Laser radar intensity information extraction and correction method

Abstract

The invention discloses a laser radar echo intensity information extraction and correction method, and belongs to the field of full-waveform laser radar echo signal processing. The method mainly comprises the following steps: full-waveform recording of digital laser radar echo signals and filtering and noise reduction processing are carried out; using an integral centroid method to extract and store intensity information of the laser radar echo signal; a deconvolution reconstruction laser radar echo signal is used to acquire a laser radar impact response, distance information of a detection point is extracted, and an inclination angle of a local micro surface element of the detection point is calculated through inversion; generating a laser radar intensity point cloud image, and calculating and correcting a normal vector of the point cloud image; the incident angle of the laser radar is calculated, a laser radar intensity information correction model is established based on the detection distance and the incident angle of the laser radar, and a corrected laser radar intensity point cloud picture is generated. The method is characterized in that the influence of the detection distance and the incident angle on the intensity information of the laser radar is effectively eliminated through the correction of the intensity information of the laser radar.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of full waveform lidar signal processing, and particularly relates to a lidar intensity information extraction and correction method. BACKGROUND

[0002] Laser radar is an active detection technology that uses a specific wavelength of laser pulse to obtain target distance, direction, reflectivity and other related information. Due to the characteristics of long detection distance, high measurement accuracy and less environmental influence, the application scenarios of laser radar have gradually expanded in recent years. In the national defense and military industry, laser radar is an important part of weapon guidance system, target recognition system, missile defense system and satellite remote sensing system. In daily life, laser radar has great application prospects in intelligent driving, robot navigation, unmanned aerial vehicle mapping and other fields.

[0003] At present, many studies show that the intensity data of laser radar has great application prospects and is applied to cultural relic protection, rock and soil layer classification, structure damage detection and water content detection scenes. However, in the process of laser radar detection, the intensity value of laser radar is also affected by atmospheric transmission characteristics, geometric structure of the measured target and other factors. Due to the influence of the above factors, the intensity values of the same object may be different, and the intensity values of different objects may be the same. Therefore, how to correct the intensity data of laser radar and extract the surface attribute information of the measured target has become a key problem in the application of laser radar intensity data and has become a research hotspot in the field of laser radar at home and abroad in recent years.

[0004] In the process of laser radar detection, the detection distance and the incident angle have a great influence on the intensity of laser echo; with the increase of the detection distance and the incident angle, the intensity of laser echo decreases. Moreover, the influence of the distance effect depends on the system characteristics of the instrument, and the influence of the incident angle effect is related to the properties of the measured target surface. Therefore, it is necessary to eliminate the influence of distance effect and incident angle effect on the intensity data of laser radar, so as to obtain the corrected intensity value of laser radar related only to the properties of the target surface. SUMMARY

[0005] The purpose of the present application is to provide a laser radar intensity information extraction and correction method.

[0006] The technical scheme for achieving the purpose of the present application is as follows: a laser radar echo intensity information extraction and correction method, comprising the following steps:

[0007] Step S1: performing filtering and noise reduction processing on the full waveform of the digitized laser radar echo signal;

[0008] Step S2: Extract the intensity information of the filtered and denoised laser radar echo signal using the integral centroid method, and store the intensity information of each detection point according to the laser radar detection sequence;

[0009] Step S3: Deconvolution reconstruction of the filtered and denoised laser radar echo signal to obtain the laser radar impulse response, extract the distance information of the detection point according to the laser radar impulse response, and calculate the local micro facet tilt angle of the detection point;

[0010] Step S4: Generate a laser radar intensity point cloud map according to the intensity information in step S2, the distance information in step S3, and the scanning angle information recorded in the laser radar detector, calculate the normal vector of the point cloud map using the K nearest neighbor algorithm, and correct the point cloud normal vector according to the tilt angle of each local micro facet of the detection point in step S3;

[0011] Step S5: Calculate the incidence angle of the laser radar according to the point cloud normal vector, the point cloud spatial coordinates, and the laser radar detector spatial coordinates calculated in step S4, establish a laser radar intensity information correction model based on the detection distance and incidence angle of the laser radar, and generate a corrected laser radar intensity point cloud map;

[0012] Preferably, the Savitzky-Golay filter method is used in step S1 to filter and denoise the digitized laser radar echo signal.

[0013] Preferably, the specific formula for extracting the intensity information of the filtered and denoised laser radar echo signal using the integral centroid method in step S2 is:

[0014]

[0015] Where P lidar (t) is the filtered and denoised laser radar echo signal, N is the number of effective sampling points of the laser radar echo signal, and t is the duration of the laser radar echo signal.

[0016] Preferably, the specific formula for deconvolution reconstruction of the laser radar echo signal is:

[0017]

[0018] In the formula, H is the matrix form of the laser radar impulse response, Y is the matrix form of the laser radar echo, X is the blur matrix composed of laser emission waves, is the objective function The laser radar impulse response matrix when the minimum value is taken, which is the final result of deconvolution reconstruction; is the residual sum of squares, ||H||1 is the L1 regularization term, ||H||2 is the L2 regularization term, μ is the regularization coefficient, χ is the mixing parameter, and is used to balance the L1 and L2 regularization terms.

[0019] Preferably, the specific method of extracting distance information based on the laser radar impulse response is:

[0020] The time point of extracting the peak position of the laser radar impulse response is taken as the timing end time t2, and the peak position of the laser radar emission pulse is taken as the timing start time t1, and the distance information is calculated, specifically: d = c * (t2-t1) / 2, wherein c is the speed of light.

[0021] Preferably, the specific method of calculating the tilt angle of the local micro facet of the detection point based on the laser radar impulse response inversion is:

[0022] The relationship between the tilt angle α of the plane of the detected object and the half peak width ω of the laser radar impulse response is established: c

[0023]

[0024] In the formula, k = ω c / θd, θ is the divergence angle of the laser radar, and d is the detection distance of the laser radar;

[0025] According to the above formula, in the subsequent laser radar point cloud normal vector calculation process, the tilt angle α of the local micro facet of the detection point is calculated according to the actual detection time laser radar impulse response half peak width ω c

[0026] Preferably, the specific method of calculating the normal vector of the point cloud graph using the K-neighbor algorithm and correcting the point cloud normal vector according to the tilt angle of the local micro facet of each detection point is:

[0027] Select K neighboring points around the calculation point, fit a plane with the smallest sum of projections of all points by least squares, and take the normal vector of the plane as the normal vector of the point. Repeat the above process for all point clouds until the calculation of the normal vector of all detection points is completed;

[0028] When the slope amplitude of the detected target surface changes by more than a threshold value, the normal vector of the point cloud is corrected by the tilt angle of the local micro facet of each detection point The specific method is:

[0029] Compare the tilt angle of the K-neighbor fitting plane with the tilt angle of the local micro facet of the detection point. When the difference between the two is below a threshold value, the normal vector calculated by the K-neighbor method is retained When the difference between the two exceeds the threshold value, the tilt angle of the local micro facet of the detection point is taken as the tilt angle of the fitting plane to calculate the normal vector of the point cloud.​​

[0030] Preferably, the incident angle θ of the laser radar i The calculation formula is as follows:

[0031]

[0032] In the formula, cos -1 is the inverse cosine function, is the laser incident vector (x, y, z) is the spatial coordinates of the detection point, (x0, y0, z0) is the spatial coordinates of the laser radar, is the normal vector of the point cloud.

[0033] Preferably, the laser radar intensity information correction model is specifically as follows:

[0034]

[0035] In the formula, I lidar is the intensity information of the laser radar echo extracted based on the integral centroid method, d is the detection distance extracted based on the deconvolution reconstruction, θ i is the actual incident angle of the laser radar, η and C are fitting parameters, θ0 and d0 are the set reference incident angle and reference detection distance respectively.

[0036] Compared with the prior art, the present application has the following advantages: the integral centroid method is used in the extraction of the laser radar echo intensity information, which significantly improves the extraction accuracy of the laser radar intensity information; the deconvolution of the integral echo waveform is used to extract the distance information, which eliminates the influence of the transmission waveform and obtains the impulse response related to the target attribute; the inclination angle of the local micro facet is used to assist in the correction of the normal vector of the point cloud when the normal vector of the point cloud is calculated, which improves the correction accuracy of the laser radar intensity information.

[0037] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the present application. The objects and other advantages of the present application can be achieved and obtained by the structure particularly pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF DRAWINGS

[0038] Figure 1 is the flowchart of the present application

[0039] Figure 2 is the schematic diagram of the extraction of the laser radar intensity information by the integral centroid method

[0040] Figure 3 is the normal vector diagram of the visualized laser radar point cloud in the plane detection scene in some specific embodiments

[0041] Figure 4 Fig. 1 is a schematic diagram of a laser radar scanning detection on the same straight line for a diffuse reflection plane target

[0042] Figure 5 Fig. 2 is a distribution range of intensity values and average intensity values of each target laser radar before and after correction of intensity information DETAILED DESCRIPTION

[0043] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments of the present application.

[0044] Examples of the embodiments are shown in the drawings, in which the same or similar symbols represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by reference to the drawings are exemplary and are intended to explain the present application, and cannot be understood as a limitation of the present application.

[0045] Referring to the flowchart of Figure 1 , the specific implementation process of the present application is as follows:

[0046] Step S1: digitizing the laser radar echo signal full waveform record and performing filtering and noise reduction processing;

[0047] In some embodiments, the laser radar echo signal filtering and noise reduction in step S1 uses the Savitzky-Golay filtering method;

[0048] Step S2: using the integral centroid method to extract the intensity information of the laser radar echo signal in step S1, and storing the intensity information of each detection point according to the laser radar detection sequence;

[0049] In some embodiments, referring to Figure 2 , the laser radar echo intensity information in step S2 is extracted based on the integral centroid method, and the laser radar intensity information I lidar The specific calculation method is represented by the following formula:

[0050]

[0051] Where P lidar (t) is the laser radar echo signal filtered in step S11, and N is the number of effective sampling points of the laser radar echo signal.

[0052] Step S3: using deconvolution to reconstruct the laser radar echo signal in step S1 to obtain the laser radar impulse response, extracting the distance information of the detection point according to the laser radar impulse response, and inversely calculating the tilt angle of the local micro facet element of the detection point;

[0053] In some embodiments, the step S3 includes the following:

[0054] S31: The specific method of deconvolution reconstruction of the laser radar echo signal in step S3 is represented by the following formula:

[0055]

[0056] In the formula, H is the matrix form of the laser radar impulse response, Y is the matrix form of the laser radar echo, X is the blur matrix composed of laser emission waves, is the objective function The laser radar impulse response matrix when taking the minimum value is the final result of deconvolution reconstruction; is the residual sum of squares, ||H||1 is the L1 regularization term, ||H||2 is the L2 regularization term, μ is the regularization coefficient, χ is the mixing parameter, which is used to balance the L1 and L2 regularization terms.

[0057] Generally, it can be solved by the following specific steps: 1. Initialize the value of H; 2. Iteratively update H, each update is along the direction of the objective function descent; 3. In each iteration, adjust H to minimize the objective function; 4. Repeat the iteration until the stopping condition is met, such as the improvement of the objective function is less than a certain threshold or the maximum number of iterations is reached.

[0058] The regularization parameter μ and the mixing parameter χ can be determined by the generalized cross-validation method, and the specific steps are as follows: first, define a parameter grid containing different μ and χ values; second, divide the data set into multiple subsets, each subset is used as a test set, and the rest is used as a training set, repeat multiple times and calculate the average performance of the model. Finally, from the cross-validation process, select the best regularization parameter μ and mixing parameter χ combination with the best average performance.

[0059] S32: The specific method of extracting distance information based on the laser radar impulse response in step S3 is to extract the time of the peak position of the laser radar impulse response as the timing end time t2, and the peak position of the laser radar emission pulse as the timing start time t1. The calculation method of the distance information is d = c*(t2-t1) / 2, where c is the speed of light.

[0060] S33: The specific method of calculating the tilt angle of the local micro facet of the detection point based on the laser radar impulse response in step S3 is:

[0061] The impulse response and echo model of the target, the half peak width of the impulse response is related to the laser divergence angle, the distance, and the target inclination angle factor, in the case where the laser divergence angle and the detection distance are known, using a standard diffuse reflection plate as the detection target, in the detection distance and the laser radar divergence angle fixed reflection plate inclination angle range of 1°-89°, the rotation step is 1°, the detection scene is collected, the laser radar echo information is reconstructed by deconvolution, and the laser radar impulse response is obtained. c The half peak width ω c of the laser radar impulse response is fitted by using a rational approximation function to fit the element inclination angle α and the half peak width ω c of the laser radar impulse response, and the specific fitting result is determined according to the following formula:

[0062]

[0063] In the formula, k=ω c / θd, θ is the laser radar divergence angle, and d is the laser radar detection distance. According to the above formula, the half peak width ω c of the laser radar impulse response is used to calculate the local micro-element inclination angle of the detection point.

[0064] Step S4: generating a laser radar intensity point cloud map according to the intensity information in step S2, the distance information in step S3, and the scanning angle information recorded in the laser radar detector, calculating the normal vector of the point cloud map using a K-nearest neighbor algorithm, and correcting the point cloud normal vector according to the inclination angle of each local micro-element of the detection point in step S3;

[0065] In some specific embodiments, the step S4 includes the following contents:

[0066] S41: The specific method for calculating the normal vector of the laser radar point cloud in step S4 is: selecting K adjacent points around the calculation point, fitting a plane with the smallest sum of all point projections by least squares, and taking the normal vector of the plane as the normal vector of the point. The above process is repeated for all point clouds until the calculation of the normal vector of all detection points is completed. In some specific embodiments, in the case of taking a Lambertian plane as the detection target, the normal vector diagram of the laser radar point cloud can be visualized as shown in Figure 3 .

[0067] S42: When the detection target surface slope has a large amplitude change or the boundary region between multiple targets, the K-nearest neighbor algorithm has a large error in the calculation of the point cloud normal vector. At this time, the local micro-element inclination angle of each detection point is used to correct the point cloud normal vector. The specific method is: comparing the inclination angle of the K-nearest neighbor fitting plane with the local micro-element inclination angle of the detection point, and when the difference between the two is below a threshold value, the normal vector calculated by the K-nearest neighbor method is retained. The normal vector of the point cloud is calculated according to the local microfacet tilt angle of the detection point when the difference between the two exceeds a threshold value

[0068] Step S5: Calculate the incident angle of the laser radar according to the normal vector of the point cloud, the spatial coordinates of the point cloud and the spatial coordinates of the laser radar detector calculated in step S4, and establish a laser radar intensity information correction model based on the detection distance and the incident angle of the laser radar and generate a corrected laser radar intensity point cloud map;

[0069] In some embodiments, step S5 includes the following contents:

[0070] The actual incident angle θ of the laser radar i is determined according to the following formula:

[0071]

[0072] In the formula, cos -1 is the inverse cosine function, is the laser incident vector (x, y, z) is the spatial coordinates of the detection point, (x0, y0, z0) is the spatial coordinates of the laser radar, is the normal vector of the point cloud.

[0073] The laser radar intensity information correction model is specifically:

[0074]

[0075] In the formula, I lidar is the intensity information of the laser radar echo extracted based on the integral centroid method, d is the detection distance extracted based on deconvolution reconstruction, θ i is the actual incident angle of the laser radar, η and C are fitting parameters, θ0 and d0 are the set reference incident angle and reference detection distance, respectively.

[0076] In some embodiments, a moving plane target is used to simulate the laser radar scanning detection on a straight line, and the relative placement position of the plane target and the laser radar is shown in Figure 4 To enhance accuracy, 8 different positions are selected to record the corresponding laser radar echo signals, and the comparison before and after the laser radar intensity information correction of each target position is shown in Figure 5 After correction, the average intensity of the laser radar of different targets tends to be uniform, and the standard deviation of the average intensity of different targets changes from 0.1163V to 0.0124V after correction.

[0077] The above embodiments only express several implementation ways of the present application, and the description is more specific and detailed, but it should not be understood as a limitation to the patent scope of the application. It should be pointed out that for ordinary skilled in the art, without departing from the concept of the present application, several modifications and improvements can be made, which are all within the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A method for extracting and correcting lidar echo intensity information, characterized in that, Includes the following steps: Step S1: Filter and reduce noise from the entire waveform of the digital lidar echo signal; Step S2: Use the integral center method to extract the intensity information of the filtered and denoised lidar echo signal, and store the intensity information of each detection point according to the lidar detection sequence; Step S3: Deconvolve the filtered and denoised lidar echo signal to reconstruct the lidar impact response, extract the distance information of the detection point based on the lidar impact response, and invert and calculate the tilt angle of the local micro-element of the detection point. Step S4: Generate a lidar intensity point cloud map based on the intensity information described in step S2, the distance information described in step S3, and the scanning angle information recorded in the lidar detector. Use the K-nearest neighbor algorithm to calculate the normal vector of the point cloud map and correct the point cloud normal vector according to the tilt angle of the local micro-surface element of each detection point described in step S3. Step S5: Calculate the incident angle of the lidar based on the point cloud normal vector, point cloud spatial coordinates and lidar detector spatial coordinates calculated in step S4. Establish a lidar intensity information correction model based on the lidar detection distance and incident angle and generate a corrected lidar intensity point cloud map.

2. The method for extracting and correcting lidar echo intensity information according to claim 1, characterized in that, In step S1, the Savitzky-Golay filtering method is used to filter and reduce noise in the digital lidar echo signal.

3. The method for extracting and correcting lidar echo intensity information according to claim 1, characterized in that, The specific formula for extracting the intensity information of the filtered and denoised lidar echo signal using the integral centroid method in step S2 is as follows: Among them, P lidar (t) represents the LiDAR echo signal after filtering and noise reduction, N represents the number of valid sampling points of the LiDAR echo signal, and t represents the duration of the LiDAR echo signal.

4. The method for extracting and correcting lidar echo intensity information according to claim 1, characterized in that, The specific formula for reconstructing lidar echo signals using deconvolution is as follows: In the formula, H is the matrix form of the lidar impact response, Y is the matrix form of the lidar echo, and X is the fuzzy matrix composed of the laser emission waves. For the objective function The lidar shock response matrix at which the minimum value is taken is the final result of deconvolution reconstruction; It is the sum of squared residuals, ||H||1 is the L1 regularization term, ||H||2 is the L2 regularization term, μ is the regularization coefficient, and χ is the mixing parameter used to balance the L1 and L2 regularization terms.

5. The method for extracting and correcting lidar echo intensity information according to claim 4, characterized in that, The specific method for extracting distance information based on the impulse response of lidar is as follows: The peak position of the lidar impact response is extracted as the timing end time t2, and the peak position of the lidar emitted pulse is taken as the timing start time t1. The distance information is calculated as follows: d = c * (t2 - t1) / 2, where c is the speed of light.

6. The method for extracting and correcting lidar echo intensity information according to claim 4, characterized in that, The specific method for calculating the tilt angle of the local micro-element at the detection point based on the impact response inversion of lidar is as follows: Establish the relationship between the tilt angle α of the detected object plane and the full width at half maximum (FWHM) ω of the lidar impulse response. c Relationship: In the formula, k = ω c / θd, where θ is the divergence angle of the lidar and d is the detection range of the lidar; Based on the above formula, in the subsequent calculation of the lidar point cloud normal vector, the half-peak width ω of the lidar impulse response during actual detection should be used. c The tilt angle α of the local micro-surface element at the detection point was calculated by inversion.

7. The method for extracting and correcting lidar echo intensity information according to claim 1, characterized in that, The specific method for calculating the normal vector of the point cloud using the K-nearest neighbor algorithm and correcting the point cloud normal vector according to the tilt angle of the local micro-element at each probe point is as follows: Select K neighboring points around the calculation point, and use least squares fitting to find the plane with the smallest sum of the projections of all points. Use the normal vector of this plane as the normal vector of the point. Repeat the above process for all point clouds until the normal vectors of all detection points are obtained. Calculation; When the slope of the target surface changes beyond a threshold, the point cloud normal vector is corrected using the tilt angle of the local micro-element at each detection point. The specific method is as follows: Compare the tilt angle of the K-nearest neighbor fitted surface with the tilt angle of the local micro-surface element at the probe point. When the difference between the two is below a threshold, retain the normal vector calculated by the K-nearest neighbor method. When the difference between the two exceeds a threshold, the tilt angle of the local micro-element of the probe point is used as the tilt angle of the fitting plane to calculate the normal vector of the point cloud.

8. The method for extracting and correcting lidar echo intensity information according to claim 1, characterized in that, The incident angle θ of the lidar i The calculation formula is: In the formula, cos -1 It is an inverse cosine function. laser incident vector (x, y, z) represents the spatial coordinates of the detection point, and (x0, y0, z0) represents the spatial coordinates of the lidar. Let be the normal vector of the point cloud.

9. The method for extracting and correcting lidar echo intensity information according to claim 8, characterized in that, The lidar intensity information correction model is specifically as follows: Among them, I lidar Here, θ represents the intensity information of the lidar echo extracted using the integral centroid method, d represents the detection range extracted based on deconvolution reconstruction, and θ represents the intensity information of the lidar echo. i η is the actual incident angle of the lidar, C is the fitting parameter, and θ0 and d0 are the set reference incident angle and reference detection distance, respectively.

Images