CN110415259A - A kind of shade tree point cloud recognition methods based on laser reflection intensity - Google Patents
A kind of shade tree point cloud recognition methods based on laser reflection intensity Download PDFInfo
- Publication number
- CN110415259A CN110415259A CN201910696187.6A CN201910696187A CN110415259A CN 110415259 A CN110415259 A CN 110415259A CN 201910696187 A CN201910696187 A CN 201910696187A CN 110415259 A CN110415259 A CN 110415259A
- Authority
- CN
- China
- Prior art keywords
- laser
- intensity
- distance
- reflection intensity
- laser reflection
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/50—Depth or shape recovery
- G06T7/521—Depth or shape recovery from laser ranging, e.g. using interferometry; from the projection of structured light
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
- G06T7/75—Determining position or orientation of objects or cameras using feature-based methods involving models
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10028—Range image; Depth image; 3D point clouds
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Abstract
A kind of shade tree point cloud recognition methods based on laser reflection intensity, method includes the following steps: S1, the range correction model for establishing laser reflection intensity and incidence angle calibration model；S2, for region to be measured, selected part shade tree obtains point cloud data as sample；S3, the laser reflection intensity according to calibration model, after calculating correction；S4, region segmentation is carried out for the laser reflection intensity after correction, obtains tree crown, the trunk laser reflection intensity of shade tree sample, generate histogram, recognition threshold is set；S5, region to be measured for entirety, scanning obtain laser reflection intensity and judge whether cloud belongs to tree crown or trunk target according to recognition rule.Recognition methods of the invention, removal distance, influence of the incidence angle to intensity, by the laser reflection intensity distribution of tree crown, trunk after analysis correction, identifies shade tree tree crown, trunk point cloud, filter out the atural objects such as building, pedestrian, lane, pavement, turf, street lamp.
Description
Technical field
The present invention relates to shade tree point cloud recognition methods, especially a kind of shade tree point cloud based on laser reflection intensity is known
Other method.
Background technique
Laser radar (light detection and ranging, LiDAR) active remote sensing technology energy quick obtaining target
Surface high resolution, high-precision three-dimensional point cloud data, the forest parameter extraction based on LiDAR have become current research hotspot
And following development trend.However, diversified city type of ground objects causes shade tree point cloud identification difficulty big, so that being based on
The shade tree parameter extraction of LiDAR becomes a complexity and challenging work, studies the trade under complicated street environment
A tree point cloud recognition methods seems very necessary and urgently.
Existing method is mainly classified according to a cloud three-dimensional coordinate computational geometry feature, it is difficult to be filtered out and trade tree shape
Similar atural object.Except measurement point range information, the laser reflection intensity of LiDAR while return measurement point.Laser reflection intensity letter
Claim " intensity ", characterize target to the reflective spectral property of laser, currently, in terrain classification, marine environment exploration, building damage
Triage survey etc. is succeeded application.The laser reflection of the artificial atural object such as shade tree and building, street lamp, electric pole, sign board
Rate is different, therefore, improves shade tree point cloud accuracy of identification using laser reflection intensity.
However, due to a variety of by scanner characteristics, propagation in atmosphere, detector and amplifying circuit noise, scan geometry etc.
The influence of factor, there are relatively large deviations between laser reflection intensity and target actual reflectance, the different spectrum of jljl, foreign matter occur with spectrum
The phenomenon that, it can not be directly used in the extraction of target reflectivity characteristics, need to eliminate the influence of various factors by correction.
Summary of the invention
The purpose of the present invention is what is be difficult to filter out for the similar ground object target of shape present in the identification of shade tree point cloud
Problem proposes a kind of method using laser reflection intensity identification shade tree point cloud.
The technical scheme is that
The present invention provides a kind of shade tree point cloud recognition methods based on laser reflection intensity, and this method includes following step
It is rapid:
S1, two-dimensional laser radar laser reflected intensity calibration model is established, is with two-dimensional laser radar scanning frame intermediate point
Calibration object, laser reflection intensity data of the extraction standard diffusing reflection plate under different distance r, incidence angle θ, establishes laser reflection
The range correction model f of intensityrWith incidence angle calibration model fθ；
S2, for region to be measured, selected part shade tree is as sample, using mobile two-dimensional laser scanning system to trade
It sets sample and carries out motion scan, obtain point cloud distance r, scanning angle α, laser reflection intensity I, calculate laser foothold to laser
The incidence angle θ of origin；
S3, according to laser reflection intensity range correction model frWith incidence angle calibration model fθ, calculate correction after laser it is anti-
Penetrate intensity Ic；
S4, for correction after laser reflection intensity IcRegion segmentation is carried out, obtains the tree crown of shade tree sample, trunk swashs
Light reflected intensity Ic, generate tree crown, trunk point cloud intensity histogram after correction；Respectively using tree crown, trunk as target, according to target
Recognition threshold corresponding with laser reflection intensity histogram setting tree crown, trunk after non-targeted correction；
S5, region to be measured for entirety carry out motion scan to all shade trees using mobile two-dimensional laser scanning system,
It obtains laser reflection intensity I and judges whether cloud belongs to tree crown or trunk target according to following recognition rules
Further, step S1 specifically:
S1-1, fixed two-dimensional laser radar incidence angle θref, distance range [r is setmin, rmax] and distance interval Δ r,
Standard diffusing reflection plate is obtained within the scope of afore-mentioned distance using two-dimensional laser radar, adjusts distance according to fixed intervals, record is each
Received laser reflection intensity data (r under distances, I (rs, θref))；
Wherein: r indicates two-dimensional laser radar to the distance of standard diffusing reflection plate；θrefIt indicates to refer to incidence angle；I indicates to swash
Light reflected intensity；S indicates the number of different measurement distances；
S1-2, according to above-mentioned distance-intensity measurement data, obtained using least square method with reference to laser reflection under incidence angle
Functional relation f of the intensity I about distance rr；
Wherein: rspIndicate fitting function waypoint, choose at aforementioned each laser reflection maximum of intensity under away from
From value；K and L representative polynomial order；akAnd blEach term coefficient of representative polynomial, the laser reflection obtained according to step S1-1 are strong
Degree is according to (rs, I (rs, θref)) obtained using least square method solution, k, l representative polynomial items number；
According to root-mean-square error RMSE1And RMSE2Obtain piecewise function frEach section model order K, L；
Wherein: s indicates that the number of different measurement distances participates in distance-intensity data number of fitting；Wherein, s1It indicates
Measure distance r≤rspNumber；s2Indicate measurement distance r > rspNumber,It is measurement distance r≤rspData amount check,It is measurement distance r > rspData amount check；For RMSE1And RMSE2, current order and next order are selected respectively
Corresponding current order is as fitting order when the difference of RMSE value≤0.5.
The scanning distance r of S1-3, fixed two-dimensional laser radarref, ranges of incidence angles [θ is setmin, θmax] and angle between
Every Δ θ, standard diffusing reflection plate is obtained in aforementioned ranges of incidence angles using two-dimensional laser radar, is adjusted according to fixed intervals incident
Angle records received laser reflection intensity data (θ under each incidence angles′, I (rref, θs′))；
Wherein: θ indicates laser beam to the incidence angle of standard diffusing reflection plate；rrefIndicate reference distance；I indicates that laser is anti-
Penetrate intensity；The number of s ' expression different incidence angles；
S1-4, according to above-mentioned incidence angle-intensity measurement data, laser reflection under reference distance is obtained using least square method
Functional relation f of the intensity I about θθ；
Wherein: M representative polynomial order；cmEach term coefficient of representative polynomial, the laser reflection obtained according to step S1-3
Intensity data (θs′, I (rref, θs′)) obtained using least square method solution, m representative polynomial items number；
F is obtained according to root-mean-square error (RMSE)θModel order M；
Wherein: the number of s ' expression different incidence angles participates in incidence angle-intensity data number of fitting；NθIt indicates to participate in
The incidence angle of fitting-intensity data number；For RMSEθ, when selecting the difference of the RMSE of current order and next order≤0.5 pair
The current order answered is as fitting order.
Further, step S2 specifically:
S2-1, using two-dimensional laser radar initial position as coordinate origin O, establish rectangular coordinate system O-xyz, x-axis direction is
The direction of motion of the two-dimensional laser radar on vehicle, y-axis direction are two-dimensional laser radar scanning depth direction, and z-axis direction is quilt
Scan target short transverse perpendicular to the ground, the three-dimensional coordinate of jth frame ith measurement point are as follows:
Wherein, r (i, j) indicates the distance of i-th of laser point in two-dimensional laser radar jth frame, and α (i) indicates two-dimensional laser
The scanning angle of i-th of laser point of radar scanning frame, x (i, j) indicate the coordinate of i-th of laser point in the x direction in jth frame, y
(i, j) indicate jth frame in i-th of laser point depth direction coordinate, z (i, j) indicate jth frame in i-th of laser foothold
In the coordinate of short transverse, Δ t indicates the scan period of two-dimensional laser radar, and v indicates vehicle movement speed；
S2-2, the normal vector for calculating point cloud obtain with P (i, j) distance less than δ's point P (i, j) each in cloud
Point, establishes spheric neighbo(u)rhood point set, and δ indicates ball domain radius；
If P (i, j) neighborhood point number more than two, establishes the covariance matrix M of P (i, j) spheric neighbo(u)rhood point set, to its into
Row Eigenvalues Decomposition, by Metzler matrix minimal eigenvalue λminAs P, (normal vector of i, j fit Plane, is denoted as corresponding feature vector
N (i, j), goes to step S2-3；Otherwise, S3 is gone to step；
S2-3, incidence angle cosine value is calculated, seeks cos θ in conjunction with measurement point normal vector n (i, j) and laser vector l (i, j)
(i, j):
Wherein, l (i, j) is the coordinate difference of laser foothold P (i, j) Yu two-dimensional laser radar；
L (i, j)=(x (i, j), y (i, j), z (i, j))T(x (i, j), 0,0)T=(0, y (i, j), z (i, j))T。
Further, step S3 specifically: for point P (i, j) each in cloud, if its neighborhood point number more than two, is adopted
Laser reflection intensity is calculated with following formula:
Otherwise
Wherein: rrefIndicate reference distance, θrefIt indicates to refer to incidence angle.
Further, in step S4, be arranged recognition threshold the step of be: first pass through visually obtain coarse threshold, then again
Threshold value is adjusted, intensity value when selection discrimination is best is as recognition threshold.
Beneficial effects of the present invention:
Shade tree point cloud recognition methods based on laser reflection intensity of the invention, utilizes laser reflection intensity correction mould
Type carries out intensity correction, removal distance, incidence angle pair to the street point cloud data that mobile two-dimensional laser radar measurement system obtains
The influence of intensity.By the laser reflection intensity distribution of tree crown, trunk after analysis correction, shade tree tree crown, trunk point cloud are identified,
Filter out the atural objects such as non-targeted object, including building, pedestrian, lane, pavement, turf, street lamp.
Other features and advantages of the present invention will then part of the detailed description can be specified.
Detailed description of the invention
Exemplary embodiment of the invention is described in more detail in conjunction with the accompanying drawings, it is of the invention above-mentioned and its
Its purpose, feature and advantage will be apparent, wherein in exemplary embodiment of the invention, identical reference label
Typically represent same parts.
Fig. 1 shows the schematic diagram of mobile two-dimensional laser radar fix system.
Fig. 2 shows the schematic diagrames that incidence angle cosine value in embodiment is sought.
Fig. 3 shows the fitting precision schematic diagram of different rank in embodiment.
Fig. 4 shows the street point cloud data schematic diagram before correcting in embodiment.
Fig. 5 shows the point cloud intensity pseudo-colours schematic diagram data after correcting in embodiment.
Fig. 6 shows visual tree comb point cloud recognition result schematic diagram in embodiment.
Fig. 7 shows visual tree in embodiment and does cloud recognition result schematic diagram.
Specific embodiment
The preferred embodiment that the present invention will be described in more detail below with reference to accompanying drawings.Although showing the present invention in attached drawing
Preferred embodiment, however, it is to be appreciated that may be realized in various forms the present invention without the embodiment party that should be illustrated here
Formula is limited.
A kind of shade tree point cloud recognition methods based on laser reflection intensity, method includes the following steps: S1, establishing two
Tie up laser radar laser reflection intensity correction model, using two-dimensional laser radar scanning frame intermediate point as calibration object, extraction standard
Laser reflection intensity data of the diffusing reflection plate under different distance r, incidence angle θ, establishes the range correction mould of laser reflection intensity
Type frWith incidence angle calibration model fθ；
S1-1, fixed two-dimensional laser radar incidence angle θref, distance range [r is setmin, rmax] and distance interval Δ r,
Standard diffusing reflection plate is obtained within the scope of afore-mentioned distance using two-dimensional laser radar, adjusts distance according to fixed intervals, record is each
Received laser reflection intensity data (r under distances, I (rs, θref))；
Wherein: r indicates two-dimensional laser radar to the distance of standard diffusing reflection plate；θrefIt indicates to refer to incidence angle；I indicates to swash
Light reflected intensity；S indicates the number of different measurement distances；
S1-2, according to above-mentioned distance-intensity measurement data, obtained using least square method with reference to laser reflection under incidence angle
Functional relation f of the intensity I about distance rr；
Wherein: rspIndicate fitting function waypoint, choose at aforementioned each laser reflection maximum of intensity under away from
From value；K and L representative polynomial order；akAnd blEach term coefficient of representative polynomial, the laser reflection obtained according to step S1-1 are strong
Degree is according to (rs, I (rs, θref)) obtained using least square method solution, k, l representative polynomial items number；
According to root-mean-square error RMSE1And RMSE2Obtain piecewise function frEach section model order K, L；
Wherein: s indicates that the number of different measurement distances participates in distance-intensity data number of fitting；Wherein, s1It indicates
Measure distance r≤rspNumber,；s2Indicate measurement distance r > rspNumber,It is measurement distance r≤rspData amount check,It is measurement distance r > rspData amount check；For RMSE1And RMSE2, current order and next order are selected respectively
Corresponding current order is as fitting order when the difference of RMSE value≤0.5；
The scanning distance r of S1-3, fixed two-dimensional laser radarref, ranges of incidence angles [θ is setmin, θmax] and angle between
Every Δ θ, standard diffusing reflection plate is obtained in aforementioned ranges of incidence angles using two-dimensional laser radar, is adjusted according to fixed intervals incident
Angle records received laser reflection intensity data (θ under each incidence angles′, I (rref, θs′))；
Wherein: θ indicates laser beam to the incidence angle of standard diffusing reflection plate；rrefIndicate reference distance；I indicates that laser is anti-
Penetrate intensity；The number of s ' expression different incidence angles；
S1-4, according to above-mentioned incidence angle-intensity measurement data, laser reflection under reference distance is obtained using least square method
Functional relation f of the intensity I about θθ；
Wherein: M representative polynomial order；cmEach term coefficient of representative polynomial, the laser reflection obtained according to step S1-3
Intensity data (θs′, I (rref, θs′)) obtained using least square method solution, m representative polynomial items number；
F is obtained according to root-mean-square error (RMSE)θModel order M；
Wherein: the number of s ' expression different incidence angles participates in incidence angle-intensity data number of fitting；NθIt indicates to participate in
The incidence angle of fitting-intensity data number；For RMSEθ, when selecting the difference of the RMSE of current order and next order≤0.5 pair
The current order answered is as fitting order.
S2, for region to be measured, selected part shade tree is as sample, using mobile two-dimensional laser scanning system to trade
It sets sample and carries out motion scan, obtain point cloud distance r, scanning angle α, laser reflection intensity I, calculate laser foothold to laser
The incidence angle θ of origin；
S2-1, as shown in Figure 1, using two-dimensional laser radar initial position as coordinate origin O, establish rectangular coordinate system O-xyz,
X-axis direction is the direction of motion of the two-dimensional laser radar on vehicle, and y-axis direction is two-dimensional laser radar scanning depth direction, z-axis
Direction is scanned target short transverse perpendicular to the ground, the three-dimensional coordinate of jth frame ith measurement point are as follows:
Wherein, r (i, j) indicates the distance of i-th of laser point in two-dimensional laser radar jth frame, and α (i) indicates two-dimensional laser
The scanning angle of i-th of laser point of radar scanning frame, x (i, j) indicate the coordinate of i-th of laser point in the x direction in jth frame, y
(i, j) indicate jth frame in i-th of laser point depth direction coordinate, z (i, j) indicate jth frame in i-th of laser foothold
In the coordinate of short transverse, Δ t indicates the scan period of two-dimensional laser radar, and v indicates vehicle movement speed；
S2-2, the normal vector for calculating point cloud obtain with P (i, j) distance less than δ's point P (i, j) each in cloud
Point, establishes spheric neighbo(u)rhood point set, and δ indicates ball domain radius；
If P (i, j) neighborhood point number more than two, establishes the covariance matrix M of P (i, j) spheric neighbo(u)rhood point set, to its into
Row Eigenvalues Decomposition, by Metzler matrix minimal eigenvalue λminAs P, (normal vector of i, j fit Plane, is denoted as corresponding feature vector
N (i, j), goes to step S2-3；Otherwise, S3 is gone to step；
S2-3, as shown in Fig. 2, calculate incidence angle cosine value, in conjunction with measurement point normal vector n (i, j) and laser vector l (i,
J) cos θ (i, j) is sought:
Wherein, l (i, j) is the coordinate difference of laser foothold P (i, j) Yu two-dimensional laser radar；
L (i, j)=(x (i, j), y (i, j), z (i, j))T(x (i, j), 0,0)T=(0, y (i, j), z (i, j))T。
S3, according to laser reflection intensity range correction model frWith incidence angle calibration model fθ, calculate correction after laser it is anti-
Penetrate intensity Ic；For point P (i, j) each in cloud, if its neighborhood point number more than two, it is anti-that laser is calculated using following formula
Penetrate intensity:
Otherwise
Wherein: rrefIndicate reference distance, θrefIt indicates to refer to incidence angle.
S4, for correction after laser reflection intensity IcRegion segmentation is carried out, obtains the tree crown of shade tree sample, trunk swashs
Light reflected intensity Ic, tree crown, trunk point cloud intensity histogram after correction are generated, respectively using tree crown, trunk as target, according to target
Recognition threshold corresponding with laser reflection intensity histogram setting tree crown, trunk after non-targeted correction；First pass through visual acquisition
Then coarse threshold adjusts threshold value again, intensity value when selection discrimination is best is as recognition threshold；
S5, region to be measured for entirety carry out motion scan to all shade trees using mobile two-dimensional laser scanning system,
It obtains laser reflection intensity I and judges whether cloud belongs to tree crown or trunk target according to following recognition rules
When specific implementation:
Two-dimensional laser radar of the experiment using the production of Hokuyo company, Japan, model UTM-30LX.This two-dimensional laser thunder
Up to the infrared ray using wavelength 905nm, the measured value for obtaining different angle is swung by motor, measures distance 0.1m-30m, is surveyed
Accuracy of measurement ± 30mm, 270 ° of scanning range, 0.25 ° of angular resolution, scan period 25ms.UTM-30LX scans acquisition 1 every time
Frame data, distance and laser reflection intensity comprising 1081 different angles, respectively with 4 bytes, 2 byte representations.Using UTM-
30LX scans intermediate point, and the 541st measurement point is tested, and establishes laser reflection intensity correction model.Table 1 and Fig. 3 provide rank
Number K, L, M take root-mean-square error when different value, choose K=3, L=5, M=1.
The root-mean-square error of 1 different rank of table
The mobile two-dimensional laser radar measurement system used is tested, with the track bottom comprising 2 driving wheels, 18 driven wheels
Disk trolley is as mobile platform, using STM32F103ZET6 as agv controller, obtains small vehicle speed by speed measuring coder,
Active wheel speed is controlled according to proportional-plus-derivative-integral (Proportion-Integral-Differential, PID) algorithm, it is real
The accurate control of existing trolley driving direction and speed.
Three-dimensional point cloud is shown using the PCLVisualizer visualization class that PCL open source point cloud library provides, and takes cloud laser anti-
Continuous three bytes of intensity are penetrated as point cloud color R, G, B, the point cloud data before correcting after Pseudocolor is as shown in Figure 4.By
The factors such as distance, incidence angle influence, and point cloud intensity can not reflect target real reflectance, it is difficult to be identified by laser reflection intensity
Tree crown, trunk point cloud.
Fig. 5 is the point cloud intensity pseudocolour picture after correction, can obviously distinguish tree crown, trunk and other ground object targets.
Tree crown, trunk intensity distribution and other targets have greater overlap before correcting, corrected to enable intensity accurately anti-
Target reflectivity is reflected, tree crown, trunk intensity and other targets have larger difference.Significantly, since the air-conditioning of building,
Window jamb is all metal material as street lamp, and building still has more be overlapped with the intensity distribution of street lamp after correction.
Tree crown, trunk point cloud strength range are set according to intensity histogram, tree crown, trunk point are identified according to bound threshold value
Cloud.Shade tree point cloud recognition effect is measured with precision ratio (precision) and recall ratio (recall):
Table 2 provides the recognition result of correction front and back tree crown, trunk point cloud.Original point cloud intensity can not be applied to tree crown, tree
Dry identification, by laser reflection intensity correction, effective compensation distance, incidence error greatly improve tree crown, the identification of trunk point cloud
Rate, tree crown, trunk point cloud precision ratio reach respectively 99.15% and 92.73%, effectively filter out other most ground object targets
Point cloud, including street lamp similar with stem form.
2 shade tree point cloud of table identifies precision ratio and recall ratio
Tree crown, trunk point cloud recognition result are visualized, as shown in Figure 6,7.Branch part and trunk due to tree crown point cloud
Point cloud has similar laser reflection intensity, and tree crown false-alarm is mainly trunk point cloud, and tree crown missing inspection is mainly branch point cloud, and trunk is empty
Alert main branch point cloud, trunk missing inspection concentrate on trunk marginal position, and partial dot cloud incidence angle is larger is not easy to measure for this, angle school
Plus effect is poor.
Various embodiments of the present invention are described above, above description is exemplary, and non-exclusive, and
It is not limited to disclosed each embodiment.Without departing from the scope and spirit of illustrated each embodiment, for this skill
Many modifications and changes are obvious for the those of ordinary skill in art field.
Claims (5)
1. a kind of shade tree point cloud recognition methods based on laser reflection intensity, which is characterized in that method includes the following steps:
S1, two-dimensional laser radar laser reflected intensity calibration model is established, is correction with two-dimensional laser radar scanning frame intermediate point
Object, laser reflection intensity data of the extraction standard diffusing reflection plate under different distance r, incidence angle θ, establishes laser reflection intensity
Range correction model frWith incidence angle calibration model fθ；
S2, for region to be measured, selected part shade tree is as sample, using mobile two-dimensional laser scanning system to shade tree sample
This progress motion scan obtains point cloud distance r, scanning angle α, laser reflection intensity I, calculates laser foothold to laser origin
Incidence angle θ；
S3, according to laser reflection intensity range correction model frWith incidence angle calibration model fθ, calculate correction after laser reflection it is strong
Spend Ic；
S4, for correction after laser reflection intensity IcRegion segmentation is carried out, tree crown, the trunk laser for obtaining shade tree sample are anti-
Penetrate intensity Ic, generate tree crown, trunk point cloud intensity histogram after correction；Respectively using tree crown, trunk as target, according to target with it is non-
Tree crown, the corresponding recognition threshold of trunk is arranged in laser reflection intensity histogram after target correction；
S5, whole region to be measured is obtained using mobile two-dimensional laser scanning system to all shade trees progress motion scan
Laser reflection intensity I judges whether cloud belongs to tree crown or trunk target according to following recognition rules
2. the shade tree point cloud recognition methods according to claim 1 based on laser reflection intensity, which is characterized in that step
S1 specifically:
S1-1, fixed two-dimensional laser radar incidence angle θref, distance range [r is setmin, rmax] and distance interval Δ r, it uses
Two-dimensional laser radar obtains standard diffusing reflection plate within the scope of afore-mentioned distance, adjusts distance according to fixed intervals, records each distance
Under received laser reflection intensity data (rs, I (rs, θref))；
Wherein: r indicates two-dimensional laser radar to the distance of standard diffusing reflection plate；θrefIt indicates to refer to incidence angle；I indicates that laser is anti-
Penetrate intensity；S indicates the number of different measurement distances；
S1-2, according to above-mentioned distance-intensity measurement data, obtained using least square method with reference to laser reflection intensity under incidence angle
Functional relation f of the I about distance rr；
Wherein: rspIt indicates fitting function waypoint, chooses the distance value at aforementioned each laser reflection maximum of intensity under；
K and L representative polynomial order；akAnd blEach term coefficient of representative polynomial, the laser reflection intensity data obtained according to step S1-1
(rs, I (rs, θref)) obtained using least square method solution, k, l representative polynomial items number；
According to root-mean-square error RMSE1And RMSE2Obtain piecewise function frEach section model order K, L；
Wherein: s indicates that the number of different measurement distances participates in distance-intensity data number of fitting；Wherein, s1Indicate measurement
Distance r≤rspNumber；s2Indicate measurement distance r > rspNumber,It is measurement distance r≤rspData amount check,It is
Measure distance r > rspData amount check；For RMSE1And RMSE2, select respectively current order and next order RMSE value it
Corresponding current order is as fitting order when difference≤0.5；
The scanning distance r of S1-3, fixed two-dimensional laser radarref, ranges of incidence angles [θ is setmin, θmax] and angle interval delta
θ obtains standard diffusing reflection plate in aforementioned ranges of incidence angles using two-dimensional laser radar, adjusts incidence angle according to fixed intervals,
Record received laser reflection intensity data (θ under each incidence angles′, I (rref, θs′))；
Wherein: θ indicates laser beam to the incidence angle of standard diffusing reflection plate；rrefIndicate reference distance；I indicates that laser reflection is strong
Degree；The number of s ' expression different incidence angles；
S1-4, according to above-mentioned incidence angle-intensity measurement data, laser reflection intensity under reference distance is obtained using least square method
Functional relation f of the I about θθ；
Wherein: M representative polynomial order；cmEach term coefficient of representative polynomial, the laser reflection intensity number obtained according to step S1-3
According to (θs′, I (rref, θs′)) obtained using least square method solution, m representative polynomial items number；
F is obtained according to root-mean-square error (RMSE)θModel order M；
Wherein: the number of s ' expression different incidence angles participates in incidence angle-intensity data number of fitting；N indicates to participate in fitting
Incidence angle-intensity data number；For RMSEθ, corresponding when selecting the difference of the RMSE of current order and next order≤0.5
Current order is as fitting order.
3. the shade tree point cloud recognition methods according to claim 1 based on laser reflection intensity, which is characterized in that step
S2 specifically:
S2-1, using two-dimensional laser radar initial position as coordinate origin O, establish rectangular coordinate system O-xyz, x-axis direction is two dimension
The direction of motion of the laser radar on vehicle, y-axis direction are two-dimensional laser radar scanning depth direction, and z-axis direction is scanned
Target short transverse perpendicular to the ground, the three-dimensional coordinate of jth frame ith measurement point are as follows:
Wherein, r (i, j) indicates the distance of i-th of laser point in two-dimensional laser radar jth frame, and α (i) indicates two-dimensional laser radar
The scanning angle of i-th of laser point of scanning frame, x (i, j) indicate jth frame in the coordinate of i-th of laser point in the x direction, y (i,
J) indicate jth frame in i-th of laser point depth direction coordinate, z (i, j) indicate jth frame in i-th of laser foothold in height
The coordinate in direction is spent, Δ t indicates the scan period of two-dimensional laser radar, and v indicates vehicle movement speed；
S2-2, the normal vector for calculating point cloud obtain the point for being less than δ with P (i, j) distance, build for point P (i, j) each in cloud
Vertical spheric neighbo(u)rhood point set, δ indicate ball domain radius；
If P (i, j) neighborhood point number more than two, establishes the covariance matrix M of P (i, j) spheric neighbo(u)rhood point set, spy is carried out to it
Value indicative is decomposed, by Metzler matrix minimal eigenvalue λminNormal vector of the corresponding feature vector as P (i, j) fit Plane, is denoted as n
(i, j) goes to step S2-3；Otherwise, S3 is gone to step；
S2-3, calculate incidence angle cosine value, in conjunction with measurement point normal vector n (i, j) and laser vector l (i, j) seek cos θ (i,
J):
Wherein, l (i, j) is the coordinate difference of laser foothold P (i, j) Yu two-dimensional laser radar；
L (i, j)=(x (i, j), y (i, j), z (i, j))T(x (i, j), 0,0)T=(0, y (i, j), z (i, j))T。
4. the shade tree point cloud recognition methods according to claim 1 or 2 based on laser reflection intensity, which is characterized in that
Step S3 specifically: for point P (i, j) each in cloud, if its neighborhood point number more than two, calculated and swashed using following formula
Light reflected intensity:
Otherwise
Wherein: rrefIndicate reference distance, θrefIt indicates to refer to incidence angle.
5. the shade tree point cloud recognition methods according to claim 1 based on laser reflection intensity, which is characterized in that step
In S4, the step of recognition threshold is arranged, is: first passing through and visually obtains coarse threshold, then adjust threshold value again, chooses discrimination most
Intensity value when good is as recognition threshold.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910696187.6A CN110415259B (en) | 2019-07-30 | 2019-07-30 | Street tree point cloud identification method based on laser reflection intensity |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910696187.6A CN110415259B (en) | 2019-07-30 | 2019-07-30 | Street tree point cloud identification method based on laser reflection intensity |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110415259A true CN110415259A (en) | 2019-11-05 |
| CN110415259B CN110415259B (en) | 2023-04-28 |
Family
ID=68364245
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910696187.6A Active CN110415259B (en) | 2019-07-30 | 2019-07-30 | Street tree point cloud identification method based on laser reflection intensity |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110415259B (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110910407A (en) * | 2019-11-25 | 2020-03-24 | 南京林业大学 | Pavement tree trunk extraction method based on mobile laser scanning point cloud data |
| CN113109832A (en) * | 2021-03-18 | 2021-07-13 | 浙江农林大学 | Bamboo degree judgment method based on laser echo intensity |
| CN113155027A (en) * | 2021-04-27 | 2021-07-23 | 中铁工程装备集团有限公司 | Tunnel rock wall feature identification method |
| CN113311408A (en) * | 2021-07-07 | 2021-08-27 | 中国地质大学（武汉） | Radiation correction method and device for hyperspectral laser radar |
| CN113419251A (en) * | 2021-05-17 | 2021-09-21 | 重庆大学 | Attitude recognition, coding and decoding and communication method based on laser reflection |
| CN113666305A (en) * | 2021-08-31 | 2021-11-19 | 杭州派珞特智能技术有限公司 | Intelligent forklift laser positioning method based on motion compensation and reflecting plate optimized sorting |
| CN115079126A (en) * | 2022-05-12 | 2022-09-20 | 探维科技(北京)有限公司 | Point cloud processing method, device, equipment and storage medium |
| CN115390051A (en) * | 2022-10-27 | 2022-11-25 | 深圳煜炜光学科技有限公司 | Laser radar calibration method, device, equipment and storage medium |
| CN116400371A (en) * | 2023-06-06 | 2023-07-07 | 山东大学 | Indoor reflective transparent object position identification method and system based on three-dimensional point cloud |
| CN115079126B (en) * | 2022-05-12 | 2024-05-14 | 探维科技(北京)有限公司 | Point cloud processing method, device, equipment and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2013088188A (en) * | 2011-10-14 | 2013-05-13 | Fuji Architect Co Ltd | Form investigation method of three-dimensional measurement subject |
| CN104808191A (en) * | 2015-05-08 | 2015-07-29 | 南京林业大学 | Tree species classification method based on full-waveform LiDAR single-tree canopy volume decomposition |
-
2019
- 2019-07-30 CN CN201910696187.6A patent/CN110415259B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2013088188A (en) * | 2011-10-14 | 2013-05-13 | Fuji Architect Co Ltd | Form investigation method of three-dimensional measurement subject |
| CN104808191A (en) * | 2015-05-08 | 2015-07-29 | 南京林业大学 | Tree species classification method based on full-waveform LiDAR single-tree canopy volume decomposition |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110910407A (en) * | 2019-11-25 | 2020-03-24 | 南京林业大学 | Pavement tree trunk extraction method based on mobile laser scanning point cloud data |
| CN110910407B (en) * | 2019-11-25 | 2023-09-15 | 南京林业大学 | Street tree trunk extraction method based on mobile laser scanning point cloud data |
| CN113109832A (en) * | 2021-03-18 | 2021-07-13 | 浙江农林大学 | Bamboo degree judgment method based on laser echo intensity |
| CN113155027A (en) * | 2021-04-27 | 2021-07-23 | 中铁工程装备集团有限公司 | Tunnel rock wall feature identification method |
| CN113419251A (en) * | 2021-05-17 | 2021-09-21 | 重庆大学 | Attitude recognition, coding and decoding and communication method based on laser reflection |
| CN113419251B (en) * | 2021-05-17 | 2023-07-18 | 重庆大学 | Gesture recognition, coding and decoding and communication method based on laser reflection |
| CN113311408B (en) * | 2021-07-07 | 2023-05-26 | 中国地质大学（武汉） | Radiation correction method and device for hyperspectral laser radar |
| CN113311408A (en) * | 2021-07-07 | 2021-08-27 | 中国地质大学（武汉） | Radiation correction method and device for hyperspectral laser radar |
| CN113666305B (en) * | 2021-08-31 | 2023-02-21 | 杭州派珞特智能技术有限公司 | Intelligent forklift laser positioning method based on motion compensation and reflecting plate optimized sorting |
| CN113666305A (en) * | 2021-08-31 | 2021-11-19 | 杭州派珞特智能技术有限公司 | Intelligent forklift laser positioning method based on motion compensation and reflecting plate optimized sorting |
| CN115079126A (en) * | 2022-05-12 | 2022-09-20 | 探维科技(北京)有限公司 | Point cloud processing method, device, equipment and storage medium |
| CN115079126B (en) * | 2022-05-12 | 2024-05-14 | 探维科技(北京)有限公司 | Point cloud processing method, device, equipment and storage medium |
| CN115390051A (en) * | 2022-10-27 | 2022-11-25 | 深圳煜炜光学科技有限公司 | Laser radar calibration method, device, equipment and storage medium |
| CN116400371A (en) * | 2023-06-06 | 2023-07-07 | 山东大学 | Indoor reflective transparent object position identification method and system based on three-dimensional point cloud |
| CN116400371B (en) * | 2023-06-06 | 2023-09-26 | 山东大学 | Indoor reflective transparent object position identification method and system based on three-dimensional point cloud |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110415259B (en) | 2023-04-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110415259A (en) | A kind of shade tree point cloud recognition methods based on laser reflection intensity | |
| Lambert et al. | Performance analysis of 10 models of 3D LiDARs for automated driving | |
| EP3491332B1 (en) | Reflectivity map estimate from dot based structured light systems | |
| Chauve et al. | Processing full-waveform lidar data: modelling raw signals | |
| Pesci et al. | Effects of surface irregularities on intensity data from laser scanning: an experimental approach. | |
| US5087916A (en) | Method of navigation | |
| CN110208771B (en) | Point cloud intensity correction method of mobile two-dimensional laser radar | |
| Tan et al. | Correction of mobile TLS intensity data for water leakage spots detection in metro tunnels | |
| Bienert et al. | Analysis of the information content of terrestrial laserscanner point clouds for the automatic determination of forest inventory parameters | |
| Briese et al. | Calibration of full-waveform airborne laser scanning data for object classification | |
| Forsman et al. | Bias of cylinder diameter estimation from ground-based laser scanners with different beam widths: A simulation study | |
| Abed et al. | Echo amplitude normalization of full-waveform airborne laser scanning data based on robust incidence angle estimation | |
| CN105526906A (en) | Wide-angle dynamic high-precision laser angle measurement method | |
| Gross et al. | Segmentation of tree regions using data of a full-waveform laser | |
| Wang et al. | Multichannel interconnection decomposition for hyperspectral LiDAR waveforms detected from over 500 m | |
| Liu et al. | Point cloud intensity correction for 2D LiDAR mobile laser scanning | |
| CN110306414B (en) | Pavement structure depth detection method | |
| Zhu et al. | Design of laser scanning binocular stereo vision imaging system and target measurement | |
| Tanaka et al. | Measurement of forest canopy structure with a laser plane range-finding method–development of a measurement system and applications to real forests | |
| CN109708570A (en) | Information collection and analysis method and device for face structural plane | |
| Ullrich et al. | Long-range high-performance time-of-flight-based 3D imaging sensors | |
| Li et al. | Damage detection for historical architectures based on TLS intensity data | |
| CN109559356B (en) | Expressway sight distance detection method based on machine vision | |
| Li et al. | Achieve accurate recognition of 3D point cloud images by studying the scattering characteristics of typical targets | |
| Anna et al. | Physical modelling of point-cloud (3D) and full-wave-form (4D) laser imaging |
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 |