CN103018728B - Laser radar real-time imaging and building characteristic extracting method - Google Patents

Abstract

The invention discloses a method for laser radar real-time imaging and building feature extraction. Map the laser feet on the scanning line obtained by the lidar to its ideal scanning line; rasterize the ideal scanning line; obtain the elevation information at each grid position on the ideal scanning line through linear interpolation; The elevation information in the ideal scan line is gray-scaled to realize imaging of the detection area corresponding to the scan line; the feature information of buildings in the ideal scan line is extracted by discrete stationary wavelet transform. Using the above method, the laser footpoints in the lidar point cloud data are processed one by one according to their respective scanning lines, so as to realize real-time imaging and building feature extraction during the lidar scanning process.

Description

A method of lidar real-time imaging and building feature extraction

【Technical field】

The invention relates to the field of remote sensing surveying and mapping, in particular to a method for laser radar real-time imaging and building feature extraction.

【Background technique】

As an important part of the 3D geographic information system, buildings are usually used as important calibration objects in lidar point cloud data due to their relatively fixed positions and relatively regular shapes. Therefore, in the process of processing lidar point cloud data, how to extract the building location information and top surface type information contained in it in real time and accurately is an important problem that needs to be solved urgently. At present, the methods for filtering and segmenting lidar point cloud data basically adopt the "post-processing" method, such as: mathematical morphology filtering method, triangular mesh filtering method, region growing method, random sampling consistency method And K-means cluster analysis method, etc. These methods are more suitable for processing flake point cloud data. Although they have high precision, they require a long calculation time, and it is difficult to meet the needs of real-time imaging and building feature extraction during the scanning process.

In the method described in the invention patent (application number: 201110337099.0) applied by Hu Xiangyun and Li Xiaokai, the point cloud data of the lidar is filtered according to each scanning line it contains. With the help of this method, point cloud data can be processed in real time during the scanning process of lidar, but this method cannot meet the processing requirements of real-time imaging and building feature extraction. In the method described in the invention patent (application number: 200710045734.1) applied by Hao Yongtao, the discrete wavelet transform is applied to the two-dimensional laser radar point cloud data processing. But in this method, the wavelet transform is only used as a denoising method, and does not involve the classification of the ground and buildings and the extraction of building features.

In the method of the present invention, for a scanning line that is scanned during the laser radar scanning process, the corresponding ideal scanning line is calculated by using the least squares fitting method, and the laser foot point on the scanning line is mapped to the ideal scanning line. Scan in a straight line. By means of rasterization and linear interpolation, the elevation values of the grid positions on each ideal scanning line are calculated according to the elevation values of the laser foot points on each ideal scanning line. The position and elevation values are converted to a regular set of sequences of elevation values. The elevation sequence is analyzed by stationary discrete wavelet transform, and the building location information and roof type information are extracted from the obtained wavelet detail coefficients. Through spatio-temporal correlation analysis, the extraction results of each scanning line are verified by using the elevation value and wavelet detail coefficient on adjacent scanning lines to ensure the accuracy of data extraction. Compared with the traditional point cloud data processing method, the method proposed in the present invention has the characteristics that the point cloud data can be processed in real time during the laser radar scanning process, and the data update frequency of the obtained processing results is equal to the line rate of the laser radar used. scanning frequency.

【Content of invention】

The invention is a method for laser radar real-time imaging and building feature extraction. The ideal straight line expression of each scanning line in the lidar point cloud data is calculated by using the least squares fitting method. The elevation value corresponding to each grid position on the ideal straight line is calculated by means of rasterization and linear interpolation. Using discrete stationary wavelet transform to extract the elevation change information of each scanning line elevation value sequence, and extract the edge position information of the building and the top surface type information of the building. Using the method of grayscale to calculate the grayscale value of the elevation value sequence and wavelet detail coefficient of each scanning line, the real-time image information of the corresponding detection area is obtained. Use the time-space correlation method to analyze the elevation value of the corresponding grid position on the adjacent scanning line, the wavelet detail coefficient, the building position information and the top surface type information of the building, and verify the obtained point cloud data extraction results to improve Accuracy of building feature extraction.

Technical scheme of the present invention is realized like this:

A kind of laser radar real-time imaging and the method for building feature extraction of the present invention, the steps are as follows:

Step 1. Linearize the laser footpoints belonging to the same scan line in the point cloud data in the xoy plane.

Under ideal conditions, the laser footpoints on a scanning line obtained by the laser radar through the rotation of the prism in the scanning mechanism should be relatively evenly distributed on the same straight line on the xoy plane in the Cartesian three-dimensional xyz coordinate system. However, in actual situations, due to the elevation changes of ground objects and the deviation of the heading angle, roll angle, and pitch angle of the lidar payload platform under the influence of complex airflow, the distance between the laser foot point on each scanning line and its ideal scanning line There is a certain distance difference between them. During the laser radar scanning process, the ideal scan line equation corresponding to the actually obtained scan line is calculated by using the least square fitting method, so that in the xoy plane, the sum of the squares of the distances from all laser foot points on the scan line to the ideal line is the smallest . Calculate the perpendicular line between each laser foot point and the ideal line in the xoy plane, and take the intersection coordinates of each vertical line and the ideal line in the xoy plane as the x-axis and y-axis of each laser foot point after linearization coordinate.

Step 2: On the ideal scanning line, perform rasterization processing of elevation values on the linearized laser footpoints.

After linearization, the laser footpoints on the ideal scanning line are distributed in a straight line, and the average value of the spacing of each laser footpoint on the straight line is used as the grid spacing value. Divide the line from the first laser foot point into several grids with this spacing value. Using the elevation value of each laser foot point on the straight line, the elevation value at each grid position on the ideal straight line is calculated by using the method of linear interpolation.

Step 3, real-time imaging of laser radar and extraction of building features based on stationary wavelet transform.

Through operations such as linearization and elevation value rasterization in step 1 and step 2, the laser footpoints on the scanning line are converted into regular discrete elevation data sequences, which are processed by discrete stationary wavelet transform. According to the obtained wavelet The feature of the detail coefficient extracts the edge position information of the building contained in it, as well as the top surface type information of the building. Through the method of grayscale, the grayscale value of the regular discrete elevation data sequence and wavelet detail coefficient after rasterization is calculated. During the scanning process of the laser radar, the laser footpoint data obtained on each scanning line is processed in real time according to the above method, so as to realize real-time imaging of the corresponding detection area.

Step 4, verifying the extraction results of each scanning line by using the method of spatio-temporal correlation analysis.

Through the operation of step 3, the building edge position information and building top surface type information contained in the laser foot points of each scanning line can be extracted in real time, and can be calculated according to the elevation information and wavelet detail coefficient of each scanning line. The corresponding detection area is imaged in real time. However, the elevation information on each scan line only comes from the scanning direction of the lidar, and does not include the information in the direction of travel of the load platform. Therefore, the method of spatio-temporal correlation analysis is used to analyze and correct the elevation value, wavelet detail coefficient and the position and top surface type information of each building in the corresponding grid position on each adjacent scanning line during the scanning process. To improve the accuracy of building feature extraction.

Beneficial effects of the present invention: By using processing methods such as plane linearization and elevation value rasterization, the disordered and irregular laser foot points on each scanning line in the point cloud data are converted into a group of regular elevation discrete sequences. Using discrete stationary wavelet to analyze the elevation discrete sequence. Compared with the traditional laser radar point cloud data processing method, the present invention can perform imaging and building feature extraction processing in real time during the laser radar scanning process. With the help of spatio-temporal correlation analysis, use the elevation value of the corresponding grid position on the adjacent scanning line and the wavelet detail coefficient to verify the information such as the location of the building on each scanning line and the type of the top surface of the building to ensure that The accuracy of laser radar point cloud data processing by adopting the method of the invention.

【Instructions attached】

Fig. 1 is a schematic diagram of a method for real-time imaging of laser radar and feature extraction of buildings according to the present invention;

Fig. 2 is a schematic diagram of a specific embodiment of processing a real lidar point cloud data by using the method of the present invention.

【Detailed ways】

A kind of laser radar real-time imaging and the method for building feature extraction of the present invention, the steps are as follows:

Step 1. Linearize the laser footpoints belonging to the same scan line in the point cloud data in the xoy plane.

利用最小二乘拟合方法计算出当目标函数sumdistL_k取最小值时理想扫描线L_k的相关参数：a和b的值。将这N+1个激光脚点的x轴和y轴坐标分别映射至该理想扫描线上，并将激光脚点P_i(x_i，y_i)在扫描线L_k上直线化的结果定义为P′_i(x′_i，y′_i)。根据P_i，P′_i和L_k之间的关系，建立方程组： $\left\{\begin{array}{c}\mathrm{dist}(L_k,P_i)=\sqrt{{(y_i-y_i^{\′})}^2+{(x_i-x_i^{\′})}^2}\\ \frac{(y_i-y_i^{\′})}{(x_i-x_i^{\′})}\·a=-1\end{array}\right.,$ 通过对方程组进行求解，计算出P′_i点的x轴和y轴坐标：(x′_i，y′_i)。对第k条扫描线上的N+1个激光脚点均采用这样的方式进行处理，完成其在xoy平面上的直线化处理过程。Under ideal conditions, the laser feet on a scanning line obtained by the lidar through the rotation of the prism in the scanning mechanism should be relatively evenly distributed on the same straight line on the xoy plane. However, in actual situations, due to the elevation change of ground objects and the deviation of the heading angle, roll angle, and pitch angle of the loading platform of the scanning mechanism under the influence of complex airflow, the distance between the laser foot point on each scanning line and its ideal scanning line There is a certain distance difference between them. Under the influence of complex external conditions, the laser footpoints on the kth scanning line are usually distributed around the ideal scanning line. The mathematical expression of the ideal scan line L _k corresponding to the kth scan line on the xoy plane can be defined as: y=ax+b. The distance between the laser footpoint P _i (i∈[0, N], and i∈Z) on the xoy plane and the ideal scanning line to which it belongs is recorded as dist(L _k , P _i ). The sum of the squares of the distances between all N+1 laser foot points on the kth scan line and the ideal scan line is defined as the objective function sumdistL _k :

The relevant parameters of the ideal scan line L _k when the objective function sumdistL _k takes the minimum value are calculated by using the least square fitting method: the values of a and b. Map the x-axis and y-axis coordinates of the N+1 laser footpoints to the ideal scanning line respectively, and define the result of linearizing the laser footpoint P _i ( _xi , y _i ) on the scanning line L _k is P′ _i (x′ _i , y′ _i ). According to the relationship between P _i , P′ _i and L _k , establish a system of equations: $\left\{\begin{array}{c}\mathrm{dist}(L_k,P_i)=\sqrt{{({\mathrm{the\; y}}_i-{\mathrm{the\; y}}_i^{\′})}^2+{(x_i-x_i^{\′})}^2}\\ \frac{({\mathrm{the\; y}}_i-{\mathrm{the\; y}}_i^{\′})}{(x_i-x_i^{\′})}\·a=-1\end{array}\right.,$ By solving the equations, the x-axis and y-axis coordinates of point P′ _i are calculated: (x′ _i , y′ _i ). The N+1 laser foot points on the k-th scanning line are all processed in this way, and the linearization process on the xoy plane is completed.

Step 2: Rasterizing the elevation value of the linearized laser footpoints.

将栅格间距grid定义为理想扫描线L_k上的N+1个激光脚点的平均点间距：

通过线性插值的方法，将理想扫描线上的激光脚点的高程值信息映射至等间隔分布的各个栅格位置处。栅格位置G_i处的横坐标为i·grid，将其高程信息定义为z′_i。G_i与在理想扫描线上其两侧最邻近激光脚点P′_i-1(X_i-1，z_i-1)和P′_i(X_i，z_i)的位置和高程信息之间的关系可以表示为：

进而求解出G_i处的高程信息z′_i：

按照该方法，对L_k上所有栅格位置的高程值进行计算，其中，栅格序列左右两端点由于不满足存在两侧最邻近激光脚点的条件，直接以其最邻近激光脚点的高程值对其进行赋值。通过上述处理方法，L_k上原本不均匀、散乱分布在三维空间中的激光脚点被映射为二维空间内的规则离散数学序列。After linearization, the horizontal distance between the laser feet on the k ideal scanning line is discrete and unevenly distributed. In order to simplify the calculation complexity and improve the calculation speed, the three-dimensional coordinates of each mapped laser foot point are simplified to two-dimensional coordinates on the section plane that contains L _k and is perpendicular to the xoy plane, and each laser foot point after linearization The included elevation information is rasterized. Define the two-dimensional coordinate form of the laser foot point P′ ₀ (x′ ₀ , y′ ₀ , z ₀ ) at the starting position of the ideal scanning line as P′ ₀ (X ₀ , z ₀ ), and use it as the grid The starting position G ₀ (0, z ₀ ) of the grid, where z ₀ is the elevation value of point P′ ₀ . By calculating the distance between other laser foot points and the starting position of the grid, their respective abscissa coordinates X _i on the section plane are obtained:

The grid spacing grid is defined as the average point spacing of N+1 laser foot points on the ideal scanning line L _k :

By means of linear interpolation, the elevation value information of the laser footpoint on the ideal scanning line is mapped to each grid position distributed at equal intervals. The abscissa at grid position G _i is i·grid, and its elevation information is defined as z′ _i . Between G _i and the position and elevation information of the nearest laser feet P′ _i-1 (X _i-1 , zi _-1 ) and P′ _i (X _i , _zi ) on both sides of the ideal scanning line The relationship can be expressed as:

Then solve the elevation information z′ _i at G _i :

According to this method, the elevation values of all grid positions on L _k are calculated, and the left and right ends of the grid sequence do not satisfy the condition of the nearest laser footpoints on both sides, so the elevation values of the nearest laser footpoints Value is assigned to it. Through the above processing method, the laser footpoints on _Lk that are originally uneven and scattered in the three-dimensional space are mapped to a regular discrete mathematical sequence in the two-dimensional space.

Step three, real-time imaging and building feature extraction based on stationary wavelet transform.

表示卷积运算，dhGk和dgGk分别表示通过DSWT变换获取的小波近似系数和小波细节系数。分析离散小波变换的性质可知，序列Gk中所包含高程变化信息主要包含于其小波细节系数dgGk内。采用本发明所述的直线化和栅格化方法对平顶面建筑物、坡顶面建筑物和人字形顶面建筑物三类建筑物顶面的激光脚点进行处理获得三组离散序列。对三组离散序列进行一阶DSWT变换，获得三组小波细节系数。对得到的小波细节系数进行分析，对应于建筑物顶面的边缘位置，小波细节系数的波形会出现冲击信号，其中，建筑物的高程上升沿位置对应负值冲击信号，高程下降沿位置对应正值冲击信号。由于DSWT变换所采用的是线性滤波器，因此可以根据冲击信号幅值反向解算出建筑物的高程值。平顶型建筑物顶面部分不存在高程变化，相应的，在负值冲击信号和正值冲击信号之间小波细节系数均为0值。坡型建筑物顶面部分存在一定斜率，若该斜率为一固定正值，相应的，在负值冲击信号和正值冲击信号之间小波细节系数为一固定负值；反之，若该斜率为一固定负值，则对应的小波细节系数为一固定正值。人字型屋顶可理解为由两斜率互为相反数的坡型顶面组成，其顶面部分的小波系数为一负值序列和一对应的正值序列组成。通过对这三种典型的建筑物顶面模型进行组合扩展，可得到大量复杂建筑物顶面模型。采用上述的方法，根据冲击信号的位置判断建筑物顶面边界，根据建筑物顶面内部区域的小波细节系数对其顶面类型进行判断，得出激光雷达点云数据中的建筑物顶面类型信息。通过灰度化的方法，计算出栅格化后的规则离散高程数据序列和小波细节系数的灰度值，实现对对应探测区域进行实时成像处理。Through operations such as linearization of the xoy plane and rasterization of elevation values in steps 1 and 2, the laser footpoints on the kth scanning line are converted into regular discrete elevation data sequences. Discrete stationary wavelet transform is used to process it, and according to the characteristics of the obtained wavelet detail coefficients, the location information of the building contained in the scan line and the top surface type information of the building are extracted. According to the shape of the top of the building, the buildings contained in the point cloud data can be divided into three categories: flat top buildings, slope top buildings and herringbone top buildings. These three types of buildings are scanned, and the obtained laser footpoints on each scanning line are converted into regular discrete sequences through linearization, rasterization and other processing methods. Discrete Stationary Wavelet Transform (DSWT) method is used to analyze the discrete sequence of laser footpoints and extract the elevation change information it contains. In addition to the fast computing capability of discrete wavelet transform, DSWT transform, compared with other discrete wavelet transforms, has the important characteristics of redundancy and translation invariance. The signal lengths of the wavelet detail coefficients at all scales obtained through DSWT transformation are equal to the signal length of the original data. Therefore, the characteristic signals shown in the wavelet detail coefficients can be quickly and accurately determined by their corresponding time-shift parameters. Its position in the original discrete sequence. The grid position data G _i (i ∈ [0, N], and i ∈ Z) calculated according to the laser footpoint P _i (i ∈ [0, N], and i ∈ Z) on the scanning line L _k The formed regular discrete sequence is named Gk, and the sequence Gk is subjected to a stationary discrete wavelet transform using the low-pass filter h and the high-pass filter g: $\left\{\begin{array}{c}\mathrm{wxya}=h\⊗K\\ \mathrm{wxya}=g\⊗K\end{array}\right.,$ in,

Indicates the convolution operation, dhGk and dgGk respectively indicate the wavelet approximation coefficient and wavelet detail coefficient obtained through DSWT transformation. Analyzing the properties of the discrete wavelet transform shows that the elevation change information contained in the sequence Gk is mainly contained in its wavelet detail coefficient dgGk. The linearization and rasterization methods of the present invention are used to process the laser footpoints on the top surfaces of three types of buildings, namely flat roof buildings, slope roof buildings and herringbone roof buildings, to obtain three sets of discrete sequences. First-order DSWT transformation is performed on three groups of discrete sequences to obtain three groups of wavelet detail coefficients. Analyzing the obtained wavelet detail coefficients, corresponding to the edge position of the top surface of the building, the waveform of the wavelet detail coefficients will appear shock signals, in which the position of the rising edge of the building’s elevation corresponds to a negative shock signal, and the position of the falling edge of the elevation corresponds to a positive value. Value shock signal. Since the DSWT transformation uses a linear filter, the elevation value of the building can be calculated inversely based on the amplitude of the shock signal. There is no elevation change on the top surface of a flat-roofed building, and correspondingly, the wavelet detail coefficients are all 0 between the negative impact signal and the positive impact signal. There is a certain slope on the top of the slope-shaped building. If the slope is a fixed positive value, correspondingly, the wavelet detail coefficient between the negative impact signal and the positive impact signal is a fixed negative value; otherwise, if the slope is A fixed negative value, the corresponding wavelet detail coefficient is a fixed positive value. The herringbone roof can be understood as being composed of two slope-shaped top surfaces whose slopes are opposite to each other, and the wavelet coefficients of the top surface part are composed of a negative value sequence and a corresponding positive value sequence. By combining and extending these three typical building top models, a large number of complex building top models can be obtained. Using the above method, the boundary of the building top surface is judged according to the position of the impact signal, and the top surface type is judged according to the wavelet detail coefficient of the inner area of the building top surface, and the building top surface type in the lidar point cloud data is obtained information. Through the method of grayscale, the grayscale value of the regular discrete elevation data sequence and wavelet detail coefficient after rasterization is calculated, and the real-time imaging processing of the corresponding detection area is realized.

Step 4, verifying the extraction results of each scanning line by using the method of spatio-temporal correlation analysis.

Through the operation of step 3, the building position information and building top surface type information contained in the laser foot points of each scanning line can be extracted in real time, and according to the elevation information and wavelet detail coefficient of each scanning line, the corresponding Real-time imaging of the detection area. However, the elevation information on each scanning line only comes from the scanning direction of the lidar, and does not include the information on the traveling direction of the load platform. The elevation value of the grid position, the wavelet detail coefficient, and the position and top surface type information of each building contained in it are analyzed and verified. Taking the wavelet coefficients as an example, the wavelet detail coefficients of each scan line are classified, and the corresponding grid positions are marked with the obtained results. In the wavelet detail coefficient, if there is an interval with a negative impact signal on one side and a positive impact signal on the other side, the interval is judged as a building. Among them, the starting position of the negative impact signal represents the rising edge of the building top, and its corresponding grid position is marked as 1; the starting position of the positive shock signal represents the falling edge of the building top, and Its corresponding grid position is marked as 2. The grid position inside the interval is classified and marked according to its corresponding wavelet detail coefficient value: when the wavelet coefficient is 0, there is no elevation change at this position, and it is marked as 3, indicating a flat top surface; when the wavelet coefficient is negative , mark it as 4, which means it is a slope top surface with positive slope; when the wavelet coefficient is positive, mark it as 5, which means it is a slope top surface with negative slope; when the wavelet coefficient is a combination of positive and negative numbers and , mark it as 4&5, which means it is a herringbone top surface. The spatial-temporal correlation analysis is carried out on the grid values of each marked scanning line, and the marked values of the corresponding grid positions of the forward and backward scanning lines are compared with the marked values of the obtained building feature extraction results, and the obtained The results of building feature extraction are verified.

The method of the present invention is further described by applying the method of the present invention to a specific embodiment of real-time processing of real laser radar point cloud data:

The real lidar point cloud data and the remote sensing image of the corresponding scanning area are shown in Fig. 2(a) and 2(b) respectively. Utilize the method described in the present invention to process the laser feet in the point cloud data according to their respective scan lines, the point cloud data elevation image obtained is as shown in Figure 2 (c), and the wavelet detail coefficient image is as shown in Figure 2 (d) shown. Utilize the method described in the present invention to analyze the wavelet detail coefficients of each scanning line obtained, obtain the building features wherein, and mark according to the method described in step 4, utilize the time-space correlation analysis method to obtain the result Check to get the building position and top surface type information in the point cloud data, as shown in Figure 2(e). From the observation of Fig. 2(e), it can be seen that the laser radar point cloud data can be well processed by using the method of the present invention, and operations such as real-time imaging and building feature extraction can be realized.

The above is only the basic scheme of the specific implementation method of the present invention, but the protection scope of the present invention is not limited thereto, and any conceivable change or replacement within the technical scope disclosed by the present invention by anyone familiar with the technical field, All should be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be determined by the protection scope of the claims. All changes that come within the equivalent meaning and range of the claims are intended to be included in the scope of the claims.

Claims (1)

1. a method for laser radar real-time imaging and building feature extraction, is characterized in that comprising following four steps: Step 1. Linearize the laser footpoints belonging to the same scanning line in the point cloud data in the xoy plane; Under ideal conditions, the laser foot points on a scanning line obtained by the lidar through the rotation of the prism in the scanning mechanism should be relatively evenly distributed on the same straight line on the xoy plane in the Cartesian three-dimensional xyz coordinate system, but in actual conditions In the process, due to the change of the elevation of ground objects and the deviation of the heading angle, roll angle and pitch angle of the lidar payload platform under the influence of complex airflow, there is a certain gap between the laser foot point on each scanning line and its ideal scanning line. Distance deviation, during the laser radar scanning process, use the least squares fitting method to calculate the ideal scan line equation corresponding to the completed scan line, so that in the xoy plane, the distance between all laser foot points on the scan line and the ideal line The sum of the squares of the distances is the smallest, calculate the perpendicular line between each laser foot point in the xoy plane and the ideal line, and take the intersection coordinates of each vertical line and the ideal line in the xoy plane as the linearized coordinates of each laser foot point x-axis and y-axis coordinates; Step 2, on the ideal scanning line, perform elevation value rasterization processing on the linearized laser foot points; After linearization, the laser footpoints on the ideal scanning line are distributed in a straight line, and the average value of the distance between each laser footpoint on the straight line is used as the grid spacing value, and the straight line is drawn from the first The laser footpoint starts to be divided into several grids, and the elevation value of each grid position on the ideal straight line is calculated by using the elevation value of each laser footpoint on the line and the method of linear interpolation; Step 3. Lidar real-time imaging and building feature extraction processing based on stationary wavelet transform; Through the linearization and elevation value rasterization operations in step 1 and step 2, the laser footpoints on the scanning line are converted into regular discrete elevation data sequences, which are processed by discrete stationary wavelet transform. According to the obtained wavelet details The characteristics of the coefficients, extract the edge position information of the buildings contained in it, and the top surface type information of the buildings, and calculate the grayscale of the regular discrete elevation data sequence and wavelet detail coefficient after rasterization by grayscale method During the scanning process of the laser radar, the laser footpoint data obtained on each scanning line is processed in real time according to the above method, so as to realize real-time imaging of the corresponding detection area; Step 4, using the method of spatio-temporal correlation analysis to verify the extraction results of each scanning line; Through the operation of step 3, the edge position information of the building and the type information of the building top surface contained in the laser foot points of each scanning line are extracted in real time, and the elevation information and wavelet detail coefficient of each scanning line are used for real-time imaging processing, but The elevation information on each scanning line only comes from the scanning direction of the lidar, and does not include the information on the traveling direction of its load platform. The elevation value of the location, the wavelet detail coefficient, and the location and top surface type information of each building contained in it are analyzed and verified to improve the accuracy of feature extraction.

Images