High-precision registration method for unmanned aerial vehicle laser point cloud and image constrained by spherical target

Abstract

The application discloses a kind of spherical target restrained unmanned aerial vehicle laser point cloud and image high-precision registration method, comprising: using unmanned aerial vehicle to collect laser radar point cloud data and image data, according to preset point arrangement spherical target, and the elevation coordinates of the spherical center plane of spherical target are collected, the spherical laser point of each spherical target is obtained from laser radar point cloud data;Get spherical target potential area, according to spherical target potential area, least square ellipse fitting is carried out in image data, and the image coordinates of target spherical center are calculated;The image coordinates of target spherical center obtained are used as observation value, the spherical laser point obtained is used as control information, unmanned aerial vehicle image joint regional network adjustment is carried out with spherical normal distance and photogrammetry spherical center back projection distance constraint, and the optimal value of camera internal parameter and image exterior orientation element is obtained, to realize the accurate registration of unmanned aerial vehicle image and laser radar point cloud. Solve the problem that point cloud and image registration is difficult in the natural scene with lack of artificial features.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of laser point cloud mapping technology, and particularly relates to a method for high-precision registration of unmanned aerial vehicle laser point cloud and image based on spherical target constraint. BACKGROUND

[0002] With the development of software and hardware devices and sensor technology, airborne laser radar technology is gradually becoming an important means of spatial remote sensing data acquisition. Compared with traditional aerial photogrammetry, laser radar has the advantages of active measurement and strong penetration, and can obtain ground point coordinates in vegetation covered areas, so it has been widely used in land surveying, railway, highway, water conservancy and other engineering surveying fields. In addition, due to the efficiency of laser radar measurement, it can directly obtain the three-dimensional information of the space target, avoiding the high complexity and high time-consuming processing of traditional photogrammetry dense matching, so the timeliness of the results is greatly improved. However, airborne laser radar also has some defects, mainly including: (1) low resolution, the average point spacing of current laser point cloud is generally between 0.1m~1m, which is much lower than the ground resolution of low-altitude aerial images represented by unmanned aerial vehicles; (2) laser radar measurement cannot directly obtain color information, so it is not easy to identify part of the ground target with unobvious shape and height such as road. Therefore, the two types of data sources of laser radar point cloud and aerial image show obvious complementarity, and joint application of the two types of data can realize complementary advantages, which can quickly obtain high-precision three-dimensional coordinate information and also obtain spectral and texture information of the ground object, which will have wide application prospects. However, due to the existence of camera installation error, GNSS positioning error and other factors, the coordinate system of the laser point cloud and image data directly collected will have certain deviation, which leads to the fact that the two types of data cannot be directly combined. For urban areas, artificial ground features such as building corners, roof edge lines or wall surfaces widely exist in laser point cloud, so the two types of data can be registered. For natural or outdoor scenes lacking artificial ground features, laser point cloud can only be interpolated into a raster image according to height or intensity, and then an image registration method is used for processing, but the height accuracy of this registration method is usually low. Therefore, how to realize high-precision registration of airborne laser point cloud and image for application scenarios lacking artificial ground features is still a difficulty. SUMMARY

[0003] Therefore, the purpose of the present application is to provide a method for high-precision registration of unmanned aerial vehicle laser point cloud and image based on spherical target constraint, which solves the problems of difficult registration of point cloud and image and unguaranteed registration accuracy in natural scenes lacking artificial ground features; uses the homonymous mapping relationship of the spherical target in the unmanned aerial vehicle image and the laser radar point cloud as the image orientation control information, constructs an image area network adjustment model constrained by spherical laser points, and realizes accurate registration of unmanned aerial vehicle image and laser point cloud.

[0004] In order to achieve the above object, the application provides a spherical target constrained unmanned aerial vehicle laser point cloud and image high-precision registration method, which comprises the following steps:

[0005] S1, collecting laser radar point cloud data and image data by using an unmanned aerial vehicle, arranging a spherical target according to a preset point, collecting a spherical center plane elevation coordinate of the spherical target, and obtaining spherical laser points of each spherical target from the laser radar point cloud data;

[0006] S2, obtaining a spherical target potential region according to the spherical laser points, performing ellipse detection in the image data according to the spherical target potential region, performing least square ellipse fitting on the detected ellipse boundary, obtaining an ellipse center coordinate, a half major axis and a half minor axis length, and calculating an image coordinate of the target spherical center;

[0007] S3, taking the image coordinate of the target spherical center obtained in S2 as an observation value, taking the spherical laser points obtained in S1 as control information, performing spherical normal distance and photogrammetry spherical center back projection distance constrained unmanned aerial vehicle image joint regional network adjustment to obtain optimal values of camera internal parameters and image exterior orientation elements, and realizing accurate registration of the unmanned aerial vehicle image and the laser radar point cloud.

[0008] Further preferably, in S1, POS assisted image aerial triangulation is further performed on the laser radar point cloud data and the image data collected by the unmanned aerial vehicle to obtain adjusted unmanned aerial vehicle image exterior orientation elements, camera internal parameters and photogrammetry connection point coordinates, image distortion correction is performed by using the camera internal parameters to obtain a non-distortion image.

[0009] Further preferably, in S2, the spherical target potential region is further obtained according to the following steps:

[0010] S201, calculating a coordinate average value of the spherical laser points based on the spherical laser points obtained in S1 as a rough position of the spherical target ;

[0011] S202, back projecting the rough position of the spherical target to the image by using the exterior orientation elements and the camera internal parameters of the unmanned aerial vehicle image to obtain a rough position of the spherical target on the image ; ;

[0012] S203, dividing the rough position as the center and according to a preset radius range to obtain the spherical target potential region.

[0013] Further preferably, according to the spherical target potential region obtained in S203, ellipse detection is performed, and least square ellipse fitting is performed on the detected ellipse boundary to obtain a long half axis length of the ellipse, a short half axis length , the coordinates of the two end points of the long axis of the ellipse are and , wherein is the end point closer to the image principal point;

[0014] Suppose that the image exterior orientation elements of the observable spherical target are , the focal length is , and the coordinates of the approximate position of the spherical target , then the coordinates of the spherical center of the spherical target on the image are as follows:

[0015]

[0016] In the above formula, and are intermediate variables of coordinate conversion.

[0017] Further preferably, in S3, the image coordinates of the target spherical center obtained in S2 are taken as observed values, the spherical laser points obtained in S1 are taken as control information, and the spherical normal distance and photogrammetric spherical center back projection distance constraint unmanned aerial vehicle image joint block adjustment is performed according to the following process:

[0018] S301, the image coordinates of the target spherical center obtained in S2 are taken as observed values, and the photogrammetric coordinates of the spherical center are calculated by using the photogrammetric least squares forward intersection method and are marked as ;

[0019] S302, a spherical normal distance constraint cost function is constructed, and the cost of the normal distance of the target laser point to the photogrammetric sphere is calculated ;

[0020] S303, the spherical normal distance and photogrammetric spherical center back projection distance constraint unmanned aerial vehicle image bundle block adjustment is optimized, the spherical target is taken as the absolute orientation control element of the unmanned aerial vehicle image block, and the connection point is used to maintain the geometric consistency in the block;

[0021] S304, based on the optimization result and the spherical normal distance cost value obtained in S302, the overall cost function of the block adjustment model is obtained as shown in the following formula:

[0022]

[0023] In the formula, represents the overall cost of the block adjustment, represents the back projection cost of the photogrammetric connection point, represents the back projection cost of the photogrammetric spherical center on the image, ​This represents the normal distance cost from the target laser point to the photogrammetric sphere.

[0024] Further preferably, in S3, the optimal values ​​of camera intrinsic parameters and image exterior orientation elements are obtained, including: performing Taylor series expansion on each term in the overall cost function of the regional adjustment model in S304, neglecting higher-order terms, obtaining the linearized error equation, and accurately solving the parameters in the error equation according to the least squares adjustment criterion to obtain the accurate exterior orientation elements and camera intrinsic parameters of the UAV image, thereby achieving overall geometric registration of UAV image and laser point cloud data.

[0025] Further preferably, in S302, a spherical normal distance constraint cost function is constructed to calculate the cost of the normal distance from the target laser point to the photogrammetric sphere. ,include:

[0026] Assuming the radius of the spherical target is The coordinates of the spherical point on the lidar target are The coordinates of the photogrammetric center of the target are: The formula for calculating the cost value of the laser point is as follows:

[0027]

[0028] The total cost of the normal distance from the target laser point to the photogrammetric sphere is as follows:

[0029]

[0030] In the formula: This represents the cost of the normal distance from the target laser point to the photogrammetric sphere. Indicates the first A spherical target; Indicates a spherical target The corresponding number One laser point; The cost function is the distance from the laser point on the sphere to the center of the sphere measured by photography.

[0031] More preferably, in S303, when optimizing the UAV image bundle method area network adjustment constrained by the spherical normal distance and the photogrammetric sphere center back projection distance, the optimization parameters are expressed by the following formula.

[0032]

[0033] in, Represents the vector of parameters to be optimized. Represents the image exterior orientation vector. This represents the object-space coordinate vector of the photogrammetric connection point. Represents the camera intrinsic parameter vector. represents the photogrammetric sphere center coordinate parameter vector; represents the photogrammetric connection point back-projection cost, represents the photogrammetric sphere center back-projection cost on the image, represents the normal distance cost of the target laser point to the photogrammetric sphere.

[0034] Further preferably, in S304, the photogrammetric sphere center back-projection cost on the image is represented by the following formula:

[0035] In the formula, is a photogrammetric back-projection function, represents the photogrammetric sphere center corresponding to the th spherical target, represents the camera interior and exterior orientation elements of the image that can observe the spherical target, represents the observation value of the th photogrammetric sphere center on the th image; represents the Euclidean distance function of the image point.

[0036] Further preferably, the photogrammetric connection point back-projection cost is calculated by the following formula:

[0037] (10)

[0038] In the formula, is a photogrammetric back-projection function, represents the th photogrammetric connection point, represents the camera interior and exterior orientation elements of the image that can observe the connection point, represents the observation value of the th photogrammetric connection point on the th image, represents the Euclidean distance function of the image point.

[0039] The spherical target constraint unmanned aerial vehicle laser point cloud and image high-precision registration method disclosed in the application has at least the following advantages compared with the prior art:

[0040] 1. The application realizes accurate registration of unmanned aerial vehicle images and laser radar point clouds based on spherical targets. The method does not depend on artificial or natural features in the scene, is also applicable to natural or outdoor scenes with relatively scarce textures, and meets the application requirements of high-precision registration of images and laser point clouds in outdoor engineering survey and topographic mapping.

[0041] 2、The application utilizes the spherical surface normal constraint of the spherical target, effectively overcomes the influence of point cloud density on target extraction precision and registration precision on the basis of simultaneously realizing geometric control in the plane and height direction, thereby effectively improving processing precision.

[0042] 3、The application designs a UAV image block adjustment model constrained by the spherical target, can guarantee geometric consistency within the image block on the basis of realizing registration of the UAV image and the laser point cloud, is convenient for subsequent development of orthophoto production, topographic map production and the like, and has strong practical application and popularization value. BRIEF DESCRIPTION OF DRAWINGS

[0043] Figure 1 A flowchart of a high-precision registration method for a UAV laser point cloud and an image constrained by a spherical target is provided for the application.

[0044] Figure 2 A schematic diagram of spherical laser point normal distance and photogrammetry center reverse projection distance constraint is provided for the application. DETAILED DESCRIPTION

[0045] The application is further described in detail below through the drawings and specific embodiments.

[0046] As shown in the drawings, Figure 1 A high-precision registration method for a UAV laser point cloud and an image constrained by a spherical target is provided for the application, which includes the following steps:

[0047] S1, laser radar point cloud data and image data are collected by a UAV, a spherical target is laid according to a preset point, the center plane elevation coordinates of the spherical target are collected, and spherical laser points of each spherical target are obtained from the laser radar point cloud data; further, in S1, POS-aided image aerial triangulation is performed on the laser radar point cloud data and the image data collected by the UAV, the exterior orientation elements of the adjusted UAV image, the camera internal parameters and the photogrammetry connection point coordinates are obtained, image distortion correction is performed by using the camera internal parameters, and non-distorted images are obtained.

[0048] Specifically, a certain number of spherical targets are laid according to the survey area, the center plane elevation coordinates are collected by using RTK, leveling or total station measurement and the like. Laser radar point cloud and images are collected by a UAV, POS-aided image aerial triangulation is performed, the exterior orientation elements of the adjusted image, the camera internal parameters and the photogrammetry connection point coordinates are obtained, image distortion correction is performed by using the camera internal parameters, and non-distorted images are obtained. Laser points on the surface of the spherical target are obtained from the laser point cloud by using the man-machine interactive picking or automatic detection mode.

[0049] S2, obtaining a spherical target potential region according to the spherical laser points, performing ellipse detection in the image data according to the spherical target potential region, performing least square ellipse fitting on the detected ellipse boundary to obtain an ellipse center coordinate, a semi-major axis length and a semi-minor axis length, and calculating an image coordinate of a spherical target center;

[0050] In S2, the spherical target potential region is obtained according to the following steps:

[0051] S201, calculating a coordinate average value of the spherical laser points as a rough position of the spherical target based on the spherical laser points obtained in S1 ;

[0052] S202, projecting the rough position of the spherical target back to the image by using the exterior orientation elements of the unmanned aerial vehicle image and the camera internal parameters to obtain a rough position of the spherical target on the image ; ;

[0053] S203, taking the rough position as the center and dividing a spherical target potential region according to a preset radius range .

[0054] Further preferably, ellipse detection is performed according to the spherical target potential region obtained in S203, and least square ellipse fitting is performed on the detected ellipse boundary to obtain a long semi-axis length , a short semi-axis length , and two end point coordinates of the long axis of the ellipse and , wherein the end point closer to the image principal point is

[0055] Suppose that the image exterior orientation elements of the observable spherical target are , the focal length is , the coordinate of the rough position of the spherical target is , and the coordinate of the center of the spherical target on the image is .

[0056] (1)

[0057] In the above formula, and are intermediate variables for coordinate conversion, which are mainly angle variables for assisting coordinate conversion, wherein is an angle between the projection light of the center and the image plane, and is an auxiliary angle based on the photogrammetry rotation angle and the fitted ellipse.

[0058] In the above formula, the calculation method of the parameter is as follows: ​

[0059] (2)

[0060] in,

[0061] (3)

[0062] in, , , These represent the three rotation angles in photogrammetry. Indicates the heading angle Indicates image rotation angle Indicates the lateral deflection angle.

[0063] parameter The calculation method is as follows:

[0064] (4)

[0065] and This represents the lengths of the major and minor axes of the fitted ellipse.

[0066] S3. Using the image coordinates of the target sphere center obtained in S2 as the observation value and the spherical laser point obtained in S1 as the control information, perform joint regional network adjustment of UAV images constrained by the spherical normal distance and the photogrammetric sphere center back projection distance to obtain the optimal values ​​of camera intrinsic parameters and image exterior orientation elements, thereby achieving accurate registration between UAV images and lidar point clouds.

[0067] In S3, the image coordinates of the target sphere center obtained in S2 are used as the observation values, and the spherical laser points obtained in S1 are used as the control information. The joint regional network adjustment of UAV images constrained by the spherical normal distance and the photogrammetric sphere center back projection distance is performed according to the following process:

[0068] S301, Assuming there is a total A spherical target For each target, there is a set of image points where the center of the target sphere lies in the UAV image. Using the image coordinates of the target sphere center obtained in S2 as the observation value, the photogrammetric coordinates of the sphere center are calculated using the photogrammetric least squares forward intersection method. ;

[0069] S302. Construct a spherical normal distance constraint cost function to calculate the cost of the normal distance from the target laser point to the photogrammetric sphere. ;

[0070] like Figure 2When the point cloud and image coordinate systems are precisely aligned, the spherical points in the laser point cloud should lie on the photogrammetric sphere, meaning the normal distance from the lidar spherical point to the photogrammetric sphere is 0. Using the normal distance from the laser point to the photogrammetric sphere as the cost, a cost equation is derived. The method is as follows: Assume the radius of the spherical target is... The coordinates of the spherical point on the lidar target are The coordinates of the photogrammetric center of the target are: The formula for calculating the cost value of the laser point is as follows:

[0071] (5)

[0072] The total cost of the normal distance from the target laser point to the photogrammetric sphere is as follows:

[0073] (6)

[0074] In the formula: This represents the cost of the distance from the spherical laser point to the photogrammetric sphere. Indicates the first A spherical target; Indicates a spherical target The corresponding number One laser point. Let be the cost function for the distance from the laser point on the sphere to the center of the sphere measured by photography. Figure 2 d p Represents the two-dimensional distance between points in the image.

[0075] S303. Optimize the UAV image bundle method regional network adjustment constrained by the spherical normal distance and the photogrammetric sphere center back projection distance. Use the spherical target as the absolute orientation control unit of the UAV image regional network and use the connection points to maintain the geometric consistency inside the regional network.

[0076] More preferably, in S303, when optimizing the UAV image bundle method area network adjustment constrained by the spherical normal distance and the photogrammetric sphere center back projection distance, the optimization parameters are expressed by the following formula.

[0077] (7)

[0078] in, Represents the vector of parameters to be optimized. Represents the image exterior orientation vector. This represents the object-space coordinate vector of the photogrammetric connection point. Represents the camera intrinsic parameter vector. Represents the vector of photogrammetric sphere center coordinate parameters; This represents the backprojection cost of the photogrammetric connection point. represents the back-projection cost of the photogrammetric center of the sphere on the image, represents the normal distance cost of the target laser point to the photogrammetric sphere.

[0079] S304、based on the optimization results and the spherical normal distance cost value obtained in S302, obtaining the overall cost function of the area adjustment model shown in the following formula:

[0080] (8)

[0081] In the formula, represents the overall cost of the area adjustment, represents the back-projection cost of the photogrammetric connection point, represents the back-projection cost of the photogrammetric center of the sphere on the image, represents the normal distance cost of the target laser point to the photogrammetric sphere.

[0082] The back-projection cost of the photogrammetric center of the sphere on the image is represented by the following formula:

[0083] (9)

[0084] In the formula, is a photogrammetric back-projection function, represents the photogrammetric center of the sphere corresponding to the th spherical target, represents the camera interior and exterior orientation elements of the th image that can observe the sphere, represents the observation value of the th photogrammetric center on the th image. represents the Euclidean distance function of the image point.

[0085] Further preferably, the back-projection cost of the photogrammetric connection point is calculated by the following formula:

[0086] (10)

[0087] In the formula, is a photogrammetric back-projection function, represents the th photogrammetric connection point, represents the camera interior and exterior orientation elements of the th image that can observe the connection point, represents the observation value of the th photogrammetric connection point on the th image, The Euclidean distance function represents the image point.

[0088] exist Figure 2 In this process, the extraction point of the target center on the image is the potential area of ​​the spherical target obtained by using spherical laser points in step S2 above. Based on this potential area, ellipse detection is performed on the image data. Least-squares ellipse fitting is then performed based on the detected ellipse boundary to obtain the ellipse center coordinates, semi-major axis length, and semi-minor axis length. The image coordinates of the target center are then calculated using the fitted coordinates. The back-projection position of the photogrammetric center point on the image is the object-space coordinate of the center obtained by photogrammetric forward intersection in step S3 above. This object-space coordinate is then back-projected back onto the image, which is obtained using formula 10. The back-projection function of photogrammetry is used to calculate the back-projection. In other words, the target sphere's center, obtained from the image through ellipse fitting and geometric back-calculation, is extracted onto the image as the observed value, used for forward intersection and regional network adjustment. The back-projected position of the photogrammetric sphere's center onto the image, obtained from the object point obtained through forward intersection, is used as the theoretical value, constructed for the back-projection error term in the adjustment. The optimal UAV parameters are obtained when the error between the observed and theoretical values ​​is minimized. This allows for precise matching between UAV imagery and radar point clouds.

[0089] Further preferably, in S3, the optimal values ​​of camera intrinsic parameters and image exterior orientation elements are obtained, including: performing Taylor series expansion on each term in the overall cost function of the regional adjustment model in S304, neglecting higher-order terms, obtaining the linearized error equation, and accurately solving the parameters in the error equation according to the least squares adjustment criterion to obtain the accurate exterior orientation elements and camera intrinsic parameters of the UAV image, thereby achieving overall geometric registration of UAV image and laser point cloud data.

[0090] The linearized error equation takes the following form:

[0091] (11)

[0092] In the formula, , and Let represent the back projection residual vector of the photogrammetric connection point, the back projection residual vector of the photogrammetric center of the target sphere, and the normal distance residual vector from the laser point of the target to the center of the photogrammetric sphere, respectively. This represents the vector of corrections to the exterior orientation elements of the UAV imagery. Corrected vector for photogrammetric connection points. This is the camera intrinsic parameter correction vector. The correction vector for measuring the center of the target sphere in a photographic test. , , , , , and are the first-order partial derivative matrices of the objective function with respect to the optimization parameters, respectively. , and are constant vectors; , and are weight matrices. According to the least square adjustment principle, the above parameters are solved accurately to obtain the accurate exterior orientation elements and camera interior parameters of the UAV image, thereby realizing the overall geometric registration of the UAV image and the laser point cloud data.

[0093] Obviously, the above embodiments are only examples for clearly illustrating the present application, but not limitation on the embodiments. Based on the above description, other different forms of changes or variations can be made by those skilled in the art. Here, all the embodiments are not required to be exhausted, and the obvious changes or variations derived therefrom are still within the protection scope of the present application.

Claims

1. A method for high-precision registration of a spherical target-restrained UAV laser point cloud and image, characterized in that, The method comprises the following steps: S1, collecting laser radar point cloud data and image data by using a UAV, arranging a spherical target according to a preset point, collecting a spherical center plane elevation coordinate of the spherical target, and obtaining spherical laser points of each spherical target from the laser radar point cloud data; S2, obtaining a spherical target potential region according to the spherical laser points, performing ellipse detection in the image data according to the spherical target potential region, performing least square ellipse fitting on a detected ellipse boundary to obtain an ellipse center coordinate, a semi-major axis, and a semi-minor axis length, and calculating an image coordinate of a target spherical center; S3, taking the image coordinate of the target spherical center obtained in S2 as an observation value, taking the spherical laser points obtained in S1 as control information, performing a spherical normal distance and photogrammetry spherical center back projection distance constraint UAV image joint block adjustment to obtain optimal values of camera internal parameters and image exterior orientation elements, and realizing accurate registration of the UAV image and the laser radar point cloud; The spherical normal distance and photogrammetry spherical center back projection distance constraint UAV image joint block adjustment is performed according to the following process: S301. Using the image coordinates of the target center obtained in S2 as the observation value, the photogrammetric coordinates of the center are calculated using the least squares forward intersection method of photogrammetry. ; S302, constructing a spherical normal distance constraint cost function, and calculating a cost of a normal distance of a target laser point to a photogrammetry sphere; S303, optimizing the spherical normal distance and photogrammetry spherical center back projection distance constraint UAV image bundle block adjustment, taking the spherical target as a UAV image block absolute orientation control element, and maintaining geometric consistency in a block by using a connection point; S304, calculating an overall cost of a block adjustment model based on the optimized result and the spherical normal distance value obtained in S302.

2. The spherical target constrained UAV laser point cloud and image high-precision registration method according to claim 1, characterized in that, In S1, POS assisted image aerial triangulation is further performed on the laser radar point cloud data and the image data collected by the UAV to obtain adjusted UAV image exterior orientation elements, camera internal parameters, and photogrammetry connection point coordinates, image distortion correction is performed by using the camera internal parameters to obtain a non-distortion image.

3. The spherical target restrained UAV laser point cloud and image high-precision registration method according to claim 2, characterized in that, In S2, the spherical target potential region is further obtained according to the following steps: S201, calculate the average value of the coordinates of the spherical laser points as the approximate position of the spherical target based on the spherical laser points obtained in S1 ; S202, using the exterior orientation elements and the camera internal parameters of the UAV image, obtaining the approximate position of the spherical target in the image obtaining the approximate position of the spherical target in the image by back projection ; S203, approximate position The potential region of the spherical target is divided according to the preset radius range with the center.

4. The spherical target restrained UAV laser point cloud and image high-precision registration method according to claim 3, characterized in that, According to the spherical target potential region obtained in S203, an ellipse is detected, and a least square ellipse fitting is performed on the detected ellipse boundary to obtain a long semi-axis length of the ellipse , a short semi-axis length , and two end point coordinates of the long axis of the ellipse and , wherein is the end point closer to the image principal point Assume that the exterior orientation elements of the image observing the spherical target are , the focal length is , the approximate position of the spherical target is , and the coordinates of the spherical target are , then the coordinates of the center of the spherical target on the image are : In the above formulae, and are intermediate variables for the coordinate transformation.

5. The spherical target constrained UAV laser point cloud and image high-precision registration method according to claim 1, characterized in that, In S304, the overall cost of the block adjustment model is calculated by using the following formula: wherein represents the total cost of network adjustment, represents the re-projection cost of the photogrammetric connection point, represents the re-projection cost of the photogrammetric sphere center on the image, represents the normal distance cost of the target laser point to the photogrammetric sphere.

6. The spherical target-restrained UAV laser point cloud and image high-precision registration method according to claim 5, characterized in that, In S3, the optimal values of the camera internal parameters and the image exterior orientation elements are obtained by performing Taylor series expansion on each term in the overall cost function of the block adjustment model in S304, omitting high-order terms to obtain a linearized error equation, accurately solving parameters in the error equation according to a least square adjustment criterion, obtaining accurate exterior orientation elements and camera internal parameters of the UAV image, and thus realizing overall geometric registration of the UAV image and the laser point cloud data.

7. The spherical target-restrained UAV laser point cloud and image high-precision registration method according to claim 5, characterized in that, In S302, a spherical method normal distance constraint cost function is constructed, and a cost of a normal distance from a target laser point to a photogrammetry sphere is calculated , comprising: Assuming the radius of the spherical target is , the coordinates of the laser radar target spherical point are , and the coordinates of the photogrammetry sphere center corresponding to the target are , the calculation formula of the value of the laser point is as follows: The overall cost of the normal distance of the target laser point to the photogrammetry sphere is as follows: where: represents the cost of the normal distance from the target laser point to the photogrammetric sphere; represents the i-th spherical target; represents the i-th spherical target corresponding i-th laser point; is the distance cost function of the spherical laser point to the photogrammetric sphere center.

8. The spherical target restrained UAV laser point cloud and image high-precision registration method according to claim 5, characterized in that, In S303, when the spherical normal distance and photogrammetry spherical center back projection distance constraint UAV image bundle block adjustment is optimized, the optimization parameters are represented by using the following formula wherein, denotes the vector of parameters to be optimized, denotes the vector of exterior orientation elements of the imagery, denotes the vector of object coordinates of the photogrammetric tie points, denotes the vector of intrinsic camera parameters, denotes the vector of photogrammetric center of sphere coordinates; denotes the re-projection cost of the photogrammetric tie points, denotes the re-projection cost of the photogrammetric center of sphere on the imagery, denotes the normal distance cost of the target laser points to the photogrammetric sphere.

9. The spherical target-restrained UAV laser point cloud and image high-precision registration method according to claim 5, characterized in that: In S304, the back-projection cost of the photogrammetry sphere center on the image This can be expressed by the following formula: In the formula, For photogrammetry, the back projection function is used. Indicates the first The photogrammetric center of a spherical target This indicates the number of observable spheres. Zhang's images include camera interior and exterior orientation elements. Indicates the first The photogrammetric center of the sphere is at the... Observations on Zhang's images; The Euclidean distance function represents the image point.

10. The spherical target-restrained UAV laser point cloud and image high-precision registration method according to claim 5, characterized in that, Re-projection cost of photogrammetric connection points is calculated using the following formula: In the formula, For photogrammetry, the back projection function is used. Indicates the first A photogrammetric connection point, This indicates the first node that can be observed as a connection point. Zhang's images include camera interior and exterior orientation elements. Indicates the first The photogrammetric connection point at the ... Observations on Zhang's image, The Euclidean distance function represents the image point.

Images