3D modeling method based on point cloud data

1. A 3D modeling method based on point cloud data for generating 3D object data corresponding to a target object ( 50 , 51 ), comprising:

(a) activating a 3D scanner ( 2 ) to obtain 2D images of the target object ( 50 , 51 ) from different angles and a plurality of depths of the target object ( 50 , 51 ), each depth corresponding to a plurality of pixels of the 2D images;

(b) generating a plurality of point cloud data ( 60 - 64 ) based on the 2D images and the depths, each of the point cloud data ( 60 - 64 ) including a plurality of point data ( 70 - 79 );

(c) performing a pre-modeling process on the plural point cloud data ( 60 - 64 ) to filter the plural point data ( 70 - 79 ) not belonging to the target object ( 50 , 51 ) out of each of the point cloud data ( 60 - 64 ); and

(d) performing a 3D modeling process on the filtered point cloud data ( 60 - 64 ) to generate the 3D object data;

wherein step (c) comprises the sub-steps of:

(c11) selecting at least three point data ( 70 - 79 ) from each of the point cloud data ( 60 - 64 );

(c12) determining a regression plane based on the selected at least three point data ( 70 - 79 );

(c13) counting the number of the plural point data ( 70 - 79 ) belonging to the regression plane in the plural point cloud data ( 60 - 64 ) to obtain the number of the plural point data ( 70 - 79 ) in an inlier;

(c14) setting the regression plane as one of a plurality of candidate planes when the number of the plural point data ( 70 - 79 ) of the inlier is greater than a critical value of the inlier;

(c15) repeatedly performing steps (c11), (c12), (c13) and (c14) until an ending condition is met; and

(c16) selecting one of the candidate planes and filtering the plural point data ( 70 - 79 ) belonging to the selected candidate plane out of each of the point cloud data ( 60 - 64 ) so as to filter the plural point data ( 70 - 79 ) belonging to a platform ( 53 ) out of each of the point cloud data ( 60 - 64 ).

2. The 3D modeling method as claimed in claim 1 , wherein step (c) further comprises the sub-steps of:

(c21) setting focus data in each of the point cloud data ( 60 - 64 );

(c22) calculating a filter range based on coordinates of the plural point data ( 70 - 79 ) of each of the point cloud data ( 60 - 64 ) wherein the coordinates of each of the point data ( 70 - 79 ) are recorded with a plane location and a depth of each of the point data ( 70 - 79 ) in the plural point cloud data ( 60 - 64 ); and

(c23) setting a filter area ( 54 ) in each of the point cloud data ( 60 - 64 ) based on the focus data and the filter range of each of the point cloud data ( 60 - 64 ), and filtering out the plural point data ( 70 - 79 ) externally of the filter area ( 54 ) so as to filter both the plural point data ( 70 - 79 ) belonging to a background ( 52 ) and the plural point data ( 70 - 79 ) belonging to the platform ( 53 ) out of each of the point cloud data ( 60 - 64 ).

3. The 3D modeling method as claimed in claim 2 , wherein in sub-step (c21) the plural point data ( 70 - 79 ) in a center of the plural point cloud data ( 60 - 64 ) are set as the focus data.

4. The 3D modeling method as claimed in claim 2 , wherein in sub-step (c21) an average of the coordinates of the plural point data ( 70 - 79 ) at a central area of the plural point cloud data ( 60 - 64 ) is calculated to obtain coordinates of the focus data.

5. The 3D modeling method as claimed in claim 2 , wherein in sub-step (c22) a standard deviation between the coordinates of the plural point data ( 70 - 79 ) of each of the point cloud data ( 60 - 64 ) is calculated, and the filter range is set based on the standard deviation.

6. The 3D modeling method as claimed in claim 1 , wherein in sub-step (c15) the ending condition is defined by repeatedly performing steps (c11), (c12), (c13) and (c14) a plurality of times and counting same until the count is equal to a predetermined count or the number of the candidate planes is equal to a predetermined number.

7. The 3D modeling method as claimed in claim 1 , wherein in sub-step (c13) an extension range is set at a periphery of the regression plane, and the number of the plural point data ( 70 - 79 ) belonging to the regression plane or the extension range of the regression plane is calculated to obtain the number of the plural point data ( 70 - 79 ) of the inlier; and wherein in sub-step (c16) a distance between each candidate plane and the plural point data ( 70 - 79 ) in the extension range of each candidate plane is calculated, a plurality of the distance is calculated as a total distance, one of the candidate planes having a minimum total distance is selected, and the plural point data ( 70 - 79 ) belonging to the selected candidate plane and the extension range of the selected candidate plane are filtered out of each of the point cloud data ( 60 - 64 ).

8. The 3D modeling method as claimed in claim 1 , wherein in sub-step (c12) a regression function is calculated based on the coordinates of the at least three point data ( 70 - 79 ) to determine the regression plane formed by the at least three point data ( 70 - 79 ), and wherein in sub-step (c13) the plural point data ( 70 - 79 ) belonging to the regression plane is determined when the coordinates of each of the plural point data ( 70 - 79 ) are complied with the regression function.

9. The 3D modeling method as claimed in claim 1 , wherein step (c) further comprises the sub-steps of:

(c31) obtaining a critical depth; and

(c32) filtering the plural point data ( 70 - 79 ) having a depth not less than the critical depth out of each of the point cloud data ( 60 - 64 ) so as to filter the plural point data ( 70 - 79 ) belonging to the background ( 52 ) out of each of the point cloud data ( 60 - 64 ).

10. The 3D modeling method as claimed in claim 1 , after step (d) further comprising the steps of:

(e1) identifying platform object data in the 3D object data corresponding to a platform object ( 82 ); and

(e2) deleting the platform object data from the 3D object data.