JP2002259945A - Three dimensional modeling method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a three dimensional model with favorable resolution from small data volume even when data volume and the number of cross-sectional images increase, and provide a process of providing the three dimensional model in a short time by a small and inexpensive computer. SOLUTION: When composing the three dimensional model by a plurality of cross-sectional images of a measured object, the plurality of cross-sectional images are divided into groups per suitable numbers, a three dimensional modeling process is carried out per each group, and three dimensional model parts m1, m2, etc., mN are respectively acquired. Then, a polygon reducing process is carried out in each three dimensional model part, and a single three dimensional model M is acquired by coalescing them. Since joint end face portions between adjacent three dimensional model parts are excluded from the polygon reducing process and they are not deformed, coalescence is smooth.

Description

DETAILED DESCRIPTION OF THE INVENTION

The present invention relates to a three-dimensional modeling method for constructing a three-dimensional model using a plurality of cross-sectional images of an object to be measured.

[0002]

2. Description of the Related Art A conventional method of this kind is disclosed, for example, in Japanese Patent Application Laid-Open No. H11-328442. This gives the brightness value (density) of each of four opposing pixels of two adjacent cross-sectional images in a plurality of cross-sectional images obtained at intervals along a certain direction of the DUT to the apex. This is a method of setting a cube or a rectangular parallelepiped, and constructing a three-dimensional model by an interpolation surface forming process using a marching cube method for generating a triangular surface based on a comparison result between the luminance value of each vertex and a threshold value.

[0003] The marching cube method,
"Marching Cubes: A Hig" by WELorensen and HECline
h Resolution 3D Surface Reconstruction Algorithm ''
(1987; Computer Graphics 21 (4); pp163-169).

According to such a three-dimensional modeling method,
Easy and simple operation only by computer processing without human intervention
There is an advantage that a dimensional model can be configured.

[0005]

However, in the above-mentioned prior art, when the number of cross-sectional images is high, the number of obtained three-dimensional models becomes large and the processing time increases in proportion to the data amount. I do. This is particularly remarkable because the marching cube method is a processing in units of one voxel (one pixel of a cross-sectional image).

Hereinafter, this will be described. Now, when a cross-sectional image is an X-ray CT image (hereinafter, abbreviated as a CT image) and one image is obtained by using 500 general CT images of 512 × 512 pixels in length and width, a three-dimensional model is obtained. It is assumed that the data amount is 250 MB.

According to this, one sheet is 2048 × 20
Using 1000 high-resolution 48-pixel CT images, 3
When a three-dimensional model is obtained, the three-dimensional model is (2
048/512) × (2048/512) × (1000
/ 500) = 32, the data amount is 32 times that of the three-dimensional model. It is assumed that the data amount representing the density of each pixel is the same.

That is, a three-dimensional model obtained by using 1000 high-resolution CT images of 2048 × 2048 pixels vertically and horizontally has a large data amount of 250 MB × 32 = 8000 MB, that is, 8 GB. This means an increase in the processing time, and in subsequent handling, for example, storage and recording of such a three-dimensional model, and reuse or processing of the model by CAD, CAM, CAE, or the like, the capacity of the recording medium. Or the time required for various calculations is increased.

Therefore, it is conceivable to perform polygon reduction on the obtained three-dimensional model. Polygon reduction is
This is a method to reduce the number of polygons that make up the model while maintaining the model shape as much as possible (GARLAND, M. a
nd HECKBERT, PS SurfaceSimplification using Qua
dric Error Metrics. Proceedings of SIGGRAPH 97.In
Computer Graphics Proceedings, Annual Conference Se
ries, 1997, ACM SIGGRAPH, pp.209-216 etc.),
Useful for reducing data volume.

However, the polygon reduction process consumes many times, for example, four times as much memory as the data amount of the original model. For this reason, an expensive computer equipped with 32 GB of memory is required to perform polygon reduction processing on an 8 GB three-dimensional model, which is not practical.

Further, in a CT image of about 4000 × 4000 pixels which can be output by a current X-ray CT apparatus,
When the data amount representing the pixel density is 2B (byte),
The data amount is 4000 × 4000 × 2 = 32 MB.
The three-dimensional model obtained by using 1000 of these is 3
The data amount becomes 2 MB × 1000 = 32 GB.
To perform polygon reduction processing on a two-dimensional model, 32
GB × 4 = 128 GB of memory is consumed. Therefore, realizing polygon reduction processing of such a scale is limited to a very large computer at present, and has not been practically feasible.

As described above, one sheet is 512.times.51
3 obtained by using 500 general CT images of 2 pixels
In the dimensional model, the data amount is about 250 MB, and the processing time is short. However, in order to improve the measurement accuracy and the like of the obtained three-dimensional model, it is necessary to increase the resolution and the number of CT images used.

For this reason, conventionally, even if the data amount of all cross-sectional images to be processed is increased by using a high-resolution, large number of cross-sectional images, in other words, a small data amount of about 250 MB
There has been a demand for the appearance of a three-dimensional modeling method that can obtain a three-dimensional model and can also process the same in a short time with a small computer.

SUMMARY OF THE INVENTION The present invention has been made in view of the above-mentioned demands, and has a relatively small data amount and is easy to handle, despite the increase in the number of cross-sectional image data used. 3 which does not greatly reduce the resolution of
It is an object of the present invention to provide a three-dimensional modeling method that can obtain a three-dimensional model and can also perform a process for obtaining the three-dimensional model in a short time by a small and inexpensive computer.

[0015]

In order to achieve the above object, according to the first aspect of the present invention, a plurality of cross-sectional images obtained at intervals along a certain direction of an object to be measured are obtained. Grouping by the number of sheets appropriately without changing the arrangement,
For each group, a three-dimensional modeling process is performed using a group of cross-sectional images constituting the group, and the obtained three-dimensional model part for each group is subjected to an end face portion related to joining with an adjacent three-dimensional model part. , Except that the three-dimensional model sections after the polygon reduction processing are united according to the arrangement before the grouping to obtain a single three-dimensional model.

According to a second aspect of the present invention, in the first aspect, a polygon reduction process is further performed on a single three-dimensional model.

According to a third aspect of the present invention, there is provided the first or second aspect.
In the invention described in (1), a three-dimensional modeling process based on a marching cube method is used for the three-dimensional modeling process.

[0018]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is an explanatory diagram of one embodiment of a three-dimensional modeling method according to the present invention, and is a flowchart showing a schematic diagram of execution contents with each step added thereto.

As shown in FIG.
In, an arbitrary number of cross-sectional images serving as a basis for three-dimensional modeling are prepared. In this case, the multiple cross-sectional images are cross-sectional images obtained at intervals along a certain direction of the DUT.

Here, 10 obtained at a predetermined interval are obtained.
00 CT images were prepared. Each CT image has a length and width of 20
It is 48 × 2048 pixels, and the data amount representing the density of each pixel is 2 bytes (the total data amount is 8.3 MB).

In step 102, the 1000 CT images prepared in step 101 are appropriately grouped into groups without changing the arrangement. Each group g
1, g2,... GN include at least as many CT images as can be processed in each group by a processing unit (not shown) used when performing subsequent processing.
If a fraction occurs in the grouping, the fractional CT image is assigned to, for example, the last group gN.

In step 103, the group of sectional images grouped in step 102 is sent to the processing means to perform a three-dimensional modeling process, and a three-dimensional model (three-dimensional model part) m1 for each of the groups g1, g2,. , M2, ... mN
Get.

Here, the group of cross-sectional images is sequentially sent to the processing means for each of the groups g1, g2,... GN, and a three-dimensional modeling process is performed for each group.
The three-dimensional modeling process is performed by parallel processing by a single processing unit having a parallel processing function, or by parallel processing by a plurality of processing units, and the processing speed is increased. In this case, parallel processing may be performed on a group of cross-sectional images forming one group, or parallel processing may be performed on a plurality of groups.

In any case, all data to be processed (8.
(3 MB × 1000 = 8.3 GB), the data is divided into groups and processed in 1 / N here. Therefore, it is possible to use a small and inexpensive computer such as a personal computer or a workstation as the processing means. Become. Note that the three-dimensional modeling process is based on a sequential process for a plurality of pixels and cross-sectional images, so that parallel processing is easy.

The three-dimensional modeling process is preferably a three-dimensional modeling process based on the marching cube method described above, that is, by the marching cube method or by applying the same. This is because three-dimensional modeling for each part on the premise of merging can be easily performed as described below.

That is, the three-dimensional modeling process forms a model surface based on a luminance distribution in a three-dimensional space (a rectangular parallelepiped formed by four opposing pixels of two cross-sectional images) arranged in a grid. This is performed by an algorithm for generating a polygonal surface, and the polygonal surface is generated on a grid basis. In the marching cube method, vertices of a generated triangular surface, which is a polygonal surface, are always arranged on a lattice (each side of a three-dimensional space arranged in a lattice). From this, according to the three-dimensional modeling process based on the marching cube method, the obtained three-dimensional model can be easily divided with a line parallel to each side of the three-dimensional space arranged in a lattice as a boundary. .

Therefore, by the reverse method, a plurality of divided three-dimensional models (three-dimensional model parts) can be easily combined. Therefore, this step 1
03, three-dimensional model parts m1, m2,... MN by three-dimensional modeling processing based on the marching cube method.
I tried to get.

As described above, since the three-dimensional modeling process for each part on the premise of uniting is easy and the parallel processing is easy, in the present embodiment, all the processes in the entire process are performed. From the beginning, high-speed processing can be easily performed with a small-sized, low-cost computer.

In step 104, the three-dimensional model parts m1, m2,... For each of the obtained groups g1, g2,.
... For each mN, the above-mentioned polygon reduction processing (processing for reducing the number of polygons constituting the model while maintaining the model shape as much as possible) is performed.

In this step 104, the adjacent 3
The end face of the joint with the dimensional model part is excluded from the target of the polygon reduction processing. FIG. 2 is an enlarged view of the joint end face portions of the adjacent three-dimensional model portions. In FIG. 2, portions indicated by meshes 21 and 22 are joint end face portions that have not been subjected to polygon reduction processing. The reason why the joint end face portions 21 and 22 of the three-dimensional model parts m1, m2,... MN are excluded from the polygon reduction processing target is that the joint assurance between adjacent three-dimensional model parts after the polygon reduction processing is ensured. That's why.

In step 105, the three-dimensional model parts m1 ', m2',... MN 'after the polygon reduction processing are combined in accordance with the arrangement before grouping in step 102 to obtain a single three-dimensional model M.

As described above, in the polygon reduction processing in step 104, the joint end face portions 21 and 22 (see FIG. 2) with the adjacent three-dimensional model are excluded from the polygon reduction processing. That is, the joining end face portions 21, 22 of the three-dimensional model parts m1 ', m2',.
Is not processed at all, and the real model (step 103
, The end face shapes of the three-dimensional model parts m1, m2,..., MN) obtained as described above. Therefore, the joining end face portions 21 and 22 of the adjacent three-dimensional model portions are smoothly and appropriately joined to each other, and the three-dimensional model portions m1 ′, m2 ′,.

In step 106, the simple three-dimensional model obtained in step 105 (the three-dimensional model immediately after merging)
Further, polygon reduction processing is performed on M. If the polygon reduction processing is performed again immediately before the completion of the three-dimensional modeling, the desired data amount, particularly the reduced three-dimensional data amount, can be obtained without impairing the shape of the single 3D model M obtained in step 105 as much as possible. A model M 'is obtained.

More specifically, small polygons are concentrated and arranged in a complicated shape, in other words, high-resolution CT
Highly accurate three-dimensional model M 'reflecting image resolution
Can be obtained with a relatively small amount of data convenient for handling, for example, a data amount of about a general CT image (250 MB).

Similar to the three-dimensional modeling process, the polygon reduction process can be easily parallelized, and the three-dimensional model portions m1 ', m2',... MN 'can be easily combined. Therefore, in this embodiment, high-speed processing can be easily performed by a small-sized and low-cost computer not only from the initial stage in the entire processing process to the three-dimensional modeling process but also to the final stage.

In the above-described embodiment, an X-ray CT image has been described as an example of a cross-sectional image. However, another cross-sectional image (tomographic image) such as an MRI image may be used. Further, the image may be a cross-sectional image stored in a storage device, or a cross-sectional image transmitted in real time from a various cross-sectional image generating device such as an X-ray CT device or a communication line.

In the present invention, since a cross-sectional image is to be processed, the type, size, and the like of the object to be measured that provide the cross-sectional image are not limited.

[0038]

As described above, according to the present invention, when constructing a three-dimensional model using a plurality of cross-sectional images of an object to be measured, the plurality of cross-sectional images are appropriately grouped into groups, A three-dimensional modeling process is performed for each group of cross-sectional images to obtain a three-dimensional model part. Thereafter, polygon reduction processing is performed on each of the three-dimensional model parts, and the three-dimensional model parts after the polygon reduction processing are united according to the arrangement before grouping to obtain a single three-dimensional model.

That is, the three-dimensional modeling process requiring a long time is performed for each of the group of grouped cross-sectional images. It can be performed in a short time by a computer of small scale (small size, low price). Also, since the polygon reduction processing is performed on a partial three-dimensional model (three-dimensional model part), similar to the three-dimensional modeling processing,
It can be done in a short time with a small computer. Furthermore, the three-dimensional modeling process and the polygon reduction process can be easily performed in parallel, and the speed of the process can be increased from this point as well. Such an effect becomes more remarkable as the resolution and the number of cross-sectional images to be processed increase.

Processing related to grouping and merging of three-dimensional models other than the three-dimensional modeling processing and the polygon reduction processing is significantly easier than the three-dimensional modeling processing and the polygon reduction processing. Therefore, according to the present invention,
High-speed processing can be easily performed on a small-scale computer throughout the three-dimensional modeling process.

Further, according to the polygon reduction processing, despite the increase in the data amount and the number of cross-sectional images to be used, a relatively small amount of data which is convenient for handling and the resolution of the cross-sectional image to be used are improved. A three-dimensional model without significant damage can be obtained. In particular, in the present invention, since the polygon reduction processing is performed except for the end face portion related to the joining with the adjacent three-dimensional model portion, the joining end face portions are smoothly and appropriately joined to each other, and the three-dimensional model portion without mutual inconsistency. Can be combined.

By further performing polygon reduction processing on a single three-dimensional model, the reduced amount of data, that is, a relatively small amount of data that is convenient for handling, is maintained without losing the shape of the single three-dimensional model as much as possible. A three-dimensional model of the data volume is obtained.

If the three-dimensional modeling process based on the marching cube method is used for the three-dimensional modeling process for each cross-sectional image group of each group, the obtained three-dimensional model portions can be easily combined. This makes it possible to further speed up the processing.

[Brief description of the drawings]

FIG. 1 is an explanatory diagram of one embodiment of a three-dimensional modeling method according to the present invention.

FIG. 2 is an enlarged view showing joint end face portions of adjacent three-dimensional model parts.

[Explanation of symbols]

g1, g2,... gN CT image group m1, m2,... mN 3D model section m1 ′, m2 ′,... mN ′ 3D model section after polygon reduction processing M 3D model M alone (immediately after merging) polygon 3D model after reduction processing

Claims (3)

[Claims] 1. A plurality of cross-sectional images obtained at intervals along a certain direction of an object to be measured are divided into groups as appropriate without changing the arrangement thereof. A three-dimensional modeling process is performed using the group of cross-sectional images constituting the group, and the obtained three-dimensional model part for each group is separately processed except for an end face part related to joining with an adjacent three-dimensional model part. A three-dimensional modeling method, comprising: performing a polygon reduction process; and combining the three-dimensional model portions after the polygon reduction process according to the arrangement before the grouping to obtain a single three-dimensional model. 2. The three-dimensional model according to claim 1, wherein a polygon reduction process is further performed on a single three-dimensional model.
Dimension modeling method. 3. The three-dimensional modeling method according to claim 1, wherein a three-dimensional modeling process based on a marching cube method is used for the three-dimensional modeling process.