JPH0625658B2 - Three-dimensional measuring device for line segment using bit plane - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Outline] A three-dimensional line segment measuring apparatus for measuring a direction or the like in a three-dimensional space of a line segment forming a stereoscopic image, and a metal appearing in a part of a frame in an image to be measured. Aiming to provide a high-speed operation type device in which the extraction of direction information is not disturbed even by reflection on the surface, an imaging means for capturing a stereoscopic image for each frame while moving the imaging position, and an imaging means Line segment extraction means for extracting S points from the image, mapping function generation means for generating a mapping function based on the extracted S point mapping points, a mapping memory for storing the generated mapping functions, and a bit memory, With the writing of the mapping function to the mapping memory, the bit plane number of the bit memory is made to correspond to the frame number for the pixel on which the mapping function is drawn, and the presence / absence information of the mapping function is stored in the bit memo. Further, it is provided with a writing means for writing each frame, and a filtering means for extracting line segment direction information by filtering with a predetermined threshold value using the contents of the bit memory.

[Industrial application field]

The present invention relates to a three-dimensional measuring device for line segments using a bit plane, which measures the direction of a line segment in a three-dimensional space forming a stereoscopic image.

For example, when operating a robot in a place where human entry is dangerous, such as a nuclear reactor, in order to operate the object of the work with the robot manipulator,
It is necessary to measure the distance to the target and the posture of the target. It is necessary that such measurement can be performed quickly and accurately.

[Conventional technology]

In order to perform such stereoscopic measurement, a camera provided on the robot side performs spherical projection of an object, and the projection image is subjected to mapping processing to obtain various three-dimensional image information such as presence / absence of line segment, direction of line segment, Techniques for extracting the distance to a line segment are known, and there are techniques proposed in, for example, JP-A-60-218183 and JP-A-59-184973.

A method for measuring the direction of a line segment in a three-dimensional space using a spherical map will be described below with reference to FIGS. 5 and 6.
As shown in FIG. 5, the line segment L is spherically projected onto a spherical surface to obtain an arc L '. Each point P ₁ , P ₂ , P ₃ on the line segment L ...
Points P ₁ ′, P ₂ ′, P ₃ ′ projected onto the arc L ′.
, As a circle, which is a mapping function with these points P ₁ ′, P ₂ ′, P ₃ ′ being the center (corresponding to the equator when each point is the north pole) l ₁ , l ₂ , l ₃ ... Draw on top.

As a result, a line segment mapping point S (hereinafter, referred to as S point) at which a plurality of great circles l ₁ , l ₂ , l ₃ ... Cross is obtained. The point S is representative of the straight line segment L, and indicates that the straight line segment L exists.

Next, the position of the camera is moved in an arbitrary linear direction by a constant interval ΔC, a line segment is photographed at each moving point, and the point S is extracted. Here, one image obtained at each shooting point will be referred to as an image of each frame. FIG. 6 shows this state. In this figure, instead of moving the camera, the camera is placed at the center of the sphere, and the line segment L is L ₁ ,
It is shown as being moved relative to L ₂ , L _3, ... Corresponding to these line segments L ₁ , L ₂ , L _3, ... S ₁ ,
It is assumed that the S point of S ₂ , S ₃ ... Is extracted.

When a plurality of S points thus obtained are further spherically mapped to draw a great circle corresponding to each S point S ₁ , S ₂ , S ₃ ...
A direction mapping point SS (hereinafter referred to as SS point) is obtained as an intersection of these great circles. The direction of the vector connecting the center 0 of the sphere and this SS point represents the direction of the line segment L.

In order to measure the three-dimensional azimuth of a line segment in this way, image data over several frames is used. That is, a great circle with the S point of each frame as a pole is drawn, and after drawing for all frames, the SS point is obtained as the convergence point of the great circle and the three-dimensional orientation is measured.

As a method of extracting the SS point as the great circle convergence point, a method of narrowing down the peak point (convergence point) in consideration of the density of the S point is known. That is, when drawing a great circle with the S point as a pole, the density information of the S point is also entered, and after drawing for all frames, the peak point at which the added value of the density is a peak is extracted as the SS point. It is a thing. The density at the point S is a value that increases with the length of the line segment or the clarity of the line segment, and is a parameter that represents the "probability" of the existence of the line segment.

In the conventional method that considers the density of the S point in the extraction of the SS point, the length of the line segment affects the density data, which is effective when processing a clean and clear image. Is the way.

On the other hand, if a metal surface or the like is reflected in the image and the light ray is strongly reflected by specular reflection or the like, the line segment is extracted with a very strong density. Such an image appears only on the screen of a part of a large number of frames, but its density is so large that it becomes a false image that is not to be measured and the image that is actually measured is desired. It confuses and interferes with the extraction of image contours.

The following is proposed as a method for solving such a problem.

First, the overall operation will be described with reference to FIG. Image data I for 10 frames is obtained by photographing the object while sequentially moving the photographing position (here, 10 positions) with the camera device.
M _{0 to} IM ₉ are obtained. These image data IM ₀ ~
The IM ₉ is processed by the contour extraction unit 2 (step S1), and as a result, the contour data P for 10 frames is obtained.
D _{0 to} PD ₉ are obtained.

These contour data PD _{0 to} PD ₉ are further processed by the line segment extraction unit 4 and S point data SD _{0 to} S of each frame are processed.
A total of 10 D ₉ can be obtained. These 10 frames of S point data SD _{0 to} SD ₉ are further input to the three-dimensional azimuth measuring unit 5, and the line segment direction measuring process is performed according to the procedure of FIG. The S point data SD _{0 to} SD ₉ are integrated to obtain SS point data SSD including SS point information indicating the direction of the line segment.

The line segment direction measurement processing procedure will be described below with reference to FIG. Here, a mapping plane SP for writing the mapping function and a cumulative plane IP for binary accumulation of the mapping function are prepared.

First, the contents of the cumulative plane IP and the mapping plane SP are cleared (steps S12, 13). Next, the contents of the first frame of the S point data SD _{0 to} SD ₉ , that is, the contents of the S point data SD ₀ are read, a great circle having the read S point as a pole is generated, and this great circle is generated. The mapping plane S
Write to P (step S14). This operation is performed for all S points in the S point data SD ₀ (step S1).
4, S15). As a result, a grayscale mapped image for the first frame is generated on the mapping plane SP.

Mapping plane S for all S points of S point data SD ₀
When the writing of the great circle to P is completed, the contents of the mapping plane SP are then scanned for all pixels, and if a great circle is present, the cumulative plane IP corresponding to that pixel.
Increments the content of the pixel of (step S1
6).

If this operation is completed for the S point data SD ₀ of the first frame, this time the S point data SD of the second frame
_The same operation as described above is repeated for ₁ , and the operation is repeated up to the S point data SD ₉ of the last frame (steps S13 to S17).

As a result, the content of each pixel of the cumulative plane IP is counted up by the number of frames in which a great circle exists for that pixel.

The SS points are extracted by using the contents of this cumulative plane IP as a frame number filter based on the number of frames (step S18). That is, for example, "5" is set as the threshold value depending on the number of frames, and among the pixels of the mapping result in which the density is taken into consideration, only the pixels of the cumulative plane IP having a value of "5" or more are left as the SS point, and the others. The cumulative value of the pixel of is erased. As a result, the high density S point that appears as a false image only in some frames due to reflection on the metal surface is removed.

[Problems to be Solved by the Invention]

The above proposed method is not efficient because it requires scanning the mapping plane SP for all pixels when incrementing the contents of the cumulative plane IP, which takes a long time.

That is, assuming that a mapping surface of 1000 × 1000 pixels is used, it is necessary to scan 1000000 pixels for each frame even if there is one great circle to be drawn. In the mapping of the line segment distance measuring method in which the “sieve” of the number of frames is effective, the number of great circles is often about 10. In this case, the number of pixels through which the great circle passes is about 200.
It is about 00 pixels, and therefore the number of pixels that need to be scanned is 5 with respect to the number of pixels that need to be incremented.
It is 0 times, so it is extremely inefficient.

Therefore, an object of the present invention is to provide a high-speed operation three-dimensional measuring device in which the extraction of direction information is not disturbed by reflection of a metal surface appearing in a part of a frame in an image of a measurement target.

[Means for Solving the Problems]

FIG. 1 is a principle block diagram according to the present invention. The three-dimensional measuring apparatus according to the present invention includes an imaging unit 11 that captures a stereoscopic image for each frame while moving the imaging position, and an imaging unit 1.
Line segment extraction means 12 for extracting a line segment mapping point S representing a line segment forming a stereoscopic image by using the image obtained in 1 as a spherical projection and a spherical mapping, and a line segment mapping point extracted by the line segment extracting means 12. Mapping function generating means 13 for generating a mapping function based on S
, A mapping memory 14 for storing the mapping function generated by the mapping function generating means 13, and a frame number having the bit plane for each bit position in the depth direction of the data and the number of frames of the mapping function drawn is the frame number. The bit memory 15 is stored for each pixel while being associated with the bit plane number, and the plane number of the bit memory 15 is associated with the frame number for the pixel for which the mapping function is drawn as the mapping function is written to the mapping memory. Then, the writing means 16 for writing the presence / absence information of the mapping function in the bit memory 15 for each frame, and the filtering means 17 for extracting the line segment direction information by filtering the contents of the bit memory 15 with a predetermined threshold value. And.

[Action]

The image pickup unit 11 picks up a stereoscopic image for each frame to obtain an image, and the line segment extraction unit 12 extracts the line segment mapping point S from this image. Then, the mapping function generating means 13 generates a mapping function based on the extracted mapping point S, and writes this in the mapping memory 14 for each frame. Therefore, the image data in which the mapping function is drawn is stored in the mapping memory 14.

When the mapping function is generated by the mapping function generating means 13 in a certain frame, the writing means 16 sets, for example, “1” at the bit position corresponding to the frame number of the bit memory 15 for the pixel on which the generated mapping function appears. Stand up. This process is performed for all frames.

As a result, in the content of the bit memory 15, if a mapping function is drawn for a certain pixel over a plurality of frames,
There are as many "1" bits in the data as there are frames in which the mapping function is drawn. .

When the writing to the bit memory 15 is completed, the filter means 17 filters by filtering the contents of the bit memory 15 with a predetermined threshold value. That is, set the number of certain frames as a predetermined threshold,
The contents of the bit memory 15 in which the number of bits of "1" does not reach the number of frames are excluded from the candidates for the direction mapping point. As a result, only pixels in which the mapping function appears more times than the set threshold number of frames remain, and line segments are referred to while referring to the contents of the mapping memory 14 based on the result of this filtering process. If the direction mapping point SS is extracted, the direction information can be extracted from the viewpoint of using the number of frames as a parameter. False images such as reflections on a metal surface are rarely seen over many frames. On the other hand, we want to extract line segments with low density that are seen over many frames. Filtering is very effective.

This makes it possible to accurately measure an actual image while eliminating the influence of a false image that appears only in some frames, such as reflection on a metal surface.

〔Example〕

 Embodiments of the present invention will be described below with reference to the drawings.

FIG. 2 is a block diagram showing the overall configuration of a three-dimensional measuring apparatus as an embodiment of the present invention. In FIG. 2, the apparatus of the present embodiment is roughly configured to include a camera device 1, a contour extracting unit 2, a host computer 3, a line segment extracting unit 4, a three-dimensional azimuth measuring unit (matching processor) 5, and the like. It

The camera device 1 is configured so as to take an image of an object to be photographed at each photographing position while moving at predetermined intervals in an arbitrary linear direction, and thereby obtain images of a plurality of frames of the object to be photographed. . The number of shooting positions is, for example, 1
About 0 position is selected.

The image captured by the camera device 1 is input to the contour extraction unit 2. The contour extraction unit 2 extracts the contour of the object to be photographed in the photographed image by a known method and supplies it to the line segment extraction unit 4.

The host computer 3 includes a multi-bus 30, a CPU 31, a command interface 32, a data interface 33, D
It is configured to include an MA controller 34, a command interface 35, etc., and controls the operation of each unit.

The line segment extraction unit 2 includes an internal bus 40, a controller board 41, a front stage interface 42, a contour extraction unit interface 43, a buffer memory board 44, and a WCS.
Control board 45, 77 mapping boards 46 (# 0) to 4
6 (# 76), a bus monitor board 47, a data monitor board 48, a rear interface 49, etc., and configures the contour of the object to be photographed based on the contour data transferred from the contour extraction unit 2. The S point, which represents a straight line segment, is extracted for each frame by a known method.
The point data is given to the three-dimensional azimuth measuring unit 5.

The three-dimensional azimuth measuring unit 5 includes an internal bus 50, a controller board 51, a front stage interface 52, a buffer memory board 53, a WCS control board 54, 77 mapping boards 55 (# 0) to 55 (# 76), and a bus monitor board 56. , Data monitor board 57, rear interface 5
It is configured to include 8 etc.

The three-dimensional azimuth measuring unit 5 is a parallel arithmetic computer that executes parallel arithmetic with 77 mapping boards 55 (# 0) to 55 (# 76), and the controller board 51 receives commands from the host computer 3 and interprets them. Then, it has a function of issuing an execution command to another board on the three-dimensional azimuth measuring unit 5. The interfaces 52 and 58 are boards for transmitting and receiving data such as S point. The WCS control board 54 is for updating the microprograms of the mapping boards 55 (# 0) to 55 (# 76). The buffer memory 53 is a board for temporarily storing data, and the data monitor board 57 is a board for displaying and monitoring the data.

A detailed configuration example of the mapping boards 55 (# 0) to 55 (# 76) is shown in FIG. In FIG. 3, 61 to 64 are communication interface units, 65 is a great circle ROM, 66 is a luminance ROM, 67 is a command latch, 68 is an auxiliary code latch, 69 and 70 are adjacent interfaces, 71 is an address calculator, and 72 is an address calculator. A data calculator 73 is a mapping plane RAM.

The mapping boards 55 (# 0) to 55 (# 76) have eight mapping planes.
Have a face. Each surface is referred to as a bank B _{1 to} B ₈ . The bank is designated by using the RAM address.

Two arithmetic units 71 and 72 are mounted on this mapping board, and the address arithmetic unit 71 is an arithmetic unit for calculating the address of the mapping RAM 73. The data calculator 72 is connected to the data port of the mapping plane RAM 73 and processes data.

The luminance ROM 66 is a look-up table when the grayscale data is used as the mapping function. Command latch 67
Is a register for holding an execution command issued by the controller board 51. The auxiliary code latch 68 is a register that holds an auxiliary code for an execution command issued by the controller board 51, and stores a bit plane number here. The adjacent interfaces 69 and 70 are communication circuits for transferring data between adjacent boards.

The operation of the apparatus of this embodiment will be described below with reference to the drawings. The overall flow of processing by the apparatus of this embodiment is the same as that shown in FIG. The line segment direction measurement processing procedure in FIG. 7 is shown in FIG.

In this embodiment, the number of bits of data for one pixel on the image plane is 1.
6 bits, this data is LSB (least significant bit)
The first bit, the second bit, ... Then, each of these bit positions is associated with a frame number. For example, the great circle of the first frame is written in the first bit of data. As a result, the mapping surface appears to have 16 1-bit surfaces. This will be referred to as a bit plane BP, and the bit plane of the nth frame will be represented by BP _n .

Now, it is assumed that the bank B ₁ which is the mapping surface of the mapping surface RAM 73 is assigned to the bit plane BP and the bank B ₂ is assigned to the mapping plane SP.

First, according to an instruction from the host computer 3, the mapping plane RA of the mapping board 55 in the three-dimensional azimuth measuring unit 5
To clear the contents of the bank _{B 1} and _{B 2} of M73 (step S32). This is due to writing data of the data calculator 72 outputs "0", the address calculator 71 while scanning the address of the bank B ₁ and B ₂ "0".

Next, in accordance with an instruction from the host computer 3, the S point data SDn of the _nth frame is received from the data way via the data way interface 52 and stored in the buffer memory 53 (step S33).

Further, the host computer 3 instructs the controller board 51 to cause the auxiliary code latch 68 of the mapping board 55 to
A bit plane number n representing the nth frame is written (step S34).

Next, the host computer 3 issues a mapping command to the mapping board 55 via the controller board 51 (step S).
35). As a result, the mapping board 55 reads the S point data SD _n word by word from the buffer memory 53, and the great circle ROM 65 and the address calculator 71 generate a great circle having a pole at the S point.

Then, of the contents of the bit plane BP _{n with} the number n designated by the auxiliary code latch 68, the value of the pixel on which the generated great circle is placed is set to “1” using the data calculator 72. This processing is performed by ORing the corresponding bit plane BP _n and the bank B ₁ . Bit plane BP _n
Since the depth of 1 is 1 bit, the bit corresponding to the frame of pixels in the bank B ₁ in which the great circle is drawn may be set to "1".

Specifically, to create the image data and the n-th bit only the "1" for pixels that great circle rests, performs an OR operation between this and the bank B ₁ stored data, the bank B ₁ The results By writing to again.

On the other hand, the generated great circle is added to the bank B ₂ as it is.

Next, it is checked whether or not there is a frame to be mapped (step S36), and if there is a frame, the above operation is repeated (steps S33 to S36). As a result, when the processing is completed for all the frames, "1" is set in the bit position corresponding to the frame number of the data of the pixel of the bank B ₁ corresponding to the frame number of the pixel in which the great circle is drawn. As a result, if the data (16 bits) in bank B ₁ is read out for a certain pixel and the number of bits of “1” is counted in that pixel, a great circle will be placed on that pixel over how many frames out of all the frames. You can know what was going on.

On the other hand, steps S33 to S for all the above-mentioned frames
When the processing of 36 is completed, the mapping plane SP is generated in the bank B ₂ .

When there are no more frames, the host computer 3 instructs the controller board 51 to write the slice level in the auxiliary code latch 68 of the mapping board 55 (step S37). This slice level is a threshold value when performing the filtering process by the number of frames, and is set to, for example, "5".

Next, the host computer 3 issues a filter processing command to the mapping board 55 via the controller board 51 (step S38). The mapping board 55 scans the data of all the pixels of the bank B ₁ which is a bit plane by this, and when the number of “1” s of the bit pattern of the read pixel is below the slice level, the mapping board 55 detects the pixel of the corresponding bank B ₂ . Set the value to 0.

After this process is completed, the host computer 3
A data output command is issued to the three-dimensional azimuth measuring unit (mapping processor) 5 (step S39). Thus mapping board 55 scans the bank B _2, when the value of the pixel is not "0", place the addresses and values of the pixel to the internal data bus. The data way interface 58 outputs the data of the internal data bus to the data way. The host computer 3 receives the mapping result via the data way.

Through the above processing, the high density S point that appears as a false image only in some frames due to reflection on the metal surface is removed.

As a result, the obtained SS point is stored as SS point data SSD. This makes it possible to extract the information indicating the direction of the line segment.

〔The invention's effect〕

According to the present invention, the extraction of the direction of the line segment is not disturbed by the false image such as the reflection of the metal surface appearing in a part of the frame in the measurement target image, and the direction information can be accurately extracted. . In addition, since it is not necessary to scan the image plane over the entire screen when filtering to remove false images, the direction measurement operation can be speeded up.

[Brief description of drawings]

FIG. 1 is a block diagram showing the principle of the present invention, FIG. 2 is a block diagram showing an apparatus according to an embodiment of the present invention, FIG. 3 is a detailed block diagram of the mapping board in FIG. 2, and FIG. FIG. 5 is a flow chart showing the processing procedure of direction measurement, FIG. 5 is an explanatory view of a conventional S point extraction method, FIG. 6 is an explanatory view of a conventional SS point extraction method, and FIG. FIG. 8 is a flowchart showing the overall processing, and FIG. 8 is a flowchart showing the procedure of the line segment direction measurement processing in FIG. In the figure, 1 ... Camera device 2 ... Contour extraction unit 3 ... Host computer 4 ... Line segment extraction unit 5 ... Three-dimensional orientation measurement unit 30 ... Multibus 31 ... Central processing unit (CPU) 32, 35 ... Command interface 33 ... Data interface 34 ... DMA controller 41 ... Controller board 42, 52 ... Stage interface 43 ... Contour extraction unit interface 44 ... Buffer memory board 45, 54 ... WCS control board 46, 55 …… Mapping board 47,56 …… Bus monitor board 48,57 …… Data monitor board 49,58 …… Post-stage interface

Claims (1)

[Claims] 1. An image pickup means (11) for picking up a stereoscopic image for each frame while moving an image pickup position, and spherical projection and spherical mapping of an image obtained by the image pickup means (11) to obtain the stereoscopic image. And a line segment mapping point (S) extracted by the line segment extracting means (12) for extracting a line segment mapping point (S) representing a line segment forming the
Mapping function generating means (13) for generating a mapping function based on
A mapping memory (14) for storing the mapping function generated by the mapping function generating means (13); and a bit plane for each bit position in the depth direction of the data. The information on the number of times corresponds to a bit memory (15) that is stored for each pixel while associating the frame number with the bit plane number, and the pixel on which the mapping function is drawn as the mapping function is written to the mapping memory. On the other hand, the bit plane number of the bit memory is made to correspond to the frame number respectively, and the writing means (16) for writing the presence / absence information of the mapping function into the bit memory for each frame, and the predetermined means using the contents of the bit memory. Filter means for filtering line segment direction information by filtering with a threshold value (1
7) A three-dimensional measuring device comprising: