TWI577344B - Calibrating method and system for three dimensional scanning - Google Patents

Description

Three-dimensional scanning correction method and system thereof

The invention relates to a method for correcting a three-dimensional scan and a system thereof, in particular to aligning a coordinate system of a through-image scanner with a coordinate system of an appearance image capturing unit, so as to make a three-dimensional reconstruction data composed of a penetrating image and Appearance images can be represented in a consistent coordinate system.

According to today, computer tomography scanners are widely used in medical and industrial fields, that is, through computer tomography scanners, information on the internal structure of the human body can be obtained, so that doctors can diagnose diseases inside the human body or obtain object structures through a computed tomography scanner. Information, so that security personnel can conduct security checks on objects.

The general computed tomography scanner includes an X-ray source, a rotating device, and an X-ray sensor, and the X-ray source and the X-ray sensor are rotated through the rotating device, or are rotated and disposed on the X-ray source and An object to be tested between the X-ray sensors to obtain a projection image of the object to be tested, and collecting the three-dimensional image of the object to be tested by collecting the projection image; reconstruction theory of the computed tomography scanner And the algorithm is still developing continuously. At present, the calibration method of geometric parameters has successively calibrated and adjusted each geometric parameter. Due to the correlation between parameters, it is often necessary to adjust the parameters one by one, making the calibration process quite complicated. Another example is the mainland patent CN102743184B "A method for calibrating geometric parameters of X-ray cone beam computer tomography system", which uses camera calibration technology to present the geometry between the X-ray source, the flat panel detector and the rotating shaft in the cone beam CT system. The positional relationship, which directly solves the geometric parameters of the cone beam CT and its error, is a systematic measurement The solution method can abstract the cone beam CT into a basic camera system model, thereby simultaneously solving multiple geometric parameters associated with each other in the system, thereby simplifying the complexity of the calibration measurement of the entire system and simultaneously improving the cone beam CT geometry. Calibration stability.

Since the accurate calibration of geometric parameters and errors is very important for the reconstruction of the computed tomography scanner, in addition to improving the accuracy of the geometric calibration of the computed tomography scanner, if the correction of subsequent images can reduce the error, the imaging quality can be improved. The image reconstruction of the computed tomography scanner can be more accurate; therefore, the invention mainly improves the error of the geometric parameters, so as to truly play the role of geometric correction and avoid reconstructed image distortion.

Therefore, the present invention provides a method for correcting a three-dimensional scan, comprising the following steps: Step A: scanning a calibration object by a penetrating image scanner to obtain a penetrating image, and defining a first coordinate system according to the penetrating image And capturing, by an appearance image capturing unit, an appearance image of the correction object, and defining a second coordinate system according to the appearance image; and step B. defining a plane in the space as ( , d ), where S ² is a plane normal vector, d For the shortest distance from the plane to the origin, the first coordinate system defines one of the first plane parameters of one of the correction planes on the correction object: , _x d ), where x represents the first coordinate system; and step C. takes the second plane parameter of the correction plane in one of the second coordinate systems: , _c d ), where c represents the second coordinate system; step D. will ( , _x d ) and ( , _c d ) association, the conversion relationship of the first coordinate system and the second coordinate system is defined as E=(R, t), where R is a rotation matrix, t is a displacement vector; , d ), a point m on the plane satisfies <m, n=d>; using a relational expression of E=(R, t) and <m, n=d>, a displacement relation is obtained And a rotational relationship Where q is a quaternion, q = v _w + v _x i+ v _y j+ v _z k, v is a matrix The minimum eigenvector; step E. repeating step D to obtain the corresponding relationship between the first plane parameter and the second plane parameter different from N ≧ 3 groups, and then bringing the first plane parameter and the second plane parameter into The displacement relationship and the rotation relationship obtain the conversion relationship E=(R, t) of the first coordinate system and the second coordinate system; and step F. use the conversion relationship E of the first coordinate system and the second coordinate system = (R, t) to obtain the penetration image and the position of the appearance image on the correction object.

Wherein, step E further includes error correction, assuming that a point m in the space = (x, y, z), the point in the first coordinate system and the second coordinate system are represented as _x m and _c m respectively , according to The conversion relationship between the first standard system and the second coordinate system is E=(R,t), so _x m = R _c m + t , and m is rewritten as a homogeneous coordinate , the former is equivalent to And then bring a point m on the aforementioned plane to satisfy <m, n=d> a plane error function The plane error function is nonlinearly optimized, thereby reducing geometric errors.

In step A, the penetrating image scanner includes a light source emitter and a sensor corresponding to the light source emitter, and the correction object is disposed between the light source emitter and the sensor, and The light source emitter and the external image capturing unit are disposed on opposite sides of the sensor, and the penetrating image scanner forms the penetrating image on the sensor, and the plane of the penetrating image is defined as a first An XY plane, and defining a first Z axis perpendicular to the first XY plane to form the first coordinate system, and the appearance image capturing unit defines a plane of the appearance image as a second XY plane, and defines a second The Z axis is perpendicular to the second XY plane to form the aforementioned second coordinate system.

The three-dimensional scanning correction system used in the above three-dimensional scanning correction method comprises: a rotating device having a contrast space for placing the correcting object; the penetrating image scanner is disposed on the rotating device and aligned The correction object; the appearance image capturing unit is disposed on the rotating device and aligned with the correction object.

The rotating device includes a rotating shaft and a rotating arm pivotally connected to the rotating shaft, the rotating arm has a first cantilever and a second cantilever, and the first cantilever is provided with the sensor penetrating the image scanner The second cantilever is provided with the light source emitter of the image capturing machine and the external image capturing unit, and the light source emitter and the external image capturing unit and the sensor are located on the rotating shaft. And rotating around the correction object centering on the rotation axis.

Wherein, the correction object is a checkerboard correction plate, and the metal blank or the adhesive metal foil is applied to the edge blank of the checkerboard correction plate.

The penetrating image scanner is a computed tomography scanner.

The appearance image capturing unit is an optical camera or a camera.

The method for correcting a three-dimensional scan of the present invention and the system thereof, by aligning a first coordinate system defined by the penetration image scanner with a second coordinate system defined by the appearance image capturing unit, so that the penetration image is scanned The three-dimensional reconstruction data generated by the machine and the optical image generated by the appearance image capturing unit can be expressed in a consistent coordinate system, thereby achieving information fusion.

The method for correcting three-dimensional scanning of the present invention and the system thereof are related to each other through an association plane ( , _c d _i )→( , _x d _i ), in the foregoing and Solve the algebraic solution E=( R , t ), and then use the algebraic solution as the initial solution, with the associated point and surface correspondence _x m _i →( , _c d _i ), right By performing nonlinear optimization and reducing geometric errors, a more accurate coordinate system conversion relationship can be obtained, thereby improving image quality.

(1)‧‧‧Rotating device

(11) ‧ ‧ angiography space

(12)‧‧‧Store platform

(13) ‧‧‧ shaft

(14)‧‧‧Rotating arm

(141)‧‧‧First cantilever

(142)‧‧‧Second cantilever

(2) ‧‧‧ checkerboard correction board

(21)‧‧‧Sheet

(3)‧‧‧ penetrating image scanner

(31)‧‧‧ Sensors

(32)‧‧‧Light source transmitter

(4)‧‧‧ Appearance image capture unit

(A) ‧‧‧ Plane that penetrates the image

(B) ‧ ‧ the plane of the appearance image

[First Figure] is a schematic view of a device according to a preferred embodiment of the present invention.

[Second figure] is a schematic structural view of the checkerboard correction plate in the preferred embodiment of the present invention.

In summary of the above technical features, the main effects of the three-dimensional scanning correction system and method of the present invention can be clearly demonstrated in the following embodiments.

Referring to the first and second figures, the apparatus of the preferred embodiment of the present invention comprises: a rotating device (1) having a contrast space (11) and a storage platform in the contrast space (11) ( 12), for placing a calibration object, which in this embodiment is a checkerboard correction plate (2), and is coated with metal paint or metal on the edge blank of the checkerboard correction plate (2). a sheet (21); a penetrating image scanner (3), such as a computed tomography scanner, disposed on the rotating device (1) and aligned with the checkerboard (2); an appearance image capturing unit (4) ), such as an optical camera or camera, disposed on the rotating device (1) and The checkerboard correction plate (2); wherein the rotating device (1) comprises a rotating shaft (13) and a rotating arm (14) pivotally connected to the rotating shaft (13), the rotating arm (14) has a first a cantilever (141) and a second cantilever (142), the first cantilever (141) is provided with a sensor (31) penetrating the image scanner (3), and the second cantilever (142) a light source emitter (32) and the appearance image capturing unit (4) disposed on the image capturing machine (3), the light source emitter (32) and the appearance image capturing unit (4) and the The sensor (31) is located at opposite ends of the rotating shaft (13), and rotates around the checkerboard (2) centering on the rotating shaft (13).

The method for correcting a three-dimensional scan using the above apparatus includes the following steps:

Step A. placing the checkerboard calibration plate on the storage platform, scanning the checkerboard correction plate with the penetrating image scanner, and obtaining a penetrating image on the sensor of the penetrating image scanner, according to the penetrating The image defines a first coordinate system, and the appearance image capturing unit of the correcting object is synchronously captured by the external image capturing unit disposed on the opposite surface of the sensor, and a second coordinate system is defined according to the image; The first standard system is defined by the plane (A) of the through image as a first XY plane, and the first XY plane is defined by a first Z axis, and the second coordinate system is formed by the appearance image. The plane (B) is defined as a second XY plane, and the second XY plane is defined by a second Z-axis. Please refer to the first figure. In the embodiment, the first coordinate system and the second coordinate system are respectively represented by { X } and { C }, so a point m = ( x , y , z ) in the space, in the first coordinate system and the first The two coordinate system is expressed as _x m and _c m .

Step B. According to the plane correspondence relationship proposed by the document "Geometric calibration of a multi-layer LiDAR system and image sensors using plane-based implicit laser parameters for texture 3-D depth reconstruction", a plane in the space is used in this embodiment. defined as( , d ), where S ² is a plane normal vector, d For the shortest distance from the plane to the origin, the first coordinate system defines one of the first plane parameters of one of the correction planes on the correction object: , _x d ), where x represents the first coordinate system; the first plane parameter first corresponds to the plurality of sets of metal dot coordinates _x m _i by the penetration image generated according to the metal foil on the correction object, and then calculates the foregoing The optimum plane parameters of the metal points are obtained.

Step C. Obtaining a second plane parameter of the calibration plane in the second coordinate system by using a Zhang's calibration method: , _c d ), where c represents the second coordinate system.

Step D. Will ( , _x d ) and ( , _c d ) association, the conversion relationship between the first coordinate system and the second coordinate system is described by a rigid transformation, defined as E=(R, t), where R SO (3) is a 3 × 3 rotation matrix, and t For a displacement vector of 3 × 1, the conversion process is as follows: _x m = R _c m + t , if m is rewritten as a homogeneous coordinate , the former is equivalent to: And its inverse transform to When a point corresponding to the N group _c m _i → _x m _i is given, if N > 3 The solution is not unique, and the least squares solution can be changed at this time: , where ∥. ∥ is the Euclidean distance, and is derived from the literature "Closed-form solution of absolute orientation using unit quaternions". Has a close-formed solution. However, the geometric point of the multi-layer LiDAR system and image sensors using plane-based implicit laser parameters for texture 3-D depth reconstruction is not easy to obtain. ”Proposed plane correspondence, the aforementioned plane definition ( , d ), a point m on the plane satisfies <m, n=d>, where <. 〉 is the inner product of the vector. According to the definition of _x m = R _c m + t and <m, n = d>, the vector _c d A point on the aforementioned plane, which is transformed by the coordinate system Due to vector After rotation Parallel, so ,will versus Subtractive collation If more than three sets of plane correspondences are known ( , _c d _i )→( , _x d _i ), then construct a set of over-determined linear systems, as follows: The S-squared solution is obtained by applying Singular Value Decomposition (SVD). Moreover, the aforementioned plane correspondence may also be applied to derive a rotation relationship between the first coordinate system and the second coordinate system, a known normal vector After rotation Parallel, so the inner product value should be 1. Considering N > 3 sets of normal vectors, the rotation matrix R should satisfy the formula: According to the document "Closed-form solution of absolute orientation using unit quaternions", versus a closed solution of the same form, for this solution, first to the matrix Perform eigendecomposition to obtain the eigenvector v corresponding to the smallest eigenvalue S ³ , converted to a quaternion form q = v _w + v _x i+ v _y j+ v _z k, and then the rotation matrix is restored to obtain a rotation relationship

Step E: repeating step D to obtain a correspondence between the first plane parameter and the second plane parameter different from the N≧3 group, and then bringing the first plane parameter and the second plane parameter into the displacement relationship and The rotation relationship obtains the conversion relationship E=(R, t) of the first coordinate system and the second coordinate system.

In step E, error correction is further included. Since the above calculation method finds that ( R , t ) is an algebraic solution, the solution cannot reflect the geometric properties of the space when the error exists. Therefore, after obtaining the linear solution, nonlinear optimization must be performed to obtain a more accurate coordinate system conversion relationship. N group known system the first coordinate point _x m _i corresponding to the second coordinate system to a plane ( , _c d _i ) point-to-face correspondence _x m _i →( , _c d _i ), according to With the definition of < m , n 〉= d , a plane error function can be constructed, as follows: And then By performing nonlinear optimization and reducing geometric errors, a more accurate conversion relationship between the first coordinate system and the second coordinate system can be obtained.

Step F. Using the conversion relationship E=(R, t) of the first coordinate system and the second coordinate system, the penetration image and the position of the appearance image on the correction object are obtained.

In view of the foregoing description of the embodiments, the operation and the use of the present invention and the effects of the present invention are fully understood, but the above described embodiments are merely preferred embodiments of the present invention, and the invention may not be limited thereto. Included within the scope of the present invention are the scope of the present invention.

(1)‧‧‧Rotating device

(11) ‧ ‧ angiography space

(12)‧‧‧Store platform

(13) ‧‧‧ shaft

(14)‧‧‧Rotating arm

(141)‧‧‧First cantilever

(142)‧‧‧Second cantilever

(2) ‧‧‧ checkerboard correction board

(3)‧‧‧ penetrating image scanner

(31)‧‧‧ Sensors

(32)‧‧‧Light source transmitter

(4)‧‧‧ Appearance image capture unit

(A) ‧‧‧ Plane that penetrates the image

(B) ‧ ‧ the plane of the appearance image

Claims (8)

A method for correcting a three-dimensional scan includes the following steps: Step A: scanning a calibration object by a penetrating image scanner to obtain a penetrating image, defining a first coordinate system according to the penetrating image, and using an appearance image The capture unit captures an appearance image of the correction object, and defines a second coordinate system according to the appearance image; step B. defines a plane in the space as ( , d ), where Is a plane normal vector, d is the shortest distance from the plane to the origin, and the first coordinate system defines one of the first plane parameters of one of the correction planes on the correction object: , _x d ), where x represents the first coordinate system; and step C. takes the second plane parameter of the correction plane in one of the second coordinate systems: , _c d ), where c represents the second coordinate system; step D. will ( , _x d ) and ( , _c d ) association, the conversion relationship of the first coordinate system and the second coordinate system is defined as E=(R, t), where R is a rotation matrix, t is a displacement vector; , d ), a point m on the plane satisfies <m, n=d>; using a relational expression of E=(R, t) and <m, n=d>, a displacement relation is obtained And a rotational relationship Where q is a quaternion, Step E: repeating step D to obtain a correspondence between the first plane parameter and the second plane parameter different from the N≧3 group, and then bringing the first plane parameter and the second plane parameter into the displacement relationship and The rotation relationship obtains the conversion relationship E=(R, t) of the first coordinate system and the second coordinate system; step F. uses the conversion relationship of the first coordinate system and the second coordinate system E=(R, t Obtaining the penetration image and the position of the appearance image on the correction object. The method for correcting a three-dimensional scan as described in claim 1, wherein the step E further includes error correction, assuming that a point in the space is m coordinate = (x, y, z), the point is in the first coordinate system And the second coordinate system is represented as _x m and _c m respectively , according to the conversion relationship E=(R, t) of the first coordinate system and the second coordinate system, so _x m = R _c m + t , m is aligned The secondary coordinates are rewritten as , the former is equivalent to And then bring a point m on the aforementioned plane to satisfy <m, n=d> a plane error function The plane error function is nonlinearly optimized, and the geometric error is also reduced. The method for correcting a three-dimensional scan according to claim 1, wherein in step A, the penetrating image scanner comprises a light source emitter and a sensor corresponding to the light source emitter, wherein the correction object is placed in the Between the light source emitter and the sensor, the light source emitter and the appearance image capturing unit are disposed on opposite sides of the sensor, and the penetrating image scanner forms the penetration on the sensor The image is defined as a first XY plane by the plane of the through image, and defines a first Z axis perpendicular to the first XY plane to form the first coordinate system, and the appearance image capturing unit is flat with the appearance image The face is defined as a second X-Y plane and defines a second Z-axis perpendicular to the second X-Y plane to form the aforementioned second coordinate system. A three-dimensional scanning correction system for use in a three-dimensional scanning correction method according to any one of claims 1 to 3, comprising: a rotating device having a contrast space for placing the correction object; The through image scanner is disposed on the rotating device and aligned with the correcting object; the external image capturing unit is disposed on the rotating device and aligned with the correcting object. The three-dimensional scanning correction system of claim 4, wherein the rotating device comprises a rotating shaft and a rotating arm pivotally connected to the rotating shaft, the rotating arm has a first cantilever and a second cantilever, The first cantilever is provided with the sensor penetrating the image scanner, and the second cantilever is provided with the light source emitter penetrating the image scanner and the appearance image capturing unit, the light source emitter and the appearance The image capturing unit and the sensor are located at opposite ends of the rotating shaft, and rotate around the correcting object around the rotating shaft. The three-dimensional scanning correction system of claim 4, wherein the correction object is a checkerboard correction plate, and a metal paint or a metal foil is applied to the margin of the checkerboard correction plate. The three-dimensional scanning correction system of claim 4, wherein the penetration image scanner is a computed tomography scanner. The three-dimensional scanning correction system of claim 4, wherein the external image capturing unit is an optical camera or a camera.