JP2017003525A - 3D measuring device - Google Patents

Abstract

In a technique for creating a three-dimensional model of a subject from a photographed image, a problem that a background becomes an obstacle to processing is solved.
[Solution]
The imaging target is provided with color targets 104 and 105 serving as orientation characteristic displays and random dot patterns 106, and a rotatable turntable 101, color targets 104 and 105, subject 103, and random dot patterns 106 constituting a background feature display. And a camera 102 set to occupy all of the three-dimensional measuring device.
[Selection] Figure 1

Description

  The present invention relates to a technique for three-dimensional measurement using an image.

  A technique for acquiring data relating to a three-dimensional shape of a plant using an image is known. For example, Patent Document 1 describes a technique for taking a photograph of a plant from a plurality of viewpoints while rotating the plant and obtaining a three-dimensional model of a subject (plant) from the obtained plurality of captured images.

JP 2004-191243 A

  In the three-dimensional measurement using an image, the background other than the target subject becomes noise. In such a background, an object of the present invention is to solve the problem that the background is an obstacle to processing in the technique of creating a three-dimensional model of a subject from a captured image.

  The invention described in claim 1 is provided with a feature display constituted by a pattern that cannot be shifted and overlapped, can be rotated, and includes a turntable on which a subject is placed, and a camera that photographs the subject. It is a characteristic three-dimensional measuring device.

  If the patterns cannot be shifted and overlapped, the pattern means a pattern in which two features having the same pattern are prepared and overlapped so that the feature display matches only one way. In this case, even if one pattern is relatively shifted from the state in which the patterns overlap, the two patterns cannot be matched.

  Examples of the pattern include dots, figures, characters, and other patterns that can be recognized as features. A combination of colors can also be used as a pattern. Also, various patterns can be used in combination.

  According to the present invention, the first still image is taken at the first rotation angle position of the turntable, the second still image is taken at the second rotation angle position, and the third rotation angle position is taken. By taking a third still image and performing operations such as..., Still images of the subject taken from a plurality of different viewpoints can be obtained. Here, for example, when specifying the correspondence between the first still image obtained from the first viewpoint and the second still image obtained from the second viewpoint, there is only one positional relationship where the feature displays match. Therefore, the correspondence relationship between the first still image and the second still image can be specified efficiently and accurately. The number of cameras is not limited to one and may be two or more. Photographing may be performed while rotating the turntable, or may be performed in a state where the rotation of the turntable is stopped.

  According to a second aspect of the present invention, in the first aspect of the present invention, the camera is focused on the setting or the subject so that the rotary base occupies the entire photographing range, and the portions other than the rotary base are This is characterized in that the image is out of focus and the captured image is blurred.

  By making the photographing field of view occupied by the turntable, the problem of background noise can be avoided. Further, even if the background appears, blurring the background can avoid extraction of feature points from the background, and the problem of background noise can be avoided.

  The invention described in claim 3 is the invention described in claim 1 or 2, wherein the feature display includes a random dot pattern. Since the random dot pattern has only one way of overlapping, the “pattern that cannot be shifted and overlapped” in the present invention can be easily obtained.

  According to a fourth aspect of the present invention, in the invention according to any one of the first to third aspects, the feature display includes a scale that gives a distance between two points. By including a scale that gives the distance between two points in the feature display, the real scale is given to the finally obtained three-dimensional model. Examples of the scale that gives the distance between two points include a method of arranging individually identifiable targets at two points, a method of displaying a straight line with a known distance, and the like. As a simple method, it is possible to place a ruler with a fixed length on a turntable. It is desirable that there are a plurality of scales that give the distance between the two points.

  According to a fifth aspect of the present invention, in the invention according to any one of the first to fourth aspects of the present invention, it extends in an inverted conical shape upward from the turntable and rotates together with the turntable. A cover member having the feature display is provided on the inner side. According to the fifth aspect of the present invention, since the field of view of the camera can be filled with the subject and the characteristic display, noise such as a background image that becomes an obstacle when matching between the captured images can be eliminated.

  The invention according to claim 6 is the invention according to any one of claims 1 to 5, wherein the feature display is a random dot pattern, and the distribution of feature points of the subject and the dots of the random dot pattern The feature display and the image of the subject are separated using a difference in distribution or a difference in density.

  What is necessary for creating a three-dimensional model of feature points is a feature point of the subject, and the role of feature display ends when matching between images is completed. Therefore, it is necessary to separate the feature points constituting the subject and the feature points extracted from the feature display before the three-dimensional model of the subject is created. According to the fifth aspect of the present invention, the feature point of the feature display and the feature point of the subject are separated by utilizing the difference in distribution and density of the feature display and the feature point of the subject.

  For example, when a subject is a plant, the extracted feature points are concentrated on the outlines of stems and leaves. On the other hand, when the random dot pattern is adopted as the feature display, the distribution of the feature points is more uniform than that of the subject even though it is random. Therefore, the uniformity of the distribution of the feature points is evaluated numerically, and it is determined based on a set threshold value whether the extracted feature points are the subject or the background feature display. Specifically, an area with high uniformity of feature point distribution is determined as background feature display, and an area with low uniformity of feature point distribution is determined as a subject.

  According to a seventh aspect of the invention, in the invention according to any one of the first to sixth aspects, the feature display and the image of the subject are separated by compressing an image taken by the camera. It is characterized by. As described in relation to the sixth aspect, the uniformity of the feature point distribution of the feature display is higher than the uniformity of the feature point distribution of the subject. By utilizing this, the dot display of the feature display is set so that when the image is compressed, the interval between the dots is narrowed and the gap between the dots is eliminated and the image becomes black. Since the feature points of the subject are concentrated on the contour portion, the feature points are less frequently concentrated on the surface, and in many cases, the feature points are distributed linearly. Therefore, the feature point of the subject can be selectively re-extracted by making the surface of the feature display portion black by using the compressed image and separating the portion.

  According to the present invention, in the technique of creating a three-dimensional model of a subject from a photographed image, the problem that the background becomes an obstacle to processing is solved.

It is a perspective view of the measurement system using invention. It is a block diagram of the three-dimensional measuring apparatus using invention. It is explanatory drawing of a back intersection method. It is explanatory drawing of the forward dating method. It is explanatory drawing of template matching. It is a flowchart which shows an example of the procedure of a process. It is a perspective view of the measurement system using invention. It is a perspective view of the measurement system using invention. 6 is a drawing-substituting photograph showing an example of a subject placed on a turntable. 6 is a drawing-substituting photograph showing an example of a subject placed on a turntable. 6 is a drawing-substituting photograph showing an example of a subject placed on a turntable.

1. First embodiment (outline)
FIG. 1 shows a three-dimensional measuring apparatus 100. The three-dimensional measuring apparatus 100 includes a turntable 101 that is rotated by driving means such as a motor. The turntable 101 has a flat upper surface, on which an orientation feature display composed of color targets 104 and 105 and a background feature display 106 composed of a random dot pattern are provided. In addition, in the center of the turntable 101, a subject 103 for which a three-dimensional model is to be created is placed. In this example, a plant is selected as the subject.

  The random dot pattern is a dot pattern in which dots are randomly distributed without regularity. Techniques related to random dot patterns are disclosed in, for example, Japanese Patent Application Laid-Open No. 11-39505, Japanese Patent Application Laid-Open No. 10-97641, Research Reports of the Institute of Electronics, Information and Communication Engineers. Etc. are described.

  The subject 103 is photographed by the camera 102, and the camera 102 is set so that the angle of view (shooting field of view) is occupied by the upper surface of the turntable 101. That is, the system is set so that the image captured by the camera 102 is occupied by the color targets 104 and 105, the background feature display 106, and the subject 103.

  The color targets 104 and 105 constituting the orientation characteristic display include a plurality of color markers 104a and 105a, respectively. The plurality of color markers 104a and 105a are configured by finer color code display (not shown). The positions of the color markers 104a and 105a are strictly determined in advance, and the correspondence between the color code and the position of the color marker is acquired in advance. The position of each color code can be read by recognizing the color code display of the color marker as an image and reading the color code. A technique relating to the color code is described in, for example, Japanese Patent No. 4828125.

  The distance between the color targets 104 and 105 and the distance between the plurality of color markers 104a and 105a are determined in advance, and are set so that an actual scale can be given by specifying the position of the color marker.

  A plurality of images are continuously captured by the camera 102 while rotating the turntable 101. In this case, relatively the same effect is obtained as when shooting is performed while the turntable 101 is stationary and the camera 102 is moved around the turntable 101, and a plurality of still images with different viewpoints are obtained. It is done. In shooting, the rotation speed of the turntable 101 and the shooting interval of the camera 102 are adjusted so that a portion where the subject 103 is not captured does not occur.

  FIG. 2 is a block diagram of the arithmetic device 200 according to the embodiment. The arithmetic device 200 processes a still image photographed by the camera 102 and creates three-dimensional model data of the subject 103. The three-dimensional measuring apparatus 200 has a function as a computer, and includes a CPU, a solid-state electronic memory, a hard disk storage device, various interfaces, and other arithmetic elements as required. FIG. 2 shows various functional units regarded as functions. Some or a plurality of the various functional units illustrated may be configured as software, or may be configured as dedicated hardware. For example, the three-dimensional measuring apparatus 200 can be obtained by realizing the functions of the functional units shown in the figure using a general-purpose computer.

  The three-dimensional measurement apparatus 200 includes a data storage unit 201, a feature point extraction unit 202, an image classification unit 204, a specific image extraction unit 205, an external orientation element calculation unit 206, a corresponding point specification unit 207, and a feature point three-dimensional position calculation unit. 208, a three-dimensional model creation unit 209, and an orientation feature display detection unit 210. The data storage unit 201 stores data input from the outside to the three-dimensional measurement apparatus 200, data obtained in the middle of the calculation, calculation result data, and an operation program necessary for the operation of the three-dimensional measurement apparatus 200. Note that the data stored in the data storage unit 201 may be temporarily stored data.

  The feature point extraction unit 202 extracts feature points from the image captured by the camera 102. The feature point is a point that can be distinguished from the surroundings, and for example, an edge portion or a portion having a color different from that of the surrounding portion is extracted as the feature point. The feature points are extracted by software processing. For the extraction of the feature points, differential filters such as Sobel, Laplacian, Prewitt, and Roberts are used.

  The image classification unit 204 separates images using the positions of feature points and the distribution state. First, a method for separating the background feature display 106 from other images based on the difference in the position of feature points will be described. As shown in FIG. 1, the background feature display 106 is a plane, and the position in the height direction is different from the subject 103. The feature points extracted from the background feature display are separated from the feature points extracted from the subject 103 using the difference in position.

  For example, paying attention to the position of the feature point, a threshold is set at the height position. When the height is equal to or lower than the threshold value, it is determined that the feature point is not a feature point constituting the subject 103 but a feature point constituting another portion (for example, the background feature display 106). Also, for example, a threshold is set for the distance from the camera 102, and when the distance from the camera 102 is equal to or greater than the threshold, the feature point is not a feature point constituting the subject but another part (for example, the background feature display 106). ).

  Next, a method of separating the background feature display from other images using the feature point density will be described. One method compresses the entire image. When the entire image is compressed, the dot interval of the random dot pattern is that is, the background image becomes black at a certain stage. Using this phenomenon, a background image composed of random dot patterns is separated.

  As another method of separating a background image composed of random dot patterns, there is a method of selectively detecting a random dot pattern region using the bias and density of the distribution of feature points. In this method, the region of interest is divided into grid sections, and the deviation of the number of feature points in each grid section is evaluated by a standard deviation to determine the distribution of the feature point distribution. On the other hand, the density of feature points in the focused area is detected. And when the value of said standard deviation is below a threshold value, and the density of a feature point is more than a threshold value, it is the background feature display (background image) comprised by the said focused area | region by a random dot pattern judge.

  Other methods include the following methods. The random dot pattern has a property that a change in dot density is relatively small when the location of the region of interest is changed. When an actual object such as a plant is targeted, the feature points are concentrated on the contour portion or the like. Therefore, when the location of the region of interest is changed to several places, the density of the feature points changes greatly at a certain timing. Using this property, a background image composed of random dot patterns is separated from other images.

  Still another method is the following method. The random dot pattern does not constitute a specific shape because the dots are irregularly distributed. On the other hand, since the subject 103 and the color targets 104 and 105 in FIG. 1 have contour lines that specify the shape, there are portions in which feature points are distributed in a linear shape. Using this fact, it is determined that the image is a background image in which a specific shape such as a straight line or a curve is not detected and a region where the density of feature points is equal to or higher than a threshold is configured by a random dot pattern.

  The orientation feature display extraction unit 210 detects the color targets 104 and 105. The images of the color targets 104 and 105 are acquired in advance, and the color targets 104 and 105 are extracted from the captured image using the images as a reference.

  The external orientation element calculation unit 206 calculates the external orientation elements (position and orientation) of the camera 102 in FIG. 1 by the backward intersection method using the orientation characteristic display in the image. FIG. 3 shows the principle of the backward intersection method. The backward intersection method is a method of observing the direction from an unknown point to three or more known points and determining the position of the unknown point as an intersection of these direction lines. Examples of the backward intersection method include single photo orientation and DLT method (Direct Liner Transformation Method). The association method is described in Basic Surveying (Denki Shoin: April 2010) 182p, p184. JP 2013-186816 discloses an example of a specific calculation method for the meeting method.

  Hereinafter, a method for obtaining the external orientation element using the backward intersection method will be described. First, in the captured image, a plurality of orientation characteristic displays are specified, and their positions are acquired as reference points. For example, a case will be described in which the color targets 104 and 105 shown in FIG. In this case, the position of each color marker can be read by recognizing the color code display of the color markers 104a and 105a as an image.

Hereinafter, an example of specific calculation will be described. In this case, for example, P1 to P3 in FIG. 3 are reference points (reference points for calculating external orientation elements) specified by the color markers. _Further, the screen coordinate value in the image p 1 ~p ₃ is focused reference points P1 to P3. Here, by setting the three directions line connecting P1 and _p 1, P2 and _p 2, P3 and _{p 3,} the position of the camera 102 to the intersection O is taken the image. Further, the extending direction of the direction line connecting the point O and the center of the screen is the direction (posture) of the camera.

  The external orientation element of the camera can also be calculated from a feature point that is not a reference point (for example, a feature point whose absolute position is extracted from a random dot pattern). In this case, the external orientation element is a relative position and orientation with respect to the feature point in the captured image. In this case, the external orientation element of the camera is calculated using the above-described backward intersection method. In this case, the relative positional relationship between the points P1 to P3 and the intersection point O is obtained, and the relative position and orientation (posture) of the camera at the point O with respect to the points P1 to P3 are obtained. In this case, by increasing the number of feature points to be used (in the above case, three points P1 to P3) as much as possible, the accuracy of the calculated value of the external orientation element of the camera is ensured. In addition, orientation may be performed prior to photographing, and processing for obtaining external orientation elements of the camera may be performed. In this case, no reference point is required. Note that the orientation elements include an external orientation element and an internal orientation element, and the process for obtaining these orientation elements is called calibration. In particular, the process for obtaining the external orientation elements is called orientation.

  The corresponding point specifying unit 207 specifies the correspondence between the feature points individually extracted from the two images taken from different viewpoints. That is, a process is performed in which the same feature point as the feature point extracted in one image is specified in the other image. The processing for specifying the correspondence between the feature points is performed using, for example, template matching shown in FIG.

Examples of template matching include a residual sequential detection method (SSDA), a cross-correlation coefficient method, and the like. Hereinafter, an example of template matching will be described. Template matching is a method in which the coordinate data of images in two coordinate systems are compared with each other, and the correspondence between the two images is obtained by the correlation between the two. In template matching, a correspondence relationship between feature points of an image viewed from two viewpoints is obtained. FIG. 5 is an explanatory diagram for explaining the principle of template matching. In this method, as shown in the drawing, a template image of N ₁ × N ₁ pixel is moved on a search range (M ₁ −N ₁ +1) ² in an input image of M ₁ × M ₁ pixel larger than that, The upper left position of the template image is calculated so that the cross-correlation function C (a, b) expressed by the following equation 1 is maximized (that is, the degree of correlation is maximized).

  The above processing is performed while changing the magnification of one image and rotating it. And the area | region where both images correspond is calculated | required on the conditions from which the correlation became the maximum, Furthermore, the corresponding point is detected by extracting the feature point in this area | region.

  By using template matching, a matching portion between two images to be compared can be specified, and the correspondence between the two images can be known. In this method, the relative positional relationship between the two images is determined so as to maximize the correlation between the two images. The correlation between the two images is determined by the feature points of both images. As a technique for obtaining a correspondence relationship between two images, for example, techniques described in Japanese Patent Application Laid-Open Nos. 2013-178656 and 2014-35702 can be used. Note that Japanese Patent Application Laid-Open No. 2013-178656 and Japanese Patent Application Laid-Open No. 2014-35702 also describe a technique for extracting feature points and a technique for obtaining a three-dimensional position of feature points. It can be used for the technology described in the specification.

  The feature point three-dimensional position calculation unit 208 calculates the three-dimensional position of the feature point extracted from the image by the feature point extraction unit 202 by the forward intersection method. By calculating the three-dimensional positions of a large number of feature points constituting the target object, three-dimensional point group position data serving as data capable of grasping the target object as a set of points can be obtained. Furthermore, a 3D model can be created based on the 3D point cloud position data.

FIG. 4 shows the principle of the forward intersection method. In the forward intersection method, a direction from a plurality of known points (two points O ₁ and O _{2 in} the case of FIG. 4) to the unknown point P is observed, and the position of the unknown point P is determined as an intersection of these direction lines. Ask.

  Hereinafter, an example of processing in the three-dimensional position calculation unit 208 will be described with an example. First, it is assumed that there is a first image photographed from the first viewpoint and a second image photographed from the second viewpoint while rotating the turntable 101. Here, the first image is a captured image from the first viewpoint, and the second image is a captured image from a second viewpoint different from the first viewpoint.

Here, P is a feature point that is common to the first image and the second image (a feature point corresponding to both images), O ₁ is the first image viewpoint (photographing position), and O ₂ is the first image. The viewpoint (photographing position) of the second image, p ₁ is the screen coordinate in the first image of the feature point P, and p ₂ is the screen coordinate in the second image of the feature point P. In this case, a direction line connecting O ₁ and p ₁ is set, on the other hand, a direction line connecting O ₂ and p ₂ is set, and by calculating the coordinates of the intersection of the two direction lines, the feature point P Three-dimensional coordinates can be calculated.

  The 3D model creation unit 209 creates a 3D model based on 3D point cloud position data obtained by analyzing a large number of captured images. For example, tin (irregular triangle network) is created using the obtained three-dimensional point group position data consisting of the coordinates of many feature points, and a three-dimensional model of the object to be imaged is created. At this time, the actual dimensions of the three-dimensional model are given by a plurality of targets. A technique for creating a three-dimensional model based on point cloud position data is described in, for example, Japanese Patent Application Laid-Open No. 2012-230594.

(Example of processing)
An example of the processing procedure will be described below. A program for determining the processing procedure is stored in the data storage unit 202, read into an appropriate storage area, and executed. In addition, a form in which the program is stored in an appropriate storage medium and provided from the storage medium is also possible. FIG. 6 is a flowchart showing an example of the processing procedure.

  First, the camera 102 continuously shoots while rotating the turntable 101 to shoot the subject 103 from a plurality of different viewpoints. The obtained image data relating to the plurality of images is stored in the data storage unit 201 (step S101). Next, in each image, the color targets 104 and 105 are extracted (step S102). By this process, a reference position whose position is known in advance is detected. This process is performed in the orientation feature display extraction unit 210.

  Next, in each image, the external orientation element of the camera 102 is calculated using the orientation image obtained in step S102 (step S103). This process is performed in the external orientation element calculation unit 206. With this process, the camera's external orientation elements (position and orientation) in each image are calculated.

  Next, feature points are extracted from each of the acquired plurality of images (step S104). This process is performed in the feature point extraction unit 202. After the feature points are extracted, the correspondence relationship between the feature points between a plurality of images is specified (step S105). This process is performed in the corresponding point specifying unit 207. For example, in this process, the feature point correspondence between two images obtained from adjacent viewpoints is specified. By performing this process on all target still images, the correspondence of feature points between the still images is specified.

  Next, the three-dimensional position of the feature point extracted in step S105 is calculated (step S106). This processing is performed in the feature point three-dimensional position calculation unit (208). In this process, the three-dimensional coordinates of the corresponding points are calculated according to the principle shown in FIG.

For example, assume that O ₁ in FIG. 4 is a _first viewpoint, O ₂ is a second viewpoint, and P is a feature point corresponding to two still images. In this case, p ₁ is a screen coordinate value of the first photo on the screen, p ₂ is the screen coordinate value in the second photograph on the screen. Here, the attitude of the camera in position and its position of o _1, the orientation of the camera in position and its position of o _2, are obtained by the process of step S103. Accordingly, the direction line connecting O ₁ and p ₁ and the direction line connecting O ₂ and p ₂ are set, and the coordinates of the intersection are obtained, whereby the position of the feature point P (three-dimensional coordinate position) is obtained.

  When the position of the feature point is obtained in step S106, classification is performed based on the position of the feature point (step S107). This process is performed in the image classification unit 204. In this process, the feature point of the subject 103 is separated from the other feature points using the difference in the position of the feature points. In this process, for example, the feature point of the subject 103 is separated from the other feature points using the difference in distance from the camera.

  Next, classification based on feature amounts is performed (step S108). This process is performed in the image classification unit 204. In this process, the number of feature points detected by the feature point detection unit 202 per unit area is counted, and information on the density distribution state of the feature points is obtained. Then, the background image is separated based on the distribution state of the feature points. In this process, a process in which the nonuniformity of the distribution of feature points is equal to or less than a threshold and the area where the density of feature points is equal to or greater than the threshold is used as the background feature display 106 (background image). By this process, the background feature display 106 is separated from the target still image. At this stage, the images of the color targets 104 and 105, which are the orientation display features detected in step S102, are also separated from the image of the subject 103.

  Next, in each image, an image to be a target for creating a three-dimensional model (extraction of an image of the subject 103) is performed (step S109). This process is performed in the specific image extraction unit 205. In this processing, the background feature display 106 and the image of the color targets 104 and 105 that are the orientation feature display that have already been separated are removed, and the image of the subject 103 is extracted.

  When the image of the subject 103 is extracted, a three-dimensional model of the subject 103 is created based on the information on the three-dimensional position of the feature point related to the subject 103 that has already been acquired in the process of step S104 (step S110). This process is performed in the 3D model creation unit 209.

(Superiority)
In the system of FIG. 1, the image capturing range of the camera is filled with an image of an object to be captured, an image for orientation, and an image for background that can be easily separated from other images. By doing so, an increase in calculation and calculation errors due to an image that becomes noise can be suppressed when a three-dimensional model is created. In particular, by making the background image easy to separate, it is possible to suppress an increase in calculation time and calculation errors related to the background image processing.

  Further, since the random dot pattern adopted as the background image can easily take the correspondence between still images taken from different viewpoints, the process of step S105 can be performed efficiently. In the above example, the image classification in step S107 using the feature point position information and the image classification in step S108 using the feature amount difference are performed, and noise other than the subject 103 is divided in two stages. The process which isolate | separates the image used as is performed. Therefore, it is possible to create a three-dimensional model of the subject 103 after separating an image that becomes noise other than the subject 103 more effectively.

2. Second Embodiment FIG. 7 shows a structure in which an inverted conical cover member 107 provided with a random dot pattern 106 ′ on the inside is attached to the turntable 101 shown in FIG. In this example, the angle of view (shooting range) of the camera 102 can be set wider. In the configuration of FIG. 7, an orientation feature display may be provided inside the cover member 107.

3. Third Embodiment In FIG. 8, only the background feature display 106 configured by a random dot pattern is arranged on the upper surface of the turntable 101 in the configuration of FIG. In this case, an external orientation element of the camera 102 is calculated using the background feature display 106, and a three-dimensional model of the subject 103 obtained based on still images photographed from a plurality of photographing positions (viewpoints) is obtained. Here, since there is no display for giving a scale in the captured image, the obtained three-dimensional model is a relative model to which no real scale is given. Further, the external orientation elements of the camera 102 have relative positions and orientations with respect to a large number of feature points. In calculating the external orientation elements of the camera 102, the calculation accuracy is ensured by using as many feature points as possible. In addition, orientation may be performed in advance to obtain an external orientation element of the camera 102.

  Even if the obtained three-dimensional model is a relative model, a scale can be given to the three-dimensional model by using a still image obtained by copying a known scale. The actual scale can be given by a method of giving a display for giving a distance between two points on the top surface of the turntable 101 or placing a scale (for example, a ruler) on the top surface of the turntable 101.

4). Fourth Embodiment FIG. 9 shows an example in which a strawberry that is a subject is arranged at the center of the upper surface of the turntable. FIG. 9 shows an example in which a random dot pattern, a numerical position designation for designating the position of a specific dot, and a color target arranged on the periphery are arranged (displayed) on the upper surface of the turntable.

  In the example of FIG. 9, it is possible to read a position display of a number or a color target and use it to calculate an external orientation element of a camera or create a three-dimensional model given an actual scale.

  In addition, the random dot pattern shown in FIG. 9, a numerical position designation designating the position of a specific dot, one or more of the color targets arranged at the periphery are regarded as simple patterns, and they are shifted and overlapped in the present invention. It can also be used as a feature display composed of patterns that cannot be performed. When a random dot pattern, a numerical position designation specifying the position of a specific dot, and all color targets arranged at the periphery are used as mere feature displays, no actual scale is given. In this case, an actual scale may be given later to the obtained three-dimensional model.

5. Fifth Embodiment There is also a method for extracting an image of the subject 103 using a still image taken with a variable focus camera. In this case, the subject 103 is photographed while changing the focus from one viewpoint, and a plurality of still images are obtained. The background image is blurred using the difference in focus, and the background image is separated using the image focused on the subject 103. Hereinafter, an example of a technique for separating the background by blurring the background will be described.

  FIG. 10 shows an image when the setting is focused on the subject strawberry and the background is blurred. In this case, since the background is blurred, feature points are not extracted and are not subject to matching. For this reason, in the process (matching process) for obtaining the correspondence between still images obtained from a plurality of different viewpoints, the background image does not become noise.

  In the case of FIG. 10, the feature display on the turntable behind the strawberry that is the subject is blurred, but the feature display of the focused part is used for matching. Therefore, no problem occurs in the process (matching process) for specifying the correspondence between a plurality of still images. FIG. 10 shows a state in which a ruler for giving a scale is placed on the turntable.

6). Sixth Embodiment FIG. 11 shows an image when the background is set to black, and the setting is focused on the subject strawberry and the background is blurred. In this case, the feature point is not extracted from the background because the background is dark (in dark color) and not in focus. For this reason, it is possible to avoid the problem that the background is not the target of matching and the background image becomes noise.

4). Others There is also a method in which the display of a specific shape is repeatedly arranged on the background. For example, there is a method of constructing a background image by repeatedly displaying polygons (triangles etc.) that are easy to recognize images. In this case, an area where a specific shape is repeated is separated as a background image.

  There is also a method of separating a background image by a difference in color as a specific color that can distinguish the background image from other regions. In addition, there is a method in which the background image is an image in which a specific color pattern is repeated so that it can be easily separated from other images. Further, a random edge pattern can be used instead of the random dot pattern. Since the random edge pattern can be erased on the compressed image, this phenomenon can be used to separate the background image.

  Further, there is a method in which the background image is reflected at a specific wavelength so that the background image stands out when taken with a hyperspectral camera. For example, the background feature display 106 in FIG. 1 is coated using elements and materials not included in the subject 103 and the color targets 104 and 105. In this case, since the reflected light from the background feature display portion has a spectral distribution different from that of the other portions, the background feature display portion can be separated from the image by the hyperspectral camera. In addition, there is a method in which the background feature display 106 in FIG.

  In the example of FIG. 1, an image captured by the camera 102 is filled with the subject 103, the color targets 104 and 105, and the background feature display 106, thereby suppressing noise generation. Here, the background feature display 106 may be filled with a color target.

  In the above description, an example in which a plant is a subject is shown, but the subject is not limited to a plant. Note that a plurality of embodiments described in this specification can be used in combination.

DESCRIPTION OF SYMBOLS 100 ... Three-dimensional measurement system, 101 ... Rotary table, 102 ... Camera, 103 ... Subject, 104 ... Color target, 104a ... Color marker, 105 ... Color target, 105a ... Color marker, 106 ... Background feature display, 107 ... Background feature Cover member provided with an indication.

Claims (7)

It has a feature display composed of patterns that cannot be shifted and superimposed, and can be rotated and a turntable on which the subject is placed;
A three-dimensional measuring apparatus comprising: a camera that photographs the subject.   The camera is set so that the turntable occupies the entire shooting range or the subject is in focus, and the part other than the turntable is not in focus and the captured image is set to be blurred. The three-dimensional measuring apparatus according to claim 1.   The three-dimensional measuring apparatus according to claim 1, wherein a random dot pattern is included in the feature display.   The three-dimensional measurement apparatus according to claim 1, wherein the feature display includes a scale that gives a distance between two points.   5. The apparatus according to claim 1, further comprising a cover member that extends in an inverted conical shape upward from the turntable, rotates together with the turntable, and includes the feature display on the inside. The three-dimensional measuring apparatus according to one item. The feature display is a random dot pattern;
6. The feature display and the image of the subject are separated using a difference in distribution or density of feature points of the subject and the random dot pattern. The three-dimensional measuring device according to claim 1. The three-dimensional measurement apparatus according to claim 1, wherein the feature display and the image of the subject are separated by compressing an image captured by the camera.

Images