CN107632308A - A kind of vehicle front barrier profile testing method based on recurrence superposition algorithm - Google Patents

Abstract

The invention discloses a kind of vehicle front barrier profile testing method based on recurrence superposition algorithm, this method includes：The geometrical relationship on radar and ground calculates barrier profile elevations h；History scans the coordinate matching with Current Scan；Consider that there is radar light beam the characteristics of normal distribution to introduce probability density function；Realize the quasi-continuous estimation of barrier profile；Accurate barrier profile elevations h is tried to achieve by way of history scanning is superimposed with the continuous recurrence of Current Scan data；By returning computed altitude deviation and pitch angle deviation；The fusion of new and old scanning.The measuring method that the present invention uses is simple and feasible, the general outline height of road ahead barrier can accurately and efficiently be detected, and overlay region constantly carries out recurrence superposition caused by new and old scanning, can remove influences caused by some interference, greatly strengthen information density.Accurate vehicle front barrier profile elevations h can be obtained.

Description

A Contour Detection Method of Obstacles in Front of Vehicle Based on Recursive Superposition Algorithm

technical field

The invention belongs to the field of automobile intelligent driving and radar technology, and relates to a method for identifying targets by radar, in particular to a method for detecting the contours of obstacles in front of a vehicle based on a recursive superposition algorithm, which is used to solve the problem that existing radars cannot quickly and accurately identify road obstacles Contour problem.

Background technique

LiDAR is a commonly used ranging sensor. It is widely used in various fields due to its advantages of high resolution and low interference from environmental factors. Lidar is divided into three types: single-line lidar, multi-line lidar and area array radar. Single-line lidar generates one scan line per scan. It has fast ranging speed, less data volume, small size, light weight, and is suitable for fast processing. Because of its high cost performance, single-line lidar is currently the most widely used.

LiDAR is a class of sensors that use light waves for distance measurement. LiDAR often uses the pulse propagation time method to measure the distance of objects. In operation, it emits a laser pulse in the form of a focused beam and measures the travel time between sending and receiving the echo, and since the laser pulse travels at the speed of light, the distance can be determined. The distance from the laser generator to a single reflection point of the target is: d=ct/2, where c is the speed of light; t is the time difference from when the ground laser radar emits the laser beam to when the echo signal is received; d is the reflection from the laser transmitter to the target point distance.

The laser beam is well focused by the optics so that not only the distance but also the exact lateral and vertical position of the target relative to the sensor can be determined. This measuring principle enables distance measurements in the millimeter range and is therefore ideal for detecting small changes in road surface irregularities and thus constructing the entire ground level profile.

When the lidar beam is projected on the surface of dust particles and raindrops, it will cause the calculation of the distance value to be wrong. Moreover, because the vehicle-mounted radar has some external interference during the motion state, the radar pulse echo is more uncontrollable, so an intelligent analysis algorithm must be used to achieve practicality.

At present, there have been many studies on the measurement of the obstacle profile in front of the vehicle:

Related Document 1: Application No. 201310063898.2, Chao Chen and Zhongchao Shi of Ricoh Corporation provided a method and system for estimating the height and shape of the road surface in a road scene by using a binocular camera. The method includes: obtaining a disparity map of the road scene; based on The disparity map detects a road surface region of interest; determines a plurality of road surface interest points based on the road surface region of interest; and estimates a road surface height shape based on the plurality of road surface interest points. Because the road scene is usually very complex, including pedestrians, vehicles, obstacles, etc., the calculation of the algorithm is very large. Generally, the scanned data needs to be processed offline, and it is difficult to detect online in real time.

Related literature 2: F.Moosmann et al. proposed an image recognition algorithm with obstacle recognition ability. This algorithm uses three-dimensional lidar to identify and segment the ground and obstacles. Since the three-dimensional lidar has not yet reached the mass production stage, it is expensive, which limits the scope of application of this method.

Related Document 3: Application No. 201610804686.9, Song Wei, Zhou Xiaolong, and Wu Bin disclosed an obstacle recognition method using convolutional neural network CNN, using deep learning algorithms to identify obstacles based on the bionic eye system, and provided The method of configuring the interface during the recognition process strengthens the communication process between the recognition process and the bionic eye system. This method has a high recognition rate, but the effective execution of the algorithm must rely on the neural network training of a large number of image model libraries. Once the obstacle information is missing in the model library, the recognition results will be greatly affected.

Contents of the invention

The object of the present invention is to provide a method for detecting the contour of an obstacle in front of a vehicle based on a recursive superposition algorithm. The laser pulse echo signal is analyzed multiple times through recursive matching of multiple continuous scans, thereby increasing the information density of the signal and accurately obtaining the vehicle front. Obstacle profile height, can be used in road traffic without restriction.

The purpose of the present invention is achieved by the following scheme:

A method for detecting the contours of obstacles in front of the vehicle based on the recursive superposition algorithm, and the laser radar installation scheme:

The lidar is installed at the height of the headlights in the front of the vehicle, and can measure the road from the position where the bumper ends, and the beam of the lidar is projected obliquely on the lane. The disadvantages caused by the flatter scanning angle are relatively slight compared with the measured length of a few meters in front of the vehicle that is lost when installed in the vehicle. Therefore, to realize the preview function, it is more appropriate to install the lidar at the height of the headlight;

The inventive method comprises the following steps:

Step 1. Calculate the obstacle contour height through the geometric relationship between the laser radar and the ground, establish a polar coordinate system, and convert the original data of the obstacle contour height collected by the laser radar into polar coordinates through trigonometric coordinate transformation;

Step 2. Match the coordinates of the past scan data and the current scan data: convert the obstacle contour height equation from the polar coordinate system to the Cartesian coordinate system through trigonometric coordinate transformation, so as to incorporate the two scans into the same coordinate system;

Step 3. Introduce the probability density function considering the normal distribution of the radar beam: Introduce the Gaussian normal distribution to obtain the normal distribution probability density function of each measurement point, and use the probability density function to characterize the obstacles obtained by the radar measurement point spot The real distribution of object silhouette height;

Step 4. Realize the quasi-continuous estimation of the obstacle contour: the probability density distribution of each scanning point is completed by the step 3, and the quasi-continuous estimation of the contour height can be carried out through the probability density curve; a coordinate system is established, and the abscissa represents the laser radar The distance from the measurement point scanned by the beam to the radar, the ordinate represents the height value of the obstacle contour, and a shift register with equidistant sampling points is introduced on the abscissa, and the height value of each scan is input through a quantized abscissa, through The distance from the lidar to the scanning point calculates the obstacle contour height value and probability density in the rasterized register; the sum of n probability density functions in one scan is used as a unified standard to obtain the obstacles of each equidistant point in each scan quasi-continuous estimation of object contours;

Step 5. Accurate obstacle contour height is obtained by continuously recursively superimposing the past scan data and the current scan data, and at the same time, a correlation coefficient is introduced to evaluate the correlation degree of the influence of the two scans on the real obstacle contour height;

Step 6. Calculate the height value deviation and pitch angle deviation through linear regression: through the step 5, a new relationship equation between the current scan data and the past scan data is obtained, and the linear regression is used to solve the problem, and multiple sets of obstacles can be obtained Profile height value deviation and pitch angle deviation, the optimal value of height value deviation and pitch angle deviation can be determined by using the least square method, and the height correction value of the new scanning data can be calculated at this time;

Step 7. Fusion of old and new scan data: Use the height correction value superimposed and overlapped in step 6 to fuse the current scan data with the previous scan data to obtain an accurate obstacle contour height value; at the same time update the obstacle contour height value, Carry out the next recursive superposition, and repeat the superposition to obtain an accurate outline of the obstacle in front of the vehicle.

The present invention provides a method for detecting the contours of obstacles in front of a vehicle based on a recursive superposition algorithm. Only one low-cost single-line laser radar is used, and the measurement cost is relatively low. All boundary conditions applied in the vehicle can be applied in road traffic without restriction. The invention can be independently and efficiently executed independently of any external model library and other constraints, and can realize real-time online detection and processing, and overcomes many problems existing in this field at present.

Description of drawings

Fig. 1 is the conception diagram of principle of the present invention

Fig. 2 is a flowchart of the present invention

Figure 3 calculates the obstacle contour height for the geometric relationship between the radar and the ground

Figure 4 is a shift register with equidistant sampling points

Figure 5 shows the quasi-continuous estimation of obstacle contours at each equidistant point in each scan

detailed description

Describe technical scheme of the present invention in detail below in conjunction with accompanying drawing:

A method for detecting the contours of obstacles in front of a vehicle based on a recursive superposition algorithm. The single-line lidar needs to be installed at the height of the headlights at the front of the vehicle, and the road can be measured from the position where the bumper ends. The laser beam will be projected obliquely on the On the lane, the inventive method comprises the following steps, as shown in Figure 2:

Step 1. Calculate the obstacle contour height through the geometric relationship between the lidar and the ground;

The current obstacle height is calculated by using the laser beam of the lidar installed in the front of the vehicle, as shown in Figure 3.

While the lidar is rotating, the laser beam also provides the distance value of the measuring point i fan-shaped within the measuring angle range. The expression of the inclination n ₀ of the pulse beam emitted by the lidar relative to the road is: Among them, n _c represents the pitch angle offset of the lidar in the installation position, n _L represents the relative pitch angle between the car body and the wheel, Represents the angle of the current lidar measurement beam relative to the sensor housing.

The laser radar measurement angle ranges from 0° to 90°. The first measurement point of a scan is at approximately a 45° downward turn from the horizontal to the road. Through trigonometric coordinate transformation, using the installation height of the laser radar and the inclination parameters of the laser beam, the absolute vertical height z ₀ from the laser radar to the ground and the distance x ₀ from the measuring point to the sensor in the x-axis direction can be calculated. The calculation formula is as follows :

x ₀ ＝d ₀ *cos(n ₀ )

z ₀ ＝zd ₀ *sin(n ₀ )

The original vertical distance z between the lidar and the lane is calculated by:

z＝z _cz +z _zd -x _s *sin(n _L )+y _s *sin(w _L )

Then it can be deduced that the calculation formula of the obstacle contour height value _z0 is as follows:

In the formula, Z _cz represents the vertical offset Z _cz of the lidar in the installation position; Z _zd , n _L , w _L represent the relative motion vibration, bump and swing between the car body and the wheel; x _s and y _s respectively Describes the vehicle's longitudinal position and lateral position, the distance between the vehicle's center of gravity and the lidar; d ₀ represents the distance between the lidar and the measurement point.

Step 2. Match the coordinates of the past scan data with the current scan data

During the actual driving process of the vehicle, the vehicle will definitely have longitudinal and lateral motions, and the vehicle body itself will also move relative to the road. The motion parameters in this process can be described as: the longitudinal speed of the vehicle body v _x , the vehicle body Relative motion vibration z _zd between body and wheel, sensor bump n _L , sensor wobble w _L .

Assuming the driving state of the vehicle is known, the two scans can be superimposed. Assuming that the two scan data before and after are known, the equation is converted from the polar coordinate representation to the Cartesian coordinate system representation through the coordinate transformation relationship of the trigonometric function. The specific conversion formula is as follows:

Past scan data:

Now scan the data:

In the formula, the past scan is represented by "past", and the subscript "p" is used to annotate, Represents the distance from a measurement point to the sensor in the x-axis direction in the past scan data, Represents the height value of the obstacle contour obtained in the past scan; the current scan is represented by "now" and annotated with the subscript "n", Represents the distance from a measurement point to the sensor in the x-axis direction in the current scan data, Represents the obstacle contour height value obtained by scanning now.

Step 3. Considering that the radar beam has a normal distribution, introduce a probability density function

In the equation of the above step 2, the measurement point is regarded as a point for research, and each distance value measured by the lidar corresponds to an obstacle contour height value. However, the real situation is that each measurement point is actually distributed in the form of spots, not a point. Within a spot, the height values are distributed with a certain probability. Applying the characteristics of the normal distribution of the laser radar beam to introduce the normal distribution probability density function, the probability density of the measurement point can be approximated by a continuous distribution function, the Gaussian normal distribution function:

In the above equation, x is a continuous random variable, which can be understood as the horizontal distance between the lidar and the measurement point in the model, and σ is the standard deviation (or variance).

The algorithms implemented in the above steps are all based on the infinitesimal propagation of the measurement points. Such assumptions are too idealized and cannot truly describe the height profile of obstacles. If the two scans have the same distance basis, then the regression analysis will achieve an overlay of the two scans. For this reason, a coordinate system as shown in Figure 4 is established, the abscissa represents the distance from the measurement point scanned by the lidar beam to the radar, and the ordinate represents the height of the obstacle contour. Introduce a shift register with equidistant sampling points on the abscissa (which can be understood as an array in the algorithm program), and the height value of each scan is input through a quantized abscissa value, so that the diffusion problem of the measurement point can be solved solve.

The shift register with equidistant sampling points has the following advantages: the planar distribution of the measuring points is basically considered; the matching of multiple scans has a common distance basis.

Figure 4 shows an example of a shift register application for measuring points in a scan. A certain abscissa value in the shift register corresponds to a discrete height value. Since the measurement point is displayed in the form of a light spot in the real situation, within the range of the light spot, the discrete measurement value may have a certain statistical probability Appearance, so the probability density of the occurrence of the measured value within the distribution range of the measurement points can be given.

After introducing the shift register, through the parameter of the distance from the laser radar scanning point to the laser radar installation position, the obstacle contour height value and probability density corresponding to this parameter can be calculated respectively.

In the shift register, every equidistant sampling point with a distance of Δx ₁ is divided on the abscissa, which is equivalent to rasterizing the distance between the lidar scanning point and the radar installation position. It can be clearly concluded from Figure 4 that:

The abscissa of the shift register covers the entire measurement range of the lidar signal. According to the grid width and the maximum scanning distance range, a shift register with m+1 discrete equidistant sampling points can be obtained. For example, if the measurement range is 0-20m and the grid width is 10cm, then the number of sampling points in the shift register is m+1=201. In the register, the height value and the probability density distribution of each measuring point are entered via the abscissa value x. Table 1 shows the results:

Table 1 Registers with equidistant sampling points for saving scan data

Assuming that a set of scans has k measurement points, the probability density of different measurement spots at each sampling point 0...m is:

The probability density function can characterize the accuracy of the obstacle height measured in the light spot. The larger the probability density peak value is, the more concentrated the probability distribution is, and the higher the measurement accuracy is. Through the probability density function, the measurement data can be continuously processed to obtain a denser obstacle contour height curve.

Step 4. Realize quasi-continuous estimation of obstacle contours

Each time a new scan is generated, k height values will be obtained, which are expressed in a discrete form in terms of distance. However, in actual situations, there will be a corresponding height value at each position of the shift register where the probability is not zero. Assuming that the current scan is composed of k measurement points, the current estimated value of the height value can be calculated by m+1 discrete grid points in the shift register, and then calculated according to the product of the probability density matrix and the vector of k height values (The sum of n probability density functions of one scan is used as a unified standard):

The probability density values at the first sampling point are ξ _0,1 , ξ _0,2 ... ξ _0,3 , and the estimated value of the corresponding height value can be calculated by standardization:

Among them, the weighted sum of the scanned data is:

The probability density values of the 2nd, 3rd... sampling points and the quasi-continuous estimation of the corresponding height values can also be obtained through standardization.

The probability density value of the mth sampling point is ξ _m,1 , ξ _m,2 ...ξ _m,k , and the estimated value of the corresponding height value can be calculated by standardization:

Among them, the weighted sum of the scanned data is:

According to the above algorithm, the quasi-continuous estimation of the obstacle contour of each equidistant point in each scan can be obtained.

Step 5. Accurate obstacle contour height is obtained by continuously recursively superimposing the past scan data and the current scan data

Using all scanned data, including current scan and past scan data, through recursive calls of the algorithm can greatly improve the signal quality of obstacles. This goal can be achieved by recursively calling the scan matching algorithm between the past scan and the current scan at each scan. The recursive superposition algorithm can be simply described by the following formula:

Now scan the recursive call:

Recursive calls for past scans:

In the formula, Represents the calculated value of the height value in the current scan data, Represents the sum of the probability density functions of the first sampling point in the current scan data; represents the computed value of the height value from the past scan data, Represents the sum of the probability density functions of the first sampling point in the past scan data; Represents the distance from the measurement point to the sensor in the x-axis direction in the current scan data, Represents the distance from the measurement point to the sensor in the x-axis direction in the past scan data.

Recursively calling the algorithm function f is the fusion process of the old and new scan data in the step seven.

Step 6. Calculate the height value deviation and pitch angle deviation by regression

In the recursive superposition algorithm, the calculation of relevant factors also has to be considered. When the current new scan and the past old scan are overlaid by the regression method, the error of the road contour height value must be taken into account. The height offset or height error in the shift register can be expressed as:

In the formula, Represents the height value of the obstacle contour scanned in the past; Represents the currently scanned obstacle contour height value; Represents the height offset or height error in the shift register.

In order to superimpose the new scan and the old scan through linear regression, it is necessary to determine how much weight to consider the obstacle contour height value error corresponding to each abscissa value in the shift register. In simple terms, errors in height values need to be considered only at locations with high normalized probability densities from current scans and past scans. Therefore, the regression analysis was considered within the minimum intersection of the probability density distributions of the two scans, since only the overlap of the two scans is relevant. In view of this consideration, the correlation coefficient R is introduced, and the correlation R can be calculated from the minimization standard of the generalized probability density function. The specific calculation formula is as follows:

In the formula, Represent the probability density function sum of the current scan data and the past scan data, respectively.

In the recursive superposition algorithm, another parameter is added: the correlation coefficient R, which also shows that the spot plane of the laser measurement point is also considered when determining the deviation of the obstacle contour height value and pitch angle deviation between the current scan and the past scan distribution, as well as factors such as the probability density distribution of height values corresponding to the corresponding measurement points. Between the current scan data and the past scan data, the following new relationship can be derived:

In the formula, R represents the correlation coefficient; Represents the distance between the measurement point and the X-axis direction of the lidar in the shift register with equidistant sampling points; Δn and Δz represent the pitch angle deviation and height value deviation between the old and new scan data, respectively;

This equation is an overdetermined system of equations, similar to Ax=b. The above equation can be solved by linear regression. Construct the generalized inverse matrix A+ of matrix A:

According to the above formula, the height value deviation and pitch angle deviation Δn of multiple groups of obstacle contours can be obtained, and the optimal value of the height value deviation Δz and the optimal value of the pitch angle deviation Δn can be determined by using the least square method. At this time, the new scan data can be calculated Altitude correction value:

z _0,cy,xz =z _0,cy,n +Δn*x _0,cy +Δz

In the formula, the new and old scans are superimposed, and z _{0, cy, xz} represent the height correction values of the new scan data.

Step 7. Fusion of old and new scan data

Through the implementation of the above steps, at this time, the current scanned data and the past scanned data can be fused by using the correction obtained from the previous recursive superimposition and superposition. After adding the new scan's data to the saved data from past scans, the generalized probability densities of all previous scans are increased by the new scan's probability density:

∑ξ _0,sum = ∑ξ _0,n + ∑ξ _0,p

In the formula, ∑ξ _0,sum represents the general probability density of all scans included in the first sampling point; ∑ξ _0,p represents the sum of generalized probability densities of past scanning data; ∑ξ _0,n represents the current scanning data The sum of the generalized probability densities for .

Taking into account the new probability density, the updated mean height value of the road profile can be calculated:

This average height value z _{0, cy, sum} is the final accurate obstacle contour height value, use z _{0, cy, sum} to replace the obstacle contour height value of this scan, and then perform the current scan and a new round of scan The recursive superposition, calculate the height value of the next scan, repeat the recursive superposition, you can get the accurate outline of the obstacle in front of the vehicle.

The present invention has the following advantages:

The new recursive superimposed obstacle contour information processing algorithm considers all boundary conditions applied in real vehicles, and can be applied in road traffic without restriction.

The information for a single scan is too incomplete and inaccurate. This algorithm makes full use of the fact that some ranges overlap in continuous scans, and to a certain extent "overlaps" the scans and increases the information density.

Through the introduction of probability density function and quasi-continuous estimation, the recursive superposition algorithm can make the radar scanning data "infinitely approach" the true value.

In order to enable those skilled in the art to better understand the present invention, the present invention will be further exemplified below with the algorithm simulation flow of MATLAB.

Based on the powerful array computing capability of MATLAB itself, in order to test the reliability of the algorithm, MATLAB can be used to build the algorithm studied in this topic for reliability analysis. The following is the algorithm for building and forming, assuming that two sets of data from the two scans of the lidar are obtained respectively. Firstly, transform the data to coordinate axes and realize the function of the equidistant shift register, then use the interpl function to realize the matching of the two scans; next, use the polyfit and polyval functions to fit the data obtained from the new scan to obtain a new regression Afterwards, the new scanning data is used to realize the standardization of the probability density function in the algorithm; then, the correlation coefficient R and the matrix A containing the correlation coefficient are obtained to obtain x, and the height value deviation and pitch angle deviation are calculated. Finally, the data fusion is used to obtain the corrected new height value, and the realization of the algorithm is completed.

Claims (8)

1. A method for detecting the contours of obstacles in front of a vehicle based on a recursive superposition algorithm, characterized in that the single-line laser radar is installed at the height of the headlights at the front of the vehicle, and the road can be measured from the position where the bumper ends, and the laser beam is inclined Projected on the lane, the method of the present invention comprises the following steps: Step 1. Calculate the obstacle contour height through the geometric relationship between the laser radar and the ground, establish a polar coordinate system, and convert the original data of the obstacle contour height collected by the laser radar into polar coordinates through trigonometric coordinate transformation; Step 2. Match the coordinates of the past scan data and the current scan data: convert the obstacle contour height equation from the polar coordinate system to the Cartesian coordinate system through trigonometric coordinate transformation, so as to incorporate the two scans into the same coordinate system; Step 3. Introduce the probability density function considering the normal distribution of the radar beam: Introduce the Gaussian normal distribution to obtain the normal distribution probability density function of each measurement point, and use the probability density function to characterize the obstacles obtained by the radar measurement point spot The real distribution of object silhouette height; Step 4. Realize the quasi-continuous estimation of the obstacle contour: the probability density distribution of each scanning point is completed by the step 3, and the quasi-continuous estimation of the contour height can be carried out through the probability density curve; a coordinate system is established, and the abscissa represents the laser radar The distance from the measurement point scanned by the beam to the radar, the ordinate represents the height value of the obstacle contour, and a shift register with equidistant sampling points is introduced on the abscissa, and the height value of each scan is input through a quantized abscissa, through The distance from the lidar to the scanning point calculates the obstacle contour height value and probability density in the rasterized register; the sum of the k probability density functions of one scan is used as a unified standard to obtain the obstacles of each equidistant point in each scan quasi-continuous estimation of object contours; Step 5. Accurate obstacle contour height is obtained by continuously recursively superimposing the past scan data and the current scan data, and at the same time, a correlation coefficient is introduced to evaluate the correlation degree of the influence of the two scans on the real obstacle contour height; Step 6. Calculate the height value deviation and pitch angle deviation through linear regression: through the step 5, a new relationship equation between the current scan data and the past scan data is obtained, and the linear regression is used to solve the problem, and multiple sets of obstacles can be obtained Profile height value deviation and pitch angle deviation, the optimal value of height value deviation and pitch angle deviation can be determined by using the least square method, and the height correction value of the new scanning data can be calculated at this time; Step 7. Fusion of old and new scan data: Use the height correction value superimposed and overlapped in step 6 to fuse the current scan data with the previous scan data to obtain an accurate obstacle contour height value; at the same time update the obstacle contour height value, Carry out the next recursive superposition, and repeat the superposition to obtain an accurate outline of the obstacle in front of the vehicle. 2. a kind of vehicle front obstacle contour detection method based on recursive superposition algorithm as claimed in claim 1, is characterized in that, described step one comprises the following process by calculating the obstacle contour height by the geometric relation of lidar and ground: The expression of the inclination n ₀ of the pulse beam emitted by the lidar relative to the road is: Among them, n _c represents the pitch angle offset of the lidar in the installation position, n _L represents the relative pitch angle between the car body and the wheel, Represents the angle of the current lidar measurement beam relative to the sensor housing; Through the coordinate transformation of trigonometric functions, using the installation height of the laser radar and the inclination parameters of the laser beam, the obstacle contour height value z ₀ and the distance x ₀ from the measurement point to the sensor in the x-axis direction can be calculated. The calculation formula is as follows: x ₀ ＝d ₀ *cos(n ₀ ) z ₀ ＝zd ₀ *sin(n ₀ ) The original vertical distance z between the lidar and the lane is calculated by: z＝z _cz +z _zd -x _s *sin(n _L )+y _s *sin(w _L ) Then it can be deduced that the calculation formula of the obstacle contour height value _z0 is as follows: In the formula, Z _cz represents the vertical offset Z _cz of the lidar in the installation position; Z _zd , n _L , w _L represent the relative motion vibration, bump and swing between the car body and the wheel; x _s and y _s respectively Describes the vehicle's longitudinal position and lateral position, the distance between the vehicle's center of gravity and the lidar; d ₀ represents the distance between the lidar and the measurement point. 3. a kind of vehicle front obstacle contour detection method based on recursive superposition algorithm as claimed in claim 1, is characterized in that, the process of the coordinate matching of scanning data in the past of described step 2 and present scanning data is: It is assumed that the driving state of the vehicle is known, so that two scans can be overlapped; Assuming that the previous and last two scan data are known, the height of the obstacle contour is converted from polar coordinates to Cartesian coordinates through the coordinate transformation relationship of trigonometric functions. The specific conversion formula is as follows: Past scan data: Now scan the data: In the formula, Represents the distance from a measurement point to the sensor in the x-axis direction in the past scan data, Represents the obstacle contour height value obtained in the past scan; Represents the distance from a measurement point to the sensor in the x-axis direction in the current scan data, Represents the obstacle contour height value obtained by scanning now. 4. a kind of vehicle front obstacle contour detection method based on recursive superposition algorithm as claimed in claim 1, is characterized in that, described step 3 considers that radar beam has the characteristic of normal distribution and introduces the process of probability density function as: Applying the characteristics of the normal distribution of the laser radar beam to introduce the normal distribution probability density function, the probability density of the measurement point can be approximated by a continuous distribution function, the Gaussian normal distribution function: <mrow><mi>&amp;xi;</mi><mrow><mo>(</mo><mi>x</mi><mo>)</mo></mrow><mo>=</mo><mfrac><mn>1</mn><mrow><mi>&amp;sigma;</mi><msqrt><mrow><mn>2</mn><mi>&amp;pi;</mi></mrow></msqrt></mrow></mfrac><msup><mi>exp</mi><mrow><mo>(</mo><mo>-</mo><mfrac><mn>1</mn><mn>2</mn></mfrac><msup><mrow><mo>(</mo><mfrac><mrow><mi>x</mi><mo>-</mo><msub><mi>x</mi><mrow><mi>r</mi><mi>e</mi><mi>f</mi></mrow></msub></mrow><mi>&amp;sigma;</mi></mfrac><mo>)</mo></mrow><mn>2</mn></msup><mo>)</mo></mrow></msup></mrow> where x is a continuous random variable, representing the horizontal distance between the lidar and the measurement point; σ is the standard deviation; Establish a coordinate system, the abscissa represents the distance from the measurement point scanned by the lidar beam to the radar, and the ordinate represents the height of the obstacle contour; A shift register with equidistant sampling points is introduced on the abscissa, and the height value of each scan is input through a quantized abscissa value: through the parameter of the distance from the laser radar scanning point to the laser radar installation position, it can be Obtain the obstacle contour height value and probability density corresponding to this parameter respectively; In the shift register, the abscissa is divided every equidistant sampling point with a distance of Δx ₁ , which is equivalent to rasterizing the distance between the lidar scanning point and the radar installation position: <mrow><msub><mover><mi>x</mi><mo>&amp;RightArrow;</mo></mover><mrow><mn>0</mn><mo>,</mo><mi>c</mi><mi>y</mi></mrow></msub><mo>=</mo><mrow><mo>(</mo><msub><mi>x</mi><mn>0</mn></msub><mo>,</mo><msub><mi>x</mi><mn>0</mn></msub><mo>+</mo><mn>1</mn><mo>*</mo><mi>&amp;Delta;</mi><mi>x</mi><mo>,</mo><mn>...</mn><mo>,</mo><msub><mi>x</mi><mn>0</mn></msub><mo>+</mo><mi>m</mi><mo>*</mo><msub><mi>&amp;Delta;x</mi><mn>1</mn></msub><mo>)</mo></mrow><mo>=</mo><mrow><mo>(</mo><msub><mi>x</mi><mrow><mn>0</mn><mo>,</mo><mi>c</mi><mi>y</mi><mo>,</mo><mi>j</mi></mrow></msub><mo>,</mo><mn>...</mn><mo>,</mo><msub><mi>x</mi><mrow><mn>0</mn><mo>,</mo><mi>c</mi><mi>y</mi><mo>,</mo><mi>m</mi></mrow></msub><mo>)</mo></mrow></mrow> <mrow><mi>j</mi><mo>=</mo><mn>0....</mn><mi>m</mi><mo>,</mo><msub><mover><mi>x</mi><mo>&amp;RightArrow;</mo></mover><mrow><mn>0</mn><mo>,</mo><mi>c</mi><mi>y</mi></mrow></msub><mo>&amp;Element;</mo><msup><mi>R</mi><mrow><mi>m</mi><mo>+</mo><mn>1</mn></mrow></msup></mrow> The abscissa of the shift register covers the entire measurement range of the lidar signal. According to the grid width and the maximum scanning distance range, a shift register with m+1 discrete equidistant sampling points can be obtained: ie 0, 1 ...m In the register, the obstacle contour height value and probability density distribution corresponding to each measurement point are input through the abscissa value x; Assuming that a set of scans has k measurement points, the probability density of different measurement spots at each sampling point 0...m is: <mrow><msub><mi>&amp;xi;</mi><mi>j</mi></msub><mrow><mo>(</mo><msub><mi>x</mi><mi>i</mi></msub><mo>)</mo></mrow><mo>=</mo><mfrac><mn>1</mn><mrow><msub><mi>&amp;sigma;</mi><mrow><mi>c</mi><mi>l</mi><mi>d</mi><mo>,</mo><mi>j</mi></mrow></msub><mrow><mo>(</mo><msub><mi>x</mi><mrow><mn>0</mn><mo>,</mo><mi>j</mi></mrow></msub><mo>)</mo></mrow><msqrt><mrow><mn>2</mn><mi>&amp;pi;</mi></mrow></msqrt></mrow></mfrac><msup><mi>exp</mi><mrow><mo>(</mo><mo>-</mo>mo><mfrac><mn>1</mn><mn>2</mn></mfrac><msup><mrow><mo>(</mo><mfrac><mrow><msub><mi>x</mi><mi>i</mi></msub><mo>-</mo><msub><mi>x</mi><mi>j</mi></msub></mrow><mrow><msub><mi>&amp;sigma;</mi><mrow><mi>c</mi><mi>l</mi><mi>d</mi><mo>,</mo><mi>j</mi></mrow></msub><mrow><mo>(</mo><msub><mi>x</mi><mrow><mn>0</mn><mo>,</mo><mi>j</mi></mrow></msub><mo>)</mo></mrow></mrow></mfrac><mo>)</mo></mrow><mn>2</mn></msup><mo>)</mo></mrow></msup><mo>,</mo><mi>i</mi><mo>=</mo><mn>1</mn><mo>,</mo><mn>2</mn><mo>,</mo><mn>...</mn><mi>k</mi><mo>;</mo><mi>j</mi><mo>=</mo><mn>0</mn><mo>,</mn>mo><mn>1...</mn><mi>m</mi></mrow> The above-mentioned probability density function can characterize the accuracy of the obstacle height measured in the light spot. The probability density function can be used to perform quasi-continuous estimation processing on the measurement data to obtain a denser obstacle contour height curve. 5. a kind of vehicle front obstacle contour detection method based on recursive superposition algorithm as claimed in claim 1, is characterized in that, according to the described quasi-continuous estimation of obstacle contour of step 4, comprises the following process: The probability density values at the first sampling point are ξ _0,1 , ξ _0,2 ... ξ _0,3 , and the estimated value of the corresponding height value can be calculated by standardization: <mrow><msub><mover><mi>z</mi><mo>&amp;RightArrow;</mo></mover><mrow><mn>0</mn><mo>,</mo><mi>c</mi><mi>y</mi></mrow></msub><mo>=</mo><mfrac><mrow><mo>&amp;Sigma;</mo><mrow><mo>(</mo><msub><mover><mi>&amp;xi;</mi><mo>&amp;RightArrow;</mo></mover><mn>0</mn></msub><mo>*</mo><msub><mover><mi>z</mi><mo>&amp;RightArrow;</mo></mover><mrow><mn>0</mn><mo>,</mo><mi>i</mi></mrow></msub><mo>)</mo></mrow></mrow><mrow><mo>&amp;Sigma;</mo><msub><mover><mi>&amp;xi;</mi><mo>&amp;RightArrow;</mo></mover><mn>0</mn></msub></mrow></mfrac><mo>,</mo><mi>i</mi><mo>=</mo><mn>1</mn><mo>,</mo><mn>2...</mn><mi>k</mi></mrow> Among them, the weighted sum of the scanned data is: The probability density values of the 2nd, 3rd... sampling points and the quasi-continuous estimation of the corresponding height values can also be obtained through standardization. The probability density value of the mth sampling point is ξ _m,1 , ξ _m,2 ...ξ _m,k , and the estimated value of the corresponding height value can be calculated by standardization: Among them, the weighted sum of the scanned data is: According to the above algorithm, the quasi-continuous estimation of the obstacle contour of each sampling point can be obtained in each scan. 6. A method for detecting the contours of obstacles ahead of a vehicle based on a recursive superposition algorithm as claimed in claim 1, wherein said step 5 obtains accurate obstacles by means of continuous recursive superposition of past scan data and current scan data Object silhouette height includes the following processes: Taking the first sampling point as an example, recursively call the scan matching algorithm between the past scan and the current scan, and the recursive superposition algorithm is briefly described by the following formula: Now the recursive call to scan the data: Recursive call for past scan data: In the formula, Represents the calculated value of the height value in the current scan data, Represents the sum of the probability density functions of the first sampling point in the current scan data; represents the computed value of the height value from the past scan data, Represents the sum of the probability density functions of the first sampling point in the past scan data; Represents the distance from the measurement point to the sensor in the x-axis direction in the current scan data, Represents the distance from the measurement point to the sensor in the x-axis direction in the past scan data. 7. a kind of vehicle front obstacle contour detection method based on recursive superposition algorithm as claimed in claim 1, is characterized in that, described step 6 comprises the following process by regression calculation height value deviation and pitch angle deviation: The height offset in the shift register in step 4 can be expressed as: <mrow><mover><mi>&amp;epsiv;</mi><mo>&amp;RightArrow;</mo></mover><msub><mi>z</mi><mrow><mn>0</mn><mo>,</mo><mi>c</mi><mi>y</mi></mrow></msub><mrow><mo>(</mo><msub><mover><mi>x</mi><mo>&amp;RightArrow;</mo></mover><mrow><mn>0</mn><mo>,</mo><mi>c</mi><mi>y</mi></mrow></msub><mo>)</mo></mrow><mo>=</mo><msub><mover><mi>z</mi><mo>&amp;RightArrow;</mo></mover><mrow><mn>0</mn><mo>,</mo><mi>c</mi><mi>y</mi><mo>,</mo><mi>p</mi></mrow></msub><mo>-</mo><msub><mover><mi>z</mi><mo>&amp;RightArrow;</mo></mover><mrow><mn>0</mn><mo>,</mo><mi>c</mi><mi>y</mi><mo>,</mo><mi>n</mi></mrow></msub></mrow> In the formula, Obstacle contour height values representing past scan data; Represents the obstacle contour height value of the current scan data; Represents the height offset or height error in the shift register; Introduce the correlation coefficient R: <mrow><mi>R</mi><mo>=</mo><mo>|</mo><mo>|</mo><mo>&amp;Sigma;</mo><msub><mover><mi>&amp;xi;</mi><mo>&amp;RightArrow;</mo></mover><mrow><mn>0</mn><mo>,</mo><mi>n</mi></mrow></msub><mo>,</mo><mo>&amp;Sigma;</mo><msub><mover><mi>&amp;xi;</mi><mo>&amp;RightArrow;</mo></mover><mrow><mn>0</mn><mo>,</mo><mi>p</mi></mrow></msub><mo>|</mo><msub><mo>|</mo><mi>min</mi></msub></mrow> In the formula, Represent the probability density function sum in the current scan and the past scan, respectively; Evaluate the degree of correlation between the influence of the two scans on the real obstacle contour height by the correlation coefficient R; Therefore, the following new relationship equation can be obtained between the current scan data and the past scan data: <mrow><mo>(</mo><mi>R</mi><mo>,</mo><mi>R</mi><mo>*</mo><msub><mover><mi>x</mi><mo>&amp;RightArrow;</mo></mover><mrow><mn>0</mn><mo>,</mo><mi>c</mi><mi>y</mi></mrow></msub><mo>)</mo><mo>*</mo><mfenced open = "(" close = ")"><mtable><mtr><mtd><mrow><mi>&amp;Delta;</mi><mi>z</mi></mrow></mtd></mtr><mtr><mtd><mrow><mi>&amp;Delta;</mi><mi>n</mi></mrow></mtd></mtr></mtable></mfenced><mo>=</mo><mi>R</mi><mo>*</mo><mover><mi>&amp;epsiv;</mi><mo>&amp;RightArrow;</mo></mover><msub><mi>z</mi><mrow><mn>0</mn><mo>,</mo><mi>c</mi><mi>y</mi></mrow></msub></mrow> In the formula, R represents the correlation coefficient; Represents the distance between the measurement point and the X-axis direction of the lidar in the shift register with equidistant sampling points; Δn and Δz represent the pitch angle deviation and height value deviation between the old and new scan data, respectively; The above equation can be solved by linear regression to construct the generalized inverse matrix A+ of matrix A: <mrow><mover><mi>x</mi><mo>^</mo></mover><mo>=</mo><msup><mi>A</mi><mo>+</mo></msup><mo>*</mo><mi>b</mi><mo>=</mo><msup><mrow><mo>(</mo><msup><mi>A</mi><mi>T</mi></msup><mo>*</mo><mi>A</mi><mo>)</mo></mrow><mrow><mo>-</mo><mn>1</mn></mrow></msup><mo>*</mo><msup><mi>A</mi><mi>T</mi></msup><mo>*</mo><mi>b</mi></mrow> <mrow><mfenced open = "(" close = ")"><mtable><mtr><mtd><mrow><mi>&amp;Delta;</mi><mi>z</mi></mrow></mtd></mtr><mtr><mtd><mrow><mi>&amp;Delta;</mi><mi>n</mi></mrow></mtd></mtr></mtable></mfenced><mo>=</mo><msup><mrow><mo>(</mo><msup><mrow><mo>(</mo><mrow><mi>R</mi><mo>,</mo><mi>R</mi><msub><mover><mi>x</mi><mo>&amp;RightArrow;</mo></mover><mrow><mn>0</mn><mo>,</mo><mi>c</mi><mi>y</mi></mrow></msub></mrow><mo>)</mo></mrow><mi>T</mi></msup><mo>&amp;CenterDot;</mo><mo>(</mo><mrow><mi>R</mi><mo>,</mo><mi>R</mi><msub><mover><mi>x</mi><mo>&amp;RightArrow;</mo></mover><mrow><mn>0</mn><mo>,</mo><mi>c</mi><mi>y</mi></mrow></msub></mrow><mo>)</mo><mo>)</mo></mrow><mrow><mo>-</mo><mn>1</mn></mrow></msup><mo>&amp;CenterDot;</mo><msup><mrow><mo>(</mo><mi>R</mi><mo>,</mo><msub><mi>Rx</mi><mrow><mn>0</mn><mo>,</mo><mi>c</mi><mi>y</mi></mrow></msub><mo>)</mo></mrow><mi>T</mi></msup><mo>&amp;CenterDot;</mo><mrow><mo>(</mo><mi>R</mi><mo>&amp;CenterDot;</mo><mover><mi>&amp;epsiv;</mi><mo>&amp;RightArrow;</mo></mover><msub><mi>z</mi><mrow><mn>0</mn><mo>,</mo><mi>c</mi><mi>y</mi></mrow></msub><mo>)</mo></mrow></mrow> According to the above formula, the height value deviation and pitch angle deviation of multiple groups of obstacle contours can be obtained, and the optimal value of the height value deviation Δz and the optimal value of the pitch angle deviation Δn can be determined by using the least square method. At this time, the height of the new scanning data can be calculated Correction value: z _0,cy,xz =z _0,cy,n +Δn*x _0,cy +Δz In the formula, the new and old scans are superimposed, and z _{0, cy, xz} represent the height correction values of the new scan data. 8. a kind of vehicle front obstacle contour detection method based on recursive superposition algorithm as claimed in claim 1, is characterized in that, the fusion of described step 7 new and old scanning data comprises the following process: After adding the data from the current scan to the saved data from the past scans, the generalized probability densities of all previous scans are increased by the probability densities of the new scans: ∑ξ _0,sum = ∑ξ _0,n + ∑ξ _0,p In the formula, ∑ξ _0,sum represents the generalized probability density of all scanning data contained in the first sampling point; ∑ξ _0,p represents the sum of the generalized probability density of past scanning data; ∑ξ _0,n represents the current scanning the sum of the summary probability densities of the data; Taking into account the new probability density, the updated mean height value of the road profile can be calculated: <mrow><msub><mi>z</mi><mrow><mn>0</mn><mo>,</mo><mi>c</mi><mi>y</mi><mo>,</mo><mi>s</mi><mi>u</mi><mi>m</mi></mrow></msub><mo>=</mo><mfrac><mrow><msub><mi>z</mi><mrow><mn>0</mn><mo>,</mo><mi>c</mi><mi>y</mi><mo>,</mo><mi>p</mi></mrow></msub><mo>*</mo><mo>&amp;Sigma;</mo><msub><mi>&amp;xi;</mi><mrow><mn>0</mn><mo>,</mo><mi>p</mi></mrow></msub><mo>+</mo><msub><mi>z</mi><mrow><mn>0</mn><mo>,</mo><mi>c</mi><mi>y</mi><mo>,</mo><mi>x</mi><mi>z</mi></mrow></msub><mo>*</mo><mo>&amp;Sigma;</mo><msub><mi>&amp;xi;</mi><mrow><mn>0</mn><mo>,</mo><mi>n</mi></mrow></msub></mrow><mrow><mo>&amp;Sigma;</mo><msub><mi>&amp;xi;</mi><mrow><mn>0</mn><mo>,</mo><mi>s</mi><mi>u</mi><mi>m</mi></mrow></msub></mrow></mfrac></mrow> This average height value z _{0, cy, sum} is the final accurate obstacle contour height value, use z _{0, cy, sum} to replace the obstacle contour height value of this scan, and then perform the current scan and a new round of scan The recursive superposition, calculate the height value of the next scan, repeat the recursive superposition, you can get the accurate outline of the obstacle in front of the vehicle.