Underwater complex target laser scanning point cloud simulation method based on ray tracing

Abstract

The simulation method of underwater complex target laser scanning point cloud based on ray tracing belongs to the technical field of computer simulation, mainly solving the problems of boundary effect and multiple reflection effect of light and target intersection which are not considered in existing methods. Firstly, the transmission process of each light in water is simulated according to the Mont-Carlo random scattering theory; secondly, the intersection of light and triangular mesh and the outer normal direction of the mesh are obtained by traversing the target surface mesh; then, the improved BRDF model is used to sample the reflected light multiple times to obtain the light parameters after multiple reflections on the target; next, the physical parameters of the light reaching the receiving field after scattering in the water again are obtained by using the Mont-Carlo random scattering theory, and the echo signal of the single point laser irradiating the target is obtained by counting the light intensity and optical path reaching the detector; then, the distance and intensity of the target are obtained by using the echo signal processing method; finally, the three-dimensional laser scanning point cloud data of the target are obtained by traversing the scanning path.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of computer simulation, and relates to a three-dimensional point cloud simulation method for underwater target laser radar, and is specifically realized based on a water Mont-Carlo random light scattering algorithm and a light ray tracking algorithm, and solves problems such as a target boundary region laser backscattering echo signal and multiple scattering caused by a complex target boundary effect to a certain extent, and can be used for laser radar scattering characteristic simulation analysis of an underwater target. BACKGROUND

[0002] With the progress of laser technology, the laser radar technology not only rapidly develops and is applied in fields such as air missile guidance and target detection, but also plays an important role in detecting and developing marine resources and accurately identifying small underwater targets as a visual measurement system of an underwater robot. At present, the robot laser radar visual system development technology applicable to underwater target detection is not mature enough in China, and a method of theoretical modeling is often used to estimate a laser radar scanning scene. Accurately establishing a laser radar simulation method for an underwater complex target has a guiding significance for the development of an underwater robot visual system based on a laser radar and three-dimensional reconstruction of an underwater target.

[0003] Most of the existing target laser scattering models are for air targets. For example, a method based on GPU programming by Zhang Jianqi and Huang Xi uses a target 3DMAX mesh model to realize laser radar 8-bit grayscale image simulation based on a light ray tracking method. However, the laser spot size is not considered in the simulation process, so that the imaging effect in a complex target surface region is greatly different from the actual situation. The actual modeling process is not given for the problem of multiple scattering. For the underwater target laser radar simulation model, Ma Yi uses a laser echo signal equation to establish a water target laser radar echo signal simulation model. Tan Yajun uses a Mont-Carlo random scattering theory to establish a water target laser echo signal simulation model. However, the three-dimensional geometric structure of the target is not considered in the modeling process, so that it is difficult to realize laser radar imaging simulation. SUMMARY

[0004] The marine environment is a strong scatterer. The spot size is expanded to a certain extent after the laser is transmitted for a distance. When the spot reaches the target, the size is close to that of a small target, especially the surface size of a complex target. When the laser is scanned to the boundary region of the target surface, on the one hand, the laser cannot completely irradiate the surface, so that the target echo signal is weakened. On the other hand, the light after the first reflection can be reflected again on the target surface, so that the echo signal of the scanned point is delayed. The above two factors will affect the discrimination of the three-dimensional point distance and intensity of the complex target surface. The present application fully considers the spot size after the scattering of the water body, the three-dimensional geometric structure of the complex target and the influence of the multiple scattering light on the target surface, and improves the underwater target laser scattering simulation precision.

[0005] The simulation method of underwater complex target laser scanning point cloud based on ray tracing utilizes the ray tracing technology to simulate the whole physical process of the transmission of laser rays in water, the discrimination of laser rays irradiating the target area and the calculation of reflection intensity, the transmission of rays in water again, and the target echo signal received by the detector, obtains the target position coordinates and echo intensity in combination with signal detection and scanning path parameters, encodes the target position coordinates and echo intensity, and obtains the.pcd format target point cloud simulation data.

[0006] 1. Randomly generating N rays according to the waist radius and divergence angle of the scanning laser beam, recording the starting position and transmission direction vector of the rays, and setting the intensity value of each ray to 1;

[0007] 2. Inputting the absorption coefficient and scattering coefficient of water, simulating the transmission process of each ray in water by using the Mont-Carlo random scattering theory, and recording the optical path, position, transmission direction vector and intensity value of the ray when the distance between the ray position and the target is less than the unit attenuation scale;

[0008] 3. Target surface scattering spot area judgment. Introducing the triangular mesh of the target surface, traversing the target surface mesh for each ray, solving and judging whether each ray intersects with the triangular mesh, if not, the ray parameters are temporarily unchanged; if yes, recording the intersection coordinates of the ray and the surface element index, finding the three node coordinates of the surface element through the surface element index, and then obtaining the outer normal vector of the surface element for the next operation;

[0009] 4. Traversing the rays intersecting with the target, obtaining the incidence angle of the rays according to the position, transmission direction vector and outer normal vector of the surface element, combining the improved BRDF model, and using the Phong reflection ray random sampling method to randomly sample the reflection rays, obtaining the scattering angle and azimuth angle of all rays on the target surface, and then obtaining the reflection ray vector, and the intensity of the ray after each reflection is multiplied by the hemispherical reflectivity under the incidence angle;

[0010] 5. Continue to traverse all rays, repeat multiple times according to the methods of 3-4, and solve the scattering ray optical path, position, transmission vector and intensity after multiple reflections on the target;

[0011] 6. According to the solving method of step 2, the scattering rays again pass through the water scattering to reach the receiving field of view, the intensity and optical path of the rays reaching the detector are counted, the echo signal of the single point laser irradiating the target is obtained, and the distance and scattering intensity of the target under the single point laser are generated by using the echo signal processing method;

[0012] 7 Repeat 1-6, using the target distance and intensity parameters obtained by echo signal processing, combined with the scanning path parameters, through the encoding of all scanned point coordinates and intensity data, obtain the target point cloud data in the format of PCL point cloud library readable Point_XYZI type *.pcd. BRIEF DESCRIPTION OF DRAWINGS

[0013] Figure 1 Method flow chart of the present application;

[0014] Figure 2 Target surface laser spot scattering area judgment result sample; Figure 2 The middle part contains the incident light direction and the target surface triangular mesh surface element, and the intersection point of the incident light and the target surface. DETAILED DESCRIPTION

[0015] Reference Figure 1 , the specific implementation process of the present application is as follows:

[0016] 1 According to the waist radius ω0, the divergence angle α of the scanning laser beam, randomly generate N light rays, record the starting position (x_0 1...N , y_0 1...N , z_0 1...N ), transmission direction vector (tx_0 1...N , ty_0 1...N , tz_0 1...N ) of the light rays, and set the intensity value of each light ray to 1 (Q_0 1...N =1);

[0017] 2 Input the absorption coefficient A and the scattering coefficient C of the water body, use the Mont-Carlo random scattering theory to simulate the transmission process of each light ray in the water body, when the light ray position and the target distance are less than the unit attenuation scale, record the optical path (L_1 1...N ), position (x_1 1...N, y_1 1...N, z_1 1...N ), transmission direction vector (tx_1 1...N , ty_1 1...N , tz_1 1...N ) and intensity value (Q_1 1...N ) of all light rays after scattering through the water body;

[0018] 3 Target surface laser spot scattering area judgment, typical judgment result as shown in Figure 2

[0019] ​3a import the target surface triangle mesh file, traverse the target mesh for each ray, use the space ray and triangle intersection theory to determine whether the ray can intersect with the surface element, if intersected, the nearest intersection point of the ray and the target surface can be determined, if not intersected, the transmission vector and the transmission position are temporarily unchanged;

[0020] Assuming that the target intersected ray is N1, the non-intersected ray space parameters are: optical path (L_2 1...N1 ), position (x_2 1…N1 , y_2 1...N1 , z_2 1...N1 ), transmission direction vector (tx_2 1...N1 , ty_2 1...N1 , tz_2 1...N1 ) and intensity value (Q_2 1...N1 );

[0021] 3b record the index of the ray intersected with the target and the intersection point coordinates of the ray and the target, find the three node coordinates of the surface element through the surface element index, and then get the outer normal vector of the surface element (n i，x , n i,y, n i，z );

[0022] 4 calculate the scattering light transmission related parameters according to the improved BRDF model

[0023] 4a improved BRDF model, the target surface BRDF is divided into mirror reflection and diffuse reflection, the expression is:

[0024]

[0025] Among them, f spec is the mirror reflection part of BRDF; f diff is the diffuse reflection part of BRDF; (F(n, k, β)) = ρ I +(1-ρ I )(1-cosβ) 5 is the Fresnel function; is a function related to the rough slope of the target surface, v is the root mean square slope; is the shadow factor; θ in is the incident angle of the ray, θ re is the reflection angle, is the incident azimuth angle, is the reflection azimuth angle; k s , k d, ρ I and c are to be measured, and k s +k d= 1, by measuring and function fitting, the target surface BRDF function f can be obtained r ;

[0026] 4b According to the light transmission vector (tx_1, ty_1, tz_1) and the facet normal vector (nx, ny, nz), the incident angle θ of the light is solved x , n y , n z ), the incident angle θ in is obtained, and the mirror reflection and diffuse reflection probability density function under the incident angle is obtained

[0027]

[0028] For any reflection angle θ re , there is a probability density function corresponding to it, and after normalization, the mirror reflection and diffuse reflection probability density space {P spec} = {p spec (θ re = -π / 2,..., π / 2)} and {P diff} = {p diff (θ re = -π / 2,..., π / 2)} can be established

[0029] 4c The reflection angle and reflection azimuth are sampled by using a similar Phong light sampling model, first set random numbers [u1, u2, u3] ∈ [0, 1];

[0030] When u1 < k d , find the θ diff corresponding to the nearest value of u2 in {P re} as the reflection angle sampling value, and the reflection zenith angle is randomly 2πu3;

[0031] When k d ≤ u1 ≤ 1, find the θ re corresponding to the nearest value of u2 in {P spec} as the reflection angle sampling value, and the reflection zenith angle is randomly 2πu3;

[0032] The reflected light intensity is the incident light intensity multiplied by the hemispherical reflectivity under the incident angle;

[0033] 5 Solve the light parameter space after multiple reflections of the target surface

[0034] 5a Traverse all the light rays intersecting the target, and obtain the reflected light vector and reflected light intensity through the reflection angle and reflection zenith angle random sampling results, while the optical path and position of the reflected light are unchanged;

[0035] 5b Continue to traverse the target intersecting light, according to the method of (3)-(4) repeated several times, after multiple reflections, the scattered light parameters are obtained;

[0036] Finally, the spatial parameters of the scattered light containing multiple reflections are obtained: optical path (L_2N 1...N ), position (x_2 N1...N , y_2 N1...N , z_2 N1...N ), transmission direction vector (tx_2 N1...N , ty_2 N1...N , tz_2 N1...N ) and intensity value (Q_2 N1...N );

[0037] 6 Add the light space parameters obtained in step 3a which do not intersect the target and the light space parameters obtained in step 5 which intersect the target to the input parameters of the echo signal transmission; again use the Mont-Carlo water scattering theory in 2 to obtain all the light parameters reaching the receiving field of view.

[0038] Statistical light intensity and optical path reaching the detector to obtain the echo signal after single point laser irradiation of the target, and use the echo signal processing method to generate the distance and scattering intensity of the target under single point laser;

[0039] 7 Repeat 1-6, use the target distance and intensity parameters obtained by echo signal processing, combine the scanning path parameters, and through encoding all the scanned point coordinates and intensity data, obtain the *.pcd format target point cloud data of PointXYZI type which can be read by PCL point cloud library.

Claims

1. A method for simulating a laser scanning point cloud of an underwater complex target based on ray tracing, characterized in that: The whole physical process of the transmission of laser light in the water body, the judgment of the irradiation of the target region by the laser light, the calculation of the reflection intensity, and the transmission of the light in the water body again is simulated by using the light tracking technology to obtain all the light parameters reaching the detector; all the light is traversed, and the position coordinates and scattering intensity of the target under the scanning path are obtained by combining the relevant parameters of the detector, and the steps are as follows: Step 1: According to the waist radius ω0 and the divergence angle α of the scanning laser beam, N light rays are randomly generated, and the initial light parameters are recorded; Step 2: The initial light is scattered in the water body for the first time; The absorption coefficient A and the scattering coefficient C of the water body are input, and the transmission process of each light ray in the water body is simulated by using the Mont-Carlo random scattering theory, and when the position of the light ray is less than the unit attenuation scale from the target, all the light parameters after scattering through the water body are recorded; Step 3: Judgment of the laser spot irradiation area on the target surface 3a: Import the target surface triangular mesh file, traverse the target mesh for each light ray, and use the space ray and triangle intersection theory to judge whether the light ray can intersect with the surface element, if it intersects, the nearest intersection point of the light ray and the target surface can be obtained, if it does not intersect, the transmission vector and the transmission position are temporarily unchanged; 3b: Record the light ray index intersecting with the target, the intersection point coordinates of the light ray and the target, and the face element index of the intersection point, find the three node coordinates of the face element through the face element index, and then obtain the outer normal vector of the face element; Step 4: According to the improved BRDF model, the reflection angle and the azimuth angle of the reflected light and the normal vector are randomly sampled, and the transmission related parameters of the scattered light are calculated; Step 5: Solve the light parameter space after multiple reflections on the target surface 5a: Traverse all the light rays intersecting with the target, and obtain the reflected light vector and the reflected light intensity through the random sampling results of the reflection angle and the reflection zenith angle, while the optical path and the position of the reflected light remain unchanged; 5b: Continue to traverse the light rays intersecting with the target, and repeat multiple times according to the method of step 3-step 4 to obtain the reflected light parameters after multiple reflections; Step 6: Add the light space parameters obtained in step 3a which do not intersect with the target and the light space parameters obtained in step 5 which intersect with the target to the input parameters of the echo signal transmission; again use the Mont-Carlo water scattering theory in step 2 to obtain all the light parameters reaching the receiving field of view; The light intensity and the optical path reaching the detector are counted to obtain the echo signal after the single-point laser irradiation of the target, and the echo signal processing method is used to generate the distance and the scattering intensity of the target under the single-point laser; Step 7: Repeat steps 1-6, use the target distance and intensity parameters obtained by processing the echo signal, combine the scanning path parameters, and encode all the scanned point coordinates and intensity data to obtain the target point cloud data in the *.pcd format of the Point XYZI type which can be read by the PCL point cloud library.

2. The method of claim 1, wherein the method is characterized by, In step 4, the improved BRDF model is used to randomly sample the scattering angle and the azimuth angle of the scattered light and the normal, and the method includes three steps: Step 1: The improved BRDF model divides the BRDF of the target surface into specular reflection and diffuse reflection, and the expression is as follows: where f spec is the BRDF specular reflection part; f diff is the BRDF diffuse reflection part; (F(n, k, sβ)) = p I + (1 - p I )(1 - cosβ) 5 is the Fresnel function; is a function related to the roughness slope of the target surface, v is the root mean square slope; is a shadowing factor; θ in is the light incident angle, θ re is the reflection angle, is the incident azimuth angle, is the reflection azimuth angle; k s , k d , p I and c are to be measured, and k s + k d = 1, by measuring and function fitting, the BRDF function f r of the target surface can be obtained. Step 2: Solve the incident angle θ of the ray according to the ray transmission vector (tx_1, ty_1, tz_1) and the face element normal vector (nx, ny, nz) x , nx y , ny z , nz in , and get the mirror reflection and diffuse reflection probability density functions under the incident angle For any reflection angle θ re Each of them has a probability density function corresponding to it, and after normalization, the specular reflection and diffuse reflection probability density spaces {P spec} = {p spec (θ re = -π / 2,..., π / 2)}, {P diff} = {p diff (θ re = -π / 2,..., π / 2)}; Step 3: Similar to the Phong lighting random sampling method, the reflection angle and the reflection azimuth angle are sampled, and three random numbers [u1, u2, u3] ∈ [0, 1] are set. When u1 <k d When, find {P diff The θ corresponding to the nearest value of u2 in} re These are sampled values ​​of the reflection angle, with the reflection zenith angle randomly set to 2πu³. When k d ≤ u1≤ 1, find the value of θ spec in {P re} corresponding to the value closest to u2, and set it as the reflection angle sample value, and the reflection zenith angle is randomly set as 2πu3; The intensity of the reflected light is the intensity of the incident light multiplied by the hemispherical reflectance at the incident angle.

Images