JP2016218920A - Information processor, information processing method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a technology for achieving position attitude estimation of a target object having no CAD model with little labor.SOLUTION: An information processor includes acquisition means for acquiring distance information up to the surface of a target object, generation means for generating an outline shape of the target object on the basis of the distance information acquired by the acquisition means, extraction means for extracting a geometric feature of the target object on the basis of the distance information acquired by the acquisition means, and arrangement means for arranging the geometric feature of the target object extracted by the extraction means to the outline shape generated by the generation means.SELECTED DRAWING: Figure 2

Description

  The present invention relates to an information processing apparatus, an information processing method, and a program.

There is a technique for estimating the position and orientation of a target object from distance information obtained by observing the target or a two-dimensional image using three-dimensional shape information of the target object. Non-Patent Document 1 discloses a technique for decomposing a CAD model of a component into basic shapes and searching for the basic shapes from distance information. According to this technique, in position and orientation estimation using a three-dimensional shape, the three-dimensional shape is referred to as a model, and the model is associated with a geometric feature from an image obtained by capturing an object. Further, in this technique, the position and orientation parameters are estimated using the distance in the three-dimensional space of the associated geometric features and the difference projected on the screen as evaluation values. With this position and orientation estimation technique, for example, the hand position for picking a part by a robot can be calculated.
As the three-dimensional shape information used as a model, a CAD model used at the time of product design is generally used.

KATSUSHI IKEUCHI, "Generating an Interpretation Tree of CAD Model for 3D-Object Recognition in Bin-Picking Tasks," International Journal of V. 1, No. 1 2, pp. 145-165, 1987.

When estimating the position and orientation, a CAD model of the target object is not always prepared. In the position and orientation estimation as in Non-Patent Document 1, in order to estimate the position and orientation parameters so as to reduce the difference between the CAD model rigidly transformed by the position and orientation parameters and the observation target, there is no CAD model itself. Cannot estimate itself.
An object of this invention is to provide the technique for implement | achieving the position and orientation estimation of the target object without a CAD model with little effort.

  Therefore, the information processing apparatus of the present invention includes an acquisition unit that acquires distance information to the surface of the target object, and a generation unit that generates a schematic shape of the target object based on the distance information acquired by the acquisition unit. Extraction means for extracting the geometric feature of the target object based on the distance information acquired by the acquisition means; and the outline generated by the generation means for the geometric feature of the target object extracted by the extraction means. Arrangement means for arranging in a shape.

  ADVANTAGE OF THE INVENTION According to this invention, the technique for implement | achieving the position and orientation estimation of the target object without a CAD model with little effort can be provided.

It is a figure which shows an example of the hardware constitutions of information processing apparatus. 2 is a diagram illustrating an example of a software configuration of the information processing apparatus according to the first embodiment. FIG. 3 is a flowchart illustrating an example of information processing according to the first exemplary embodiment. 6 is a diagram illustrating an example of a software configuration of an information processing apparatus according to a second embodiment. FIG. 10 is a flowchart illustrating an example of information processing according to the second exemplary embodiment. FIG. 11 illustrates an example of a software configuration of an information processing apparatus according to a third embodiment. It is a figure which shows an example of a system configuration.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

<Embodiment 1>
FIG. 1 is a diagram illustrating an example of a hardware configuration of the information processing apparatus 100.
In FIG. 1, the CPU 101 controls various devices connected to the system bus 104.
The ROM 102 stores a BIOS program and a boot program.
The RAM 103 is used as a main storage device of the CPU 101.
A keyboard controller (KBC) 105 performs processing related to input of information and the like from a pointing device (PD) 109 such as a mouse and a keyboard (KB) 110.
A display control unit (CRTC) 106 has a video memory therein and draws the video memory in accordance with an instruction from the CPU 101, and outputs image data drawn in the video memory to the CRT 111 as a video signal. Although the CRT 111 is illustrated as the display device in FIG. 1, the type of the display device such as a liquid crystal display device is not limited.
A disk controller (DKC) 107 accesses a hard disk (HDD) 112.
A network interface card (NIC) 108 is connected to a network and performs information communication via the network.
The HDD 112 stores an OS program, programs of various applications operating on the OS, and the like. The HDD 112 is an example of a storage area.
In the above configuration, when the information processing apparatus 100 is powered on, the CPU 101 reads the OS program from the HDD 112 into the RAM 103 according to the boot program stored in the ROM 102, and executes the processing, thereby realizing the function of each apparatus. To do.
That is, the CPU 101 of the information processing apparatus 100 executes processing based on the program, thereby realizing the software configuration of the information processing apparatus 100 and the processing of the flowchart described below.

FIG. 2 is a diagram illustrating an example of a software configuration and the like of the information processing apparatus 100 according to the first embodiment.
The information processing apparatus 100 includes, as at least a software configuration, a distance information acquisition unit 10, a two-dimensional image acquisition unit 20, a schematic shape generation unit 30, a geometric feature extraction unit 40, a geometric feature selection unit 50, and a geometric feature. An arrangement unit 60 and a viewpoint-specific feature holding unit 70 are included.
The distance information acquisition unit 10 acquires distance information from a distance image including the target object acquired by a projector or the like to the surface of the target object. For example, a projector projects a structural pattern onto a target object and acquires an image reflected on the target surface. The distance information acquisition unit 10 identifies the projection position of the projector by decoding the pattern from this image. And the distance information acquisition part 10 calculates the distance of a target object by triangulation from the relationship between an imaging position and a projection position, and acquires distance information. However, this does not limit the present embodiment, and the distance information acquisition unit 10 may acquire the distance information using another method. For example, the distance information acquisition unit 10 may measure the laser reflection position when the object is linearly moved using a line laser, and acquire the distance information converted into three-dimensional position information. .
The two-dimensional image acquisition unit 20 acquires image data of a two-dimensional image including the target object imaged by the imaging device. The imaging device is, for example, a camera using an imaging element such as a CCD or CMOS in which camera parameters such as lens distortion are calibrated in advance. The two-dimensional image is a color image using, for example, three color filters. If the information processing apparatus 100 can communicate with two or more cameras, a stereo camera can be configured by calibrating the relative position and orientation between the cameras in advance. By using images from a plurality of viewpoints, it is possible to reduce the influence due to shielding between components.

  The schematic shape generation unit 30 reconstructs the three-dimensional shape of the target object from the distance information obtained by the distance information acquisition unit 10. If the distance information is a distance value to a target arranged in a grid pattern, the approximate shape generation unit 30 can generate a mesh having apex at adjacent grid points. Further, the schematic shape generation unit 30 may integrate distance information obtained as a result of imaging from a plurality of viewpoints. In this case, the schematic shape generation unit 30 can calculate the relative position and orientation of the distance point groups of the distance information by using an algorithm such as ICP (Iterative Closest Points). The schematic shape generation unit 30 can use a method of generating a mesh by using voxel lattice points, a method of generating a mesh by Delaunay triangle for a point group, or the like. When the shape mesh data of the target object is available in advance, the approximate shape generation unit 30 may generate the approximate shape using the shape mesh data. For example, the schematic shape generation unit 30 downloads and acquires the shape mesh data of the target object via a network or the like, stores it in the HDD 112 or the like, and uses it.

  The geometric feature selection unit 50 selects a geometric feature extraction method in a range where a target object included in a two-dimensional image or distance information is captured. For example, the geometric feature selection unit 50 selects a preset extraction method or an extraction method set by the user via the CRT 111, the KB 110, or the like as the extraction method used by the geometric feature extraction unit 40. There are the following methods for extracting geometric features from a two-dimensional image. These include a Harris operator that extracts corner points, a Sobel or Canny operator that extracts edges, a method that extracts texture density and direction, a method that uses a color histogram or a local area gradient histogram, and the like. In addition, methods for extracting geometric features from distance information include identification using a histogram when normal vectors around a region of interest are projected onto a cylinder, a method using the difference in depth between two points, and the curvature of a local region. There is a method to calculate. The geometric feature selection unit 50 may select a plurality of extraction methods.

  The geometric feature extraction unit 40 performs image processing on the two-dimensional image and distance information in which the target object is imaged based on the geometric feature extraction method selected by the geometric feature selection unit 50. Then, the geometric feature extraction unit 40 extracts a plurality of geometric features as two-dimensional position information of the geometric feature if the result is a two-dimensional image, and as three-dimensional position information of the geometric feature if the result is distance information. By using a plurality of geometric feature extraction methods, two-dimensional images and distance information can be handled together, and the accuracy of geometric feature association can be improved.

  The geometric feature placement unit 60 associates the position information of the geometric features extracted by the geometric feature extraction unit 40 with the schematic shape generated by the schematic shape generation unit 30. For example, when an imaging device captures an object, a pedestal portion with a pattern drawn on a plane or a structure serving as an index captures images together. In this way, the position and orientation of the imaging device at the time of imaging can be obtained from the captured image. The geometric feature placement unit 60 performs rigid body conversion between the position and orientation of the imaging device at the time of imaging and the schematic shape generated by the schematic shape generation unit 30 without changing the environment in which the index is arranged and the target object does not move. It can be carried out. Here, the rigid transformation is coordinate transformation that does not change the shape even if the position and direction of the object (object) are changed. The three-dimensional position information of the geometric feature extracted from the distance information is a coordinate system at the center of the imaging device. Therefore, the geometric feature arrangement unit 60 can convert the reference coordinate system of the target part by multiplying the three-dimensional position information by the inverse transformation of the position and orientation of the imaging device at the time of imaging. Thereby, the three-dimensional position information of the geometric features extracted from the distance information captured from various viewpoints can be integrated as the three-dimensional position information based on the model coordinate system of the target object. The geometric feature extracted from the two-dimensional image has a feature point on a straight line connecting the position on the screen and the origin of the imaging device. Therefore, the geometric feature placement unit 60 has an intersection between the straight line connecting the origin of the imaging device and the position of the geometric feature on the screen and each polygon surface of the mesh model having the approximate shape generated by the approximate shape generation unit 30. A determination is made as to whether or not a position closest to the origin of the imaging device among the intersections is obtained. The mesh model is arranged according to the position and orientation of the imaging device and the model coordinate system. Since the intersection with the polygon plane is an imaging device coordinate system (camera coordinate system), it can be converted into a position in the model coordinate system by multiplying the inverse transformation of the position and orientation at the time of imaging of the imaging device. Hereinafter, a geometrical shape having a geometric shape arranged therein is also referred to as a geometric model.

  The viewpoint-specific feature holding unit 70 registers and manages a database holding the extracted geometric features in the HDD 112 or the like. The viewpoint-specific feature holding unit 70 constructs a database that uses the position and orientation of the imaging device used in the geometric feature placement unit 60 as a table key and lists the three-dimensional position information of each geometric feature converted into the model coordinate system. And manage. By holding information on the relative position and orientation between the imaging device and the target object, it is possible to efficiently register only the geometric features that are actually observed. The viewpoint-specific feature holding unit 70 makes the position and orientation of the imaging device at the time of registration as small as possible.

Hereinafter, an example of information processing according to the first embodiment executed by the information processing apparatus 100 is illustrated in FIG. 3.
In step S10, the distance information acquisition unit 10 acquires a distance image including the target object, and acquires distance information from the distance image to the surface of the target object. For example, the distance information acquisition unit 10 identifies each position of the pattern based on the pattern light projected on the target object, calculates the distance of the target object by triangulation from the correspondence with the position captured by the imaging device, Get distance information.
In step S20, the two-dimensional image acquisition unit 20 acquires image data of a two-dimensional image including the target object imaged by the imaging device.
In step S <b> 30, the schematic shape generation unit 30 reconstructs the three-dimensional shape of the target object from the distance information obtained by the distance information acquisition unit 10. Since the distance information from one direction includes a surface where the target object is not visible, the schematic shape generation unit 30 generates a schematic shape by combining the distance information acquired from the distance images captured from different viewpoints.
In step S40, the geometric feature selection unit 50 selects distance information and a method for extracting a geometric feature from a two-dimensional image. For example, the geometric feature selection unit 50 selects a preset extraction method or an extraction method set by the user via the CRT 111, the KB 110, or the like as the extraction method used by the geometric feature extraction unit 40. Here, the information processing apparatus 100 may extract only the geometric features used for position and orientation estimation. Therefore, the geometric feature selection unit 50 selects a geometric feature extraction method that extracts only the geometric features necessary for position and orientation estimation.

In step S50, the geometric feature extraction unit 40 extracts a two-dimensional geometric feature from the two-dimensional image in which the target object is captured based on the geometric feature extraction method selected by the geometric feature selection unit 50. Hereinafter, for simplification of description, a point on an edge and its direction will be described as an example as a two-dimensional geometric feature.
In step S60, the geometric feature extraction unit 40 extracts a three-dimensional geometric feature from the distance image including the target object based on the geometric feature extraction method selected by the geometric feature selection unit 50. Hereinafter, as a three-dimensional geometric feature, a point on the surface and its normal will be described as an example for the sake of simplicity.

Here, when performing alignment, since the position is converted in accordance with the estimated position and orientation, the edge must be a three-dimensional point instead of a two-dimensional one.
Therefore, in step S70, the geometric feature arrangement unit 60 calculates a three-dimensional position when the position on the image of the edge extracted in step S50 is projected onto the schematic shape generated in step S30. More specifically, the geometric feature arrangement unit 60 is a three-dimensional outline of the edge when the position of the edge on the image of the target object is projected into a three-dimensional outline shape with reference to an imaging device that images the target object. 3D position information on the shape is obtained.
In step S80, the viewpoint-specific feature holding unit 70 uses the position / orientation information (position / orientation information) of the imaging device as a table key and the 3D geometric feature extracted in step S60 and the 3D position information in step S70. The obtained two-dimensional geometric feature is registered.

  According to the processing of the present embodiment, a geometric feature can be generated for extracting a position and orientation by extracting a geometric feature from a two-dimensional image of a target object and distance information. That is, since a model capable of restoring the target object is not generated, a technique for realizing the position and orientation estimation of the target object without a CAD model can be provided with less effort.

<Embodiment 2>
FIG. 4 is a diagram illustrating an example of a software configuration and the like of the information processing apparatus 100 according to the second embodiment.
The information processing apparatus 100 includes a viewpoint-specific feature selection unit 80 and a position / orientation estimation unit 90 in addition to the configuration of the first embodiment as a software configuration.
In the position / orientation estimation in the position / orientation estimation unit 90 described later, the position / orientation of the target object in the image is obtained using the generated geometric model. When the target object has a three-dimensional shape, it is impossible to observe the object itself or to observe the opposite surface. Therefore, by enumerating only the feature quantities observed from the observation viewpoint in advance, the efficiency of position and orientation estimation Can be improved.
When the position / orientation estimation is executed, when the position / orientation information of the imaging device and the target object is obtained, the viewpoint-specific feature selection unit 80 performs registration from the table held in the viewpoint-specific feature holding unit 70. The degree of similarity with the key, which is the position and orientation information of is calculated. Then, the viewpoint-specific feature selection unit 80 selects position and orientation information having a high degree of similarity as a candidate. The viewpoint-specific feature selection unit 80 performs the following processing as the similarity calculation. That is, the viewpoint-specific feature selection unit 80 obtains a vector from the model origin to the imaging device origin among the inquired position / orientation information and the position / orientation information that is a key registered in the table. Calculate the dot product. Then, the viewpoint-specific feature selection unit 80 obtains at least one having a larger value. When using a plurality of pieces of position / orientation information with high similarity, the viewpoint-specific feature selection unit 80 can perform linear interpolation of the geometric features by assigning the weight of the geometric features according to the inner product.

Alternatively, the viewpoint-specific feature selection unit 80 may select a registered geometric feature using a computer graphics drawing function. When using data of other viewpoints held as a table, the viewpoint-specific feature selection unit 80 uses a mesh at the imaging device viewpoint in accordance with the position and orientation information for estimating the approximate shape generated by the approximate shape generation unit 30. Drawing is performed, and a figure as an index is drawn at the position of the geometric feature. When a figure serving as an index of each geometric feature is shielded by the back side of the mesh or other part, the viewpoint-specific feature selection unit 80 is observed by using the fact that it cannot be observed by graphics depth determination. Only the most likely geometric features can be extracted from the table.
As a result, the geometric feature used in the position / orientation estimation in the position / orientation estimation unit 90 can be used in accordance with the visible state. Furthermore, the geometric feature often defines a local plane, and the detection position may shift depending on the change in the observation direction. As a method for correcting these positions, for example, the viewpoint-specific feature selection unit 80 can use a bundle adjustment method. The bundle adjustment method corrects the position / orientation information of the imaging device and the three-dimensional position of the attention point so as to minimize the projection error of the attention point observed from a plurality of viewpoints onto the screen. Since there is information observed from a plurality of viewpoints also in the processing of this embodiment, the bundle adjustment method can be applied. Thereby, the ambiguity of the three-dimensional position of the geometric feature can be reduced, and the accuracy as a geometric model for position and orientation estimation can be improved.

  The position / orientation estimation unit 90 determines the position of the target object so that the difference between the plurality of geometric features selected by the viewpoint-specific feature selection unit 80 and the geometric feature of the target object corresponding to each of the geometric features is reduced. Estimate posture. That is, the position / orientation estimation unit 90 is based on the position of the geometric feature extracted by the geometric feature extraction unit 40 and the geometric feature of the model coordinate system selected by the viewpoint-specific feature selection unit 80 based on the approximate position and orientation of the target. Then, optimization calculation is performed using the position and orientation information of the imaging device coordinate system of the geometric model as a parameter. The position / orientation estimation unit 90 can convert the position of the three-dimensional geometric feature into the imaging device coordinate system by multiplying the six parameters of the estimated position / orientation information by matrix expression. The position / orientation estimation unit 90 uses the difference from the geometric feature extracted by the geometric feature extraction unit 40 as an evaluation value, and estimates the parameters of the position / orientation information so as to reduce the difference. As an estimation method, a Gauss-Newton method or the like that is a nonlinear optimization method may be applied. If the correspondence between the three-dimensional positions is known, the position and orientation information can be obtained by a method using eigenvalue decomposition such as DLT (Direct Linear Transform). In addition, when noise is included, the estimation accuracy of the least-square estimation is lowered, so it is also effective to perform weighting by dispersion of measurement information by a robust estimation method such as M estimation. Furthermore, when multiple geometric features can be used, a combination of three or more points for estimating the position and orientation information parameters is selected at random, and the parameter with the fewest residuals is used from among the multiple sets of parameters. It is also effective to use the RANSAC method.

Hereinafter, an example of information processing according to the second embodiment executed by the information processing apparatus 100 is illustrated in FIG.
The processing in steps S10 to S80 in FIG. 5 is the same as the processing in steps S10 to S80 in FIG.
In step S90, the viewpoint-specific feature selection unit 80 calculates the approximate value of the observation position / orientation information of the target object using the distribution direction of the geometric feature, the luminance distribution around the geometric feature, and the like, and the table registered in step S80. The record similar to the position and orientation information which is the key of is searched. It is assumed that the record stores a plurality of pieces of information on the two-dimensional geometric feature when the geometric feature is registered and the geometric feature related to the distance information.
In step S100, the position / orientation estimation unit 90 converts the coordinates of each three-dimensional position of the geometric feature list referred to in step S90 using the value of the position / orientation information being estimated. In order to associate with the geometric feature obtained from the two-dimensional image, the position / orientation estimation unit 90 may perform projective transformation so as to match the three-dimensional space with the coordinate system of the camera. The position / orientation estimation unit 90 associates the geometric features that have undergone the projective transformation with the geometric features extracted in step S50 on the screen. The same applies to the distance information, and the position / orientation estimation unit 90 associates the distance information. The registered geometric feature and the observed geometric feature associated with each other can estimate the parameters of the position and orientation information using an index that reduces the distance between the corresponding points. These algorithms may use a disclosed technique generally called ICP. In general, since coordinate transformation is nonlinear, a solution can be obtained by a Gauss-Newton method or the like in which linear approximation is performed and an iterative solution is performed.

  According to the processing of this embodiment, the position and orientation of the target object can be estimated using the generated geometric model even without a CAD model.

<Embodiment 3>
When automating production of robots, etc., it is convenient if new parts are added, changes are made if the shape is slightly different from lot to lot, or shape models for position and orientation estimation are automatically added / updated. improves.
FIG. 6 is a diagram illustrating an example of a software configuration of the information processing apparatus 100 according to the third embodiment.
The information processing apparatus 100 includes a geometric feature correspondence search unit 95 as a software configuration in addition to the configuration of the second embodiment.
The geometric feature correspondence search unit 95 determines whether the part category is an unknown one or a geometric feature from a viewpoint that has not been registered so far. The geometric feature correspondence search unit 95 creates a new position / orientation estimation geometric model if it is a new part, and updates the existing position / orientation measurement geometric model if it can be corrected in the past. Control. Therefore, the geometric feature correspondence search unit 95 performs correlation calculation with all the registered geometric features, calculates a value normalized by the quantity or the like as a likelihood, and exceeds a threshold set by the user or the system. Process by whether or not. The geometric feature correspondence search unit 95 calculates a difference between the geometric features registered in the viewpoint-specific feature holding unit 70 from the viewpoint-specific feature selection unit 80 and the arrangement of the three-dimensional positions. If the difference when using the geometric feature with the smallest difference is larger than the set threshold value, the geometric feature correspondence search unit 95 assumes that the position / orientation feature point data is new or updated and the viewpoint The detected geometric feature is additionally registered in the separate feature holding unit 70. The geometric feature correspondence search unit 95 generates a new model structure for position and orientation estimation in the case of a component in an unregistered category other than the target so far. Register with it. When the position / orientation information is not registered in advance, the geometric feature correspondence search unit 95 newly registers the position / orientation information as a key, and adds the detected three-dimensional position of the geometric feature as data. Further, when the angle formed by the line of sight is large with respect to the line-of-sight direction already registered as the observation direction, the geometric feature correspondence search unit 95 assumes that the shape of the part has been changed, and obtains geometric feature data having a large difference. Delete and add the 3D position of the detected new geometric feature. As a result, the data can be updated.

  According to the processing of the present embodiment, the data can be updated using the distance information and the two-dimensional image for which the position / orientation estimation is being performed, so that even if the user does not give an instruction to generate a model, the processing is automatically performed. Thus, the position and orientation estimation model can be updated and labor can be reduced. Further, since the accuracy of the position / orientation estimation model due to the shape difference can be reduced, the failure of position / orientation estimation due to the shape difference can be reduced, and the system stop time can be shortened.

<Embodiment 4>
By using the processing of the above-described embodiment as a system for bulk picking and assembly work by a robot, it becomes possible to pick a target part without using design information by conventional CAD, reducing teaching effort, and gripping Stability with respect to operation can be improved.
FIG. 7 shows a configuration diagram of a system using a robot arm as a positioning mechanism of the present embodiment.
The information processing apparatus 100 is connected to an imaging device 510 and a projector 520 that are mounted on the end effector unit of the robot arm 600. The projector 520 projects a pattern figure including a spatially coded pattern and a code identified by a projection position such as a random dot pattern onto the target object 700. The projected reflected image is captured by the imaging device 510. In addition, the imaging device 510 captures a reflection image when a pattern is not projected or is fully lit as a two-dimensional image. The information processing apparatus 100 controls the imaging device 510 and the projector 520. By registering the table of the viewpoint-specific feature holding unit 70 in the external storage device, data can be used permanently. Since the hand position of the robot arm can be calculated by using the robot arm encoder information, the calculation result can be used as the position and orientation information of the imaging device.
The parameters of the position / orientation information of the target object estimated by the position / orientation estimation unit 90 are obtained by calibrating the relative position / orientation between the hand unit of the end effector and the imaging device in advance, thereby obtaining the hand position of the robot arm. Can be converted. The information processing apparatus 100 can grasp the target component by moving the robot arm from the current position by transmitting the converted hand position of the robot arm as the hand position command value of the robot arm 600. When the shape of the measurement target is unknown, the information processing apparatus 100 moves the robot arm position so that the target object can be observed from a plurality of viewpoints, acquires distance information and a two-dimensional image, and performs geometry for position and orientation estimation. A table of features can be generated. Thereby, even when there is no CAD model, position and orientation estimation can be performed by a series of procedures.

<Other embodiments>
The present invention supplies a program that realizes one or more functions of the above-described embodiments to a system or apparatus via a network or a storage medium. It can also be realized by a process in which one or more processors in the computer of the system or apparatus read and execute the program. It can also be realized by a circuit (for example, ASIC) that realizes one or more functions.

According to the processing of each embodiment described above, position and orientation estimation is possible even when the CAD model cannot be used as the three-dimensional information of the target object when the target object alignment technique is used on the production line site. Is possible. For example, even if the CAD model of the target part cannot be used in the first place, such as when there is only a two-dimensional drawing of an off-the-shelf general-purpose part or when the design information of the design department cannot be used confidentially in the manufacturing department, Can make models. Further, even when CAD data is already present, it can be used as mesh data in the schematic shape generation unit without changing the configuration of the apparatus. In addition, if the 3D shape of the CAD model at the time of design is different, such as when multiple parts are assembled to the target part, the model in the combined state can be redesigned using the CAD design tool. It takes time and effort to relocate. Therefore, it is possible to eliminate the trouble by registering as a single position / orientation estimation model combined by the method of the above-described embodiment.
In addition, shape reconstruction by reverse engineering requires expensive imaging equipment to be brought into the field, procedures for imaging and application proficiency are required, and it is difficult to use it universally on the production line. That is solved.
Therefore, it is possible to provide a technique for realizing the position and orientation estimation of a target object without a CAD model with less effort.

100 Information processing apparatus 101 CPU
112 HDD

Claims (9)

An acquisition means for acquiring distance information to the surface of the target object;
Generating means for generating a schematic shape of the target object based on the distance information;
Image acquisition means for acquiring an image of the target object;
Extracting means for extracting a geometric feature of the target object based on the distance information, and extracting a geometric feature of the target object based on the image;
Arrangement means for arranging geometric features of the target object extracted by the extraction means in the approximate shape generated by the generation means;
An information processing apparatus. The generation unit generates a three-dimensional schematic shape of the target object based on the distance information,
The extraction means extracts a three-dimensional geometric feature of the target object based on the distance information;
The information processing apparatus according to claim 1, wherein the arranging unit arranges a three-dimensional geometric feature of the target object in the three-dimensional schematic shape. The image acquisition means acquires a two-dimensional image of the target object,
The extraction means extracts a two-dimensional geometric feature of the target object based on the two-dimensional image;
The arranging means is configured to project the two-dimensional geometric feature of the target object onto the three-dimensional schematic shape based on an imaging device that images the target object. The information processing apparatus according to claim 2, wherein three-dimensional position information is obtained, and the two-dimensional geometric feature is further arranged in the three-dimensional schematic shape according to the three-dimensional position information. A selection means for selecting a geometric feature extraction method by the extraction means;
The information processing apparatus according to claim 1, wherein the extraction unit extracts a geometric feature of the target object according to an extraction method selected by the selection unit.   5. The storage device according to claim 1, further comprising a storage unit configured to store, in a storage area, a plurality of geometric features arranged in the schematic shape by the arrangement unit in association with position and orientation information of a device that captures the target object. Information processing device. Geometric feature selection means for selecting a plurality of geometric features stored in the storage area in association with position and orientation information similar to the position and orientation information based on position and orientation information of a device that captures a target object;
Estimating means for estimating the position and orientation of the target object so as to reduce the difference between the plurality of geometric features selected by the geometric feature selecting means and the geometric features of the target object corresponding to each of the geometric features;
The information processing apparatus according to claim 5, further comprising:   When the difference between the plurality of geometric features selected by the geometric feature selection means and the geometric feature of the target object corresponding to each of the geometric features is larger than a threshold, the plurality of geometric features of the target object are The information processing apparatus according to claim 6, further comprising registration means for registering in the storage area in association with position / orientation information of a device that images the target object. An information processing method executed by an information processing apparatus,
An acquisition step of acquiring distance information to the surface of the target object;
Generating a schematic shape of the target object based on the distance information;
An image acquisition step of acquiring an image of the target object;
Extracting the geometric feature of the target object based on the distance information and extracting the geometric feature of the target object based on the image;
An arrangement step of arranging the geometric feature of the target object extracted by the extraction step in the schematic shape generated by the generation step;
An information processing method including:   The program for functioning a computer as each means of the information processing apparatus of any one of Claims 1 thru | or 7.

Images