CN108681525B - Road surface point cloud intensity enhancing method based on vehicle-mounted laser scanning data - Google Patents

Road surface point cloud intensity enhancing method based on vehicle-mounted laser scanning data Download PDF

Info

Publication number
CN108681525B
CN108681525B CN201810464773.3A CN201810464773A CN108681525B CN 108681525 B CN108681525 B CN 108681525B CN 201810464773 A CN201810464773 A CN 201810464773A CN 108681525 B CN108681525 B CN 108681525B
Authority
CN
China
Prior art keywords
intensity
laser
point cloud
road surface
point
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.)
Active
Application number
CN201810464773.3A
Other languages
Chinese (zh)
Other versions
CN108681525A (en
Inventor
方莉娜
黄志文
罗海峰
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.)
Fuzhou University
Original Assignee
Fuzhou University
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 Fuzhou University filed Critical Fuzhou University
Priority to CN201810464773.3A priority Critical patent/CN108681525B/en
Publication of CN108681525A publication Critical patent/CN108681525A/en
Application granted granted Critical
Publication of CN108681525B publication Critical patent/CN108681525B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/15—Correlation function computation including computation of convolution operations

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Computational Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Algebra (AREA)
  • Databases & Information Systems (AREA)
  • Software Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Length Measuring Devices By Optical Means (AREA)
  • Traffic Control Systems (AREA)
  • Optical Radar Systems And Details Thereof (AREA)

Abstract

The invention relates to a road surface point cloud intensity enhancing method based on vehicle-mounted laser scanning data, which comprises the following steps of: s1: acquiring a road surface point cloud, a driving track line and the distance from a laser point in the road surface point cloud to the driving track line based on vehicle-mounted laser scanning data, and segmenting the road surface point cloud; s2: taking the road surface point cloud segment as a processing unit, and performing the following steps: 21: constructing an original distance model function and a reference distance model function; 22: correcting the intensity of the laser spot by adopting a ratio method and a difference method; 23: carrying out intensity conversion on the laser spot; 24: and (3) denoising the intensity of the laser point by adopting a multi-filter integration method to enhance the intensity of the laser point of each road point cloud segment in the road point cloud. Compared with the prior art, the method obviously improves the strength difference of different target materials of the road surface and eliminates the strength noise in local homogeneous areas.

Description

Road surface point cloud intensity enhancing method based on vehicle-mounted laser scanning data
Technical Field
The invention relates to the field of vehicle-mounted laser scanning point cloud data processing, in particular to a road surface point cloud intensity enhancing method based on vehicle-mounted laser scanning data.
Background
The vehicle-mounted laser point cloud intensity is used as the backscattering echo intensity of the target to the emitted laser, the reflection spectrum characteristic of the measured target to the laser is represented, and the method is widely applied to the research field of road marking extraction. However, the point cloud reflection intensity obtained by the vehicle-mounted laser scanning system is not only related to the surface reflectivity of the surface of the ground object medium, but also affected by factors such as the distance between the ground object and the laser scanner, the incident angle of laser, the vehicle running speed and the like, and the ground object of the same medium has uneven reflection intensity distribution due to different angles and distances; meanwhile, due to the influence of factors such as roughness and different degrees of old and new road surface abrasion, phenomena of same object, different spectrum and same spectrum of foreign matters exist widely, a large amount of salt and pepper noise and Gaussian noise exist in vehicle-mounted laser point cloud, and the accuracy of target classification and identification by using intensity information is seriously influenced. Therefore, how to eliminate the influence of the laser scanning distance, the road surface abrasion, the degree of old and new and the point density on the laser point cloud intensity and enhance the laser point cloud intensity is one of the difficult works of classifying and identifying the vehicle-mounted laser point cloud target based on the intensity information at present.
At present, there are two main methods for enhancing the laser point cloud intensity: a point cloud intensity correction method and a point cloud intensity denoising method. The point cloud intensity correction method mainly utilizes a model driving method or a data driving method to analyze various factors such as temperature and humidity in a scanning environment and the influence of sensor output energy on the point cloud intensity [ Alireza G.Kashani,2015 ]. The model driving method is based on a laser radar formula, and establishes a corresponding correction formula according to the attenuation process of laser in atmospheric transmission and the surface characteristics of a target to correct the intensity value, and the method is mainly suitable for an airborne laser radar scanning system (QiongDing, 2013); the data driving method is mainly used for a ground laser scanning system [ Fangwei, 2015] by selecting appropriate homogeneous data and fitting the relation between the laser intensity and various influence factors to obtain a corresponding correction formula. However, the data driving method needs to acquire homogeneous data in a man-machine interaction manner, and the automation degree of correction processing is limited.
At present, the laser point cloud intensity correction method mainly aims at a ground laser radar scanning system and an airborne laser radar scanning system, and research on correction of intensity data acquired by a vehicle-mounted laser radar scanning system is less. In a vehicle-mounted laser radar scanning system, an intensity characteristic image [ Yang, Bisheng,2012] is generated based on an IDW principle, or a strategy [ Yu, Yongtao,2015] based on parallel track line blocking processing weakens the influence of geometric factors on intensity to a certain extent, and the intensity correction effect is achieved. However, the influence of geometric factors on the intensity is difficult to describe and eliminate by only constructing a correction model of a simple linear relation or an inverse proportional relation, and a single intensity correction cannot suppress the noise of the point cloud intensity caused by road wear and other factors. Some scholars remove high-frequency noise in the intensity information and smooth the intensity information by an image filtering method, and filter noise data. And filtering the noise of the point cloud intensity [ Bisheng Yang,2017] [ Li Yan,2016 ]. Although the method has a certain effect of inhibiting the intensity noise of the discrete distribution in the road point cloud intensity, the method is difficult to be applied to the enhancement of the point cloud intensity in a complex road scene because the influences of the uneven distribution of the point cloud density and the geometrical factors on the point cloud intensity are not fully considered.
The patent with the patent publication number of CN105139032A and the invention name of 'a rock identification and classification method and system' provides an intensity correction method based on laser ranging influence factors for the point cloud intensity of ground laser radar. Because the method considers the relation of the distance and the intensity as an ideal inverse proportion relation, the method is difficult to be suitable for point cloud intensity correction of complex scenes.
The patent with patent publication number CN104502919A and invention name "method for extracting urban vegetation three-dimensional coverage by using airborne laser radar point cloud" provides a method for intensity enhancement by combining median filtering and intensity correction based on laser ranging influence factors for airborne laser radar point cloud intensity.
The patent with patent publication number CN106097423A and invented name "LiDAR point cloud intensity correction method based on k neighbors" proposes search based on k neighbors for ground LiDAR point cloud intensity, and takes the mean value of the intensity of point clouds in the neighborhood as the intensity value after correction of the central reference point. Although the method can ensure the consistency of the point cloud intensity of the homogeneous region to a certain extent, the method can damage the detail characteristics of the edges of the ground objects, and does not consider the influence of factors such as distance, incidence angle and the like on the intensity.
The patent with the patent publication number of CN104197897A and the invention name of 'urban road marking automatic classification method based on vehicle-mounted laser scanning point cloud' is directed at vehicle-mounted laser radar point cloud, a segmentation processing strategy is adopted, the road surface point cloud is divided into different areas in parallel with a track line in a man-machine interaction mode, and in the same area, the influence of geometrical factors on the intensity of each laser point is consistent, so that the relative correction of the intensity of the point cloud in a local area is achieved, and the influence of laser incidence angle and laser ranging on the intensity is weakened. However, the method does not substantially eliminate the influence of geometric factors on the intensity, the segment size is difficult to determine, and the influence of factors such as road wear and different degrees of freshness on the point cloud intensity is not considered.
Disclosure of Invention
The invention aims to solve the problems that the intensity of road point cloud in vehicle-mounted laser scanning data is easily influenced by factors such as scanning distance, point cloud density distribution and the like, the material characteristics of the surface of a target cannot be accurately represented, and the road target segmentation is difficult to be carried out by using the intensity, and provides a road point cloud intensity enhancing method based on the vehicle-mounted laser scanning data, the method has the advantages that the point cloud intensity is enhanced, the influence of distance geometric factors on the point cloud intensity is eliminated, the intensity noise caused by inconsistent road surface roughness is restrained, the intensity distribution range is wide, the spatial distribution is dense, salt and pepper noise and Gaussian noise in the point cloud intensity are filtered, the effective restraint of intensity difference in homogeneous point clouds is finally realized, the intensity difference among heterogeneous point clouds is enlarged, and accurate intensity characteristic expression is provided for later-stage point cloud processing such as road marking point cloud extraction.
The purpose of the invention can be realized by the following technical scheme:
a road surface point cloud intensity enhancing method based on vehicle-mounted laser scanning data comprises the following steps:
s1: acquiring a road surface point cloud, a driving trajectory line and the distance from a laser spot in the road surface point cloud to the driving trajectory line based on vehicle-mounted laser scanning data, and segmenting the road surface point cloud to obtain a road surface point cloud segment;
s2: taking the road surface point cloud segment as a processing unit, and performing the following steps:
21: aiming at the point cloud segment of the road surface, a homogeneous data set is established by adopting a median statistical method, a cubic polynomial function describing the distance-intensity relation of the close range effect is established, model parameters in the cubic polynomial function are estimated based on the homogeneous data set, and an original distance model function and a reference distance model function are established by the model parameters;
22: based on the original distance model function and the reference distance model function, performing intensity correction on the laser point by adopting a ratio method and a difference method;
23: aiming at the laser spot with the corrected intensity, obtaining an intensity probability distribution characteristic value of the laser spot according to the global probability density value of each intensity interval, and carrying out intensity conversion on the laser spot by combining the corrected intensity;
24: and (3) aiming at the laser points with the converted intensities, carrying out intensity denoising on the laser points by adopting a multi-filter integration method according to uneven laser point density distribution, and completing laser point intensity enhancement of each road point cloud segment in the road point cloud.
The step S1 specifically includes:
11: determining a breakpoint of a scanning line according to a scanning angle difference between front and rear scanning laser points in the vehicle-mounted laser scanning data, thereby extracting the scanning line;
12: based on the scanning lines extracted in the step 11, extracting a road surface point cloud by adopting a moving window method;
13: counting the density distribution of the laser points on each scanning line, extracting the laser points corresponding to the maximum density distribution on each scanning line as the traveling track points, constructing continuous traveling track lines by the traveling track points, and acquiring the distance from the laser points to the traveling track lines;
14: and (3) sampling the driving trajectory at equal intervals to obtain pseudo trajectory points, and segmenting the road surface point cloud based on the pseudo trajectory points to obtain a road surface point cloud segment.
The step 21 specifically comprises:
211: according to the distance from the laser spot to the traffic trajectory line, carrying out equal-distance subdivision on the road surface point cloud segment in the direction parallel to the traffic trajectory line to obtain a distance interval;
212: the intensity of the laser points in each distance interval is counted, and a median statistical method is adopted to extract the laser points corresponding to the intensity median in each distance interval to construct a homogeneous data set (R)k,Ik)，RkIs the distance between the kth laser point and the scanner in the homogeneous dataset, IkThe intensity of the kth laser point in the homogeneous data set is defined, and k is the serial number of the laser point in the homogeneous data set;
213: according to an approximation theorem, a cubic polynomial function describing the distance-intensity relation of the near distance effect is established, and the following formula is satisfied:
Figure BDA0001661769710000041
wherein eta is0、η1、η2、η3Is a model parameter;
214: based on homogeneous data set, using least square method to model parameter eta0、η1、η2、η3Carrying out estimation;
215: establishing an original distance model function f (R)j) And a reference distance model function f (R)s) The following formula is satisfied:
Figure BDA0001661769710000042
Figure BDA0001661769710000043
wherein R isjIs the distance between the jth laser point and the scanner, j is the laser point number in the road surface point cloud segment, RsFor a set reference distance, Rs＝min(Rj)。
The step 22 specifically includes:
221: the intensity of the laser point is roughly corrected by adopting a ratio method, and the following formula is satisfied:
I_Ratj＝Ij·Ratj
Figure BDA0001661769710000044
wherein, I _ RatjThe rough correction result of the intensity of the jth laser point is obtained, j is the laser point number in the pavement point cloud segment, IjIs the intensity of the jth laser spot, RatjTo correspond to the jthCorrection coefficient by ratio method of laser spot, f (R)j) As a function of the original distance model, f (R)s) Is a reference distance model function;
222: the self-adaptive selection difference method is used for carrying out fine correction on the rough intensity correction result of the laser spot, and the following formula is satisfied:
Figure BDA0001661769710000045
Difj＝f(Rs)-f(Rj)
ΔIj＝I_Ratj-Ij
wherein, I _ DistjFor the final intensity correction of the jth laser spot, Δ IjIs the difference in intensity before and after the rough correction of the jth laser spot, θIFor the set threshold value, DifjThe coefficients are corrected for the difference corresponding to the jth laser spot.
The step 23 specifically includes:
231: obtaining an intensity histogram of the intensity-corrected road surface point cloud segment according to the final intensity correction result of each laser point in the road surface point cloud segment obtained in the step 22, counting the global probability density value p of each intensity interval in the intensity histogram, and satisfying the following formula:
Figure BDA0001661769710000046
wherein N is the number of laser points in the intensity interval, and N is the number of laser points in the pavement point cloud segment;
232: obtaining the probability distribution characteristic value of each laser point, and satisfying the following formula:
Fj＝I_Distj·pj
wherein, FjIs the probability distribution characteristic value of the jth laser point, j is the laser point number in the road surface point cloud segment, I _ DistjFor the final intensity correction of the jth laser spot, pjThe global probability density value of the intensity interval where the jth laser point is located;
233: optimizing the probability distribution characteristic value to satisfy the following formula:
Figure BDA0001661769710000051
wherein, Fj' is the probability distribution characteristic value of the optimized j laser point;
234: and carrying out intensity conversion on the final intensity correction result of each laser spot, wherein the following formula is satisfied:
Figure BDA0001661769710000052
wherein, I _ tranjThe result of the intensity transformation of the jth laser spot.
The step 24 specifically includes:
241: performing neighborhood search by taking each laser point as a central reference point, and dividing the road surface point cloud segment after intensity conversion into a low-density edge area and a high-density central area according to the density distribution of the laser points;
242: aiming at the low-density edge area, based on the spatial proximity and the intensity value similarity between the central reference point in the neighborhood and other laser points in the field, the intensity conversion result of the laser points is subjected to intensity denoising by adopting a bilateral filtering method, and the following formula is satisfied:
Figure BDA0001661769710000053
Figure BDA0001661769710000054
wherein, I _ densejIs the intensity enhancement result of the jth laser point, j is the laser point number in the pavement point cloud segment,
Figure BDA0001661769710000055
is the first _ p in the corresponding domainjWeight coefficient of individual laser points, ner _ pjNumbering of laser spots in the field of the jth laser spot, ner _ njThe number of laser points in the field of the jth laser spot,
Figure BDA0001661769710000056
is the inner-field first _ pjAs a result of the intensity transformation of the individual laser spots,
Figure BDA0001661769710000057
to the center reference point and the first _ p in the fieldjThe similarity of intensity values between the individual laser points,
Figure BDA0001661769710000058
to the center reference point and the first _ p in the fieldjSpatial proximity between individual laser points;
aiming at the high-density central area, intensity denoising is carried out on the intensity transformation result of the laser spot by adopting a neighborhood intensity difference analysis method, and the following formula is satisfied:
Figure BDA0001661769710000061
Figure BDA0001661769710000062
wherein the content of the first and second substances,
Figure BDA0001661769710000063
is the inner-neighborhood first _ pjSum of intensity differences between one laser spot and other laser spots in the neighborhood, ner _ qjThe numbering of the laser spot in the field of the jth laser spot,
Figure BDA0001661769710000064
is the inner third of the fieldjThe intensity of each laser spot is transformed.
The spatial proximity and the intensity value similarity between the central reference point in the neighborhood and other laser points in the neighborhood in the step 242 satisfy the following formulas:
Figure BDA0001661769710000065
Figure BDA0001661769710000066
wherein, I _ tranjAs a result of intensity conversion of the jth laser spot, σ _ IjIs the standard deviation of the distribution of intensity differences between the in-domain central reference point of the jth laser spot and other laser spots in the domain,
Figure BDA0001661769710000067
to the center reference point and the first _ i in the fieldjDistance between laser spots, σ _ DjIs the standard deviation of the distance distribution between the in-domain central reference point of the jth laser spot and other laser spots in the domain.
Compared with the prior art, the invention has the following advantages:
1) the method corrects and transforms the intensity of the cloud of the pavement points, enhances the intensity difference of different medium targets, improves the intensity consistency of the same medium targets, realizes accurate intensity feature expression, can improve the subsequent segmentation precision of the pavement targets, and provides reliable intensity information for classification and identification of the targets based on the intensity features, such as extraction of road marking lines.
2) Establishing an original distance model function by adopting a cubic polynomial function, describing the relation between the distance and the intensity in the road point cloud, and establishing a homogeneous data set estimation model parameter by a median statistical method; a plurality of intensity correction models are constructed by fusing a ratio method and a difference method to realize intensity correction of the road surface point cloud, and the geometric influence of the intensity weakening along with the increase of the distance is eliminated.
3) Counting the probability distribution density of the intensity of each laser point according to the intensity fine correction result, and calculating the probability distribution characteristic of the intensity of each point cloud; the probability distribution characteristics of the point cloud are utilized, an intensity optimization model is selected in a self-adaptive mode, the intensity probability distribution characteristics in the local homogeneous region are optimized, intensity transformation is achieved by combining the optimized probability distribution characteristics and the corrected point cloud intensity, and the intensity difference of the local homogeneous region caused by the fact that the roughness and the old and new degree of the road surface are different is restrained.
4) Based on the point cloud after intensity conversion, the method of multi-filter integration is adopted to carry out point cloud intensity denoising, smooth the homogeneous region, filter the ubiquitous salt and pepper noise and Gaussian noise in the vehicle-mounted laser road surface point cloud intensity, and enhance the road surface point cloud intensity.
5) The invention firstly proposes to automatically construct a homogeneous data set by using a median statistical method and improve the automation degree of intensity enhancement processing.
Drawings
FIG. 1 is a flow chart of the method of the present invention;
FIG. 2 is a schematic diagram of a scan line extraction based on the angular difference between laser points scanned before and after;
FIG. 3 is a schematic diagram of a road surface point cloud segment extraction process;
FIG. 4 is a schematic diagram of intensity correction based on distance factors;
FIG. 5 is a schematic diagram of a road surface point cloud intensity enhancement process;
FIG. 6 is a comparison graph of point cloud intensity value distributions on road surface scan lines before and after enhancement of road surface point cloud intensity.
Detailed Description
The invention is described in detail below with reference to the figures and specific embodiments. The present embodiment is implemented on the premise of the technical solution of the present invention, and a detailed implementation manner and a specific operation process are given, but the scope of the present invention is not limited to the following embodiments.
As shown in fig. 1, a road surface point cloud intensity enhancing method based on vehicle-mounted laser scanning data includes the following steps:
s1: the method comprises the steps of obtaining road surface point cloud, a driving track line and the distance between a laser point in the road surface point cloud and the driving track line (short for point cloud distance characteristics) based on a scanning line of vehicle-mounted laser scanning data, and segmenting the road surface point cloud based on the driving track line to obtain a road surface point cloud segment. The method specifically comprises the following steps:
11: and determining the breakpoint of the scanning line according to the scanning angle difference between the front and back scanning laser points in the vehicle-mounted laser scanning data, thereby extracting the scanning line.
In this embodiment, as shown in fig. 2, on the same scanning line S, the scanning angle difference values of the consecutive laser spots are all non-negative numbers, so the break points of the scanning line are detected based on the strategy that the scanning angle difference values of the consecutive laser spots are negative numbers:
Figure BDA0001661769710000071
in fig. 2, θ is the scanning angle of the laser point pt,
Figure BDA0001661769710000072
is a scanning line Si-1The last point of the first and second points of the first,
Figure BDA0001661769710000073
for the next scanning line SiStarting point (pt)uAnd ptu+12 consecutive laser spots recorded for the system),
Figure BDA0001661769710000074
and
Figure BDA0001661769710000075
is the scan angle of 2 consecutive laser spots. In the same way, the method for preparing the composite material,
Figure BDA0001661769710000076
is a scanning line SiThe last point of the first and second points of the first,
Figure BDA0001661769710000077
for the next scanning line Si+1Starting point (pt)uAnd ptu+12 consecutive laser spots recorded for the system),
Figure BDA0001661769710000078
and
Figure BDA0001661769710000079
is a scanning angle of 2 successive laser spots, an
Figure BDA0001661769710000081
12: and (4) extracting the road surface point cloud by adopting a moving window method based on the scanning line generated in the step (11).
The moving window method in the embodiment includes 3 windows in total, and the positioning of the road edge point is realized by presetting the threshold of the cloud height difference of the points in each window. The window 1 detects a road edge point, the window 2 and the window 3 detect a road surface point, when the elevation difference of the window 1 is between 0.1m and 0.3m, and the elevation difference of the window 2 and the window 3 is less than or equal to 0.1m, the road edge point is determined to be located in the window 1, a point at a density mutation position in the window 1 is taken as a final road edge point, and a point cloud at one side of the road edge point close to the center is taken as a road surface point cloud, so that the extraction of the road surface point cloud is realized.
13: the method comprises the steps of counting the density distribution of laser points on each scanning line, inverting traveling track points according to the characteristic that the closer the distance between the laser points on the scanning lines and a scanner is, the denser the local point cloud distribution is, specifically, extracting the laser point corresponding to the maximum density distribution (namely, the laser point corresponding to the maximum density value) on each scanning line as the traveling track point during the operation of the vehicle-mounted laser scanning system, constructing a continuous traveling track line by the traveling track points, and obtaining the distance R from the laser points to the traveling track line.
14: and (3) sampling the driving track line at equal intervals to obtain pseudo track points, and making a partition line perpendicular to the road surface trend through the pseudo track points to realize road surface point cloud segmentation to obtain a road surface point cloud segment. In this embodiment, the sampling interval distance of the pseudo track points is 15m, and the segmented road surface point cloud segment is as shown in fig. 3.
S2: taking the road surface point cloud segment as a processing unit, and performing the following steps:
21: and (3) aiming at the road surface point cloud segment obtained in the step (14), establishing a homogeneous data set by adopting a median statistic method, establishing a cubic polynomial function describing the distance-intensity relation of the near distance effect, estimating model parameters in the cubic polynomial function based on the homogeneous data set, and establishing an original distance model function and a reference distance model function by using the model parameters. The method specifically comprises the following steps:
211: and (4) according to the distance R from the laser point to the traffic trajectory line obtained in the step (13), performing equal-distance subdivision on a section of road surface point cloud in a direction parallel to the traffic trajectory line to obtain a distance interval, wherein the distance interval is 0.5m in the embodiment.
212: the intensity of the laser points in each distance interval is counted, and a median statistical method is adopted to extract the laser points corresponding to the intensity median in each distance interval to construct a homogeneous data set (R)k,Ik)，RkIs the distance between the kth laser point and the scanner in the homogeneous dataset, IkThe intensity of the kth laser spot in the homogeneous dataset, k the number of the laser spot in the homogeneous dataset, k ═ 1,2,3Homo}，NHomoThe number of laser points in a homogenous data set.
213: according to the approximation theorem, a cubic polynomial function is established to describe the distance-intensity relation of the near distance effect, and the following formula is satisfied:
Figure BDA0001661769710000082
wherein eta is0,η1,η2,η3Are model parameters.
214: based on the homogeneous data set constructed in step 212 and the cubic polynomial function constructed in step 213, a least square method is adopted to carry out on the model parameter eta0,η1,η2,η3And (6) estimating.
215: according to the model parameter eta obtained in step 2140,η1,η2,η3Fitting of laser point intensity in road surface point cloud segment is realized, and an original distance model function f (R) is constructedj) Characterizing the distance-intensity relationship including the proximity effect:
Figure BDA0001661769710000091
wherein R isjIs the distance between the jth laser point and the laser scanner (laser head), j is the laser point number in the road surface point cloud segment, f (R)j) The physical meaning of the numerical value is the intensity value of the jth laser point in the road surface point cloud segment constructed according to the model.
Setting a reference distance RsAs shown in FIG. 4, a reference distance model function f (R) is establisheds)：
Figure BDA0001661769710000092
Wherein R iss＝min(Rj) The min function is a function for finding the minimum.
22: based on the original distance model function and the reference distance model function constructed in step 21, intensity correction of the laser spot is performed by using a ratio method and a difference method. The method specifically comprises the following steps:
221: and roughly correcting the intensity of the laser point by adopting a ratio method. With reference distance model function f (R)j) And the original distance model function f (R)s) Ratio of (3) to construct a correction coefficient Rat by a ratio methodjCalculating the intensity of each laser spot IjAnd the corresponding correction coefficient RatjThe product of (A) is used as the rough correction result I _ Rat of the intensity of each laser spotjThe following formula is satisfied:
I_Ratj＝Ij·Ratj
Figure BDA0001661769710000093
wherein, I _ RatjFor the result of the rough correction of the intensity of the jth laser spot, IjIs the intensity of the jth laser spot, RatjThe coefficients are corrected for the ratio corresponding to the jth laser spot.
222: and carrying out fine correction on the intensity coarse correction result of the laser spot by using a self-adaptive selection difference method. Calculating the intensity difference DeltaI before and after the rough correction of each laser point in step 221jJudging the intensity difference value delta I before and after the rough correctionjWhether the judgment condition is met: if the difference between the strength before and after the rough correctionΔIjLess than a set threshold value thetaIIf so, the intensity correction is finished; otherwise, performing intensity fine correction by adopting a difference method again to obtain a laser point cloud intensity fine correction result (as shown in a2, b2 and c2 in fig. 5):
Figure BDA0001661769710000094
Difj＝f(Rs)-f(Rj)
ΔIj＝I_Ratj-Ij
wherein, I _ DistjFor the final intensity correction of the jth laser spot, Δ IjDif is the difference in intensity before and after the rough correction of the jth laser spotjTo correct the coefficient by the difference method corresponding to the jth laser spot, θ in this embodimentIThe set value is 1000.
23: and (4) acquiring an intensity probability distribution characteristic value of the laser spot according to the global probability density value of each intensity interval based on the point cloud intensity corrected in the step (22), and performing intensity conversion by combining the corrected laser point cloud intensity to stretch the intensity difference among different ground objects. The method specifically comprises the following steps:
231: obtaining the intensity-corrected point cloud I _ Dist according to the final intensity correction result of each laser point in the road surface point cloud segment obtained in the step 22jAnd the histogram is used for counting the global probability density value p of each intensity interval in the intensity histogram and satisfying the following formula:
Figure BDA0001661769710000101
wherein N is the number of laser points in the intensity interval, N is the number of laser points in the road surface point cloud segment, and the intensity interval in the embodiment is 10.
232: calculating the probability distribution characteristic value F of each laser point in the road surface point cloud segment according to the global probability density value obtained in the step 231jThe following formula is satisfied:
Fj＝I_Distj·pj
wherein, FjIs the probability distribution characteristic value of the jth laser spot, I _ DistjFor the final intensity correction of the jth laser spot, pjThe global probability density value of the intensity interval where the jth laser point is located.
233: optimizing probability distribution eigenvalues FjIf the probability distribution characteristic value FjIf smaller, optimizing, if probability distribution characteristic value FjLarger remains unchanged:
Figure BDA0001661769710000102
wherein, Fj' is the probability distribution characteristic value of the j laser point after optimization.
234: and carrying out intensity conversion on the final intensity correction result of each laser spot, wherein the following formula is satisfied:
Figure BDA0001661769710000103
wherein, I _ tranjThe result of the intensity transformation of the jth laser spot (as shown in a3, b3, c3 of fig. 5).
24: for the laser points with the converted intensities, intensity denoising of the laser points is performed by adopting a multi-filter integration method according to uneven laser point density distribution, and laser point intensity enhancement of each road point cloud segment in the road point cloud is completed, as shown in a4, b4 and c4 of fig. 5. The method specifically comprises the following steps:
241: and performing neighborhood search by taking each laser point as a central reference point, and dividing the road surface point cloud segment after intensity conversion into a low-density edge area and a high-density central area according to the density distribution of the laser points.
In the embodiment, firstly, neighborhood search with the radius of 0.1m is carried out on a central reference point, and if the number of the searched point clouds is more than 9, the central reference point is regarded as a laser point of a high-density central area; otherwise, the center reference point is considered to be the laser point of the low density edge region.
242: 1) aiming at a low-density edge area, performing neighborhood search by adopting a K neighbor search method, and performing point cloud intensity denoising treatment by adopting a bilateral filter based on the space proximity and the intensity value similarity between a central reference point in the neighborhood and other laser points in the field to obtain a denoising result I _ noisejIn this embodiment, K is 9, and satisfies the following formula:
Figure BDA0001661769710000111
wherein, I _ densejThe intensity enhancement result for the jth laser spot, ner _ pjNumbering of laser spots in the field of the jth laser spot, ner _ njIs the number of laser points in the field of the jth laser spot, ner _ n in this embodimentjIs a number of 9, and the number of the grooves is,
Figure BDA0001661769710000112
is the inner-field first _ pjAs a result of the intensity transformation of the individual laser spots,
Figure BDA0001661769710000113
is the first _ p in the corresponding domainjWeight coefficient of each laser spot, weight coefficient
Figure BDA0001661769710000114
Depending on the product of the intensity value similarity and the spatial proximity:
Figure BDA0001661769710000115
wherein the content of the first and second substances,
Figure BDA0001661769710000116
to the center reference point and the first _ p in the fieldjThe similarity of the intensity values among the laser points is represented by the intensity difference between the central reference point and each point in the neighborhood,
Figure BDA0001661769710000117
to the center reference point and the first _ p in the fieldjThe spatial proximity between the laser points is characterized by the distance between the central reference point and each point in the proximity.
Figure BDA0001661769710000118
Figure BDA0001661769710000119
Wherein, I _ tranjAs a result of intensity conversion of the jth laser spot, σ _ IjIs the standard deviation of the distribution of intensity differences between the in-field central reference point of the jth laser spot and other laser spots in the field, σ _ DjIs the standard deviation of the distribution of the distances between the in-domain center reference point of the jth laser spot and the other laser spots in the domain,
Figure BDA00016617697100001110
to the center reference point and the first _ i in the fieldjThe distance between the laser points satisfies the following formula:
Figure BDA00016617697100001111
wherein (x)j,yj,zj) Is the coordinate of the jth laser spot (namely the in-field central reference point of the jth laser spot) in the XYZ coordinate system,
Figure BDA0001661769710000121
is the in-field inner-ner _ i of the jth laser spotjCoordinates in XYZ coordinate system of the individual laser spots.
2) Aiming at the high-density central area, a neighborhood intensity difference analysis method is adopted to select the minimum in the local neighborhood
Figure BDA0001661769710000122
The corresponding laser point intensity value is taken as the intensity value of the central reference pointSmoothing the intensity of the homogeneous point cloud, and keeping the main edge characteristics to obtain a denoising result I _ noisejThe following formula is satisfied:
Figure BDA0001661769710000123
Figure BDA0001661769710000124
wherein the content of the first and second substances,
Figure BDA0001661769710000125
the function represents the variable x that is found so that f (x) takes its minimum value,
Figure BDA0001661769710000126
is the inner-neighborhood first _ pjSum of intensity differences between one laser spot and other laser spots in the neighborhood, ner _ qjThe numbering of the laser spot in the field of the jth laser spot,
Figure BDA0001661769710000127
is the inner third of the fieldjThe intensity of each laser spot is transformed.
The point cloud intensity enhancement treatment can well inhibit the intensity noise in the homogeneous point cloud in the road surface point cloud, simultaneously keep the intensity difference between the heterogeneous point clouds, and the road surface point cloud intensity distribution after the point cloud enhancement treatment is as follows: the transition from high overlap to the presence of a more distinct demarcation. Referring to fig. 6, before the intensity of the point clouds is enhanced, the intensity distribution of the point clouds on the road surface scanning line is relatively scattered, the intensity difference in the homogeneous point clouds is relatively large, and the intensity distribution among the non-homogeneous point clouds is overlapped; after the point cloud intensity enhancement processing, the distribution of point cloud intensity values on a road surface scanning line is clear, the intensity distribution in homogeneous point clouds is basically consistent, and the intensity difference exists between non-homogeneous point clouds.

Claims (5)

1. A road surface point cloud intensity enhancing method based on vehicle-mounted laser scanning data is characterized by comprising the following steps:
s1: acquiring a road surface point cloud, a driving trajectory line and the distance from a laser spot in the road surface point cloud to the driving trajectory line based on vehicle-mounted laser scanning data, and segmenting the road surface point cloud to obtain a road surface point cloud segment;
s2: taking the road surface point cloud segment as a processing unit, and performing the following steps:
21: aiming at the point cloud segment of the road surface, a homogeneous data set is established by adopting a median statistical method, a cubic polynomial function describing the distance-intensity relation of the close range effect is established, model parameters in the cubic polynomial function are estimated based on the homogeneous data set, and an original distance model function and a reference distance model function are established by the model parameters;
22: based on the original distance model function and the reference distance model function, performing intensity correction on the laser point by adopting a ratio method and a difference method;
23: aiming at the laser spot with the corrected intensity, obtaining an intensity probability distribution characteristic value of the laser spot according to the global probability density value of each intensity interval, and carrying out intensity conversion on the laser spot by combining the corrected intensity;
24: aiming at the laser points with the converted intensities, intensity denoising of the laser points is carried out by adopting a multi-filter integration method according to uneven laser point density distribution, and laser point intensity enhancement of each road point cloud segment in the road point cloud is completed;
the step S1 specifically includes:
11: determining a breakpoint of a scanning line according to a scanning angle difference between front and rear scanning laser points in the vehicle-mounted laser scanning data, thereby extracting the scanning line;
12: based on the scanning lines extracted in the step 11, extracting a road surface point cloud by adopting a moving window method;
13: counting the density distribution of the laser points on each scanning line, extracting the laser points corresponding to the maximum density distribution on each scanning line as the traveling track points, constructing continuous traveling track lines by the traveling track points, and acquiring the distance from the laser points to the traveling track lines;
14: the method comprises the steps of sampling a driving trajectory line at equal intervals to obtain pseudo trajectory points, and segmenting a road surface point cloud based on the pseudo trajectory points to obtain a road surface point cloud segment;
the step 21 specifically comprises:
211: according to the distance from the laser spot to the traffic trajectory line, carrying out equal-distance subdivision on the road surface point cloud segment in the direction parallel to the traffic trajectory line to obtain a distance interval;
212: the intensity of the laser points in each distance interval is counted, and a median statistical method is adopted to extract the laser points corresponding to the intensity median in each distance interval to construct a homogeneous data set (R)k,Ik)，RkIs the distance between the kth laser point and the scanner in the homogeneous dataset, IkThe intensity of the kth laser point in the homogeneous data set is defined, and k is the serial number of the laser point in the homogeneous data set;
213: according to an approximation theorem, a cubic polynomial function describing the distance-intensity relation of the near distance effect is established, and the following formula is satisfied:
Figure FDA0002754047970000021
wherein eta is0、η1、η2、η3Is a model parameter;
214: based on homogeneous data set, using least square method to model parameter eta0、η1、η2、η3Carrying out estimation;
215: establishing an original distance model function f (R)j) And a reference distance model function f (R)s) The following formula is satisfied:
Figure FDA0002754047970000022
Figure FDA0002754047970000023
wherein R isjThe distance between the jth laser point and the scanner,j is the laser point number in the road surface point cloud segment, RsFor a set reference distance, Rs＝min(Rj)。
2. The method for enhancing the intensity of the road surface point cloud based on the vehicle-mounted laser scanning data according to claim 1, wherein the step 22 specifically comprises:
221: the intensity of the laser point is roughly corrected by adopting a ratio method, and the following formula is satisfied:
I_Ratj＝Ij·Ratj
Figure FDA0002754047970000024
wherein, I _ RatjThe rough correction result of the intensity of the jth laser point is obtained, j is the laser point number in the pavement point cloud segment, IjIs the intensity of the jth laser spot, RatjCorrection of the coefficients by a ratio corresponding to the j-th laser spot, f (R)j) As a function of the original distance model, f (R)s) Is a reference distance model function;
222: the self-adaptive selection difference method is used for carrying out fine correction on the rough intensity correction result of the laser spot, and the following formula is satisfied:
Figure FDA0002754047970000025
Difj＝f(Rs)-f(Rj)
ΔIj＝I_Ratj-Ij
wherein, I _ DistjFor the final intensity correction of the jth laser spot, Δ IjIs the difference in intensity before and after the rough correction of the jth laser spot, θIFor the set threshold value, DifjThe coefficients are corrected for the difference corresponding to the jth laser spot.
3. The method for enhancing the intensity of the road surface point cloud based on the vehicle-mounted laser scanning data according to claim 1, wherein the step 23 specifically comprises:
231: obtaining an intensity histogram of the intensity-corrected road surface point cloud segment according to the final intensity correction result of each laser point in the road surface point cloud segment obtained in the step 22, counting the global probability density value p of each intensity interval in the intensity histogram, and satisfying the following formula:
Figure FDA0002754047970000031
wherein N is the number of laser points in the intensity interval, and N is the number of laser points in the pavement point cloud segment;
232: obtaining the probability distribution characteristic value of each laser point, and satisfying the following formula:
Fj＝I_Distj·pj
wherein, FjIs the probability distribution characteristic value of the jth laser point, j is the laser point number in the road surface point cloud segment, I _ DistjFor the final intensity correction of the jth laser spot, pjThe global probability density value of the intensity interval where the jth laser point is located;
233: optimizing the probability distribution characteristic value to satisfy the following formula:
Figure FDA0002754047970000032
wherein, F'jThe probability distribution characteristic value of the optimized j laser point is obtained;
234: and carrying out intensity conversion on the final intensity correction result of each laser spot, wherein the following formula is satisfied:
Figure FDA0002754047970000033
wherein, I _ tranjThe result of the intensity transformation of the jth laser spot.
4. The method for enhancing the intensity of the road surface point cloud based on the vehicle-mounted laser scanning data according to claim 1, wherein the step 24 specifically comprises:
241: performing neighborhood search by taking each laser point as a central reference point, and dividing the road surface point cloud segment after intensity conversion into a low-density edge area and a high-density central area according to the density distribution of the laser points;
242: aiming at the low-density edge area, based on the spatial proximity and the intensity value similarity between the central reference point in the neighborhood and other laser points in the field, the intensity conversion result of the laser points is subjected to intensity denoising by adopting a bilateral filtering method, and the following formula is satisfied:
Figure FDA0002754047970000041
Figure FDA0002754047970000042
wherein, I _ densejIs the intensity enhancement result of the jth laser point, j is the laser point number in the pavement point cloud segment,
Figure FDA0002754047970000043
is the first _ p in the corresponding domainjWeight coefficient of individual laser points, ner _ pjNumbering of laser spots in the field of the jth laser spot, ner _ njThe number of laser points in the field of the jth laser spot,
Figure FDA0002754047970000044
is the inner-field first _ pjAs a result of the intensity transformation of the individual laser spots,
Figure FDA0002754047970000045
to the center reference point and the first _ p in the fieldjOf a laser spotThe similarity of the intensity values between the two,
Figure FDA0002754047970000046
to the center reference point and the first _ p in the fieldjSpatial proximity between individual laser points;
aiming at the high-density central area, intensity denoising is carried out on the intensity transformation result of the laser spot by adopting a neighborhood intensity difference analysis method, and the following formula is satisfied:
Figure FDA0002754047970000047
Figure FDA0002754047970000048
wherein the content of the first and second substances,
Figure FDA0002754047970000049
is the inner-neighborhood first _ pjSum of intensity differences between one laser spot and other laser spots in the neighborhood, ner _ qjThe numbering of the laser spot in the field of the jth laser spot,
Figure FDA00027540479700000410
is the inner third of the fieldjThe intensity of each laser spot is transformed.
5. The method as claimed in claim 4, wherein the spatial proximity and the intensity value similarity between the central reference point in the neighborhood and other laser points in the neighborhood in step 242 satisfy the following formulas:
Figure FDA00027540479700000411
Figure FDA0002754047970000051
wherein, I _ tranjAs a result of intensity conversion of the jth laser spot, σ _ IjIs the standard deviation of the distribution of intensity differences between the in-domain central reference point of the jth laser spot and other laser spots in the domain,
Figure FDA0002754047970000052
to the center reference point and the first _ p in the fieldjDistance between laser spots, σ _ DjIs the standard deviation of the distance distribution between the in-domain central reference point of the jth laser spot and other laser spots in the domain.
CN201810464773.3A 2018-05-16 2018-05-16 Road surface point cloud intensity enhancing method based on vehicle-mounted laser scanning data Active CN108681525B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810464773.3A CN108681525B (en) 2018-05-16 2018-05-16 Road surface point cloud intensity enhancing method based on vehicle-mounted laser scanning data

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810464773.3A CN108681525B (en) 2018-05-16 2018-05-16 Road surface point cloud intensity enhancing method based on vehicle-mounted laser scanning data

Publications (2)

Publication Number Publication Date
CN108681525A CN108681525A (en) 2018-10-19
CN108681525B true CN108681525B (en) 2020-12-25

Family

ID=63806498

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810464773.3A Active CN108681525B (en) 2018-05-16 2018-05-16 Road surface point cloud intensity enhancing method based on vehicle-mounted laser scanning data

Country Status (1)

Country Link
CN (1) CN108681525B (en)

Families Citing this family (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11592524B2 (en) * 2018-11-02 2023-02-28 Waymo Llc Computation of the angle of incidence of laser beam and its application on reflectivity estimation
EP3735678A4 (en) * 2018-11-13 2021-03-24 Beijing Didi Infinity Technology and Development Co., Ltd. Methods and systems for color point cloud generation
CN109614966B (en) * 2019-03-07 2019-07-26 纽劢科技(上海)有限公司 It is a kind of based on information fusion Lidar sensor efficient road surface and curb detection method
CN110322467B (en) * 2019-05-24 2021-07-16 麦恒智能科技(湖州)有限公司 Algorithm for improving point cloud density of 3D contour sensor on calculated plate surface
CN110308438B (en) * 2019-07-15 2021-08-10 南京林业大学 Method for correcting reflection intensity of laser radar by using multi-echo single-station scanning data
CN110598541B (en) * 2019-08-05 2021-07-23 香港理工大学深圳研究院 Method and equipment for extracting road edge information
CN110532963B (en) * 2019-08-30 2022-08-12 北京亚太文融数据技术研究院 Vehicle-mounted laser radar point cloud driven road marking accurate extraction method
CN110515060B (en) * 2019-09-05 2021-05-07 北京智行者科技有限公司 Method for calibrating reflectivity of multi-line laser radar
CN110619650A (en) * 2019-09-18 2019-12-27 嘉兴学院 Edge point extraction method and device based on line structure laser point cloud
CN112955778B (en) * 2019-09-26 2023-08-04 深圳市速腾聚创科技有限公司 Correction method, correction device, storage medium and multichannel laser radar
CN116698680B (en) * 2023-08-04 2023-09-29 天津创盾智能科技有限公司 Automatic monitoring method and system for biological aerosol
CN117541799B (en) * 2024-01-09 2024-03-08 四川大学 Large-scale point cloud semantic segmentation method based on online random forest model multiplexing

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103390169A (en) * 2013-07-19 2013-11-13 武汉大学 Sorting method of vehicle-mounted laser scanning point cloud data of urban ground objects

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103390169A (en) * 2013-07-19 2013-11-13 武汉大学 Sorting method of vehicle-mounted laser scanning point cloud data of urban ground objects

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
Study of surface brightness from backscattered laser intensity: calibration of laser data;S. Kaasalainen等;《 IEEE Geoscience and Remote Sensing Letters》;20050718;第2卷(第3期);255-259页 *
基于DBN的车载激光点云路侧多目标提取;罗海峰 等;《测绘学报》;20180228;第47卷(第2期);234-246页 *
激光强度值改正模型与点云分类精度;谭凯 等;《同济大学学报》;20140131;第42卷(第1期);131-135 *

Also Published As

Publication number Publication date
CN108681525A (en) 2018-10-19

Similar Documents

Publication Publication Date Title
CN108681525B (en) Road surface point cloud intensity enhancing method based on vehicle-mounted laser scanning data
CN109858460B (en) Lane line detection method based on three-dimensional laser radar
CN111079611B (en) Automatic extraction method for road surface and marking line thereof
CN106780524B (en) Automatic extraction method for three-dimensional point cloud road boundary
CN109188459B (en) Ramp small obstacle identification method based on multi-line laser radar
US20210350149A1 (en) Lane detection method and apparatus,lane detection device,and movable platform
CN108828621A (en) Obstacle detection and road surface partitioning algorithm based on three-dimensional laser radar
CN115761550A (en) Water surface target detection method based on laser radar point cloud and camera image fusion
CN113537316B (en) Vehicle detection method based on 4D millimeter wave radar point cloud
CN112986964B (en) Photon counting laser point cloud self-adaptive denoising method based on noise neighborhood density
CN113269889B (en) Self-adaptive point cloud target clustering method based on elliptical domain
WO2023207845A1 (en) Parking space detection method and apparatus, and electronic device and machine-readable storage medium
CN113917487A (en) Laser radar-based method for detecting road edge and drivable area of closed road
CN115690081A (en) Tree counting method, system, storage medium, computer equipment and terminal
CN113111707A (en) Preceding vehicle detection and distance measurement method based on convolutional neural network
CN116109601A (en) Real-time target detection method based on three-dimensional laser radar point cloud
CN112597839A (en) Road boundary detection method based on vehicle-mounted millimeter wave radar
Yao et al. Automatic extraction of road markings from mobile laser-point cloud using intensity data
CN114842166A (en) Negative obstacle detection method, system, medium, and apparatus applied to structured road
US11668799B2 (en) Histogram based L-shape detection of target objects
Jin et al. Pavement crack detection fused HOG and watershed algorithm of range image
CN113066041A (en) Pavement crack detection method based on stack sparse self-coding deep learning
Schomerus et al. Camera-based lane border detection in arbitrarily structured environments
CN111738094B (en) Automatic extraction method and device for remote sensing image road
Wang et al. Lane line extraction in raining weather images by ridge edge detection with improved MSR and Hessian matrix

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
GR01 Patent grant
GR01 Patent grant