042
And SiCalculating the height difference threshold value HiThe calculation formula is as follows:
Figure BDA0003407042520000043
s224, calculating CiThe elevation difference H between each data point and the seed point is calculated, and the elevation difference H is compared with the elevation difference HiAnd (3) comparison:
if h is>HiThen the data point is discarded;
otherwise, adding the data point into a ground point data set;
s225, update Si；
S226、i＝i+1；
A square unit Ci-1Extension to CiWherein, CiWith Ci-1Is central, CiHas an area of C i-12 times the area;
judgment CiWhether out of range of normal point data sets:
if yes, exiting iteration, and generating a ground point data set at the moment;
otherwise, the process returns to S222.
Preferably, step S225 includes:
a1, selection of CiThe maximum point p of elevation inmaxAnd elevation minimum point pminTo find CiMean value of ground points pave；
A2 finding pmaxRelative to paveIs inclined atmaxAnd p isaveRelative to pminIs inclined atmin；
A3, calculating SiThe calculation formula is as follows:
Si＝(Smax+Smin)/2。
preferably, step S3 includes:
s31, identifying the data of the ground target based on the ground point data set, and filtering non-ground target data to obtain a ground target data set;
and S32, based on the ground target data set, segmenting and summarizing data points of each ground target, and dividing data points related to the same ground target into the same subset to obtain the ground target clustering set.
Preferably, step S31 includes:
setting a ground target height threshold according to the ground-off relative height of the ground target, and then sequentially comparing each data point in the ground point data set with the ground target height threshold:
adding a data point to the ground target data set if the elevation of the data point is within the ground target elevation threshold;
otherwise, the data point is discarded.
Preferably, step S32 is implemented by a euclidean clustering algorithm.
Preferably, step S4 includes:
setting a discrimination rule for each interested target, identifying the interested target for each data subset in the ground target clustering set based on the discrimination rule, and finally outputting the suspected interested target;
wherein, the rule of judgement includes: the discrimination feature of the object of interest and a discrimination threshold for each of the discrimination features.
In summary, compared with the prior art, the method for processing the ground target point cloud data of the laser three-dimensional imaging radar has the following beneficial effects:
the method solves the technical problems that the traditional laser three-dimensional imaging radar point cloud data processing method cannot effectively describe the space geometric characteristics of the target, loses a large amount of information such as space normal vectors, space curvatures, space connection relations and the like, or the data post-processing method applied to the surveying and mapping field has large calculation amount and complex algorithm, and solves the processing problem when the point cloud data are irregularly arranged, has more three-dimensional information of the target, improves the calculation speed, can process the point cloud data in real time, and can quickly realize engineering.
Drawings
FIG. 1 is a flow chart of a laser three-dimensional imaging radar ground target point cloud data processing method of the invention;
FIG. 2 is a diagram illustrating a k-d tree construction according to the present invention;
FIG. 3 is a flow chart of an improved extended window elevation threshold filtering algorithm of the present invention.
Detailed Description
The laser three-dimensional imaging radar ground target point cloud data processing method provided by the invention is further described in detail below with reference to the accompanying drawings and the specific implementation mode. The advantages and features of the present invention will become more apparent from the following description. It should be noted that the drawings are simplified in form and not to precise scale, and are only used for convenience and clarity to assist in describing the embodiments of the present invention, but not for limiting the conditions of the embodiments of the present invention, and therefore, the present invention is not limited by the technical spirit, and any structural modifications, changes in the proportional relationship, or adjustments in size, should fall within the scope of the technical content of the present invention without affecting the function and the achievable purpose of the present invention.
It is to be noted that, in the present invention, relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus.
With reference to fig. 1 to 3, the present embodiment provides a method for processing laser three-dimensional imaging radar ground target point cloud data, which is used for performing ground target detection and identification on the laser three-dimensional imaging radar point cloud data and outputting a suspected interested target, and includes the steps of:
t1, obtaining a time and space registered point cloud data set, comprising the steps of:
t11, acquiring a time-registered point cloud data set:
acquiring Position and attitude data of each laser light emitting time by a POS (Position and Orientation System) System in a laser radar measuring System, wherein the POS System outputs GPS (Global Positioning System) time with TAI (times atomic International) second as a time reference;
acquiring laser echo data through a laser scanning system of a laser radar measuring system;
and the POS system and the laser scanning system are synchronized through pulse per second signals, and the measurement time of all the position, attitude data and laser echo data is uniformly registered to a GPS time coordinate system to obtain a time-registered point cloud data set.
T12, spatial registration of the time registered point cloud data set:
the position, attitude data and laser echo data are converted into a WGS-84 Coordinate System (World Geodetic System-1984 Coordinate System, World Geodetic Coordinate System in 1984, international standard Coordinate System at present) from respective local Coordinate systems to complete the spatial registration of the point cloud data, and a time and space registered point cloud data set is obtained.
T2, constructing a point cloud data set of the K-d tree by adopting a K-d tree (K-dimensional tree, a high-dimensional index tree data structure, and is commonly used for carrying out nearest neighbor search and approximate nearest neighbor search in a large-scale high-dimensional data space) algorithm based on a point cloud data set of time and space registration;
the step is to improve the subsequent data processing efficiency, because the laser radar measurement system adopts a push-scanning mode for imaging, the imaging in the horizontal direction is scanned by a scanning mirror of the laser scanning system, and the imaging in the pitching direction is scanned by the motion of a carrying platform of the laser radar measurement system, a zigzag scanning path is formed, so that the data in the acquired time and space registration point cloud data set are irregularly arranged, the subsequent data processing process is very complex, the data processing amount is very large, and the processing efficiency is very low; after the k-d tree point cloud data set is constructed through the k-d tree algorithm, the k-d tree point cloud data set can be subjected to data processing by adopting an efficient near point search strategy, such as commonly-used bounding sphere query, rectangular region query, k nearest neighbor query and the like, so that the data processing efficiency is greatly improved.
The principle of the k-d tree algorithm is as follows: the k-d tree is an expansion of a binary tree, k represents a space dimension, a space formed by a data set is divided into two subspaces by a hyperplane (such as a line of a two-dimensional space or a plane of a three-dimensional space) with the k-1 dimension parallel to a coordinate axis, about half of nodes fall into the space on one side of the hyperplane, and the other half of nodes fall into the space on the other side of the hyperplane; and dividing the two subspaces into four lower-level subspaces again, repeating the steps until the number of nodes in one lowest-level space is less than a set value, and finishing the construction of the k-d tree.
As shown in fig. 2, the k-d tree algorithm is briefly explained by taking two-dimensional data as an example:
in two-dimensional space, a data structure based on a k-d tree is constructed for a data point set { (2, 2), (3, 4), (4, 9), (7, 6), (8, 8), (10, 5) }, setting a value of 1:
(1) determining a separation domain: respectively calculating the data variance of each data point in the x direction and the y direction, and taking the larger one as a separation domain; the variance in the x direction is 3.14, the variance in the y direction is 2.58, and x > y, so the x direction is taken as the separation domain;
(2) determining a hyperplane: selecting a median value of 7 according to the sorting of values 2, 3, 4, 7, 8 and 10 of the data points in the x direction, and determining a hyperplane as a straight line x which passes through the nodes (7 and 6) and is perpendicular to the x axis to be 7;
(3) determining a left subspace and a right subspace: the hyperplane x-7 divides the whole space into two parts, wherein x <7 is the left subspace, including { (2, 2), (3, 4), (4, 9) }; x >7 is a right subspace, the right subspace comprising { (8, 8), (10, 5) };
(4) and respectively repeating the steps on the left subspace and the right subspace and dividing the subspaces into subspaces of the next level until only 1 data point is contained in one lowest level subspace, and at the moment, finishing the construction of the data structure based on the k-d tree.
T3, based on the k-d tree point cloud data set, filtering noise points and abnormal points through a k nearest neighbor query algorithm to obtain a normal point data set;
in the k-d tree point cloud data set, besides normal points, some noise points and abnormal points are also included and need to be filtered by an algorithm;
the principle of k nearest neighbor query algorithm filtering is as follows: for any given point in a k-d tree point cloud data set, k adjacent points corresponding to the given point can be quickly inquired through a k-d tree data structure, then Euclidean space distances between the given point and the k adjacent points are respectively calculated, then the mean value of the k Euclidean space distances is calculated to obtain an Euclidean mean distance, and finally whether the point is a noise point or an abnormal point which should be filtered is determined through judging the Euclidean mean distance; because the point cloud data of the normal points are distributed more densely and uniformly, the Euclidean average distance is smaller; the noise points and the abnormal points are usually isolated points, and the Euclidean average distance between the noise points and the abnormal points is larger, so that the noise points and the abnormal points can be quickly filtered and removed by setting a proper distance threshold for judgment, and a normal point data set is generated.
T4, filtering non-ground point data through an improved expanded window elevation threshold filtering algorithm based on the normal point data set to obtain a ground point data set;
in the normal point data set, ground point data and non-ground point data are included, and the non-ground point is filtered by an algorithm;
the principle of the traditional extended window elevation threshold filtering algorithm is as follows: presetting a fixed gradient threshold S according to terrain experience based on the terrain, dividing data in a normal point data set into k square units, only retaining height information of each data point in each square unit, and calculating based on the side length and S of each square unit to obtainHeight difference threshold H in each square unitk(ii) a Selecting the point with the minimum height from each square unit as a first ground point, calculating the height difference of each point in the square unit relative to the first ground point, and then calculating the height difference and the H in the square unitkComparing, the height difference is larger than HkDiscarding data points of (1), height difference less than HkThe data points of (a) are retained; enlarging the size of the square unit, and repeating the process until the square unit exceeds the size of the whole point cloud data set, so as to obtain a ground point data set; the S value of the traditional algorithm has a large influence on a filtering result, the S value is difficult to accurately grasp, a ground point with too small S value is wrongly divided into target points, and the S value is too large, so that a plurality of target points are lost;
the improved filtering algorithm for expanding the window elevation threshold dynamically adjusts the gradient threshold, so that the non-ground point filtering effect is better, and the adaptability is better; as shown in fig. 3, the improved extended window elevation threshold filtering algorithm comprises the steps of:
t41, initialization:
setting an iteration ordinal number: i is 1;
setting the first square cell C in the normal point data seti: selecting the data point with the minimum elevation (height above ground) in the normal point data set, and taking the data point as CiAnd then the height of the vehicle object of interest is set to CiLength of side of
Figure BDA0003407042520000091
Thereby dividing into Ci；
Set up CiS of the initial slope valuei: based on CiThe terrain of the user sets an initial inclination value S according to terrain experienceiSuch as 0.05;
t42, setting CiSeed point of (1): at CiSelecting a point with the minimum elevation as a seed point;
t43 based on CiLength of side of
Figure BDA0003407042520000092
And SiCalculating to obtain a height difference threshold value HiThe calculation formula is as follows:
Figure BDA0003407042520000093
t44, calculation CiThe elevation difference H between each data point and the seed point is calculated, and the elevation difference H is compared with the elevation difference HiAnd (3) comparison:
if h is>HiThen the data point is discarded;
otherwise, adding the data point into a ground point data set;
t45, update SiThe method comprises the following steps:
t451, selection of CiThe maximum point P of elevation inmaxAnd elevation minimum point PminTo find CiMean value of ground points PaveI.e. to PmaxAnd PminRespectively averaging the three-axis coordinate values to obtain PaveThe coordinate values of (a);
t452, P solutionmaxRelative to PaveIs inclined atmaxAnd PaveRelative to PminIs inclined atminThe calculation method comprises the following steps:
for any two points p1(X1，Y1，Z1) And p2(X2，Y2，Z2)，p2Relative to p1The formula for calculating the inclination S of (a) is:
Figure BDA0003407042520000094
t453, calculate SiThe calculation formula is as follows:
Si＝(Smax+Smin)/2；
T46、i＝i+1；
c is to bei-1Extension to CiWherein, CiWith Ci-1Is central, CiHas an area of C i-12 times the area;
judgment CiWhether out of range of normal point data sets:
if yes, exiting iteration, and generating a ground point data set at the moment;
otherwise, T42 is returned.
T5, setting a ground target height threshold value according to the type of the ground target to identify the ground target based on the ground point data set to obtain a ground target data set;
the ground target height threshold is set according to the relative height from the ground of a ground target (such as a ground vehicle, vegetation, house, scarp, etc.), for example: the value of the steep threshold point mainly falls within 0-0.5 m; the value of a target point of the ground large and medium-sized vehicle is mainly 0.5-4 m; the value of the tree point falls mainly between 4 and 20 m; the house point values fall mainly between 4 and 100 m;
sequentially comparing each data point in the ground point data set with a ground target height threshold value:
adding the data point to a ground target data set if the elevation of the data point is within a ground target elevation threshold;
otherwise, the data point is discarded.
T6, based on the ground target data set, dividing and summarizing each ground target data point through a Euclidean clustering algorithm to obtain a ground target clustering set;
and classifying and dividing the data points in the ground target data set through a Euclidean clustering algorithm, so that the data points associated with the same ground target are divided into the same subset, and each subset forms a ground target clustering set.
T7, setting a judgment rule for each interested target, identifying the targets based on the ground target clustering set, and outputting each suspected interested target;
the discrimination rule is discrimination feature + discrimination threshold, firstly setting discrimination features (such as length, width, height, aspect ratio, distribution entropy and the like) of the interested target, then setting the discrimination threshold of each discrimination feature, and generating the discrimination rule;
and based on the discrimination rules, carrying out interested target identification on each data subset in the ground target cluster set, and finally outputting the suspected interested target.
In summary, the method for processing the laser three-dimensional imaging radar ground target point cloud data provided by the invention solves the technical problems that the traditional method for processing the laser three-dimensional imaging radar point cloud data cannot effectively describe the space geometric characteristics of a target, and loses a large amount of information such as space normal vectors, space curvatures, space connection relations and the like, or the data post-processing method applied to the surveying and mapping field has large computation amount and complex algorithm, and solves the processing problem when the point cloud data is irregularly arranged, so that more three-dimensional information of the target is kept, the computation speed is improved, the point cloud data can be processed in real time, and the engineering can be quickly realized.
While the present invention has been described in detail with reference to the preferred embodiments, it should be understood that the above description should not be taken as limiting the invention. Various modifications and alterations to this invention will become apparent to those skilled in the art upon reading the foregoing description. Accordingly, the scope of the invention should be determined from the following claims.

Claims (10)

1. A laser three-dimensional imaging radar ground target point cloud data processing method is used for carrying out ground target detection and identification on laser three-dimensional imaging radar point cloud data and outputting a suspected interesting target, and is characterized by comprising the following steps:
s1, acquiring a time and space registration point cloud data set based on the laser three-dimensional imaging radar point cloud data;
s2, filtering non-ground point data based on the point cloud data set of the time and space registration to obtain a ground point data set;
s3, extracting, dividing and summarizing data of each ground target based on the ground point data set, and clustering and dividing the data into respective data subsets to obtain a ground target clustering set;
s4, identifying the interested target based on the ground target clustering set, and outputting the suspected interested target.
2. The laser three-dimensional imaging radar ground target point cloud data processing method as claimed in claim 1, wherein the step S2 includes:
s21, based on the point cloud data set of the time and space registration, filtering noise points and abnormal points to obtain a normal point data set;
and S22, filtering non-ground point data based on the normal point data set to obtain the ground point data set.
3. The laser three-dimensional imaging radar ground target point cloud data processing method of claim 2, wherein the step S21 includes:
s211, constructing a k-d tree point cloud data set by adopting a k-d tree algorithm based on the time and space registration point cloud data set;
s212, based on the k-d tree point cloud data set, noise points and abnormal points are filtered out, and the normal point data set is obtained.
4. The laser three-dimensional imaging radar ground target point cloud data processing method as claimed in claim 3, wherein the step S212 is implemented by a k-nearest neighbor query algorithm.
5. The method for processing the laser three-dimensional imaging radar ground target point cloud data according to claim 2, wherein the step S22 is implemented by an improved extended window elevation threshold filtering algorithm, and comprises the steps of:
s221, initialization:
setting an iteration ordinal number: i is 1;
setting the first square cell C in the normal point data seti: selecting the data point with the minimum height in the normal data set, and taking the data point as CiAnd then the height of the vehicle object of interest is set to CiLength of side of
Figure FDA0003407042510000021
Set up Ci；
Set up CiS of the initial slope valuei；
S222, at CiSelecting a point with the minimum elevation as a seed point;
s223, based on CiLength of side of
Figure FDA0003407042510000022
And SiCalculating the height difference threshold value HiThe calculation formula is as follows:
Figure FDA0003407042510000023
s224, calculating CiThe elevation difference H between each data point and the seed point is calculated, and the elevation difference H is compared with the elevation difference HiAnd (3) comparison:
if h is>HiThen the data point is discarded;
otherwise, adding the data point into a ground point data set;
s225, update Si；
S226、i＝i+1；
A square unit Ci-1Extension to CiWherein, CiWith Ci-1Is central, CiHas an area of Ci-12 times the area;
judgment CiWhether out of range of normal point data sets:
if yes, exiting iteration, and generating a ground point data set at the moment;
otherwise, the process returns to S222.
6. The laser three-dimensional imaging radar ground target point cloud data processing method of claim 5, wherein the step S225 comprises:
a1, selection of CiThe maximum point p of elevation inmaxAnd elevation minimum point pminTo find CiMean value of ground points pave；
A2 finding pmaxRelative to paveIs inclined atmaxAnd p isaveRelative to pminIs inclined atmin；
A3, calculating SiThe calculation formula is as follows:
Si＝(Smax+Smin)/2。
7. the laser three-dimensional imaging radar ground target point cloud data processing method of claim 2, wherein the step S3 includes:
s31, identifying the data of the ground target based on the ground point data set, and filtering non-ground target data to obtain a ground target data set;
and S32, based on the ground target data set, segmenting and summarizing data points of each ground target, and dividing data points related to the same ground target into the same subset to obtain the ground target clustering set.
8. The laser three-dimensional imaging radar ground target point cloud data processing method of claim 7, wherein the step S31 includes:
setting a ground target height threshold according to the ground-off relative height of the ground target, and then sequentially comparing each data point in the ground point data set with the ground target height threshold:
adding a data point to the ground target data set if the elevation of the data point is within the ground target elevation threshold;
otherwise, the data point is discarded.
9. The laser three-dimensional imaging radar ground target point cloud data processing method as claimed in claim 7, wherein the step S32 is implemented by Euclidean clustering algorithm.
10. The laser three-dimensional imaging radar ground target point cloud data processing method of claim 1, wherein the step S4 includes:
setting a discrimination rule for each interested target, identifying the interested target for each data subset in the ground target clustering set based on the discrimination rule, and finally outputting the suspected interested target;
wherein, the rule of judgement includes: the discrimination feature of the object of interest and a discrimination threshold for each of the discrimination features.
CN202111520246.8A 2021-12-13 2021-12-13 Laser three-dimensional imaging radar ground target point cloud data processing method Pending CN114200477A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111520246.8A CN114200477A (en) 2021-12-13 2021-12-13 Laser three-dimensional imaging radar ground target point cloud data processing method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111520246.8A CN114200477A (en) 2021-12-13 2021-12-13 Laser three-dimensional imaging radar ground target point cloud data processing method

Publications (1)

Publication Number Publication Date
CN114200477A true CN114200477A (en) 2022-03-18

Family

ID=80653190

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111520246.8A Pending CN114200477A (en) 2021-12-13 2021-12-13 Laser three-dimensional imaging radar ground target point cloud data processing method

Country Status (1)

Country Link
CN (1) CN114200477A (en)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114359876A (en) * 2022-03-21 2022-04-15 成都奥伦达科技有限公司 Vehicle target identification method and storage medium
CN114879217A (en) * 2022-07-12 2022-08-09 中国工程物理研究院应用电子学研究所 Target pose judgment method and system
CN114966634A (en) * 2022-07-11 2022-08-30 高德软件有限公司 Laser ranging system calibration method, device and computer program product
CN115236674A (en) * 2022-06-15 2022-10-25 北京踏歌智行科技有限公司 Mining area environment sensing method based on 4D millimeter wave radar
CN115375713A (en) * 2022-10-25 2022-11-22 杭州华橙软件技术有限公司 Ground point cloud segmentation method and device and computer readable storage medium
CN115755901A (en) * 2022-11-14 2023-03-07 杭州蓝芯科技有限公司 Mobile robot obstacle stopping control method and device
CN116538996A (en) * 2023-07-04 2023-08-04 云南超图地理信息有限公司 Laser radar-based topographic mapping system and method

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114359876A (en) * 2022-03-21 2022-04-15 成都奥伦达科技有限公司 Vehicle target identification method and storage medium
CN115236674A (en) * 2022-06-15 2022-10-25 北京踏歌智行科技有限公司 Mining area environment sensing method based on 4D millimeter wave radar
CN115236674B (en) * 2022-06-15 2024-06-04 北京踏歌智行科技有限公司 Mining area environment sensing method based on 4D millimeter wave radar
CN114966634A (en) * 2022-07-11 2022-08-30 高德软件有限公司 Laser ranging system calibration method, device and computer program product
CN114879217A (en) * 2022-07-12 2022-08-09 中国工程物理研究院应用电子学研究所 Target pose judgment method and system
CN114879217B (en) * 2022-07-12 2022-10-21 中国工程物理研究院应用电子学研究所 Target pose judgment method and system
CN115375713A (en) * 2022-10-25 2022-11-22 杭州华橙软件技术有限公司 Ground point cloud segmentation method and device and computer readable storage medium
CN115755901A (en) * 2022-11-14 2023-03-07 杭州蓝芯科技有限公司 Mobile robot obstacle stopping control method and device
CN116538996A (en) * 2023-07-04 2023-08-04 云南超图地理信息有限公司 Laser radar-based topographic mapping system and method
CN116538996B (en) * 2023-07-04 2023-09-29 云南超图地理信息有限公司 Laser radar-based topographic mapping system and method

Similar Documents

Publication Publication Date Title
CN114200477A (en) Laser three-dimensional imaging radar ground target point cloud data processing method
CN110781827B (en) Road edge detection system and method based on laser radar and fan-shaped space division
US10049492B2 (en) Method and apparatus for rendering facades of objects of interest from three-dimensional point clouds
US9846946B2 (en) Objection recognition in a 3D scene
CN110533722A (en) A kind of the robot fast relocation method and system of view-based access control model dictionary
US7995055B1 (en) Classifying objects in a scene
CN110992341A (en) Segmentation-based airborne LiDAR point cloud building extraction method
CN111553409B (en) Point cloud identification method based on voxel shape descriptor
CN112101278A (en) Hotel point cloud classification method based on k nearest neighbor feature extraction and deep learning
CN107330875B (en) Water body surrounding environment change detection method based on forward and reverse heterogeneity of remote sensing image
CN105976376B (en) High-resolution SAR image target detection method based on component model
CN111781608A (en) Moving target detection method and system based on FMCW laser radar
CN111738332B (en) Underwater multi-source acoustic image substrate classification method and system based on feature level fusion
CN107909002B (en) Sea-land segmentation method of infrared remote sensing image based on coastline matching
Yu et al. Automated detection of three-dimensional cars in mobile laser scanning point clouds using DBM-Hough-Forests
CN109034065B (en) Indoor scene object extraction method based on point cloud
CN110807781A (en) Point cloud simplification method capable of retaining details and boundary features
CN113888461A (en) Method, system and equipment for detecting defects of hardware parts based on deep learning
CN114821358A (en) Optical remote sensing image marine ship target extraction and identification method
CN114266947A (en) Classification method and device based on fusion of laser point cloud and visible light image
Fan et al. Fresco: Frequency-domain scan context for lidar-based place recognition with translation and rotation invariance
CN114332172A (en) Improved laser point cloud registration method based on covariance matrix
Liu et al. Keypoint matching by outlier pruning with consensus constraint
Manno-Kovacs et al. Automatic target classification in passive ISAR range-crossrange images
Zhang et al. Object-based classification of urban airborne LiDAR point clouds with multiple echoes using SVM

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