JP2012177671A - Fine aperiodic pattern projection device and method and three-dimensional measuring device using the same - Google Patents

Abstract

PROBLEM TO BE SOLVED: To measure three-dimensional shape of an object that has a smooth surface by projecting an aperiodic pattern that has a unique feature.SOLUTION: An aperiodic pattern composed of two sorts of rhombuses is projected on a work piece and measured with a stereoscopic camera.

Description

  The present invention relates to a three-dimensional shape identification technique. In particular, the present invention relates to a pattern projection / measurement technique for increasing the speed and accuracy of the identification process of the same point in the three-dimensional shape measurement of an object that can be deformed by a camera.

In the linen supply industry, a large number of sheets and towels are washed in a laundry factory, and are rented to customers and collected after use. Currently, many processes such as washing and finishing processes are performed in large-scale washing machines, unfolding finishing machines, roll ironers, folding finishing machines, and the like.
However, the process of first putting the laundry into the machine in the washing line and finishing line processes is performed manually because the technology for handling the cloth has not been established. Therefore, since the work environment in the laundry factory is harsh due to the danger of infection and high temperatures, it is desired to automate the work.
In the past, handling of industrial products without deformation measures the posture of the workpiece to be handled by image recognition using a vision sensor, and the robot hand grips the workpiece and moves it to the desired location according to the posture of the workpiece. There is a system.
However, in the case of measuring the shape in order to develop or classify flexible objects such as towels or sheets such as towels and sheets that have a variable sheet shape to handle, it becomes a standard for deforming cloth. Automatic handling was not possible due to lack of shape data.

Therefore, in order to automate this part, attempts were made to classify the laundry to be put into the laundry line, classify it on the loading conveyor, and automatically put towels and sheets into the finishing line. There are few examples. The cause is that the deformation pattern of the cloth is infinite and cannot be handled only by the set processing method.
For example, when trying to expand a regular towel from a piled-up state, as shown in Fig. 2, a person identifies the end of one cloth, and then pulls out one side from there to hold the entire end with the short ends. It is put into a finishing machine that spreads and folds. To identify the near end, take out one piece and take out the hem part. To expand from there, hold the side around the end and unfold the side. The whole surface can be expanded by lowering.

In order to perform these operations, in Patent Document 1, a combination of a movable holding hand or conveyor for handling a cloth work by the applicant to which the inventor belongs and a passage confirmation sensor such as a photoelectric sensor, it seems to be an end. An operation means for expanding the cloth such as a towel by causing it to appear is disclosed.
As another method, in the past, the inventors identified a part for handling a feature part of a cloth work in a spatial region using visual recognition by means described in Patent Document 2 or Patent Document 3, and the shape of the part. The towel is held and deployed in response to
Specifically, pull up a pile of towels one by one, make the end appear, measure the shape of the appearing end with a stereo camera, grasp a part of the side around the end, and spread the entire towel from there I'm out.
In particular, it recognizes the end shape that has appeared, and when the hand grips and pulls it out, and also holds the periphery of the other side of the long side separately and hangs it, the short side of the other side appears. And the shape of the side around the end is measured to determine the grip pose of the robot hand. The robot hand unfolds and hangs a piece to expand the entire surface.

These means measure the shape of the cloth, recognize the shape, perform the handling operation according to the shape, develop it, and put it into the finishing machine to automate the entire finishing line.
The visual sensor of the visual recognition unit mounted on the system that performs the corresponding operation based on the visual recognition result as described above needs the grip position coordinate information of the workpiece in the three-dimensional space for generating the motion necessary for the robot motion. Become.
Specifically, the accuracy of recognition for operation, robustness, high speed of measurement for productivity, or compact configuration from the cost aspect, and accurate and accurate features of flexible deformations It is required to measure well.
The inventor made and experimented with a robot handling system based on visual shape recognition for the described and other uses. In these systems, a three-dimensional measuring means is used to acquire gripping position information.

As a prior art that can be used for the above-mentioned purpose, Patent Document 4 by the inventor scans slit light with a stereo camera of a tilt optical system and continuously captures the slit light while taking a stereo image by the method described in Patent Document 2. A technique for measuring by scanning is disclosed.
Also disclosed is a means for projecting pattern light having a rough dot arrangement and a fine lattice arrangement described in Patent Document 5 by the inventors onto a workpiece and measuring with a stereo camera. See FIG. 3 (a and b).
Here, (1) the position of the dot on the reference image and the reference image on either the left or right side of the stereo camera is identified, the corresponding points are assigned, and the height is calculated from the parallax.
(2) Using the dot position as a reference position, the position interval between the dot and the surrounding grid line is measured in the reference image and the reference image for each image, and the relative parallax is calculated.
(3) A method for calculating the height at high speed by calculating the height of the reference dot and the relative height (depth) from the reference dot due to parallax.
With this, the deployment device measures the 3D shape of the towel around the edge you want to measure and installs a visual recognition sensor in the deployment device that outputs the 3D information required for gripping the edges. We were able to put it into a finishing machine that automatically folded and finished. In other words, the towel development system has become possible by the feature identification technology of the cloth and the handling technology according to Patent Documents 2 and 3.

  Patent Document 6 is a prior application of a projection technique for three-dimensionally measuring a measurement object having a smooth surface. Although the cloth is not limited to the object here, a random pattern is generated and projected in order to measure a smooth surface measurement object three-dimensionally with a stereo camera.

  JP 2010-222725 A   JP 2009-279700 A   JP 2010-273732 A   JP 2007-225403 A   JP 2009-270915 A   JP 2001-147110 A

In the cloth unfolding method according to Patent Document 1, the deformation of the cloth is irregular, and the unfolding may not be successful when the deformation to the desired unfolding is not performed, and it is difficult to exceed the efficiency of the operator. In the cloth unfolding means according to Patent Document 2 and Patent Document 3, as described above, the appearance position of the workpiece feature that recognizes the image appears at an unexpected position, or the irregular shape changes more than expected. Since it has not succeeded, it has been desired to expand the shape change correspondence range that can be used for image recognition.
In Patent Document 4, since a three-dimensional shape can be measured relatively stably because measurement is performed from a large number of stereo combination images, a recognition sensor for returning a determination result at high speed with a long imaging time required for one recognition. As a result, the number of processes per hour does not increase due to the increase in measurement time.
In Patent Document 5, when the shape of the cloth changes greatly (within a small range), for example, it was not possible to measure the change within the lattice interval, and the dots are shown in both the left and right camera images. Otherwise, shape measurement in the vicinity of the area could not be performed. See FIG.
In addition, when the dot spacing and the lattice spacing were reduced to increase the measurement density, other lattices periodically existed on the search parallax line, which caused mismatching.

Therefore, when trying to maximize the density of measurement points in order to cope with the large number of deformation patterns of the cloth, dots and grids themselves have areas. Dots were registered as templates, and a method of pattern matching on the left and right camera images was studied and summarized below. There is a limit to the number of points that can be measured when a marker having a size m × n, such as a dot or a grid, is arranged among M × N pixels constituting an image. That is, as shown in FIG. 5, M / m or N / n is the maximum number that can be arranged.
This means that a maximum number of arrangements cannot be made when designing a light projection pattern that matches a preset template. That is, there is a problem that the density of the measurement points cannot be further increased. In the case where the edge that exists in the image exists effectively as a solution to such a problem, a partial image cut out from either the left or right image is used as a template. A technique for measuring a parallax by calculating an interphase value on a parallax line of a counterpart image corresponding to a part is known, but it is not possible to measure a smooth surface shape portion as a problem of the present invention as it is. In order to cope with this, in Patent Document 6, a non-periodic pattern is generated by converting a code generated by random number generation into a two-dimensional pattern, but a repetitive pattern that occurs by chance at the stage of random number generation is removed (such as test correspondence). Statistical correspondence) algorithm is required. Also, the random pattern projector is complicated and difficult to make compact.

Therefore, the present invention solves the following problems.
Even a softly deformed object, for example, a smooth sheet-like object such as a cloth, can be used to identify the three-dimensional shape of the surface of the object at high speed and high density with a camera. A unique pattern is projected onto the area within the measurement area.
To make it possible to identify the position of the same point on the left and right images with a stereo camera, project pattern light with uniqueness for each measurement location.
Identifying the shape of a characteristic part to be gripped and handled from the image by the camera, and solving the problem that the robot hand grasps the gripped position and shapes the cloth.

The technical idea of the present invention is to project a pattern to measure a measurement object including a smooth surface when performing three-dimensional shape measurement,
The pattern is to perform a unique pattern projection on the minute surface around the measurement point for any measurement point regardless of the size of the measurement space range.
Filling even a space that extends infinitely in a planar space is called planar filling. The basic figure to be spread is called a tile or cell. There are known types of tiles that can be filled in a single plane, such as squares and parallelograms, but these have periodicity, and the pattern shape of each partial area is not unique among limited partial planes. .
Thus, planar filling of non-periodic patterns using a plurality of tiles is known. Since the direction of the tile is non-periodic, it has uniqueness in a limited partial plane if it has an appropriate size. If a unique mark is created by projecting an aperiodic pattern onto the workpiece surface with a projection device, a smooth workpiece surface can be labeled with a unique measurement position.

Non-periodic patterns are known and there are several types, but Sir. A pattern called Penrose tiling announced by Roger Penrose is generated under the following settings and has a characteristic of an aperiodic pattern.
[Basic properties of Penrose tiling]
(1) As shown in FIG. 6 (a), rhombuses (referred to as thin) and (2/5) π [rad] whose inner angles are (1/5) π [rad] and (4/5) π [rad]. ] And (3/5) π [rad] rhombus (referred to as fat), the space filling on the plane is possible.
(2) The connection method is determined as shown in FIG. When diamonds are combined with other diamonds, solid lines or broken lines are connected with their directions aligned.
(3) [Division] [Integration] [Stepwise refinement] As shown in FIGS. 6B and 6C, one diamond cell can be divided into several smaller diamond cells. Therefore, an infinitely fine pattern can be obtained by repeating the division.
(4) On the contrary, if the coupling is repeated, an infinite space can be filled in the plane.
(5) Fat on an infinite plane: Mathematically known to have non-periodicity because the ratio of the number of two rhombuses of thin is an irrational golden ratio φ = (1 + √5) / 2 It is. FIG. 6D shows that the area ratio when the half of thin is divided into half of fat and thin is the golden ratio. A thin produces one fat and one thin, and a fat generates two fats and one thin. When this is repeated indefinitely, the number ratio is the same as the ratio of two adjacent terms in the Fibonacci sequence converges to the golden ratio. Since the area ratio of one fat: sin is φ: 1, the area ratio in the infinite space converges to φ ² : 1.
(6) There is no translational symmetry (they do not completely match even if they are translated. For example, when a plane is filled by connecting triangles, quadrangles, and hexagons, there is a part that is completely bulky when translated)
(7) The distance between the intersections of the straight lines on the same plane as the Penrose tiling that fills the plane is aperiodic. FIG. 7A is a part of the Penrose tiling pattern of 109 × 52 pixels. Yes, a straight line segment of 35 pixels is set horizontally in the center, and the luminance value of the image on the line segment is plotted. FIG. 7B shows the line segment shifted 2 pixels above the center. FIG. 3c is a graph when the line segment is centered, and FIG. 7C is a graph when the line segment is shifted 2 pixels below the center. (8) Penrose tiling has two colors, the inside of the cell and the edge of the boundary, because it shows a different pattern even with a small amount of movement, so the matching can be done even if the minute region including the measurement point has a linear shape. Of course, as shown in FIG. 8, even if there are four or three colors and no color at the border, all cells can be painted with three colors. FIG. 8A shows an example in which each rhombus is divided into two colors and arranged in a total of four colors, and FIG. 8B shows an example in which the diamonds are arranged in three colors.
(9) [Line] [Intersection] The directions of the lines constituting the cell are (1/5) π, (2/5) π, (3/5) π, (4) / 5) Since there are four types of π, the angle of the boundary line generated from now is 2π / ((1/5) π) at the maximum, so there are ten types.
(10) At the intersection, the tip points of the corners of 3 to 10 cells overlap.

Many of the above features were projected onto the object from the camera that captured the measurement points on the measurement object, which was a problem of triangulation, which was a problem in 3D measurement with previous cameras. By identifying a unique fine pattern, the position of the measurement point on the image can be accurately identified, thereby having a feature that enables light cutting or stereo matching without error.
The present invention uses the characteristics of this non-periodic pattern for three-dimensional measurement, in particular, the pattern projected into the space for measurement can be refined, the refinement can be performed in stages, and unique even if miniaturized. This provides a floodlight device that maintains the fineness and uniqueness required for actual measurement, and in combination with existing 3D measurement devices, it can measure the 3D shape of an object at high speed and with high accuracy. It is characterized by.

Means for solving the problems based on the above technical idea will be described below.
[1] The first invention is
Projecting a pattern onto a three-dimensional shape measurement example figurine, wherein the pattern is a non-periodic pattern, and a unique non-periodic pattern is projected onto a minute surface area around the three-dimensional measuring point Is a non-periodic pattern projector.
[2] The second invention is
The non-periodic pattern described in the first invention is a pattern which is plane-filled with two types of rhombuses having inner angles of 1 / 5π and 4 / 5π and rhombuses having inner angles of 2 / 5π and 3 / 5π. This is a characteristic non-periodic pattern projection apparatus.
[3] The third invention is
The non-periodic pattern described in the first invention or the second invention is divided into a rhombus boundary line portion having a certain width and the inside of the rhombus in the two types of rhombus, one being a light transmitting portion and the other being a light shield A non-periodic pattern, wherein the non-periodic pattern is arranged in series so that the non-periodic pattern on the photomask is imaged on the surface of the measurement object. Projector.
[4] The fourth invention is:
Measurement for obtaining a three-dimensional measurement by acquiring a stereo image with a pattern of a measurement object using the stereo camera, a pair of image sensors constituting a stereo camera, and the stereo camera described in the first to third inventions A three-dimensional shape of an object to be measured by projecting a unique aperiodic pattern onto a minute surface area including a point and performing stereo image processing for assigning corresponding points between a standard image and a reference image of the stereo camera It is a three-dimensional shape measuring apparatus provided with the control part which measures this, It is a three-dimensional shape measuring apparatus characterized by the above-mentioned.
[5] A fifth aspect of the invention is that a pattern is projected onto a measurement object having a three-dimensional shape, the pattern is a non-periodic pattern, and a non-periodic unique to a surrounding micro surface area including a measurement point to be three-dimensionally measured A non-periodic pattern projection method characterized in that a pattern is projected.

The measurement of the three-dimensional shape here means that the computer measures the three-dimensional coordinates of the surface shape in the coordinate system of the camera by a method based on triangulation from the image captured by the camera. The three-dimensional measuring device is a three-dimensional measuring device including a computer having a calculation unit, a control unit, a communication unit, an image input unit, a camera, and the like.
Measurement points for 3D measurement are measurement points on the image of the object surface determined by 3D measurement preprocessing. The surrounding micro surface area including measurement points is a 3D shape measurement point. This is a very small surface area with the area and shape necessary to give unique pattern features to identify where the measurement points are.
A cell is a basic figure constituting a non-periodic pattern, for example, two basic rhombus shapes.
“Unique” means that a pattern extracted from a minute surface area around any point in the search range is unique.
The search range is, for example, in the reference image when the height of the measurement point from the camera is changed without changing the position on the standard image within the range of the measurement space in a known stereo camera in the past. This is the locus of the measurement point and the range near it.
Or in measurement with a single eye such as the light section method, the measurement point trajectory in the form of a camera image when the projection angle is fixed and the height of the surface of the measurement object is changed from the camera. And its vicinity.
When the correlation value is calculated from the point group constituting the search line on the reference image, the image of the minute surface area in the reference image extracted by being unique in the search range shows the highest correlation value. The position is the position where the object measurement point is shown on the reference image.

According to the present invention, a unique pattern can be projected onto an object surface including a smooth portion.
A three-dimensional shape can be identified by measuring the pattern image with a camera and an image processing unit.
Smooth surface can be measured at high density in 3D.
By measuring only the necessary parts in three dimensions, it is possible to speed up the shape identification necessary for object handling.
FIG. 16 shows a sample image when measuring the shape of the end of the suspended towel as an example of the measuring method. (A) and (b) are the right reference image and left reference image of the stereo camera.
In (d), in order to measure the measurement points near the end, a minute region including the measurement points is stored as a template from the image. (C) shows that a point having a high correlation value is calculated by a matching algorithm such as normalized correlation on an epipolar line that is a search range on the reference image.
(F) is an enlarged image of a template image including measurement points extracted from the reference image.
(G) is a distribution of correlation values on the epipolar line. Here, it was shown that the correlation value at the hardest white point was about 0.8.

The present invention effectively functions as a visual recognition sensor mounted on a handling robot. Especially for handling flexible objects, it is suitable for measuring the shape of flexible objects at high-speed and high-density measurement points. A visual recognition sensor of a robot that handles a freely deforming object has its own origin and coordinate system, and calculates the coordinate value of a measurement point for three-dimensional measurement.
According to the present invention, if a fine non-periodic pattern is projected regardless of a complicated relative distance measurement method, the position of a corresponding point is directly identified at a desired location even on a smooth surface with a considerable success rate. Therefore, measurement can be performed at high speed and with a high success rate, which is preferable as an embodiment.

In another case, the customer wants to grasp the highest point in order to pick up laundry that is thrown in piles, or in recent years, a laundry factory wants to classify collected laundry into different types and put it in a washing machine. There is an expectation that the laundry taken out one by one as shown in FIG. 10 is flowed on the conveyor, and the three-dimensional shape is measured and the volume is calculated at the time of passing to determine the classification based on the volume. In such a case, it is required to measure the three-dimensional shape of the entire laundry that is stacked or separated into one sheet.
In these cases, the surface of the measurement object has a smooth surface and edges due to steps, and the stepped part including the edges that can be matched without special projection is the third order of stereo cameras. Although the original measurement is possible, it is more preferable that the present invention measures the height at a high density even if the measurement point is on a smooth surface even if the surface area around the measurement point is small.

The most preferred embodiment of the present invention is a three-dimensional measuring device, a presentation handling unit that presents the three-dimensional measuring device so that the three-dimensional measuring device can measure a portion presented by deforming the taken-out cloth, and FIG. As shown in FIG. 9 (d), the cloth unfolding apparatus is configured to grasp the cloth taken out by the handling robot based on the measurement result recognized by the three-dimensional measuring apparatus and to unfold the cloth as shown in FIG. It is a three-dimensional measuring apparatus as shown in FIG.
The basic elements constituting the present invention in such an embodiment will be described below. In addition, although this invention is a projection apparatus and a three-dimensional measuring apparatus, in order to demonstrate the most preferable embodiment, it demonstrates with the periphery part.
[Projector]
The projection apparatus characterizes the present invention. In order to project the non-periodic pattern, measurement can be performed by connecting to a personal computer or the like, or by using a projector product that projects an image stored inside alone.
As a more preferred example, the projection apparatus of the third invention may be used in which a projection pattern is prepared as a photomask and a light source and a projection lens are arranged in series in order to reduce or simplify the projection apparatus.

[Projector, LCD, DLP]
When the present invention is implemented with a projector product, a commercially available projector capable of projecting a Penrose tiling image stored in a personal computer by liquid crystal or DLP (not shown) is used. For the image to be projected, prepare multiple Penrose tiling images with different fineness before and after processing by dividing the Penrose tiling projected on the workpiece to be measured, and select the image to be projected on the computer For example, the fineness of the cell can be changed.
[Color pattern]
As another advantage, a color pattern can be projected, so that when a color camera is used as a measurement camera, a minute region can be featured by a color feature. As shown in Fig. 8 (b), Penrose tiling is used to generate a pattern in which all cells are painted in three colors, for example, red, green, and blue, and a stereo image with a pattern captured by a color camera is red. In the image separated into each component of (R), green (G), and blue (B), the area around the measurement point as shown in FIG. 8 (c, d, e) is associated with each RGB image ( Matching) can be performed more accurately than the matching using the gray image by the transmission part and the light shielding part in FIG.
[shutter]
In recent years, some projectors are capable of projecting independently without being connected to a personal computer, and this may be used. A shutter may be prepared separately if necessary.

[light source]
As shown in FIG. 11, as a projection means, more preferably, as a visual recognition sensor, a high-intensity light source, a photomask, and a projection lens are arranged in series for miniaturization. 11 (a) or 11 (b) connected to a power supply unit, or a metal halide lamp as shown in FIG. 11 (d), a xenon flash lamp (not shown), an electrodeless plasma light source (LIFI light source) as shown in FIG. Etc. are used.
In particular, when using LEDs, one or more LEDs are arranged in a ring shape, and the ball lens is placed on the light emitting surface of the LED so that the light emitted from each LED efficiently passes through the mask and enters the projection lens. It is advisable to adopt an arrangement that is close to the center of the lens and slightly shifted toward the center of the lens.
The LED drive section preferably has its own power supply and a control section for causing strobe light emission in accordance with the imaging timing of the camera. Preferably, a camera having an external signal for strobe control is used, and a trigger signal is sent to the LED drive power supply unit in accordance with the imaging timing to capture an image during light emission.
If the light source has no directivity due to the directivity of the light source, it is arranged so as to cover the light source with a condensing mirror, preferably a parabolic mirror as shown in FIG. The incoming light flux may be increased. Or you may arrange | position the lens for condensing between a light source and a photomask.
Since the light source and the light source drive power supply section are likely to have heat, a cooling fin or fan is attached around the light source to dissipate the heat by passing outside air in order to dissipate heat.
Preferably, a high-intensity LED is combined with a control unit that performs lighting control in accordance with the imaging timing of the camera and a power source unit in extracting pattern light described later.

[Photomask]
The photomask is a Penrose tiling pattern composed of a light-shielding part and a transmission part. For example, a chrome deposited on a glass substrate is projected through a projection lens. Since the photomask has heat resistance, there is no problem even if heat is generated using the light source or the like.
If there is no problem such as heat, for example, a PT printed on a commercially available 35 mm film camera and slide film may be imaged and the developed slide may be used instead of the photomask for convenience.
[pattern]
In addition to the combination of fat and thin, there are two types of non-periodic Penrose tilings, called dirt and kite, which are flat-filled.
In the image processing, when a cell projected on the surface is taken out as a template and stored in a memory, a smaller template size is better for speeding up the processing. Therefore, Penrose tiling with a combination of a simple tile and a cell-shaped fat and thin is more preferable.
[Cell, border]
The pattern to be deposited is produced as CAD data. Either the boundary or the inside of the cell line of the Penrose tiling may be used as the light shielding part. Actually, since the contrast varies depending on the degree of reflection intensity of the background and pattern light workpiece, it is preferable that the boundary portion is made of a light shielding portion and a transmission portion so that the selection can be made.
[Origin marker]
It is preferable to place an origin coordinate marker indicating a reference point for projection in addition to Penrose tiling and other points near the center.
As a result, the measurement origin of the three-dimensional shape measurement apparatus can be recognized as a point that is separated by a specified distance from the camera and the origin coordinate marker is reflected, for example, after being attached to the robot system. Even when the measurement is performed in a space that is not on the reference plane, the coordinate values are easy to grasp. There is an advantage that it can be easily performed when the coordinate system of the sensor itself and the coordinate system of the robot are matched or coordinate-transformed.

[Projection lens]
As for the positional relationship between the photomask and the projection lens, the photomask is arranged at a position where the projection lens forms a real image from the measurement object.
Projection lens specifications are bright and have few aberrations.
If the image circle of the lens is larger than the photomask, the focal length of the lens, the effective size of the photomask, the working distance, and the field of view in the measurement area have the following relationship.
When projecting a pattern onto a work with the projection lens aperture open, the pattern becomes blurred as the work position moves away from the focal point. Similarly, the imaging lens attached to the camera is out of focus. Even in such a case, PT expresses a different pattern while being blurred, so it is preferable to enlarge a measurement region for measuring a smooth surface, particularly a region in the depth direction of the camera.

[camera]
A camera is used that obtains an image by an image sensor, communicates with a computer to transfer image data, and communicates with a control device to control an imaging timing.
Examples of the image sensor include a CCD (charge coupled device) and a CMOS. As a means for communication with a computer, a digital camera or an analog camera such as NTSC that communicates image data according to a communication standard such as camera link, USB, Firewire (IEEE 1394), or GIG-E is used. The camera arranged in the vicinity of the robot preferably has a digital camera specification in order to eliminate the influence of noise. Camera links are preferable when capturing a large number of images at high speed. Depending on the communication mode of the camera, a frame grabber board is connected to the personal computer, and image data from the camera is stored on the grabber board (in the case of a camera link or analog camera). An image processing unit or the like that can be directly connected to a camera may also be used.

[Stereo camera]
The camera used as the three-dimensional shape measuring apparatus is preferably a stereo camera as shown in FIG. An aperiodic pattern projector is disposed between cameras having two image sensors, and a unique pattern is projected onto the surface of a measurement object for shape measurement with a stereo camera. Here, it is not particularly necessary to integrate, and measurement is possible if the pattern is easy to hit.
Further, there is an advantage that the influence on the measurement error is small even if the position and orientation of the projection apparatus are changed.
[Integrated light source]
The advantage of integration is the compactness of the equipment. The light source and the camera may be integrated, and one of the cameras may not be used, and the non-periodic pattern may be used as a light projecting device in the light cutting method. In particular, there is little change in the tilt of the workpiece surface, and it is possible to measure a measuring object moving in translation without using a stereo camera.

[Light cutting method]
In general, the light cutting method irradiates a spot light such as a laser or line light with a small constant width, and increases the angle between the optical axis of the camera and the line light, etc., and the reflection position in the image form such as the line light measured by the camera. Measure the depth (depth), but describe the difference.
In rice projection lens the light source behind assuming for example I ₁ shown in FIG. 3 (b) and photomask aperiodic pattern in the present invention, to project the non-periodic pattern of small area near unique point P1 replacing the light projecting device can be calculated when searching the search range between Q ₁ Q ₂ 'on the image corresponding to the I ₂ height (depth). This is because the ratio of the inter-image distance between the measurement point and Q ₁ and Q ₂ is proportional to the distance between the measurement point in the depth direction and Z ₁ and Z ₂ .
In the past light cutting method, the spot light or the line light is searched. In the present invention, a partial region of the aperiodic pattern projected onto the work is used as a template. The template used for the search is a small area around the measurement point of the measurement object to be measured, and the image to be searched is obtained by projecting an aperiodic pattern onto a plane obtained by imaging in advance on the reference plane of z1 and z2 points. It is an image.
The template matched with the measurement image and the comparison image projected on the near or far calibration reference plane corresponds to the parallax of the measurement by the stereo camera, so the same method as calculating the height from the parallax You can calculate the height and depth from the camera.

Here, a calibration image in which steps are provided between the far and near calibration reference planes and other calibration reference planes are equidistantly stored in advance in the measurement region.
The non-periodic pattern in the minute area around the P1 point projected at an angle with the optical axis of the camera is projected on the far calibration reference plane from the near calibration reference plane to the far calibration reference plane in one step. If you look at the position on the camera image when moving, it will move linearly. That is, all points move.
Here, since the line width is provided in the pattern, the following conditions are satisfied if the parallax corresponding to the depth movement amount and the line width in the parallax direction of the pattern are made the same in providing steps at equal intervals.
(1) Acquire an AND image of the calibration image and measurement work image provided for each binarized stage. (2) If a diamond is formed, the work is in the vicinity of the position of the calibration image provided on that stage sheet. There is a measurement point on the surface. (3) If the diamond is missing, there is a measurement point on the workpiece surface near the position of the calibration image provided on that stage. It is possible to classify and measure by shape height hierarchy.

It becomes a compact three-dimensional measurement system by integration.
Integration makes it easy to determine the measurement coordinate system and origin of the camera itself. In cases where the robot system is mounted on a robot system, once mounted, the reference plane for calibration may not be set by the robot or other peripheral devices. If they are integrated, a calibration measuring stand can be separately provided and configured there, and there is an advantage that conversion to robot coordinates is easy even when the robot is mounted with the origin marker or the like.
The advantage of using the light cutting method is that only one camera is required, but in order to achieve the same error as a stereo camera, it is necessary to project at an angle so that the height change of the measurement object can be easily measured. For this reason, it is desirable that the angle between the optical axis of the camera and the optical axis of the projection device be set equal to the angle between the two lens optical axes of the stereo camera.

[trigger]
The imaging timing of the camera is controlled by a trigger signal given from a computer via the communication means. Alternatively, imaging may be started and stopped by giving a trigger timing to the input terminal of the camera from another control device for giving imaging timing.
For example, the control device needs to take an image with a certain interval when measuring a workpiece flowing on the conveyor with an area camera. An encoder is arranged on the conveyor, and pulses generated by the belt movement of the conveyor from the encoder are generated. Counting is performed by the control device, and when the predetermined number of times is counted up, a trigger signal is transmitted to the camera to perform one-time imaging and transfer data. By repeating this, imaging at regular intervals may be performed. A PLC (programmable logic controller, sequencer) or the like may be used as the control device.

[Image size of camera]
As shown in FIG. 14, the image plane size and the number of pixels of the image sensor are determined from the size of the area to be measured and the size of the template.
If the field of view is W _FOV , the working distance is WD, and the focal length of the imaging lens is f, the image plane size of the imaging element is W _FOV × f / WD.
"Camera resolution"
The number of pixels of the camera is determined from the required error. The object plane size corresponding to the image plane size of 1 pixel is a criterion for the required error. When the XYZ axes are set in the camera coordinate system as shown in FIG. 14, the XY direction is set to have an error of 1 pixel or less in the image. The error in the Z direction is set such that a value obtained by multiplying the error per pixel by the ratio of the binocular distance and wd of the stereo camera is an approximate measurement error.
[Template size]
The size of the aperiodic pattern is determined from the template size. Corresponds to an image area of M × N pixels when the template is unique. The size on the object plane at the working distance is m × n (mm). Although the preferred number of cells included in the search area varies depending on the area of the search area, the number of cells is increased when the accuracy of the search is expected. To reduce the calculation amount for speeding up, the number of cells is reduced.
In this case, it is needless to say that the size is required to distinguish between the inside of the cell and the boundary portion because of M × N pixels.
When the pattern projected from the projector is a color pattern, it is preferable to use a color camera to identify color information.
When the pattern projected from the projector is a monochrome pattern, it is preferable to use monochrome in order to measure a shadow with high sensitivity.

[Computer]
The camera is connected to a computer such as a personal computer or an image processing unit.
The computer includes a CPU, a storage, a memory, and the like, performs an operation related to image processing on an image stored in the memory, and transmits an identification result to a robot or a development processing device. At this time, the data may be transmitted to another control device, and the other control device may convert the data necessary for the robot operation and transmit the data to the robot.
[Image processing unit]
The computer may be an image processing unit that performs dedicated image processing. The image processing unit has the same function as that of the computer, and has a function of communicating with other control controllers, so that the reliability is improved as compared with a computer whose operation at high temperature operation is not stable.

[Integration options]
It is preferable to integrate the camera with a camera using a small image processing unit, because the area occupied by the control unit is eliminated by changing to a computer of the control unit, thereby reducing the size.
At this time, the image processing unit separately transmits image information and image processing according to a command from the integrated control apparatus, and transmits transmission information from the shape measurement result to the control apparatus or the robot controller.
When three-dimensional measurement is performed with a stereo camera, the light projecting device may be integrated or may be arranged at another position.
When integrated, a high-brightness LED or a xenon flash lamp may be used as the light source. The LED may be always turned on, or normally, the heat generation is suppressed by projecting light in synchronization with an imaging trigger without projecting light. The synchronization signal may be used to drive a xenon flash lamp

[PLC, PC]
The computer, the robot, the developing device, and the peripheral device that perform the image recognition may directly provide communication means. In addition, it may be connected to a PLC (sequencer) or a control personal computer that controls these in an integrated manner [robot] [peripheral equipment] [deployment device]
The robot or the developing device performs handling in cooperation with peripheral devices based on the image recognition data from the computer. About a connection method, it is good to carry out by the method of patent document 5 by an inventor.
Examples of the present invention will be described below, but the present invention is not limited to the examples.

The first embodiment relates to a handling system for a cloth such as a washed towel which operates at an observation field of view of 780 × 540 mm on a near calibration reference plane and a working distance of 1300 to 1500 mm.
In the developing device shown in FIG. 9 of the past inventors, the slider robot hand grips the inner end of the four corners of the towel, and pulls it out obliquely while holding the vicinity of the counterpart on the long side of the gripped end. As a result, the periphery of the opposite end of the short side of the gripped end is developed and suspended. The projection device according to the present invention can project aperiodic pattern light on the surface around the end portion. The projection device is integrated with the camera, making it compact by eliminating the space for the projector. The short side is grasped from the shape information of the corner necessary for grasping the cloth edge from the measurement result of the cloth edge shape from the three-dimensional measuring apparatus according to the present invention. Then, both ends of the short side are squeezed out and unfolded and put into a conveyor.
The imaging device uses a visual sensor composed of a stereo camera as shown in FIG. 13 using a stereo camera. A tilt optical system stereo camera in which the lens optical axes are arranged in parallel is preferable because it eliminates the need to remove perspective distortion by software.
In addition, it is preferable to have the adjusting mechanism described in FIGS. 13A to 13C because the corresponding scoring can be performed in the horizontal direction of the image. The interval between the imaging elements of the stereo camera is 500 mm.
As the imaging lens, a lens having an image circle of about 10 mm and a focal length f = 8 mm was employed.

Two planes that are the farthest and nearest to the measurement range are defined in advance. The two planes are parallel. The plane is arranged vertically on the imaging lens on the reference image side.
In the tilt optical system, the two imaging lenses are arranged in parallel. There is an advantage that the effective accuracy increases because the area where the fields of view of the left and right cameras that allow stereo measurement in the measurement area overlap increases.
As an example of the calibration method, a calibration pattern or a checkered pattern in which dots are arranged at intersections of a square lattice is temporarily arranged on a near plane (not shown). The tilt camera is adjusted so that when the near plane is captured by the left and right cameras, the same image is obtained. Alternatively, by cutting unnecessary portions around the periphery and selecting a part to be measured, the images after selection are made the same.
The tilt optical system camera used in the present embodiment employs a camera with three adjustment mechanisms described in Patent Documents 4 and 5, that is, a shift adjustment in the horizontal direction, a rise adjustment in the vertical direction, and a rotation adjustment in the rotation direction. In the present invention, one or either camera may be made shiftable, but here both cameras are made shiftable and adjusted so that both shift amounts are the same on the near reference calibration plane.
In order to obtain a tilt optical system camera having such a calibration, it is preferable to use a board camera provided at a board level with an image sensor mounted on a substrate.
The board camera uses a USB 2.0 communication standard digital camera, but it may be a separate communication standard. The image plane size is 4.8 mm × 3.6 mm, and the resolution is 640 × 480 pixels.
When the camera cannot be adjusted in this way, affine transformation is performed so that the same image is obtained by image processing on one calibration reference plane, and processing is performed using the image converted with the conversion parameter.

[Maximum search range length]
The number of pixels of the difference between the dot appearance positions of the left and right cameras when the dot pattern is moved to the nearest plane is set as the maximum search length.
To summarize the basic conditions of this embodiment, the working distance is 1300 to 1500 mm, the visual field is 780 mm at the near calibration reference plane, the visual field is 900 mm at the far calibration reference plane, the binocular distance is 500 mm, and the near calibration reference plane per pixel is Since the corresponding size is 1.21 mm, the pixel size searched by the reference image when one camera is used as the standard image is the appearance of the parallax from the standard point position of the micro area including the measurement point on the image extracted from the standard image. It is within the range of the position moved about 52 pixels in the direction. In the case of a tilt optical system, if there is no lens aberration, matching only in the horizontal direction may be performed, but a rectangular search range having a width of several pixels is set in consideration of aberration.
From the search range, the height can be calculated from the length of the epipolar line connecting the measurement points on the standard image and the measurement points on the reference image. The positional relationship between the near reference calibration surface and the measurement coordinate system of the camera is known. Since the parallax when the measurement point is on the far calibration reference plane is 52 pixels, if the parallax of the measurement point between the 200 mm distance between the near and far calibration reference planes is K (K / 52) × 200 is the height away from the near reference plane. (See Patent Document 5)
In the optical system of the present invention, even if the size on the XY plane is the same size depending on the distance of the measurement object from the camera, the size on the image varies proportionally with the distance. The size in the XY direction can be calculated by proportionally calculating the scale in the XY direction from the camera coordinate origin from the height of the measurement point obtained by the above formula. For example, a field of view of 780 mm on the near calibration reference plane has a size of 900 mm on the far calibration reference plane, and therefore the X and Y coordinate values can be calculated by correcting the distance from the origin by the ratio.
These are the advantages of the method using the optical system as described above. However, in other cases, measurement may be performed using correction by calculation if conversion such as affine transformation is performed.
In the above-mentioned Patent Document 5, a predetermined dot pattern is first searched for by three-dimensional measurement using a stereo camera. In the present invention, the following method is preferable.

A point to be measured for the three-dimensional shape is identified in the range of the measurement site from the reference image. As described in the above-mentioned patent document, the shape obtained by separating the background and the work area by binarization processing in the preprocessing is used to track the contour line while acquiring the curvature to identify a portion having an acute angle close to 90 degrees. Thus, the position of the end gripping point can be identified.
Further, a point separated by the hand grip interval (short side grip point) on the short side connected to the identified end point is identified. Next, similarly, a point that is set apart on the long side (grip direction reference point) 1) or on the normal of the line connecting the short side gripping point and the end gripping point, which originates from the short side gripping point, the short side is a point (grip direction reference point 2) that is a set length away from the point in time. Measure the position. You may use the well-known method of the patent document 5 reference.

For the measurement points on the reference image determined by the preprocessing, the three measurement point positions and the minute surface area around the measurement points are stored as a template image.
The template shape is calibrated with a square or cross shape made up of (n × m) pixels. The template size can be dealt with by the presence of a plurality of cell boundary portions of the aperiodic pattern existing in the template.
At this time, in the sensor coordinate system in which the surface of the measurement work expresses the three-dimensional measurement value of the stereo camera shown in FIG. 14, any template shape may be used when the normal direction of the minute region faces the camera direction (+ Z). .
For example, when there are many cases where the normal direction is rotated about the X axis from the direction in the change in the shape of the workpiece, a quadrilateral close to a horizontally long linear shape may be used.
Similarly, for example, when the normal direction often rotates around the Y axis from the direction, a square that is close to a vertically long line may be used.
The template size is set to a size that can maintain uniqueness in matching points within the maximum search length of the epipolar line.

The size of the pattern projected on the template part was determined so that the image projected on the object plane by the transmissive part and the light-shielding part of the pattern gave enough contrast to be identified by the camera. Here, 0.154 mm is set as a side size, the thickness of the line is 0.04 mm, and the photomask is flat-filled with two types of rhombus Penrose tilings of fat and thin.
The length of one side of all the rhombuses in the tiling image projected on the object plane at this time is 23.1 mm on the near reference calibration plane.
In the optical condition, it is preferable that a part of cells (tiles) constituting a pattern in this range per pixel include a plurality, 3 to 6, or more.

According to this embodiment, the maximum search length is 52 pixels or more, and the length of one pixel on the projection plane is 1.21 mm. Since the size of the one side on the near calibration reference plane is 23.1 mm, at least one side needs to exceed 19 pixels in order to obtain a template size exceeding this. If the template size has 2 or 3 times the number of pixels on at least one side as the number of pixels completely covering one cell, the uniqueness can be maintained in matching between the 52 pixels of the maximum search range. .
In order to increase the speed, 1 / n scale conversion of the image may be performed in the preprocessing and the measurement may be performed.
At this time, the length of one pixel on the projection plane is multiplied by n by the 1.21 mm scale conversion. Therefore, when searching with the same template size, the correlation value increases and more accurate measurement can be performed. If the size is reduced by the scale, the speed is increased. FIGS. 16A to 16G illustrate the processing method of the sample image described so far.

In the above-described matching (template matching), if a known subpixel method is used, the measurement error can be reduced to a unit equal to or less than the pixel interval.
In the embodiment, a light projecting light source integrated with a stereo camera is preferable for compactness, but the present invention is not limited to this.
This is because even if the projection device is placed away from the camera so that the pattern projected on the camera can be easily captured due to the positional relationship between the camera position and the surface of the measurement object, there is an advantage that the calibration of the camera is not affected. .

In Patent Document 5, a preprocessing method for acquiring only a pattern projected onto a measurement object is used.
This is a method of picking up only two types of pattern light by taking two types of images, a pattern projected onto a stationary measurement object and an image without a pattern, and extracting the difference between them, and extracting only the pattern light reflection part by binarization. . If a computer and a projector are connected and the imaging trigger of the camera is also controlled, an image obtained by projecting an aperiodic pattern is first captured, and then the projection of pattern light is interrupted to capture the measurement object. The difference image can be acquired. As a result, more robust pattern light can be extracted.
When there is a problem that the measurement time increases due to an increase in the number of times of imaging when acquiring the difference image of the pattern light, the dynamic threshold method is used. In the dynamic threshold method, only a pattern projection image is used. Therefore, the number of times of imaging is one. This method is effective when the contrast of the captured pattern image is strong. The processing process is as follows.
(1) Acquisition of patterned work image and saving as image 1 in memory (2) Applying averaging filter and saving as image 2 in memory (3) Differential processing of image 1 from image 2 and saving as image 3 in memory (4) Pattern 3 is extracted by binarizing image 3

The second embodiment is a system for performing classification / counting for each type before or after washing. Here, the sheets, pillow covers, and futon covers that are placed on the conveyor in a piled state are taken out one by one, and in addition to the type determination based on the load change when lifting, they are transferred to the conveyor after being lifted. Classification is also performed by measuring the volume of one piece of laundry when passing through. At that time, the color of the laundry is identified by a camera or the like.
The classification judgment based on force and the classification judgment based on vision, such as load change and inertia change, determine in advance which classification judgment is to be weighted for each item, and the probability in each classification judgment. Add the numerical value of to make a comprehensive judgment. Alternatively, each measurement data is held as vector information, and the determination is made based on the distance between vectors with the teaching data for each item.
For measurement by force, the load cell is measured at a short sampling rate in order to measure the load change at the time of ascent. In order to make the measured value of the load cell accurate, the pick-up / pull arm is provided with a robot mechanism, which is a parallelogram link machine groove, so that the measured value does not change depending on the position of the workpiece.
In this way, if the item type can be known in advance based on the determination based on the force, it is classified based on that alone. For items that do not change in weight and are difficult to understand, comprehensive judgment is made by measuring with a camera on the conveyor. When it is possible to accurately determine the measurement items that can be visually determined, such as the difference in color, the determination is made.
Furthermore, when it is still unclear, in such a method, the projection device and the three-dimensional measurement device according to the present invention image the items passing on the conveyor at regular intervals and perform stereo measurement. When imaging on a conveyor that moves at such a high speed, it is possible to determine the measurement points for measuring the three-dimensional shape by identifying the position of the workpiece. If you simply add the heights of the points, the approximate volume can be calculated at high speed.
If the volume is known, the specific gravity and the like can be known together with the weight data, so that it is possible to classify those that cannot be classified by weight alone.

The work on the conveyor is picked up when it is moved at a fixed interval, and transmitted to a computer that stores the image. Here, a camera-link digital camera was used. A stereo camera was used with the optical axes set parallel to each other, and measurements were taken using an optical measurement system. Calibration is performed in the same manner as in the first embodiment, and pattern images on both the far and near sides are acquired and the search range in the reference image is set. The reference image was acquired with the right camera and the reference image with the left camera.
The computer has a frame grabber for connecting to a camera link digital camera, and stores image information from the camera. The measurement points to be imaged are set as intersections of lattices with a constant interval in the XY direction on the conveyor as measurement points on the reference image, and matching is performed on the reference image using a peripheral region including the measurement points as a template. The dots shown in FIG. 15 are sample results in which the position of the measurement point and the measured height are represented by shading.
The height data is output as 0 on the conveyor. Since the volume of each grid area is approximately proportional to the height of the measurement point, the approximate volume can be obtained by calculating the sum of the height data in the measurement range and applying a conversion count.
Each sheet is taken out and passed through the measurement conveyor, and is stored on the flap downstream of the conveyor while the determination is made based on the finally measured data. Classification is possible by moving the conveyor so that the two slide conveyors flapped according to the judgment result can be dropped or ejected toward the classification car for each item, and rotating the conveyor so that the rotation direction is directed to the car Become.

Although the present invention was able to make the latter method by image recognition more compact, faster and more reliable, the applicable fields of 3D measurement are wide-ranging and limited to shape measurement technology for fabric handling or robot handling Not what you want.
For example, measurement of buildings, human bodies, industrial parts, agricultural products, etc. is one of the application fields.

It is a schematic diagram of the present invention. It is a figure which shows an example of an operator's cloth towel expansion | deployment method. FIG. 11 is a schematic diagram of a stereo camera described in Patent Document 5. (A) is a perspective view of a stereo camera and a light source. (B) is a schematic diagram of a depth measurement method using parallax. It is explanatory drawing of the subject of the measurement in patent document 5. (a) is a reference | standard image. (B) is a reference image. It is explanatory drawing of the density limit of the measurement point in patent document 5. FIG. It is a schematic diagram of Penrose tiling with two types of rhombuses that are non-periodic patterns. (A) is a figure which shows a rhombus shape and connection prescription | regulation. The left figure is fat and the right figure is thin. (B) It is a figure which shows the division | segmentation of fat and thin. (C) It is a figure which shows the division | segmentation and expansion of Penrose tiling. (D) It is explanatory drawing that the area ratio of Penrose tiling becomes an irrational number. It is a figure which shows the uniqueness of the intersection distribution of Penrose tiling and a line segment. It is explanatory drawing of the pattern which color-coded Penrose tiling. (A) It is explanatory drawing of the method which divides | segments fat and thin into each and paints with four colors. (B) It is explanatory drawing divided with three colors of red (R) green (G) blue (B). (C) It is the figure which extracted the red area | region. (D) It is the figure which extracted the green area | region. (E) It is the figure which extracted the blue area | region. It is explanatory drawing of the automatic expansion | deployment system of a cloth. (A) It is a perspective view of the automatic expansion | deployment system of a cloth. (B) It is a front view of the automatic expansion | deployment system of a cloth. (C) It is explanatory drawing of the holding | grip operation | movement of the robot by the measurement result of a three-dimensional shape measuring apparatus. (D) It is explanatory drawing of the discharge process by the cloth expansion | deployment operation | movement of a robot. It is explanatory drawing of the laundry classification system. (A) It is an aspect of the laundry classification system. (B) It is input side explanatory drawing of the laundry classification system. (C) It is classification discharge side explanatory drawing of the laundry classification system. It is explanatory drawing of the projection apparatus by this invention. (A) A light projecting device using LEDs. (B) It is a cross section of a light projection apparatus using LED, and is perspective explanatory drawing. (C) It is a cross section of a light projection device using an electrodeless plasma light source, and a perspective explanatory view. (D) It is a cross section of a light projection device using a metal halide light source, and is a perspective explanatory view. 1 is a photomask which is an aperiodic pattern according to the present invention. (A) It is a layout drawing of a photomask. (B) It is an origin marker part arrangement | positioning figure of the center part of a photomask. It is explanatory drawing of the imaging and projection apparatus used by this invention. (A) It is a cross section and a perspective view of the power supply of a camera, a projection apparatus, and a projection apparatus. (B) It is a top view of the power supply of a camera, a projection apparatus, and a projection apparatus. (C) It is explanatory drawing of a tilt optical system camera. It is explanatory drawing of the optical condition setting of this invention, especially parallax, a search range, etc. It is a sample image in the laundry classification system. It is a measurement sample image of the automatic development system of cloth. (A) Left camera image (reference image). (B) Right camera image (reference image). (C) It is a figure which shows the search range of a left camera image. (D) It is a figure which preserve | saves the template of the non-periodic pattern of the peripheral region containing the measurement point of a right side camera image, and shows the position of a measurement point. (E) An image showing the distribution of normalized correlation values in the template (f) on the reference image. (F) A template image cut out from (d). (G) It is a figure which shows distribution of the normalized correlation value on the line segment (epipolar line) set as a search range of (e).

Claims (5)

  Projecting a pattern onto a measurement object having a three-dimensional shape, wherein the pattern is an aperiodic pattern, and a unique aperiodic pattern is projected onto a minute surface area around a measurement point including a measurement point to be measured in three dimensions A non-periodic pattern projector.   The non-periodic pattern is a plane-filled pattern with two types of rhombuses having inner angles of 1 / 5π and 4 / 5π and rhombuses having inner angles of 2 / 5π and 3 / 5π. The aperiodic pattern projection device described.   The non-periodic pattern is a non-periodic pattern arranged on a photomask so as to be divided into a rhombus boundary line portion having a certain width and the inside of the rhombus, one being a light transmitting portion and the other being a light shielding portion. The aperiodic pattern projection according to claim 1, wherein a photomask, a light source, and a projection lens are arranged in series so that the aperiodic pattern on the photomask forms an image on the surface of the measurement object. apparatus. A measurement point for acquiring a stereo image with a pattern of a measurement object by the stereo camera and a three-dimensional measurement using the aperiodic pattern projection device according to claim 1, a pair of imaging elements constituting a stereo camera, and the stereo camera. 3D shape of the measurement object by projecting a unique non-periodic pattern onto the surrounding minute surface area including the image and performing stereo image processing for assigning corresponding points between the standard image of the stereo camera and the reference image A three-dimensional shape measuring apparatus comprising: an arithmetic unit for measuring     Projecting a pattern onto a measurement object having a three-dimensional shape, wherein the pattern is a non-periodic pattern, and a unique non-periodic pattern is projected onto a minute surface region around a measurement point to be measured in three dimensions A non-periodic pattern projection method.

Images