Three-dimensional image and laser data combined irregular object high-precision three-dimensional modeling method and system

Abstract

The invention discloses a stereoscopic image and laser data combined irregular object high-precision three-dimensional modeling technology. A modeling camera is used for collecting stereoscopic images, with the overlapping degree, of a sequence of a to-be-modeled target object; the accurate distance between the platform and the surface of the to-be-modeled target object is obtained through measurement data of a laser altimeter, and three-dimensional coordinates of a laser footprint in a world coordinate system are obtained according to a laser footprint coordinate calculation formula; a laser footprint camera is used for collecting laser spots of the laser footprint on the surface of the irregular to-be-modeled target object, and a laser footprint image coordinate extraction algorithm is used for obtaining image coordinates of the laser footprint; and processing the optical image of the to-be-modeled target object by using the bundle block adjustment model to obtain an accurate exterior orientation element of each image so as to realize high-precision three-dimensional modeling of the to-be-modeled target object. According to the method, the geometric accuracy of shape modeling is improved by combining stereoscopic images with a laser height measurement technology and integrating the advantages of two technical means of images and laser.

Description

Technical Field

[0001] The present invention relates to the fields of aerospace and surveying and mapping technology, and in particular to a high-precision three-dimensional modeling method and system for irregular objects combining stereoscopic images with laser data. Background Art

[0002] Existing shape modeling technologies are mainly aimed at rigid objects, which often have uniform shape features, which makes it possible for commonly used shape modeling technologies to achieve modeling purposes. However, in real life, there are many non-rigid objects with huge shape differences within the same category and no obvious geometric features. These make existing shape modeling technologies encounter two difficulties when modeling non-rigid objects: first, feature extraction cannot extract rich surface geometric features of irregular objects, and second, the geometric accuracy of the built shape model is low. For example, there are a large number of asteroids of various sizes in the solar system, which are irregular target objects. Three-dimensional shape modeling is the process of using three-dimensional production software to construct a model with three-dimensional data through a virtual three-dimensional space. Three-dimensional shape modeling is the process of converting objects or scenes in the real world into three-dimensional digital models.

[0003] In the field of 3D shape modeling, the existing technologies generally include stereophotogrammetry, SPC (Stereo-photo-clinometry) and LiDAR-laser 3D scanning. The above methods have the following technical defects:

[0004] (1) The geometric accuracy of the irregular shape model reconstructed by stereo photogrammetry is affected by the short baseline between the stereo image pairs. The geometric accuracy in the Z direction (elevation direction) is low and the accuracy is affected by the spatial division resolution. It is difficult to handle scenes with high accuracy and large scale.

[0005] (2) The photometric method is easily affected by uneven illumination and albedo models, has strict requirements on the environment, and requires the surface of the object to be measured to be a strictly smooth diffuse reflector Lambertian body. It is difficult to apply to objects with abrupt changes or complex reflective characteristics.

[0006] (3) The LiDAR system includes a laser altimeter and a 3D laser scanner. The disadvantage of the laser altimeter is that the point cloud density obtained is not high enough and is very sparse; the disadvantage of the laser 3D scanner is that the data volume is large, the data transmission resource usage is high, the scanning field of view is small, and the scanning distance needs to be close to the surface of the target object, which may cause damage to the target object or affect the measurement accuracy. Summary of the invention

[0007] The purpose of the present invention is to provide a high-precision three-dimensional modeling method and system for irregular objects combining stereo images and laser data. The method and system utilize stereo images combined with laser height measurement technology, and integrate the advantages of both imaging and laser technical means to overcome the shortcomings of existing methods such as low accuracy, large consumption of computing resources, and short observation distance requirements, thereby improving the geometric accuracy of shape modeling.

[0008] The objective of the present invention is achieved through the following technical solutions:

[0009] A high-precision three-dimensional modeling method for irregular objects combining stereo images and laser data, the method comprising:

[0010] Step 1: First, use a modeling camera to collect a sequence of overlapping stereo images of the target object to be modeled;

[0011] Step 2: Use the laser altimeter to measure the data to obtain the precise distance from the platform to the surface of the target object to be modeled, and obtain the three-dimensional coordinates of the laser footprint in the world coordinate system according to the laser footprint coordinate calculation formula;

[0012] Step 3, using a laser footprint camera to collect the laser spot of the laser footprint hitting the surface of the irregular target object to be modeled, and using a laser footprint image coordinate automatic extraction algorithm to obtain the image coordinates of the laser footprint;

[0013] Step 4: Based on the stereo image obtained in step 1 and the image coordinates of the laser footprint obtained in step 3, the optical image of the target object to be modeled is processed using the bundle method regional block adjustment model to obtain accurate exterior orientation elements of each image, thereby achieving high-precision three-dimensional modeling of the target object to be modeled.

[0014] A high-precision three-dimensional modeling system for irregular objects combining stereoscopic images and laser data, the system comprising:

[0015] A modeling camera, used to collect a sequence of overlapping stereo images of the target object to be modeled;

[0016] Laser altimeter, used to measure the precise distance from the data acquisition platform to the surface of the target object to be modeled, and obtain the three-dimensional coordinates of the laser footprint in the world coordinate system according to the laser footprint coordinate calculation formula;

[0017] The laser footprint camera is used to collect the laser spot of the laser footprint hitting the surface of the irregular target object to be modeled, and obtain the image coordinates of the laser footprint using the automatic extraction algorithm of the laser footprint image coordinates;

[0018] The 3D modeling processing module is used to process the optical image of the target object to be modeled based on the obtained stereo image and the image coordinates of the laser footprint using the bundle method regional block adjustment model to obtain the precise exterior orientation elements of each image, thereby realizing high-precision 3D modeling of the target object to be modeled.

[0019] It can be seen from the technical solution provided by the present invention that the above method and system utilize stereoscopic images combined with laser height measurement technology, comprehensively utilizing the advantages of both imaging and laser techniques, to overcome the shortcomings of existing methods such as low accuracy, large consumption of computing resources, and short observation distance requirements, thereby improving the geometric accuracy of shape modeling. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0021] Figure 1 A schematic flow chart of a method for high-precision three-dimensional modeling of irregular objects by combining stereoscopic images with laser data provided by an embodiment of the present invention;

[0022] Figure 2 A schematic diagram of the laser emission pointing angle according to an embodiment of the present invention;

[0023] Figure 3 The figure is a schematic diagram of the structure of the system according to the embodiment of the present invention. DETAILED DESCRIPTION

[0024] The following is a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments, which does not constitute a limitation of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0025] like Figure 1 The figure is a schematic flow chart of a method for high-precision three-dimensional modeling of irregular objects by combining stereoscopic images and laser data provided by an embodiment of the present invention, the method comprising:

[0026] Step 1: First, use a modeling camera to collect a sequence of overlapping stereo images of the target object to be modeled;

[0027] Step 2: Use the laser altimeter to measure the data to obtain the precise distance from the platform to the surface of the target object to be modeled, and obtain the three-dimensional coordinates of the laser footprint in the world coordinate system according to the laser footprint coordinate calculation formula;

[0028] In this step, a laser beam is emitted by a laser altimeter transmitter and then reflected by the surface of the target object to be modeled and received by a laser altimeter receiver;

[0029] Then calculate the time interval t between laser emission and reception, the propagation speed of light is c, then the laser one-way transmission distance ρ = (c*t) / 2; combine the position and attitude of the data acquisition platform obtained by the inertial measurement unit IMU installed on the data acquisition platform to obtain the three-dimensional coordinates of the laser footprint in the world coordinate system; in the specific implementation, the data acquisition platform includes a modeling camera, a laser altimeter, an inertial measurement unit IMU and a laser footprint camera;

[0030] like Figure 2 FIG. 1 is a schematic diagram of the laser emission pointing angle according to an embodiment of the present invention. When the laser is emitted, there is a certain angle between the pointing direction and the modeling camera coordinate system. Assume that the negative angle between the pointing direction of the laser emission and the Z axis of the modeling camera coordinate system is θ, and the angle between the projection on the XOY plane of the modeling camera coordinate system and the positive X axis is α;

[0031] If the laser ranging value is ρ, the coordinates of the laser emission point in the modeling camera coordinate system are:

[0032] ΔL=[L x ,L y ,L z ] T ,

[0033] Among them, L x Indicates the translation of the laser emission point along the X axis in the modeling camera coordinate system; L y Indicates the translation of the laser emission point along the Y axis in the modeling camera coordinate system; L z Indicates the translation of the laser emission point along the Z axis in the modeling camera coordinate system;

[0034] Then the three-dimensional coordinates of the laser footprint point in the modeling camera coordinate system are:

[0035]

[0036] Among them, Body represents the coordinate system of the laser emission point.

[0037] Step 3, using a laser footprint camera to collect the laser spot of the laser footprint hitting the surface of the irregular target object to be modeled, and using a laser footprint image coordinate automatic extraction algorithm to obtain the image coordinates of the laser footprint;

[0038] In this step, the SIFT operator is used to perform feature matching between the modeling image and the laser footprint image, where:

[0039] First, establish the DOG function Gaussian difference pyramid, expressed as:

[0040] L(x,y,σ)=G(x,y,σ)*I(x,y) (2)

[0041] D(x,y,σ)=G(x,y,kσ)-G(x,y,σ)*I(x,y) (3)

[0042] Where σ represents the scale space coordinate; k represents the ratio of Gaussian blur coefficients between adjacent layers; G(x, y, kσ) represents the Gaussian pyramid; I(x, y) represents the original image; L(x, y, σ) is defined as the convolution of a Gaussian function G(x, y, σ) with a varying scale and the original image I(x, y); D(x, y, σ) represents the Gaussian difference pyramid, and * represents the convolution calculation;

[0043] In order to improve the stability of key points, it is necessary to perform curve fitting on the scale space DOG function D(x, y, σ). The Taylor expansion (fitting function) of the DOG function in the scale space is:

[0044]

[0045] Where X = (x, y, σ) T , take the derivative and let the equation equal to zero to get the offset of the extreme point for:

[0046]

[0047] The corresponding extreme points are:

[0048]

[0049] Then, the feature matching function knnMatch based on the k-nearest neighbor algorithm is used for nearest neighbor matching, and the best matching point is screened out through the ratio test for K-nearest neighbor matching. K=2 is usually set, that is, the two points closest to each feature point are found, and the matching distance is set to the Euclidean distance d between the two points (x1, x2) and (y1, y2). 12 , expressed as:

[0050]

[0051] The best matching point is selected through the ratio test. If the ratio of the nearest neighbor distance to the next nearest neighbor distance is less than the set threshold (usually less than 0.75), the match is considered valid.

[0052] The random sample consensus algorithm RANSAC is used to calculate the homography matrix M, and the weighted average method is used to fuse the left image to the right image to generate a fused image;

[0053] In the fused image, the ellipse fitting algorithm is used to fit the edge of the laser footprint spot, and then the center of the ellipse is calculated as the image coordinates of the laser footprint. The image control points (ICPs) are constructed using the three-dimensional coordinates of the laser footprint in the world coordinate system and the corresponding image coordinates.

[0054] Step 4: Based on the stereo image obtained in step 1 and the image coordinates of the laser footprint obtained in step 3, the optical image of the target object to be modeled is processed using the bundle method regional block adjustment model to obtain accurate exterior orientation elements of each image, thereby achieving high-precision three-dimensional modeling of the target object to be modeled.

[0055] In this step, the initial value of the exterior orientation element comes from the prototype position and attitude obtained by the inertial measurement unit IMU, and is converted into the exterior orientation element of the modeling camera through the installation matrix; the exterior orientation element refers to the rotation (φ, ω, κ) and translation (X, Y, Z) parameters from the camera coordinate system to a certain spatial coordinate system of the object space;

[0056] The image coordinates of the laser footprint obtained in step 3 are used as control points and added into the bundle method block adjustment model. Under the rotation system with the Y axis as the main axis and the camera model based on perspective projection, the collinear relationship among the object point, image point and photography center is expressed by the collinear equation, which is:

[0057]

[0058] In the formula, (x, y) represents the image point coordinates; (x0, y0) represents the image principal point coordinates; (Δx, Δy) represents the distortion difference; (f x ,f y ) represents the principal distance; (X, Y, Z) represents the coordinates of the object point in the object coordinate system; (X S ,Y S ,Z S ) represents the coordinates of the photographic center in the object coordinate system;

[0059] In the formula, the definition Indicates the use of The rotation matrix of the ω,κ rotation system; represents the angle of rotation around the Y axis; ω represents the angle of rotation around the X axis; κ represents the angle of rotation around the Z axis; the parameters in the matrix R are:

[0060]

[0061] The specific expression of distortion difference (Δx, Δy) is as follows:

[0062] Δx=(x-x0)(K1r 2 +K2r 4 +...)+P1[r 2 +2(x-x0) 2 ]+2P2(x-x0)(y-y0)

[0063] Δy=(y-y0)(K1r 2 +K2r 4 +...)+P2[r 2 +2(y-y0) 2 ]+2P1(x-x0)(y-y0)(10)

[0064]

[0065] Where, K1, K2, ... represent radial distortion coefficients; P1, P2 represent tangential distortion coefficients; r represents the radial direction;

[0066] Then, the results of the bundle method block adjustment model of multiple images with known internal and external parameters are used as input, and after the seed point generation, expansion, filtering, and reconstruction process, a dense point cloud of the target object to be modeled is obtained, thereby realizing high-precision 3D modeling of the target object to be modeled. The specific implementation includes the following processes:

[0067] Feature extraction: For each input image, feature points are first extracted; commonly used feature point algorithms include: SIFT, SURF, etc. The selection of feature points should have good robustness and scale invariance.

[0068] Feature matching: Match feature points between each pair of images; a descriptor-based method or a geometric relationship-based method can be used. i And use O(I i ) represents the optical center of the corresponding camera. i For each feature f detected in , we collect a set F of features f' of the same type in the other images that are within two pixels of the corresponding epipolar line and triangulate the 3D points associated with the point of the same name (f, f'). We then follow the same procedure as in O(I i ) are considered as potential patch centers in the order of increasing distance, and the patch is generated point by point using the following process. Given a pair of features (f, f'), first construct a candidate patch p with center c(p), normal vector n(p) and reference image R(p) initialized as:

[0069] c(p)←{Triangulate the points f and f' with the same name} (11)

[0070]

[0071] R(p)←I i (13)

[0072] Extension: Given a patch p, first identify a set of adjacent image cells C(p) that meet certain criteria, and then perform extension on each of them. In the specific implementation, C(p) is initialized by collecting adjacent image cells in each visible image, expressed as:

[0073] C(p)={C i (x',y')|p∈Q i (x,y),|x-x'|+|y-y'|=1} (14)

[0074] C i (x',y')∈C(p) represents the image unit containing the patch p'; Q i (x,y) represents each image unit C i The set of patches onto which (x, y) is projected; As in the feature matching step, α is set to 0.6 and 0.3 before and after optimization, respectively, and these two values ​​are increased by 0.2 after each expansion iteration; α represents the photometric difference threshold with the reference image.

[0075] Filtering: Remove the wrong patches through the filter. The filter relies on visible consistency. Let U(p) represent the set of patches p' that are inconsistent with the current visibility information, that is, p and p' are not adjacent, but stored in the same unit of one of the images; then if the following inequality holds, p is deleted as an outlier:

[0076]

[0077] Preliminary 3D reconstruction: Using the matching relationship of feature points, the camera pose and the preliminary position of scene points are calculated; this can be achieved by solving the fundamental matrix or the essential matrix. The fundamental matrix represents the geometric relationship between the two images, while the essential matrix takes into account the intrinsic parameters of the camera.

[0078] Eliminate false matches: Eliminate false matches by checking the geometric constraints of the matching points; commonly used methods include the Ransac algorithm, which can robustly estimate model parameters and exclude outliers.

[0079] Motion recovery: By jointly optimizing the camera poses of multiple images, more accurate camera motion can be obtained; commonly used methods include bundle adjustment-based methods or graph optimization-based methods.

[0080] Dense reconstruction: After obtaining an initial estimate of the camera pose and scene points, a dense reconstruction of the scene is obtained through triangulation.

[0081] Based on the above method, the embodiment of the present invention also provides a high-precision three-dimensional modeling system for irregular objects combining stereoscopic images and laser data, such as Figure 3 FIG. 1 is a schematic diagram of the structure of a system according to an embodiment of the present invention, wherein the system comprises:

[0082] A modeling camera, used to collect a sequence of overlapping stereo images of the target object to be modeled;

[0083] Laser altimeter, used to measure the precise distance from the data acquisition platform to the surface of the target object to be modeled, and obtain the three-dimensional coordinates of the laser footprint in the world coordinate system according to the laser footprint coordinate calculation formula;

[0084] The laser footprint camera is used to collect the laser spot of the laser footprint hitting the surface of the irregular target object to be modeled, and obtain the image coordinates of the laser footprint using the automatic extraction algorithm of the laser footprint image coordinates;

[0085] The 3D modeling processing module is used to process the optical image of the target object to be modeled based on the obtained stereo image and the image coordinates of the laser footprint using the bundle method regional block adjustment model to obtain the precise exterior orientation elements of each image, thereby realizing high-precision 3D modeling of the target object to be modeled.

[0086] The specific implementation of each module of the above system is described in the above method embodiment.

[0087] In summary, the method and system described in the embodiments of the present invention combine stereo photography 3D modeling technology (SPG) with laser 3D modeling technology (LiDAR), and combine two different source data from the bottom layer to comprehensively improve the accuracy of 3D modeling; not only can the surface shape model of irregular target objects be reconstructed, eliminating worries for the practical application of irregular target objects, but also can be applied to fine terrain reconstruction in harsh environments on the earth, such as underwater scenes, emergency rescue scenes, or underground mine scenes.

[0088] It is worth noting that the contents not described in detail in the embodiments of the present invention belong to the prior art known to professional and technical personnel in the field.

[0089] In addition, a person skilled in the art can understand that all or part of the steps in the above-mentioned embodiment method can be implemented by instructing related hardware through a program, and the corresponding program can be stored in a computer-readable storage medium. The above-mentioned storage medium can be a read-only memory, a disk or an optical disk, etc.

[0090] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by any technician familiar with the technical field within the technical scope disclosed in the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims. The information disclosed in the background technology section of this article is only intended to deepen the understanding of the overall background technology of the present invention, and should not be regarded as an admission or in any form that the information constitutes prior art known to those skilled in the art.

Claims

1. A high-precision three-dimensional modeling method for irregular objects combining stereo images and laser data, characterized in that: The method comprises: Step 1: First, use a modeling camera to collect a sequence of overlapping stereo images of the target object to be modeled; Step 2: Use the laser altimeter to measure the data to obtain the precise distance from the platform to the surface of the target object to be modeled, and obtain the three-dimensional coordinates of the laser footprint in the world coordinate system according to the laser footprint coordinate calculation formula; Step 3, using a laser footprint camera to collect the laser spot of the laser footprint hitting the surface of the irregular target object to be modeled, and using a laser footprint image coordinate automatic extraction algorithm to obtain the image coordinates of the laser footprint; Step 4: Based on the stereo image obtained in step 1 and the image coordinates of the laser footprint obtained in step 3, the optical image of the target object to be modeled is processed using the bundle method regional block adjustment model to obtain accurate exterior orientation elements of each image, thereby achieving high-precision three-dimensional modeling of the target object to be modeled.

2. The high-precision three-dimensional modeling method for irregular objects combining stereoscopic images and laser data according to claim 1, characterized in that: In step 2, a laser altimeter transmitter emits a laser beam, which is reflected by the surface of the target object to be modeled and then received by a laser altimeter receiver; Then calculate the time interval t between laser emission and reception, the propagation speed of light is c, and the laser one-way transmission distance ρ = (c*t) / 2; combine the position and attitude of the data acquisition platform obtained by the inertial measurement unit IMU installed on the data acquisition platform to obtain the three-dimensional coordinates of the laser footprint in the world coordinate system; wherein the data acquisition platform includes a modeling camera, a laser altimeter, an inertial measurement unit IMU and a laser footprint camera; When the laser is emitted, there is a certain angle between the pointing direction and the modeling camera coordinate system. Assume that the negative angle between the pointing direction of the laser and the Z axis of the modeling camera coordinate system is θ, and the angle between the projection on the XOY plane of the modeling camera coordinate system and the positive X axis is α; If the laser ranging value is ρ, the coordinates of the laser emission point in the modeling camera coordinate system are: ΔL=[L x ,L y ,L z ] T Among them, L x Indicates the translation of the laser emission point along the X axis in the modeling camera coordinate system; L y Indicates the translation of the laser emission point along the Y axis in the modeling camera coordinate system; L z Indicates the translation of the laser emission point along the Z axis in the modeling camera coordinate system; Then the three-dimensional coordinates of the laser footprint point in the modeling camera coordinate system are: Among them, Body represents the coordinate system of the laser emission point.

3. The high-precision three-dimensional modeling method for irregular objects combining stereoscopic images and laser data according to claim 1, characterized in that: In step 3, the SIFT operator is used to perform feature matching between the modeling image and the laser footprint image, where: First, establish the DOG function Gaussian difference pyramid, expressed as: L(x,y,σ)=G(x,y,σ)*I(x,y) (2) D(x,y,σ)=G(x,y,kσ)-G(x,y,σ)*I(x,y) (3) Where σ represents the scale space coordinate; k represents the ratio of Gaussian blur coefficients between adjacent layers; G(x, y, kσ) represents the Gaussian pyramid; I(x, y) represents the original image; L(x, y, σ) is defined as the convolution of a Gaussian function G(x, y, σ) with a varying scale and the original image I(x, y); D(x, y, σ) represents the Gaussian difference pyramid, and * represents the convolution calculation; In order to improve the stability of key points, it is necessary to perform curve fitting on the scale space DOG function D(x, y, σ). The Taylor expansion of the DOG function in the scale space is: Where X = (x, y, σ) T , take the derivative and let the equation equal to zero to get the offset of the extreme point for: The corresponding extreme points are: Then, the feature matching function knnMatch based on the k-nearest neighbor algorithm is used for nearest neighbor matching, and the best matching point is screened out through the ratio test for K-nearest neighbor matching. K=2 is set, that is, the two points closest to each feature point are found, and the matching distance is set to the Euclidean distance d between the two points (x1, x2) and (y1, y2). 12 , expressed as: The best matching point is selected through ratio test. If the ratio of the nearest neighbor distance to the next nearest neighbor distance is less than the set threshold, the match is considered valid. The random sample consensus algorithm RANSAC is used to calculate the homography matrix M, and the weighted average method is used to fuse the left image to the right image to generate a fused image; In the fused image, the ellipse fitting algorithm is used to fit the edge of the laser footprint spot, and then the center of the ellipse is calculated as the image coordinates of the laser footprint. The image control points (ICPs) are constructed using the three-dimensional coordinates of the laser footprint in the world coordinate system and the corresponding image coordinates.

4. The high-precision three-dimensional modeling method for irregular objects combining stereoscopic images and laser data according to claim 1, characterized in that: In step 4, the initial value of the exterior orientation element is derived from the prototype position and attitude obtained by the inertial measurement unit IMU, and is converted into the exterior orientation element of the modeled camera through the installation matrix; the exterior orientation element refers to the rotation (φ, ω, κ) and translation (X, Y, Z) parameters from the camera coordinate system to a certain spatial coordinate system of the object space; The image coordinates of the laser footprint obtained in step 3 are used as control points and added into the bundle method block adjustment model. Under the rotation system with the Y axis as the main axis and the camera model based on perspective projection, the collinear relationship among the object point, image point and photography center is expressed by the collinear equation, which is: In the formula, (x, y) represents the image point coordinates; (x0, y0) represents the image principal point coordinates; (Δx, Δy) represents the distortion difference; (f x ,f y ) represents the principal distance; (X, Y, Z) represents the coordinates of the object point in the object coordinate system; (X S ,Y S ,Z S ) represents the coordinates of the photographic center in the object coordinate system; definition Indicates the use of The rotation matrix of the ω,κ rotation system; represents the angle of rotation around the Y axis; ω represents the angle of rotation around the X axis; κ represents the angle of rotation around the Z axis; the parameters in the matrix R are: The specific expression of distortion difference (Δx, Δy) is as follows: Where, K1, K2, ... represent radial distortion coefficients; P1, P2 represent tangential distortion coefficients; r represents the radial direction; Then, the results of the bundle method regional block adjustment model of multiple images with known internal and external parameters are used as input. After the seed point generation, expansion, filtering and reconstruction process, the dense point cloud of the target object to be modeled is obtained, thereby realizing high-precision three-dimensional modeling of the target object to be modeled.

5. A high-precision three-dimensional modeling system for irregular objects combining stereoscopic images and laser data, characterized in that: The system comprises: A modeling camera, used to collect a sequence of overlapping stereo images of the target object to be modeled; Laser altimeter, used to measure the precise distance from the data acquisition platform to the surface of the target object to be modeled, and obtain the three-dimensional coordinates of the laser footprint in the world coordinate system according to the laser footprint coordinate calculation formula; The laser footprint camera is used to collect the laser spot of the laser footprint hitting the surface of the irregular target object to be modeled, and obtain the image coordinates of the laser footprint using the automatic extraction algorithm of the laser footprint image coordinates; The 3D modeling processing module is used to process the optical image of the target object to be modeled based on the obtained stereo image and the image coordinates of the laser footprint using the bundle method regional block adjustment model to obtain the precise exterior orientation elements of each image, thereby realizing high-precision 3D modeling of the target object to be modeled.

Images