CN103760569A - Drivable region detection method based on laser radar - Google Patents

Abstract

The invention discloses a drivable region detection method based on a laser radar. The method includes the steps that the laser radar is used for scanning an actual traffic scene, and flat and rugged conditions of a current scene are obtained; scanning data returned by the laser radar are divided into a plurality of fragments through the height variation condition, an iterative linear fitting algorithm is used for extracting appropriate linear features from all the segments, flat regions are found out from the fragments which meet constraint conditions, and drivable regions are determined from the flat regions according to the requirements of the vehicle width; if a multi-line laser exists, corresponding drivable regions can be determined from the flat regions reflected by the line data on the basis of a drivable region of former line data from the near to the distant. According to the method, detection of the drivable regions in the actual traffic scene can be effectively completed, safe passage regions of an automobile are described accurately, and the method has good real-time performance, usability and robustness.

Description

A kind of wheeled method for detecting area based on laser radar

Technical field

The invention belongs to the traffic scene perception field in pilotless automobile technology, be specifically related to a kind of laser radar that uses and as Data Source, complete the method for wheeled region perception in actual traffic scene.

Background technology

In recent years along with the high speed development of automobile industry, traffic hazard has become global problem, the dead and wounded Population size estimation of the annual traffic hazard in the whole world surpasses more than 50 ten thousand people, therefore integrates the unmanned application of the technology such as automatic control, artificial intelligence, pattern-recognition and gives birth to.For in real time reliable perception environmental information, on pilotless automobile, equipped various active sensors and passive sensor, comprise camera, laser radar, millimetre-wave radar and GPS, it is one of key component in unmanned technology that wheeled region is detected.All the time, vision technique is the research direction of researchist's first-selection normally, and this is because its information content is large, cost is low, operate power is low, sweep time is short.But topmost deficiency of vision technique is exactly stricter for the requirement of external environment.For occurring in environment that shade, driving environment are comparatively complicated, roadside mark (or sign) is lost, illumination is not good, the situation such as low visibility or bad weather, the image information that vision technique (video camera) obtains often signal to noise ratio (S/N ratio) is very low, makes the method for feature extraction be difficult to deal with data.Thereby be difficult to the regional in image to be cut apart accurately, also just wheeled region cannot be detected accurately.

Summary of the invention

The object of the present invention is to provide a kind of efficient, in real time, the wheeled method for detecting area based on laser radar of robust.

For achieving the above object, the present invention has adopted following technical scheme.

This wheeled method for detecting area comprises the following steps:

First Calibration of Laser radar, then carries out coordinate conversion by the laser radar data collecting; Then according to the principle of work of laser radar, cutting apart laser radar data is some fragments; Data in fragment are done to assorted point is eliminated and the true value with data estimator is processed in filtering; Then carry out the feature extraction of laser radar data; Then according to wheeled range constraint and body width constraint condition, from the feature of extracting, determine wheeled region.

Described wheeled method for detecting area specifically comprises the following steps:

1) in order to complete the description for wheeled region, first determine the height that laser radar is installed on vehicle, calculate the downward-sloping angle of laser radar simultaneously, then the laser radar data collecting is transformed into local 3 dimension rectangular coordinate systems from 2 dimension polar coordinate systems;

2) laser radar data is cut apart: if adjacent laser data scans on same object, the difference in height between two data points is less; Contrary adjacent laser data scans on different objects, difference in height between two data points is larger, so obtain after laser radar data when be tilted to lower scanning road surface with laser radar, adopt the height difference of adjacent point-to-point transmission as the standard of Data Segmentation, laser radar data is divided into several fragments;

3) assorted point is eliminated and a data filtering: in laser radar data, can have all larger data points of some and adjacent both sides difference in height, in the present invention, this class data point is called to assorted point, it does not reflect the coordinate of true spatial location, so through step 2) after, the fragment that only comprises single laser radar data point is removed, then remaining laser radar data fragment is carried out to filtering processing removal noise and obtain the data slot for feature extraction, owing to being subject to the impact of system noise and random noise, there is deviation with actual corresponding locus in the locus of laser radar data reflection, therefore need to carry out filtering processing to laser radar data, to obtain, real space position is described more accurately,

4) feature extraction: after step 3), adopt the Algorithm of fitting a straight line of iteration to extract corresponding linear feature from the data slot for feature extraction, make laser radar data become the plane in corresponding actual scene;

5) determine wheeled region: the linear feature extracting from step 4), select the linear feature that meets the smooth constraint in wheeled region, wheeled region continuity constraint and body width constraint as wheeled region.

Described step 1) comprises following concrete steps:

1.1) take laser radar installation site as the local 3 dimension rectangular coordinate systems of reference point definition, then measure laser radar apart from the setting height(from bottom) H of surface level, and calculate laser radar and be tilted to lower scanning plane with respect to the angle α of vertical direction;

1.2) set up Laser Radar Scanning polar coordinates to the transformational relation of coordinate in local 3 dimension rectangular coordinate systems, according to transformational relation, obtain laser radar data point P _i3 corresponding dimension rectangular coordinate values, each Laser Radar Scanning polar data point comprises emission angle β _iwith scanning distance d _itwo parameters, P _i=(d _i, β _i) coordinate figures of corresponding local 3 dimension rectangular coordinate systems are P _i=(x _i, y _i, z _i), Equation is as follows:

x _i=sinα·cosβ _i·d _i

y _i=sinβ _i·d _i

z _i=H-cosα·cosβ _i·d _i

Described step 2) comprise following concrete steps:

According to the principle of laser radar sequential scanning, can there is not too large deviation in the height value that scans the adjacent two data point of same object, the place that difference in height is larger represents that two data points scan on different objects, so whole laser radar datas is divided into some fragment S according to the difference of consecutive number strong point height value _k:

S _k={P _i},(i=N _sk,…,N _ek)

s.t.|z _s-z _s-1|≥T _z,|z _e-z _e+1|≥T _z

Wherein, P _irepresent laser radar data point, z _sthe height that represents the initial laser radar data point of fragment, z _ethe height that represents fragment end laser radar data point, T _zrepresent difference in height threshold value.

Described filtering is processed and is comprised following concrete steps:

For data slot, according to current data point P _iselecting it to set data point in window carries out mean filter and carrys out estimative figure strong point P _itrue value.

Described step 4) comprises following concrete steps:

4.1) data slot for feature extraction is divided near X-axis and the close large class of Y-axis two;

4.2) according to the data slot for feature extraction, near different coordinate axis, select different characteristic straight line equations, for the data slot near X-axis, adopt the matching mode of y=kx+b to determine its characteristic of correspondence parameter (k, b), data slot near Y-axis adopts the matching mode of x=ky+b to determine its characteristic of correspondence parameter (k, b);

4.3) calculate each for the square error of the data slot of feature extraction and its characteristic straight line equation, if square mean error amount is more than or equal to square error threshold value, illustrate that this data slot can not be described by a straight-line equation, need two or more straight-line equations to describe this data slot, find this data slot middle distance characteristic straight line equation point farthest, with this this data slot of naming a person for a particular job, be divided into two, then two new data slots adopted respectively to step 4.2) method carry out fitting a straight line until square mean error amount is less than square error threshold value.

Described step 5) comprises following concrete steps:

5.1) select linear feature to meet the data slot of the smooth constraint in wheeled region, delete discontented data slot that can the smooth constraint of running region, the smooth constraint in described wheeled region refers to that Y-axis angle in straight line and local 3 dimension rectangular coordinate systems is more than or equal to the threshold value T of setting _y;

5.2) through step 5.1) after, merge the data slot that meets wheeled region continuity constraint, be regarded as a horizontal zone, as the candidate target of describing wheeled region; Described wheeled region continuity constraint refers to the height difference of adjacent data fragment and the corresponding threshold value T that horizontal range difference is less than respectively setting _z, T _d;

5.3) through step 5.2) after, select in candidate target, to meet body width constraint and allow candidate target that vehicle does transverse movement as far as possible less for describing wheeled region.

Beneficial effect of the present invention is embodied in:

The present invention utilizes Laser Radar Scanning actual traffic scene, obtains the smooth of current scene and fluctuating situation; Then the scan-data by height change situation, laser radar being returned is divided into several fragments, use subsequently the Algorithm of fitting a straight line of iteration from fragment, to extract suitable linear feature, then from meet the fragment of constraint condition, find flat site, more then according to overall width, require to determine wheeled region from flat site; If multi-thread laser, can be from the close-by examples to those far off, take the wheeled region of last line data is basis, from the flat site of this line data reflection, determine corresponding wheeled region, the present invention can effectively complete the detection in wheeled region in actual traffic scene, describe out accurately the safe passing region of automobile, and the present invention have good real-time and ease for use.

The present invention utilizes laser radar as Data Source, by steps such as coordinate system conversion, Data Segmentation, data filtering, feature extraction and definite wheeled regions, complete the function that wheeled region is detected, because the principle of work of laser radar is different from camera etc., the present invention can be good at overcoming the impact of the environmental factors such as weather, illumination, so can realize the function that wheeled region robust detects.

Accompanying drawing explanation

Fig. 1 is abstract road model;

Fig. 2 is that laser radar is demarcated and local coordinate system;

Fig. 3 is that the general frame is detected in laser radar wheeled region;

Fig. 4 is that laser radar data is cut apart process flow diagram;

Fig. 5 is laser radar single line Data Segmentation result;

Fig. 6 is that laser radar data true value is estimated process flow diagram;

Fig. 7 is laser radar data fragment feature extraction process flow diagram;

Fig. 8 is laser radar data fragment fitting a straight line mode zoning;

Fig. 9 is iteration fitting a straight line schematic diagram;

Figure 10 is iteration fitting a straight line result;

Figure 11 is the result that actual traffic scene 4 line laser radar wheeled regions are detected;

Figure 12 is actual traffic scene 64 line laser radar wheeled area detection result.

Embodiment

Below in conjunction with accompanying drawing, the present invention is elaborated.

For traffic scene around perception that can be real-time, stable, the present invention has provided a kind of actual traffic scene wheeled method for detecting area based on laser radar data, specifically comprise laser radar demarcation, Data Segmentation, the elimination of assorted point and data filtering, feature extraction and five, definite wheeled region part, as shown in Figure 3.The method is specifically carried out according to the following steps:

Step 1.1: the function detecting in order to realize laser radar wheeled region, laser radar need to be installed on to headstock dead ahead, oblique lower scanning, guarantees that itself and road surface are crossing.

Now define local space rectangular coordinate system as shown in Figure 2, measure laser radar apart from local space rectangular coordinate system surface level (X-Y, wherein Y-direction is the direction of headstock dead ahead indication) setting height(from bottom) H and oblique lower scanning plane with respect to the angle α of vertical direction, determine that laser radar installation site is in the position of local space rectangular coordinate system.

Step 1.2: because laser radar is the active sensor of scan-type, its data are to take laser radar as limit, distance and the angle of the polar coordinate system that dead ahead or other specific directions are pole axis, and calculating below all realizes based on rectangular coordinate system, so need to set up Laser Radar Scanning data polar coordinate system to the transformational relation of local space rectangular coordinate system coordinate, obtain rectangular coordinate value corresponding to laser spots.

Each laser data point comprises emission angle β _iwith scanning distance d _itwo parameters, i.e. P _i=(d _i, β _i), local space rectangular coordinate system needs x _i, y _i, z _i3 coordinate figures, i.e. P _i=(x _i, y _i, z _i).Position according to the laser radar installation site shown in Fig. 2 in local space rectangular coordinate system, can complete polar coordinates to the conversion of rectangular coordinate by trigonometric function, and conversion equation is as follows:

x _i=sinα·cosβ _i·d _i

y _i=sinβ _i·d _i

z _i=H-cosα·cosβ _i·d _i

Above formula will only have the polar coordinates value of two numerical value to be transformed into the rectangular coordinate system that comprises 3 numerical value, the height value z comprising data point in local space rectangular coordinate system _i, this is quite important for processing below.

Step 2.1: according to the installation site of laser radar, as shown in Figure 2, its plane of scanning motion is a plane under oblique in local space rectangular coordinate system, so its height value is basic identical when same object is arrived in adjacent two data spot scan, can there is not too large deviation, otherwise scan height value on different objects and have larger deviation, so incite somebody to action all laser radar datas according to the difference of consecutive number strong point height value, be divided into some fragment S _k.The flow process that laser radar data point is cut apart as shown in Figure 4, is carried out according to the following steps:

1) all data points are arranged according to the order of X value ascending order, guarantee that X value from left to right becomes large successively;

2) create a new fragment S ₁, wherein only comprise first data point P ₁;

3) read successively the data point P in storage sequence _i;

4) calculate P _iwith previous data point P _i-1difference in height, i.e. Δ z=|z _i-z _i-1|

5) if difference in height is less than threshold value T _z, i.e. Δ z<T _z, by P _ibe deposited in order current fragment S _kin, then turn back to 3);

6) if difference in height is not less than threshold value T _z, i.e. Δ z>=T _z, create a new fragment S _k+1, by P _ibe deposited in order S _k+1in, then turn back to 3);

By cycle calculations above, can obtain a fragment sequence, referring to Fig. 5, the empty circles in Fig. 5 represents the end points of data slot, wherein each fragment has following expression-form:

S _k={P _i},(i=N _sk,…,N _ek)

s.t.|z _s-z _s-1|≥T _z,|z _e-z _e+1|≥T _z

Wherein, z _sthe height that represents the initial laser radar data point of fragment, z _ethe height that represents fragment end laser radar data point.

Step 3.1: the operating characteristic of laser radar causes having the existence of some noise datas, in order to reflect more really actual scene, need to dispose these noise datas.The number of data points that the laser radar data fragment obtaining above by judgement in the present invention comprises realizes this function, only comprises the fragment of a laser radar data point, is considered to reflect the assorted point of scene actual conditions, is eliminated.

Step 3.2: measuring noise and system noise is ubiquitous two large noise datas in DATA REASONING, need to carry out filtering processing to the measurement data obtaining in order to access the true value of data, eliminates the interference of noise.The true value that the present invention selects classical mean filter method to complete measurement data in each data slot estimates, as shown in Figure 6, and for each laser radar data fragment, according to current data P _iselect data point c (i) in its regulation length of window N to estimate the true value of this point, as data P _iwhen left side or right side data amount check are less than regulation length of window, adjust the regulation length of window N of wave filter.The window of selective filter is as follows herein:

${\hat{x}}_i=\frac{1}{N_{\mathrm{ek}}-N_{\mathrm{sk}}+1}\underset{j=N_{\mathrm{ek}}}{\overset{N_{\mathrm{sk}}}{\mathrm{\&Sigma;}}}\left(x_j\right)$

${\hat{y}}_i=\frac{1}{N_{\mathrm{ek}}-N_{\mathrm{sk}}+1}\underset{j=N_{\mathrm{ek}}}{\overset{N_{\mathrm{sk}}}{\mathrm{\&Sigma;}}}\left(y_j\right)$

${\hat{z}}_i=\frac{1}{N_{\mathrm{ek}}-N_{\mathrm{sk}}+1}\underset{j=N_{\mathrm{ek}}}{\overset{N_{\mathrm{sk}}}{\mathrm{\&Sigma;}}}\left(z_j\right)$

Wherein, N _skthe index that participates in first data point of filtering calculating, N _ekto participate in the index that last data point is calculated in filtering, by above formula, can be so that the coordinate figure of data point approaches true value more.

Step 4.1: because the Main Means of feature extraction in the present invention is fitting a straight line, and the fitting a straight line equation of diverse location is different, conventionally near the data of X-axis, can using X-axis as independent variable, near the meeting of Y-axis, using Y-axis as independent variable, like this matching straight line out feature of data of description point more accurately.So all data slots are divided near X-axis and the close large class of Y-axis two as shown in Figure 8.

Step 4.2: the division according to previous step to data slot, the fitting a straight line mode of employing least mean-square error completes the feature extraction of data slot.Realization flow as shown in Figure 7, according to data slot, near different coordinate axis, select different fitting functions, for the data slot near X-axis, adopt the matching mode of y=kx+b to determine its characteristic of correspondence parameter (k, b), data slot near Y-axis adopts the matching mode of x=ky+b to determine its characteristic of correspondence parameter (k, b).

Step 4.3: although adopted the matching mode of least mean-square error, the method for least mean-square error can only guarantee there is minimum error for current straight-line equation, can not illustrate the just feature of data of description fragment accurately of straight line.Therefore when square error is larger, illustrate that straight line can not describe the feature of this data slot fully, need to introduce again one or more straight line and carry out the more careful feature of portraying data slot.The implementation procedure of this function is as follows:

1) whether the square error that judges fitting a straight line is less than threshold value, if be less than threshold value, carries out the extraction of lower one piece of data fragment feature;

2), if be more than or equal to threshold value, as shown in Figure 9, find the some P of the air line distance maximum that this data slot middle distance matching obtains _m;

3) from P _mplace is divided into two data slot to obtain fragment

and fragment ${\underset{\&OverBar;}{S}}_{k+1}=\{P_m,\&CenterDot;\&CenterDot;\&CenterDot;,P_n\};$

4) at fragment S _kthe fragment that rear insertion is new

use fragment

replace fragment S _k;

5) from fragment

start to carry out feature extraction, return to 1);

By processing above, can guarantee that all data slots all find its corresponding straight-line equation to describe its feature, as shown in figure 10, these features just more objectively reflect the situation of actual traffic scene, for next step wheeled region, detect stable Data support is provided.Corresponding each data slot has following descriptor format:

S _i={{f,k _i,b _i},{P _m,…,P _n}}

Wherein f is used for mark matching mode, k _i, b _ifor data characteristics, P _m..., P _nthe data point comprising for this fragment.

Step 5.1: the wheeled region in objective environment must be a smooth region, can not have rugged situation, and this feature is called as flatness principle in the present invention.Actual traffic model of place as shown in Figure 1, so select feature to meet the data slot of flatness principles and requirements as the candidate target in wheeled region.System of selection is as follows:

1) order reading out data fragment S _i;

2) angle theta of calculated characteristics straight line and Y-axis _i;

3) if this straight line and Y-axis angle are less, | θ _i-90|<T _y, think that this segment data spot scan, to non-horizontal surface, is not wheeled region, by its deletion;

4) if this straight line and Y-axis angle are larger, | θ _i-90|>=T _y, think that this segment data spot scan, to surface level, is wheeled region, by its reservation;

By processing above, the region that does not meet road surface in traffic scene model can be deleted, only retain the plane that meets flatness principle.

Step 5.2: wheeled region is a continuous flat site, there is not larger interruption and difference in height in centre, and this feature is called as continuity principle in the present invention.Fragment obtained above all represents a plane, but do not show whether adjacent segment meets continuity principle, in fact two adjacent and there is no a fragment of obvious interval and difference in height, description be same plane domain, so merged better to describe this plane domain.Performing step is as follows:

1), since second fragment, order reads fragment S _i;

2) calculate fragment S _iwith last fragment S _i-1the poor Δ d of the difference in height Δ z of 2 abutting end points and horizontal range;

3) if difference in height and horizontal range are poor, there is one to be not less than given threshold value, i.e. Δ d>=T _dor Δ z>=T _z, what think these two fragments descriptions is not same plane, returns to 1);

4) if poor given threshold value, i.e. the Δ d<T of being all less than of difference in height and horizontal range _dand Δ z<T _z, what think these two fragments descriptions is same plane, by fragment S _iwith S _i-1merge, then return to 1);

By processing above, all data slots are merged according to the plane of its description, the data slot after merging is like this exactly the monoblock plane in road model.

Step 5.3: all data slots are all described a continuous plane now, but not entirely wheeled region, because it is too far away that some region is less than width, some region distance vehicle of vehicle body, therefrom select most suitable data slot as wheeled region.The present invention has adopted one to allow vehicle do less the hypothesis of transverse movement as far as possible, select vehicle can by and the maximum region that is positioned at headstock the place ahead as wheeled region, implementation is as follows:

1) reading out data fragment successively, the horizontal range d of computational data fragment starting point and terminating point _s;

2) if horizontal range is less than overall width, i.e. d _s<W _v, think that this region can not meet vehicle by requiring, be traveling-prohibited area, return to 1).

3) if horizontal range is more than or equal to overall width, i.e. d _s>=W _v, think and meet body width constraint condition, then calculate the absolute value theta of this data slot central point and the straight line of true origin formation and the angle of Y-axis _s;

4) if current θ _sbe less than present feasible and sail the angle theta in region _c, i.e. θ _s< θ _c, using current data fragment as new wheeled region, and make θ _c=θ _s, return to 1).

Through processing above, from all fragments, select the plane domain that is positioned at headstock dead ahead that meets wide constraint as wheeled region, if there is multi-thread laser data, take the wheeled region of last line is basis, successively to external expansion, determine the wheeled region in its data slot, referring to Figure 11 and Figure 12, wherein the more shallow gray areas of color is the wheeled region detecting.

Claims (7)

1. the wheeled method for detecting area based on laser radar, is characterized in that: this wheeled method for detecting area comprises the following steps:

First Calibration of Laser radar, then carries out coordinate conversion by the laser radar data collecting; Then cutting apart laser radar data is some fragments; Fragment is done to assorted point is eliminated and the feature extraction of laser radar data is carried out in filtering after processing; Then according to wheeled range constraint and body width constraint condition, from the feature of extracting, determine wheeled region.

2. a kind of wheeled method for detecting area based on laser radar according to claim 1, is characterized in that: described wheeled method for detecting area specifically comprises the following steps:

1) first determine the height that laser radar is installed, calculate the downward-sloping angle of laser radar simultaneously, then the laser radar data collecting is transformed into local 3 dimension rectangular coordinate systems from 2 dimension polar coordinate systems;

2) laser radar data is cut apart: adopt the height difference of adjacent point-to-point transmission as the standard of Data Segmentation, laser radar data is divided into several fragments;

3) assorted point is eliminated and a data filtering: through step 2) after, the fragment that will only comprise single laser radar data point is removed, and then remaining laser radar data fragment is carried out to filtering processing removal noise and obtains the data slot for feature extraction;

4) feature extraction: after step 3), adopt the Algorithm of fitting a straight line of iteration to extract corresponding linear feature from the data slot for feature extraction, make laser radar data become the plane in corresponding actual scene;

5) determine wheeled region: the linear feature extracting from step 4), select the linear feature that meets the smooth constraint in wheeled region, wheeled region continuity constraint and body width constraint as wheeled region.

3. a kind of wheeled method for detecting area based on laser radar according to claim 2, is characterized in that: described step 1) comprises following concrete steps:

1.1) take laser radar installation site as the local 3 dimension rectangular coordinate systems of reference point definition, then measure laser radar apart from the setting height(from bottom) H of surface level, and calculate laser radar and be tilted to lower scanning plane with respect to the angle α of vertical direction;

1.2) set up Laser Radar Scanning polar coordinates to the transformational relation of coordinate in local 3 dimension rectangular coordinate systems, according to transformational relation, obtain laser radar data point P _i3 corresponding dimension rectangular coordinate values, each Laser Radar Scanning polar data point comprises emission angle β _iwith scanning distance d _itwo parameters, P _i=(d _i, β _i) coordinate figures of corresponding local 3 dimension rectangular coordinate systems are P _i=(x _i, y _i, z _i), Equation is as follows:

x _i=sinα·cosβ _i·d _i

y _i=sinβ _i·d _i

z _i=H-cosα·cosβ _i·d _i。

4. a kind of wheeled method for detecting area based on laser radar according to claim 2, is characterized in that: described step 2) comprise following concrete steps:

According to the difference of consecutive number strong point height value, whole laser radar datas is divided into some fragment S _k:

S _k={P _i},(i=N _sk,…,N _ek)

s.t.|z _s-z _s-1|≥T _z,|z _e-z _e+1|≥T _z

Wherein, P _irepresent laser radar data point, z _sthe height that represents the initial laser radar data point of fragment, z _ethe height that represents fragment end laser radar data point, T _zrepresent difference in height threshold value.

5. a kind of wheeled method for detecting area based on laser radar according to claim 2, is characterized in that: described filtering is processed and comprised following concrete steps:

For data slot, according to current data point P _iselecting it to set data point in window carries out mean filter and carrys out estimative figure strong point P _itrue value.

6. a kind of wheeled method for detecting area based on laser radar according to claim 2, is characterized in that: described step 4) comprises following concrete steps:

4.1) data slot for feature extraction is divided near X-axis and the close large class of Y-axis two;

4.2) according to the data slot for feature extraction, near different coordinate axis, select different characteristic straight line equations, for the data slot near X-axis, adopt the matching mode of y=kx+b to determine characteristic of correspondence parameter (k, b), data slot near Y-axis adopts the matching mode of x=ky+b to determine characteristic of correspondence parameter (k, b);

4.3) calculate each for the square error of the data slot of feature extraction and its characteristic straight line equation, if square mean error amount is more than or equal to square error threshold value, find data slot middle distance characteristic straight line equation point farthest, with this data slot of naming a person for a particular job, be divided into two, then two new data slots adopted respectively to step 4.2) method carry out fitting a straight line until square mean error amount is less than square error threshold value.

7. a kind of wheeled method for detecting area based on laser radar according to claim 2, is characterized in that: described step 5) comprises following concrete steps:

5.1) select linear feature to meet the data slot of the smooth constraint in wheeled region, delete discontented data slot that can the smooth constraint of running region, the smooth constraint in described wheeled region refers to that Y-axis angle in straight line and local 3 dimension rectangular coordinate systems is more than or equal to the threshold value T of setting _y;

5.2) through step 5.1) after, the data slot that meets wheeled region continuity constraint merged, as the candidate target of describing wheeled region; Described wheeled region continuity constraint refers to the height difference of adjacent data fragment and the corresponding threshold value T that horizontal range difference is less than respectively setting _z, T _d;

5.3) through step 5.2) after, select in candidate target, to meet body width constraint and allow candidate target that vehicle does transverse movement as far as possible less for describing wheeled region.

Images