A target point cloud optimization method based on spot-sphere model

Abstract

The invention discloses a target point cloud optimization method based on a spot-spherical model, and the specific steps are: taking the angle between an incident light ray and a line connecting a scanner to a sphere center, the comparison between the area of ​​a spot and a target sphere, and the emission of the spot as independent variables, constructing a spot area model under different conditions; constructing a theoretical echo energy model in combination with an echo energy model and a spot area model; calculating a theoretical echo intensity, establishing a relationship model between an actual echo intensity and a theoretical echo intensity, and screening out a theoretical point cloud that meets the actual distribution of a target sphere; combining the screened theoretical point cloud coordinates that meet the actual distribution of a target sphere, constructing a target sphere model, and establishing a parametric equation about the coordinates of the sphere center and the radius; performing parameter linearization processing on the constructed parametric equation; using a least squares fitting method to process the parametric equation after the parameter linearization processing, so as to calculate the optimized target sphere parameters; and improving the accuracy, reliability and efficiency of point cloud registration.

Description

Technical Field

[0001] The present invention belongs to the technical field of laser radar data processing, and more specifically, relates to a target point cloud optimization method based on a light spot-spherical model. Background Art

[0002] In ground-based 3D laser scanning technology, high-precision scanning of large scenes or fixed targets is a key task. In this process, it is necessary to set up multiple scanning stations at different locations and then stitch the data together. Among them, although the automatic stitching algorithm can achieve the stitching of point cloud data to a certain extent, they have high requirements on the initial posture of the point cloud, and the efficiency of registration of tens of millions of point clouds is low; therefore, using target spheres for multi-station registration is a common practice in engineering measurement. However, due to the size of the target sphere, the distance between the scanner and the target sphere should not be too far, otherwise the target sphere fitting failure will occur.

[0003] In order to solve the problem of long-distance and large-scale point cloud registration, the existing technical means provide solutions by increasing the number of stations, reducing the distance between the target ball and the scanner, and not using the target ball as the key point for stitching. However, as the scanning distance becomes longer, the scanner accuracy becomes higher, and the scale of the target to be stitched becomes larger, the automatic registration method becomes inefficient and has high requirements for the initial posture. Summary of the invention

[0004] In view of the above defects or improvement needs of the prior art, the present invention aims to solve the problem of efficiency and accuracy in point cloud registration and splicing in the prior art.

[0005] In order to achieve the above object, the present invention provides a target point cloud optimization method based on a spot-spherical model, comprising the following steps:

[0006] S10, taking the angle between the incident light and the line connecting the scanner to the center of the sphere, the comparison between the area of ​​the light spot and the target sphere, and the emission of the light spot as independent variables, construct a light spot area model under different conditions;

[0007] S20, constructing a theoretical echo energy model by combining the echo energy model with the spot area model on the target sphere;

[0008] S30, calculating the theoretical echo intensity of each point, establishing a relationship model between the actual echo intensity and the theoretical echo intensity, and then screening out a theoretical point cloud that conforms to the actual distribution of the target sphere;

[0009] S40, combining the theoretical point cloud coordinates obtained in step S30 and meeting the actual distribution of the target ball, constructing a target ball model and establishing the coordinates of the ball center x 0 , y 0 ,z 0 and radius R 0 parametric equations;

[0010] S50, performing parameter linearization processing on the parameter equation constructed in step S40;

[0011] S60, using the least squares fitting method to process the parameter equation after the parameter linearization processing in step S50, so as to calculate the optimized target ball center coordinates x 0 , y 0 , z 0 and radius R 0 parameter.

[0012] Furthermore, constructing the spot area models under different conditions in S10 includes:

[0013] When the spot area is smaller than the target area, the spot area model on the target is:

[0014] When shooting straight:

[0015] ,

[0016] In the formula, A s is the spot area on the target sphere, D is the spot diameter, , is the angular resolution of the scanner, R is the distance from the target to the radar;

[0017] When it is not orthogonal, the light spot does not emit:

[0018] ,

[0019] In the formula, is the length of the major semi-axis after deformation;

[0020] When it is not orthogonal and the light spot is emitted:

[0021] ,

[0022] In the formula, k It is the ratio of the light spot area shining on the target ball.

[0023] Furthermore, constructing the spot area model under different conditions in S10 further includes: when the spot area is larger than the target sphere area, the spot area model on the target sphere is:

[0024] When the light spot is not emitted:

[0025] ,

[0026] In the formula, is the radius of the target sphere

[0027] When the light spot is emitted:

[0028] .

[0029] Further, in said S20,

[0030] For the diffuse reflectance For diffuse reflection targets, the echo energy model is:

[0031] ,

[0032] in, is the laser power reaching the detector, is the laser emission power, is the optical system efficiency, is the effective receiving area of ​​the optical lens, is the angle between the normal of the illuminated part and the incident light, is the atmospheric attenuation coefficient, R is the distance from the target to the radar;

[0033] Combining the echo energy model and the spot area model on the target sphere, the theoretical echo intensity model is obtained:

[0034] ,

[0035] In the formula, is the theoretical area of ​​the light spot at the corresponding distance, P rl is the theoretical echo value.

[0036] Furthermore, the method for constructing the relationship model in S30 is:

[0037] The theoretical echo intensity was calculated and the actual echo intensity data was measured. The two sets of values ​​were compared and analyzed to establish a relationship model between the theoretical echo intensity and the actual echo intensity data regarding the spot area, distance, divergence angle and laser intensity.

[0038] Furthermore, in step S40, the target sphere model is constructed by combining the theoretical point cloud coordinates obtained in step S30 that conform to the actual distribution of the target sphere, and establishing the coordinates of the sphere center. x 0 , y 0 , z 0 and radius R 0The parametric equation method is:

[0039] The Gauss-Newton method is used to iteratively solve the coordinates of the target sphere center, and the equation of the space sphere is expanded to obtain the formula:

[0040] ,

[0041] ,

[0042] Construct a least squares model, where ( x , y , z ) is the theoretical point cloud coordinate that conforms to the actual distribution of the target sphere, and the parameter is required to be the sphere center coordinate x 0 , y 0 , z 0 and radius R 0 .

[0043] Further, in the step S50, the method of performing parameter linearization processing on the parameter equation constructed in the step S40 is:

[0044] make , , , , then the parameters to be determined become A, B, C, D, and we can get the following by substituting the parameters:

[0045] x 2 + y 2 + z 2 - Ax - By - C + D =0.

[0046] Furthermore, the specific method of step S60 is:

[0047] The least squares series method is used to obtain the target sphere parameters. The idea of ​​least squares fitting is to ensure that the residual is minimized, that is, to satisfy:

[0048] ,

[0049] Among them, n is the total number of target points involved in fitting, i is the serial number of the target point involved in fitting, and x i ,y i , z i are the horizontal, vertical and vertical coordinates of the target point involved in the fitting, and i=1, 2, …, n;

[0050] After taking partial derivatives of the above formula and transforming it into a matrix, we get:

[0051] ,

[0052] Multiply both sides of the above equation by the left Then we get:

[0053] ,

[0054] Then we get the following formula and calculate the parameters A, B, C, and D:

[0055] ,

[0056] Iterate the calculation parameters A, B, C, D to get the target ball parameters x 0 , y 0 , z 0 ,R 0 .

[0057] In general, the above technical solutions conceived by the present invention can achieve the following beneficial effects compared with the prior art:

[0058] 1. The target point cloud optimization method based on the spot-spherical model of the present invention improves the accuracy of point cloud registration by constructing spot area models under different conditions and combining the actual echo intensity to construct a relationship model.

[0059] 2. The target point cloud optimization method based on the light spot-sphere model of the present invention improves the accuracy and reliability of point cloud matching by setting multiple independent variables such as the angle between the incident light and the line connecting the scanner to the center of the sphere, the comparison between the light spot and the target sphere area, and the light spot emission situation, and correcting the relationship model again through the target sphere parameters.

[0060] 3. The target point cloud optimization method based on the spot-sphere model of the present invention can quickly screen out the theoretical point cloud that conforms to the actual distribution of the target sphere by establishing a relationship model between the actual echo intensity and the theoretical echo intensity, thereby improving the efficiency of the alignment. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 A schematic diagram of a flow chart of a target point cloud optimization method based on a spot-sphere model according to an embodiment of the present invention;

[0062] Figure 2 A schematic diagram of the relationship between the light spot of a target point cloud optimization method based on a light spot-spherical model according to an embodiment of the present invention, in which the light spot is smaller than the target sphere and the light spot and the target sphere when orthogonally incident;

[0063] Figure 3 A schematic diagram of the relationship between the light spot and the target sphere when the light spot of a target point cloud optimization method based on a light spot-spherical model according to an embodiment of the present invention is smaller than the target sphere, is non-orthogonal, and the light spot is not emitted;

[0064] Figure 4 A schematic diagram of the relationship between the light spot and the target sphere when the light spot is smaller than the target sphere, non-orthogonal, and the light spot is emitted according to a target point cloud optimization method based on a light spot-spherical model according to an embodiment of the present invention;

[0065] Figure 5 A schematic diagram of the relationship between a light spot and a target sphere when a light spot of a target point cloud optimization method based on a light spot-spherical model according to an embodiment of the present invention is larger than a target sphere and the light spot is not emitted;

[0066] Figure 6 This is a schematic diagram of the relationship between the light spot and the target sphere when the light spot is larger than the target sphere and when the light spot is emitted in a target point cloud optimization method based on a light spot-spherical model according to an embodiment of the present invention.

[0067] In all the drawings, the same reference numerals represent the same technical features, specifically: 1-target ball, 2-light spot. DETAILED DESCRIPTION

[0068] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0069] For long-distance and large-scale point cloud registration, the existing technical means provide solutions by increasing the number of stations, reducing the distance from the target ball to the scanner, and not using the target ball as the key point for stitching. As the scanning distance becomes longer and longer, the scanner accuracy becomes higher and higher, and the scale of the target to be stitched becomes larger and larger, the automatic registration method is inefficient and has high requirements for the initial posture. A reliable method that can use a spherical target as a key point for registration at a long distance is needed. This patent aims to solve the contradiction between efficiency and accuracy in large-scale point cloud registration and stitching.

[0070] Please refer to Figure 1 The present invention relates to a target point cloud optimization method based on a light spot-spherical model, comprising the following steps:

[0071] S10, taking the angle between the incident light and the line connecting the scanner to the center of the sphere, the comparison between the area of ​​the light spot and the target sphere, and the emission of the light spot as independent variables, construct a light spot area model under different conditions;

[0072] Generally speaking, the laser spot directly incident on the plane is a perfect circle. However, as the angle between the incident light and the line connecting the scanner to the center of the sphere increases, the spot is no longer a perfect circle, but a gradually elongated ellipse. When the spot partially exceeds the target sphere, its reflection area is the intersection of an elongated ellipse and the perfect circle of the target sphere. When the spot is larger than the target sphere, there are two cases to be discussed: one is that the spot completely covers the target sphere, and the other is that the spot does not completely cover the target sphere.

[0073] During scanning, when the distance between the scanner and the target ball is close, the point cloud density of the target ball is high, and even if there is a small amount of noise, it can meet the needs of point cloud registration and splicing in engineering measurement. Based on the above analysis, when the distance between the scanner and the target ball is far, according to the comparison of the size of the light spot and the target ball area, it is divided into two models:

[0074] When the spot area is smaller than the target area, the spot area model on the target is:

[0075] Please refer to Figure 2 , when orthophoto:

[0076] ,

[0077] In the formula, A s is the spot area on the target sphere, D is the spot diameter, , is the angular resolution of the scanner, R is the distance from the target to the radar;

[0078] Please refer to Figure 3 , when it is not orthogonal, the light spot has no emission:

[0079] ,

[0080] In the formula, is the length of the major semi-axis after deformation;

[0081] Please refer to Figure 4 , when it is not orthogonal and the light spot is emitted:

[0082] ,

[0083] In the formula, k It is the ratio of the light spot area shining on the target ball.

[0084] When the spot area is smaller than the target area, the spot area model on the target is:

[0085] Please refer to Figure 5 , when the light spot is not emitted:

[0086] ,

[0087] In the formula, is the radius of the target sphere

[0088] Please refer to Figure 6 , when the light spot is emitted:

[0089] .

[0090] S20, constructing a theoretical echo energy model by combining the echo energy model with the spot area model on the target sphere;

[0091] For the diffuse reflectance For diffuse reflection targets, the echo energy model is:

[0092] ,

[0093] in, is the laser power reaching the detector, is the laser emission power, is the optical system efficiency, is the effective receiving area of ​​the optical lens, is the angle between the normal of the illuminated part and the incident light, is the atmospheric attenuation coefficient, R is the distance from the target to the radar;

[0094] After determining the area of ​​the light spot on the target, the theoretical echo intensity model is obtained by combining the echo energy model:

[0095] ,

[0096] In the formula, is the theoretical area of ​​the light spot at the corresponding distance, P rl is the theoretical echo value.

[0097] S30, calculating the theoretical echo intensity of each point, establishing a relationship model between the actual echo intensity and the theoretical echo intensity, and then screening out a theoretical point cloud that conforms to the actual distribution of the target sphere;

[0098] By establishing a theoretical model of the theoretical echo intensity and the distribution of the point cloud spot on the target sphere surface, and screening the point cloud based on the model, it is possible to improve the existing algorithm, which has a large measurement error and cannot be applied to engineering measurement when the distance between the target sphere and the scanner is too large. The construction method of the relationship model is specifically as follows:

[0099] The theoretical echo intensity was calculated and the actual echo intensity data was measured. The two sets of values ​​were compared and analyzed to establish a relationship model between the theoretical echo intensity and the actual echo intensity data regarding the spot area, distance, divergence angle and laser intensity.

[0100] S40, combining the theoretical point cloud coordinates obtained in step S30 and meeting the actual distribution of the target ball, constructing a target ball model and establishing the coordinates of the ball center x 0 , y 0 , z 0 and radius R 0 parametric equations;

[0101] The Gauss-Newton method, a classic iterative optimization algorithm, is used to solve the problem of solving the center coordinates of the target ball in three-dimensional space. First, the equation of the space ball is expanded to obtain the calculation formula:

[0102] ,

[0103] Construct a least squares model, where ( x , y , z ) is the theoretical point cloud coordinate that conforms to the actual distribution of the target sphere, and the parameter is required to be the sphere center coordinate x 0 , y 0 , z 0 and radius R 0 .

[0104] S50, performing parameter linearization processing on the parameter equation constructed in step S40;

[0105] make , , , , then the parameters to be determined change from target sphere parameters to A, B, C, D, and the parameters are substituted into: (where at least four known points are required to fit the sphere parameters)

[0106] x 2 + y 2 + z 2 - Ax - By - C + D =0.

[0107] S60, using the least squares fitting method to process the parameter equation after the parameter linearization processing in step S50, so as to calculate the optimized target ball center coordinates x 0 , y 0 ,z 0 and radius R 0 parameter:

[0108] The least squares series method is used to obtain the target sphere parameters, that is, to find a set of parameters (i.e., the coordinates of the sphere center and the radius) so that the residual between the sphere described by these parameters and the actual observed point cloud data is minimized, that is, to satisfy:

[0109] ,

[0110] Among them, n is the total number of target points involved in the fitting, i is the serial number of the target point involved in the fitting, and x i ,y i , z i are the horizontal, vertical and vertical coordinates of the target point involved in the fitting, and i=1, 2, …, n;

[0111] After taking partial derivatives of the above formula and transforming it into a matrix, we get:

[0112] ,

[0113] Multiply both sides of the above equation by the left Then we get:

[0114] ,

[0115] Then we get the following formula and calculate the parameters A, B, C, and D:

[0116] ,

[0117] Iterate the calculation parameters A, B, C, D to get the target ball parameters x 0 , y 0 , z 0 ,R 0 The target sphere parameters calculated by this method take into account the relationship between the light spot and the target sphere surface, overcoming the problem of large measurement error of the target sphere edge point at a long distance, which leads to large target sphere fitting error.

[0118] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A target point cloud optimization method based on a spot-sphere model, characterized in that: The following steps are involved: S10, taking the angle between the incident light and the line connecting the scanner to the center of the sphere, the comparison between the area of ​​the light spot and the target sphere, and the emission of the light spot as independent variables, construct a light spot area model under different conditions; S20, constructing a theoretical echo energy model by combining the echo energy model with the spot area model on the target sphere; S30, calculating the theoretical echo intensity of each point, establishing a relationship model between the actual echo intensity and the theoretical echo intensity, and then screening out a theoretical point cloud that conforms to the actual distribution of the target sphere; S40, combining the theoretical point cloud coordinates obtained in step S30 and meeting the actual distribution of the target ball, constructing a target ball model and establishing the coordinates of the ball center x 0 , y 0 , z 0 and radius R 0 parametric equations; S50, performing parameter linearization processing on the parameter equation constructed in step S40; S60, using the least squares fitting method to process the parameter equation after the parameter linearization processing in step S50, so as to calculate the optimized target ball center coordinates x 0 , y 0 , z 0 and radius R 0 parameter.

2. The target point cloud optimization method based on the spot-spherical model according to claim 1 is characterized in that: The light spot area models under different conditions constructed in S10 include: When the spot area is smaller than the target area, the spot area model on the target is: When shooting straight: ， In the formula, A s is the spot area on the target sphere, D is the spot diameter, , is the angular resolution of the scanner, R is the distance from the target to the radar; When it is not orthogonal, the light spot does not emit: ， In the formula, is the length of the major semi-axis after deformation; When it is not orthogonal and the light spot is emitted: ， In the formula, k It is the ratio of the spot area shining on the target ball.

3. The target point cloud optimization method based on the spot-spherical model according to claim 2 is characterized in that: The light spot area model constructed in different situations in S10 also includes: when the light spot area is larger than the target ball area, the light spot area model on the target ball is: When the light spot is not emitted: ， In the formula, is the radius of the target sphere When the light spot is emitted: 。 4. The target point cloud optimization method based on the spot-spherical model according to claim 3 is characterized in that: In the S20, For the diffuse reflectance For diffuse reflection targets, the echo energy model is: ， in, is the laser power reaching the detector, is the laser emission power, is the optical system efficiency, is the effective receiving area of ​​the optical lens, is the angle between the normal of the illuminated part and the incident light, is the atmospheric attenuation coefficient, R is the distance from the target to the radar; Combining the echo energy model and the spot area model on the target sphere, the theoretical echo intensity model is obtained: ， In the formula, is the theoretical area of ​​the light spot at the corresponding distance, P rl is the theoretical echo value.

5. The target point cloud optimization method based on the spot-spherical model according to claim 4 is characterized in that: The method for constructing the relationship model in S30 is: The theoretical echo intensity was calculated and the actual echo intensity data was measured. The two sets of values ​​were compared and analyzed to establish a relationship model between the theoretical echo intensity and the actual echo intensity data regarding the spot area, distance, divergence angle and laser intensity.

6. The target point cloud optimization method based on the spot-sphere model according to claim 5, characterized in that: In step S40, the target sphere model is constructed by combining the theoretical point cloud coordinates obtained in step S30 and matching the actual distribution of the target sphere. x 0 , y 0 , z 0 and radius R 0 The parametric equation method is: The Gauss-Newton method is used to iteratively solve the coordinates of the target sphere center, and the equation of the space sphere is expanded to obtain the formula: ， ， Construct a least squares model, where ( x , y , z ) is the theoretical point cloud coordinate that conforms to the actual distribution of the target sphere, and the parameter is required to be the sphere center coordinate x 0 , y 0 , z 0 and radius R 0 .

7. The target point cloud optimization method based on the spot-sphere model according to claim 6, characterized in that: The method for performing parameter linearization processing on the parameter equation constructed in step S40 in step S50 is: make , , , , then the parameters to be determined become A, B, C, D, and we can get the following by substituting the parameters: x 2 + y 2 + z 2 - Ax - By - C + D =0。 8. The target point cloud optimization method based on the spot-sphere model according to claim 7, characterized in that: The specific method of step S60 is: The least squares series method is used to obtain the target sphere parameters. The idea of ​​least squares fitting is to ensure that the residual is minimized, that is, to satisfy: ， Among them, n is the total number of target points involved in fitting, i is the serial number of the target point involved in fitting, and x i ,y i , z i are the horizontal, vertical and vertical coordinates of the target point involved in the fitting, and i=1, 2, …, n; After taking partial derivatives of the above formula and transforming it into a matrix, we get: ， Multiply both sides of the above equation by the left Then we get: ； Then we get the following formula and calculate the parameters A, B, C, and D: ， Iterate the calculation parameters A, B, C, D to get the target ball parameters x 0 , y 0 , z 0 ,R 0 .

Images