LASER ENHANCED RECONSTRUCTION OF 3D SURFACE

Abstract

A method for reconstructing a surface of a three-dimensional object (41) involves a projection of a laser spot pattern (12, 14) onto the surface of the three-dimensional object (41) by a laser (11), and a generation of a series of endoscopic images (24) as an endoscope (21) is translated and/or rotated relative to the three-dimensional object (41). Each endoscopic image (24) illustrates a different view (23) of a laser spot array (13, 15) within the laser spot pattern (12, 14) as projected onto the surface of the three-dimensional object (41) by the laser (11). The laser spot array (13, 15) may be identical to or a subset of the laser spot pattern (12, 14). The method further involves a reconstruction of the surface of the three-dimensional object (41) from a correspondence of the different views (23) of the laser spot array (13, 15) as illustrated in the endoscopic images (24).

Claims

1. A system for reconstructing a surface of a three-dimensional object, the system comprising: an endoscope configured to generate a series of endoscopic images as the endoscope is translated and/or rotated relative to the three-dimensional object, wherein each endoscopic image illustrates a different view of a laser spot array within a laser spot pattern projected onto the surface of the three-dimensional object; and a processor, in communication with the endoscope, configured to reconstruct the surface of the three-dimensional object from a correspondence of each different view of the laser spot array as illustrated in the endoscopic images.

2. The system of claim 1, wherein, to reconstruct the surface of the three-dimensional object, the processor is further configured to: generate a fundamental matrix that relates the different views of the laser spot array as illustrated in the endoscopic images; and reconstruct three-dimensional object points as a function of the fundamental matrix and the different views of the laser spot array.

3. The system of claim 1, wherein, to reconstruct the surface of the three-dimensional object, the processor is further configured to: generate a fundamental matrix that relates the different views of the laser spot array as illustrated in the endoscopic images; detect surface features of the object as illustrated in the endoscopic image; and reconstruct three-dimensional object points as a function of the fundamental matrix and the surface features of the object detected in the endoscopic images.

4. The system of claim 1, wherein, to reconstruct the surface of the three-dimensional object, the processor is further configured to: generate a fundamental matrix that relates the different views of the laser spot array as illustrated in the endoscopic images.

5. The system of claim 4, wherein, to reconstruct the surface of the three-dimensional object, the processor is further configured to: generate an essential matrix that relates the different views of the laser spot array as illustrated in the endoscopic images, the essential matrix being a function of the fundamental matrix and a camera calibration matrix associated with a camera calibration of the endoscope.

6. The system of claim 5, wherein, to reconstruct the surface of the three-dimensional object, the processor is further configured to: generate a translation vector and a rotation matrix as a function of the essential matrix.

7. The system of claim 6, wherein, to reconstruct the surface of the three-dimensional object, the processor is further configured to: generate a projection matrix for each view of the laser spot array as a function of the translation vector and the rotation matrix, each projection matrix being a linear transformation of an associated view of the laser spot array.

8. The system of claim 7, wherein, to reconstruct the surface of the three-dimensional object, the processor is further configured to: reconstruct three-dimensional object points as a function of each projection matrix and associated views of the laser spot array.

9. The system of claim 7, wherein, to reconstruct the surface of the three-dimensional object, the processor is further configured to: detect surface features of the object as illustrated in the endoscopic images for each view of the laser spot array; and reconstruct three-dimensional object points as a function of each projection matrix and each surface feature of the object detected in the endoscopic images.

10. The system of claim 1, wherein the endoscope is intra-operatively calibrated from at least two of the endoscopic images.

11. The system of claim 1, further comprising: a laser configured to project the laser spot pattern onto the surface of the three-dimensional object.

12. A method for reconstructing a surface of a three-dimensional object, the method comprising: generating a series of endoscopic images as an endoscope is translated and/or rotated relative to the three-dimensional object, wherein each endoscopic image illustrates a different view of a laser spot array within a laser spot pattern projected onto the surface of the three-dimensional object; and reconstructing the surface of the three-dimensional object from a correspondence of each different view of the laser spot array as illustrated in the endoscopic images.

13. The method of claim 12, further comprising: generating a fundamental matrix that relates the different views of the laser spot array as illustrated in the endoscopic images; and reconstructing three-dimensional object points as a function of the fundamental matrix and the different views of the laser spot array.

14. The method of claim 12, further comprising: generating a fundamental matrix that relates the different views of the laser spot array as illustrated in the endoscopic images; detecting surface features of the object as illustrated in the endoscopic image; and reconstructing three-dimensional object points as a function of the fundamental matrix and the surface features of the object detected in the endoscopic images.

15. The method of claim 12, further comprising: generating a fundamental matrix that relates the different views of the laser spot array as illustrated in the endoscopic images; and generating an essential matrix that relates the different views of the laser spot array as illustrated in the endoscopic images, the essential matrix being a function of the fundamental matrix and a camera calibration matrix associated with a camera calibration of the endoscope.

16. The method of claim 15, further comprising: generating a translation vector and a rotation matrix as a function of the essential matrix.

17. The method of claim 16, further comprising: generating a projection matrix for each view of the laser spot array as a function of the translation vector and the rotation matrix, each projection matrix being a linear transformation of an associated view of the laser spot array.

18. The method of claim 17, further comprising: reconstructing three-dimensional object points as a function of each projection matrix and associated views of the laser spot array.

19. The method of claim 17, further comprising: detecting surface features of the object as illustrated in the endoscopic images for each view of the laser spot array; and reconstructing three-dimensional object points as a function of each projection matrix and each surface feature of the object detected in the endoscopic images.

20. A non-transitory computer-readable storage medium having stored a computer program comprising instructions, the instructions, when the computer program is executed by a process, cause the processor to: generate a series of endoscopic images as an endoscope is translated and/or rotated relative to a three-dimensional object, wherein each endoscopic image illustrates a different view of a laser spot array within a laser spot pattern projected onto a surface of the three-dimensional object; and reconstruct the surface of the three-dimensional object from a correspondence of each different view of the laser spot array as illustrated in the endoscopic images.

Description

[0013] The foregoing form and other forms of the present invention as well as various features and advantages of the present invention will become further apparent from the following detailed description of various embodiments of the present invention read in conjunction with the accompanying drawings. The detailed description and drawings are merely illustrative of the present invention rather than limiting, the scope of the present invention being defined by the appended claims and equivalents thereof.

[0014] FIG. 1 illustrates an exemplary embodiment of a 3D image reconstruction system in accordance with the present invention.

[0015] FIG. 2 illustrates one exemplary series of different views of a laser spot array in accordance with the present invention.

[0016] FIG. 3 illustrates another exemplary series of different views of the laser spot array shown in FIG. 2 in accordance with the present invention.

[0017] FIG. 4 illustrates a flowchart representative of an exemplary embodiment of an endoscopic surgical method in accordance with the present invention.

[0018] FIG. 5 illustrates an exemplary endoscopic application of the method shown in FIG. 4 by the system shown in FIG. 1.

[0019] FIG. 6 illustrates a flowchart representative of an exemplary embodiment of a 3D surface reconstruction method in accordance with the present invention.

[0020] An implementation of 3D surface reconstruction algorithms by the present invention is accomplished by a laser projecting a laser spot pattern on a 3D object and an endoscope generating a series of 2D endoscopic images of a laser spot array within the laser spot pattern. The laser spot pattern serves as a reproducible and precise feature as projected on the 3D object to facilitate a correspondence of the laser spot array among the endoscopic images.

[0021] For example, as shown in FIG. 1, a laser 11 powered by a laser source 10 projects a laser spot pattern 12 having a 7×7 matrix arrangement of circular dots. Within laser spot pattern 12 is a laser spot array 13 having a 3×3 matrix arrangement of circular dots. During a minimally invasive endoscopic surgery, an endoscope 21 is focused on an entirety or a portion of laser spot pattern 12 whereby a field of view 22 of endoscope 21 encircles laser spot array 13.

[0022] More particularly, FIG. 2 illustrates a sequence of endoscopic views 23a-23c generated by endoscope 21 as endoscope 21 is translated in a direction of the 3D object. As such, laser spot pattern 12 enlarges in the different endoscopic views 23a-23c with laser spot array 13 being identifiable in each endoscopic view 23a-23c. The enlargement of laser spot pattern across endoscopic view 23a-23c serves as a motion of laser spot array 13 for purposes of implementing the 3D surface reconstruction algorithms as will be further explained herein.

[0023] FIG. 3 illustrates an additional sequencing of endoscopic views 23d-23f generated by endoscope 21 as endoscope 21 is further translated in a direction of the 3D object. Again, laser spot pattern 12 enlarges across the endoscopic views 23d-23f with laser spot array 13 being identifiable in each endoscopic view 23d-23f, and the enlargement of laser spot pattern across endoscopic views 23d-23f serves as a motion of laser spot array 13 for purposes of implementing the 3D surface reconstruction algorithms as will be further explained herein. FIGS. 2 and 3 are provided to emphasize the inventive principle of a laser spot pattern as a reproducible and precise feature as projected on a 3D object to facilitate a correspondence of a laser spot array among the endoscopic images. In practice, preferably the laser spot pattern includes nine (9) or more laser spots, the number of endoscopic images is two (2) or more, and the distance between the endoscopic camera and the laser spot pattern facilitates an identification of the entire laser spot pattern across all of the endoscopic images whereby the laser spot pattern itself serves as the laser spot array. Nonetheless, FIGS. 2 and 3 highlight that a laser spot array, whether an entirety or a portion of the laser spot pattern, must be identifiable among all of the endoscopic images.

[0024] Referring back to FIG. 1, an image reconstruction device 20 processes the generated images of the laser spot array to reconstruct a 3D image of the surface of the object. For purposes of the present invention, an image reconstruction device is broadly defined herein as any device structurally configured for generating a 3D reconstruction of a surface of an object by processing endoscopic images in accordance with 3D reconstruction algorithms (e.g., a programmed computer), and the term “generating” as used herein is broadly defined to encompass any technique presently or subsequently known in the art for creating, computing, supplying, furnishing, obtaining, producing, forming, developing, evolving, modifying, transforming, altering or otherwise making available information (e.g., data, text, images, voice and video) for computer processing and memory storage/retrieval purposes, particularly image datasets and video frames.

[0025] FIG. 4 illustrates a flowchart 30 representative of an endoscopic surgical method of the present invention. Flowchart 30 includes pre-operative stages S32 and S33, and intra-operative stages S34-S37. The term “pre-operative” as used herein is broadly defined to describe any activity occurring or related to a period or preparations before an endoscopic application and the term “intra-operative” as used herein is broadly defined to describe as any activity occurring, carried out, or encountered in the course of an endoscopic application (e.g., operating the endoscope). Examples of an endoscopic application include, but are not limited to, an arthroscopy, a bronchoscopy, a colonscopy, a laparoscopy, a brain endoscopy, and an endoscopic cardiac surgery. Examples of an endoscopic cardiac surgery include, but are not limited to, endoscopic coronary artery bypass, endoscopic mitral and aortic valve repair and replacement.

[0026] Pre-operative stage S31 encompasses a selection of a laser for projecting the laser spot pattern on the 3D object. In practice, a Lasiris™ SNF laser may be used for endoscopic applications whereby the laser has a wavelength approximately 600 nm and a power less than 100 mW. Further, the laser preferably projects laser spot pattern a green or blue 7×7 matrix of circular dots whereby eight (8) or more of the circular dots may serve as the laser spot array. Further, the circular dots may have a 0.5 mm diameter with a 4 mm spacing between the circular dots. To specify a fan angle (FA) of ninety (90) degrees or less, an object size (L) and an operating distance (D) must be know in accordance with the following equation [1]:

FA=2*arcsin(L/(2*D))  [1]

[0027] FIG. 5 illustrates an example of an arthroscopic application involving laser 11 being a distance D from tissue 41 of a knee 40 with an object size L.

[0028] Referring again to FIG. 4, pre-operative stage S32 encompasses a known camera calibration of the endoscope. In one embodiment, the laser projects the laser spot pattern on a contrasting planar surface (e.g., a white planar surface) with the laser spot pattern having a uniform matrix (e.g., 7×7 matrix of circular dots) or a matrix with different number of laser spots in the two dimensions (e.g., 6×7 matrix of circular dots). The calibration parameters as well as radial distortion are estimated by the acquiring of images of the laser spot pattern on the planar surface under two (2) or more different orientations of the endoscope relative to the planar surface, and by a detection of the laser points in the images.

[0029] Intra-operative stage S33 encompasses a generation of the laser spot pattern on the surface of the 3D object. For example, as shown in FIG. 5, laser 11 is inserted within a surgical instrument port 43 of knee 40 to thereby project a laser spot pattern 14 of a 5×5 matrix of circular dots onto tissue 41.

[0030] An execution of intra-operative stage S34 is dependent of whether stage S32 was not executed during the pre-operative phase, or if a re-calibration of the endoscope is required. If executed, intra-operative stage S32 encompasses an endoscope taking images of the laser spot pattern projected onto the 3D object under two (2) or more different orientations of the endoscope relative to the laser spot pattern. For example, as shown in FIG. 5, an endoscope 21 is inserted in a visual port 42 of knee 40 to thereby generate an image of laser spot pattern 14 (e.g., an image 24). The endoscope 24 can be moved around port 42 (i.e., a pivot point) at any direction and rotation to generate images of the same laser spot array from different viewing angles and directions. Thereafter, a detection of the laser spot pattern 14 as the laser spot array within the images or a detection of a laser spot array 15 within the images would enable an estimation of the camera parameters.

[0031] Detection of laser spots can be performed with any algorithm known in art, such as color thresholding. Result of the detection is x[x,y].sup.T position of the spot in a coordinate system of each image.

[0032] Intra-operative stage S35 encompasses a generation of a series of two (2) or more images of the laser spot pattern on the 3D object as the endoscope is translated and/or rotated relative to the 3D object and the port 42. For example, as shown in FIG. 5, endoscope 21 is inserted in port 42 of knee 40 to thereby generate images of laser spot pattern 14 (e.g., an image 24) as endoscope 21 is translated in a direction of tissue 41 as shown by the arrow.

[0033] Intra-operative stage S36 encompasses a 3D reconstruction of the surface of the object from the endoscopic images acquired during stage S35 and the calibration of the endoscope obtained during pre-operative stage S32 or intra-operative stage S34. In practice, any 3D reconstruction algorithm may be implemented during stage S36 to achieve the 3D reconstruction of the object. In one embodiment, a shown in FIG. 6, a flowchart 50 representative of a 3D surface reconstruction that may be implemented during stage S36.

[0034] Referring to FIG. 6, a stage S51 of flowchart 50 encompasses a generation of a fundamental matrix (F) for relating the different views of the laser spot array across the endoscopic mages. In one embodiment, for the same laser spot in two different views (x) and (x′), fundamental matrix (F) is a 3×3 matrix and is defined in accordance with the following known equation [2]:

X.sup.T*F*x′=0  [2]

[0035] For N laser spots in two different views, a set of N equations is defined:

[00001]$\begin{array}{cc}{x}_1^T*F*x_1^{\prime }=0\mathrm{.Math.}{x}_N^T*F*x_N^{\prime }=0& \left[3\right]\end{array}$

[0036] The unknown (F) from equations [3] may be computed using an Eight-point algorithm if the laser spot array has eight (8) laser spots (N=8), or may be computed using an iterative method (e.g., RANSAC) if the laser spot array includes nine (9) or more laser spot.

[0037] Stage S52 encompasses a generation of an essential matrix (E) or relating the different views of the laser spot array across the endoscopic mages. In one embodiment, the essential matrix (E) is computed from the following known equation [4]:

E=K.sup.T*F*K=0  [4]

[0038] Calibration matrix (K) is a 3×3 matrix representative of the pre-operative or intra-operative calibration of the endoscope.

[0039] Stage S53 encompasses a generation of a translation vector (T) and a rotation matrix (R) (if the endoscope was rotated) as a function of the essential matrix (E). In one embodiment, a translation vector (T) and a rotation matrix (R) are derived from the following known equation [5]:

E=U*Σ*V.sup.T=0  [5]

[0040] Stage S54 encompasses a generation of a projection matrix for each view of the laser spot array. In one embodiment for two (2) views of the laser spot array, a projection matrix P.sub.1 for a view associated with spots (x) and a projection matrix P.sub.1 for a view associated for spots (x′) are computed from the following known equations [6] and [7]:

P.sub.1=K*[I|0]  [6]

P.sub.2=K.sup.T*[R|T]*K  [7]

[0041] Stage S55 encompasses a 3D object point reconstruction from the laser spot array or salient features of the object (e.g., edges) in the endoscopic images. In one embodiment, using a pinhole camera model for two (2) views, a 3D object point X is computed from the following known equations [8] and [9]:

x=P.sub.1*X  [8]

x′=P.sub.2*X  [9]

[0042] The computed 3D object point X may be reconstructed using triangulation and equations [8] and [9].

[0043] For points x and x′, two sets could be used for stage S55.

[0044] In a first embodiment, laser spots x and x′ can be used as features. These are strong features, because they are highly precise and reliable. This embodiment would result in a very sparse 3D model having as many points as the associated laser spot array.

[0045] In second embodiment, weak object surface features (e.g., edges) detected using feature detection methods known in art (e.g., a SIFT method) may be used with projection matrixes P.sub.1 and P.sub.2 computed from points x and x′. This method would result in a dense surface with lower precision of points x and x′, but maintaining high precision of projection matrixes P.sub.1 and P.sub.z.

[0046] While various embodiments of the present invention have been illustrated and described, it will be understood by those skilled in the art that the embodiments of the present invention as described herein are illustrative, and various changes and modifications may be made and equivalents may be substituted for elements thereof without departing from the true scope of the present invention. In addition, many modifications may be made to adapt the teachings of the present invention without departing from its central scope. Therefore, it is intended that the present invention not be limited to the particular embodiments disclosed as the best mode contemplated for carrying out the present invention, but that the present invention includes all embodiments falling within the scope of the appended claims.