JP2014173922A - Imaging apparatus, rotation angle estimation method, and rotation angle estimation program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To estimate, from a range image, an angle at which an apparatus axis defined by a structure of an imaging apparatus capturing the range image is rotated from a spatial axis defined by a structure or the like in a real space.SOLUTION: An imaging apparatus 6 includes a TOF camera 4 and an information processor 5. The TOF camera 4 captures a range image. The information processor 5 includes a reference setting unit 17 and an estimation unit 18. The reference setting unit 17 sets two reference points on the range image on the basis of a platform end part IC and acquires a positional relation on the range image between the two reference points. Assuming that the two reference points have a parallel positional relation with one of three spatial axes on the range image, the estimation unit 18 estimates such angles of rotation around three apparatus axes as to align three apparatus axes with three spatial axes respectively.

Description

  The present invention relates to an imaging apparatus that captures a distance image. More specifically, an imaging apparatus that estimates, based on a distance image, a rotation angle at which an apparatus axis defined based on the structure of the imaging apparatus rotates from a spatial axis defined based on a structure in real space, The present invention relates to a rotation angle estimation method and a rotation angle estimation program.

  Conventionally, a TOF (Time Of Flight) camera is known as an imaging device that captures a distance image (see, for example, Patent Document 1). The TOF camera includes a light projecting unit that projects sinusoidal modulated light (infrared light), and a plurality of light receiving elements arranged in a matrix on the light receiving surface. The TOF camera receives the diffused light (reflected light) reflected from the projected infrared light by each light receiving element, thereby associating the distance image in which the distance is associated with each pixel and the received light intensity in each pixel. The received light intensity image is captured. The distance image detects the phase difference between the infrared light projected for each pixel and the reflected light received by each pixel, and shows the time (flight time) from when the infrared light is projected until the reflected light is received. It can be obtained by associating with a pixel.

  Such a TOF camera is used for applications such as a sensor for detecting the presence or absence of an object, and is installed diagonally downward or diagonally upward at the end of the target area so that the entire target area can be imaged. When the TOF camera is arranged in this way, the device axis defined by the structure of the TOF camera is in a state of rotating from the spatial axis defined on the basis of the structure in the real space. It becomes difficult to grasp the exact position of the object in the real space from the image.

  Therefore, normally, when the TOF camera is installed, the rotation angle at which each device axis rotates from the corresponding space axis is measured by a worker using a jig or the like and registered as a parameter in the TOF camera. The TOF camera grasps the exact position of the object in the real space using these parameters.

Special table 2010-534000 gazette

  Conventionally, it is necessary for the operator to master the operation in which the operator measures the rotation angle at which each device axis rotates from the corresponding space axis and registers it as a parameter in the TOF camera. In addition, this procedure increases the procedure and time required for installing the TOF camera, which makes the installation operation of the TOF camera complicated.

  Accordingly, an object of the present invention is to provide a technique for estimating a rotation angle at which each device axis rotates from a corresponding space axis based on a distance image.

  The imaging device according to the present invention includes an imaging unit, a reference setting unit, and an estimation unit.

  The imaging unit captures a distance image obtained by capturing the target area projected on the light receiving surface. The imaging unit has, for example, a known TOF camera that captures a received light intensity image obtained by capturing the same target area as the distance image together with the distance image, or a configuration that captures the distance image by scanning laser light.

  The reference setting unit sets two reference points on the distance image captured by the imaging unit. Note that the reference setting unit may set the reference point in the distance image based on the distance feature point due to a step or the like existing in the real space, for example, or the received light intensity feature point by the marker or the line tape existing in the real space. May be extracted from the received light intensity image, and a reference point corresponding to the distance image may be set.

  The estimation unit adjusts the rotation angle about each device axis to match the three device axes with the three space axes, and the two reference points set by the reference setting unit are one of the three space axes on the distance image. It is estimated that the positional relationship is parallel to the axis. Here, the space axis is defined so as to be orthogonal to each other with reference to two reference points. The device axis is defined so as to be orthogonal to each other with respect to the light receiving surface. For example, one of the device axes is an axis perpendicular to the light receiving surface, and the other two of the device axes are The axis along the direction in which the pixels are arranged on the light receiving surface.

  In this way, by setting two reference points on the distance image and using them for the positional relationship of the two reference points on the distance image, the three device axes are aligned around the three spatial axes. The rotation angle can be estimated based on the distance image.

  In addition, before the estimation unit tries to estimate the rotation angle around each device axis, the positional relationship is parallel to the spatial axis to be combined with a device axis different from the device axis for which the rotation angle is to be estimated. It is preferable to set two reference points. Further, the reference setting unit is configured based on the positional relationship corrected by the rotation angle around the device axis previously obtained by the estimation unit before the estimation unit tries to estimate the rotation angle around each device axis. It is preferable to set a reference point. In addition, when estimating the rotation angle around each device axis, the estimation unit estimates the rotation angle around each device axis based on the positional relationship corrected by the rotation angle around the device axis obtained previously. It is preferable. Further, it is preferable that the estimation unit repeats the derivation of the rotation angle around each device axis until the rotation angle around each device axis converges to a certain range.

  By configuring the reference setting unit and the estimation unit as described above, even if an error based on the resolution included in the distance image is large, the influence of the error included in the distance image is suppressed, so that each device axis is a spatial axis. It is possible to accurately estimate the rotation angle of the rotation.

  The rotation angle estimation method according to the present invention is an invention that causes an information processing apparatus to execute processing corresponding to the configuration of the imaging unit, the reference setting unit, and the estimation unit described above.

  Furthermore, the rotation angle estimation program according to the present invention is an invention that, when installed in an information processing apparatus, causes the information processing apparatus to execute processing corresponding to the configuration of the imaging unit, the reference setting unit, and the estimation unit described above.

  According to these inventions, by setting two reference points in the distance image and using the positional relationship between the two reference points on the distance image, the rotation angle at which each device axis rotates from the space axis can be determined. Can be estimated based on the distance image. Thereby, it becomes possible to simplify the operation | work at the time of installation of an imaging device.

It is the schematic which shows the platform of the station in which the TOF camera is installed. It is a schematic diagram which shows the structure of the principal part of an imaging device. It is a flowchart which shows an example of a rotation angle estimation process. It is a flowchart which shows an example of an imaging step. It is a flowchart which shows an example of a reference | standard setting step. It is a flowchart which shows the other example of a reference | standard setting step. It is a flowchart which shows an example of an estimation step.

  Hereinafter, an imaging device, a rotation angle estimation method, and a rotation angle estimation program according to embodiments of the present invention will be described.

  An example of the imaging apparatus described in this embodiment is provided with a space near the end of a platform of a station as a target area, and detects the presence or absence of an object located in the detection target area. Then, the imaging device outputs a signal that tells the detection of the object to the external device and causes the external device to perform an alarm or the like, thereby preventing a passenger from falling on the track from the platform.

  FIG. 1 is a conceptual diagram illustrating a state in which an imaging device is installed on a platform. FIG. 1A is a side view of an imaging area captured by the imaging apparatus as viewed from the right hand side. FIG. 1B is a plan view of an imaging area captured by the imaging apparatus as viewed from above. FIG. 1C is a side view of an imaging area captured by the imaging apparatus viewed from the back side. In FIG. 1A, spatial axes X, Y, and Z that are defined with reference to the platform, which is a structure in real space, and device axes x, y, and Z that are defined with reference to the structure of the light receiving surface of the imaging device. z is displayed. The space axes X, Y, and Z may be defined based on any structure as long as it is a structure in real space. The device axes x, y, and z may be defined based on any structure as long as the structure of the light receiving surface of the imaging device.

  The station 1 has a platform 1A and a track 1B. The platform 1A is approximately horizontal and has a platform end 1C facing the track 1B.

  The platform 1 </ b> A has a fall prevention fence 2. The fall prevention fence 2 is installed away from the platform end 1C on the side opposite to the track 1B. Moreover, the fall prevention fence 2 is arrange | positioned in parallel with the platform edge part 1C. The fall prevention fence 2 has a slide door 2A and a housing 2B. The housing 2B functions as a door pocket of the slide door 2A. The slide door 2A is provided at a position facing each door of the train that stops on the track 1B. The sliding door 2A is slidably attached to the housing 2B, and is slid in the left-right direction from the closed state shown in FIG. It becomes a state to do.

  A support column 3 and a TOF camera 4 are installed on the platform 1A. The support column 3 is installed so as to be close to the housing 2 </ b> B of the fall prevention fence 2. Moreover, the support | pillar 3 is arrange | positioned rather than the housing | casing 2B of the fall prevention fence 2 at the platform end part 1C side. The TOF camera 4 constitutes an imaging unit according to this embodiment. The TOF camera 4 is attached to the column 3 and images an obliquely lower side from the upper end of the column 3 to obtain an image of an area between the slide door 2A and the track 1B projected onto the light receiving surface. The TOF camera 4 may be directly attached to the housing 2B of the fall prevention fence 2. It is preferable that a plurality of TOF cameras 4 and support columns 3 are provided so as to image a space between each door of the train that stops on the track 1B and each slide door 2A.

  Here, the space axes X, Y, and Z are defined with reference to a platform that is a structure in real space. The space axis Z forms a horizontal straight line in the real space, and is defined as a straight line parallel to the platform end 1C. The space axis X is a straight line orthogonal to the space axis Z, and is defined here as a horizontal straight line orthogonal to the platform end 1C in real space. The space axis Y is a straight line orthogonal to the space axis Z and the space axis X, and is defined here as a vertical straight line orthogonal to the platform end 1C in real space. Here, the spatial axes X, Y, and Z are determined to intersect at the center of the light receiving surface of the TOF camera 4.

  The device axes x, y, and z are defined based on the structure of the TOF camera 4. The device axis z is a straight line perpendicular to the light receiving surface of the TOF camera 4. The device axis x is a straight line parallel to the row direction of the arrangement of the light receiving elements constituting the TOF camera 4 as will be described in detail later. The device axis y is a straight line parallel to the column direction of the arrangement of the light receiving elements, details of which will be described later. Here, the apparatus axes x, y, and z are determined so as to intersect at the center of the light receiving surface of the TOF camera 4. The TOF camera 4 is arranged so that the device axes x, y, and z are slightly inclined from the spatial axes X, Y, and Z, respectively.

  FIG. 2A is a block diagram of an imaging apparatus 6 that includes the TOF camera 4 as part of its configuration. FIG. 2B is a conceptual diagram showing the configuration of the light receiving surface of the TOF camera 4.

  The imaging device 6 includes a TOF camera 4 and an information processing device 5.

  The TOF camera 4 operates in accordance with a control signal input from the information processing device 5 and captures a distance image in which a distance is associated with each pixel and a received light intensity image in which a received light intensity is associated with each pixel. Are output to the information processing apparatus 5. The TOF camera 4 includes a light projecting unit 12, a light receiving unit 13, and a distance calculating unit 14. The light projecting unit 12 irradiates the imaging area with infrared light when a control signal is input. As shown in detail in FIG. 2B, the light receiving unit 13 includes n × m light receiving elements 13A arranged in a matrix. In the row direction of the arrangement of the light receiving elements 13A, that is, in the direction along the device axis x shown in FIG. 1, n light receiving elements 13A are arranged. In the column direction of the arrangement of the light receiving elements 13A, that is, the direction along the apparatus axis y shown in FIG. 1, m light receiving elements 13A are arranged. Each light receiving element 13A receives the reflected light of the infrared light irradiated to the reflector in the imaging area. The distance calculation unit 14 calculates, for each light receiving element 13A, the time (flight time) from when the light receiving element 13A receives reflected light after irradiating the imaging area with infrared light based on the phase difference of the light. To do. Thus, the TOF camera 4 captures a received light intensity image by associating the received light intensity of the reflected light with each pixel corresponding to each light receiving element 13A, and captures a distance image by associating the distance with each pixel. . The TOF camera 4 can capture about 5 to 10 frames of these images per second.

  The imaging unit may not be able to obtain a received light intensity image as long as a distance image can be generated, and may adopt another configuration instead of the TOF camera 4. For example, the structure which scans a laser beam may be sufficient.

  The information processing apparatus 5 is installed with an object detection program for executing object detection processing and a rotation angle estimation program for executing rotation angle estimation processing. The object detection process is a process for detecting the presence or absence of an object located in the detection target area using a distance image received from the TOF camera 4. The detection target area is a part of the imaging area captured by the TOF camera 4. The rotation angle estimation process is a process for estimating the rotation angle at which each device axis rotates from the space axis using the distance image received from the TOF camera 4.

  For this reason, the information processing apparatus 5 includes a control unit 11 and a storage unit 15. The control unit 11 controls the entire information processing apparatus 5 to execute object detection processing and rotation angle estimation processing. The storage unit 15 stores various program codes including an object detection program and a rotation angle estimation program, distance images and received light intensity images received from the TOF camera 4, various parameters including area setting parameters and rotation angle parameters, and the like.

  Here, an outline of an object detection process for detecting the presence or absence of an object in the detection target area at a normal time when a train or the like is operated at the station 1 shown in FIG. 1 will be briefly described.

  The control unit 11 outputs a control signal to the TOF camera 4 repeatedly at a constant cycle. The TOF camera 4 receives a control signal, captures a distance image and a received light intensity image, and outputs the obtained image to the information processing device 5. When the control unit 11 receives the distance image and the received light intensity image from the TOF camera 4, the control unit 11 stores the distance image and the received light intensity image in the storage unit 15, and reads the area defining parameter and the rotation angle parameter registered in advance. The area defining parameter is a parameter representing the range of the detection target area in the spatial axes X, Y, and Z shown in FIG. The rotation angle parameter is a parameter obtained by estimating the rotation angle at which the device axes x, y, and z shown in FIG. 1 are rotating from the spatial axes X, Y, and Z, respectively. The control unit 11 detects the presence / absence of an object in the detection target area based on the parameter read from the storage unit 15 and the distance image. And the control part 11 will output a detection signal to external devices, such as a control apparatus of a fall prevention fence, a central control apparatus, and an alarm device, if an object is detected in a detection object area. The external device to which the detection signal is input performs processing such as notification by a warning sound or lighting of a warning lamp in order to notify a station staff or the like that an object is detected in the detection target area.

  As shown in FIG. 1, the TOF camera 4 is installed such that the apparatus axes x, y, and z are deviated from the spatial axes X, Y, and Z. For this reason, the position of the object detected from the image by the control unit 11 is not directly associated with the spatial axes X, Y, and Z, but is associated with the device axes x, y, and z. On the other hand, the area defining parameters are associated with the spatial axes X, Y, and Z. Therefore, it is difficult for the control unit 11 to accurately detect the presence or absence of an object in the detection target area using the image as it is.

Therefore, the control unit 11 uses the rotation angle parameters that estimate the rotation angles at which the device axes x, y, and z are rotating from the space axes X, Y, and Z, respectively, on the space axes X, Y, and Z. Get the position of the object. For example, the position of the object on the device axes x, y, and z is converted to the position of the object on the space axes X, Y, and Z by the following conversion formula. Note that θ _x , θ _y , and θ _z used in the following equations are a rotation angle parameter around the device axis x, a rotation angle parameter around the device axis y, and a rotation angle parameter around the device axis z, respectively. .

  The control unit 11 estimates the position of the object on the spatial axes X, Y, and Z using the above formula, and accurately detects the presence or absence of the object in the detection target area as compared with the detection target area. .

  Next, when the TOF camera 4 is installed on the platform 1A at the station 1 shown in FIG. 1, the rotation angles at which the device axes x, y, and z are rotated from the spatial axes X, Y, and Z, respectively. An outline of the rotation angle estimation process estimated from the image will be described. The rotation angle estimation process may be performed only at the beginning of the installation of the TOF camera 4 or may be performed at an arbitrary timing by receiving an instruction from the outside, and automatically every time a certain period elapses. May be implemented.

  The control unit 11 includes an image acquisition unit 16, a reference setting unit 17, and an estimation unit 18 as functional blocks for performing the rotation angle estimation process. The control unit 11 executes the rotation angle estimation process by causing the image acquisition unit 16, the reference setting unit 17, and the estimation unit 18 to function.

  FIG. 3 is a schematic flowchart of the rotation angle estimation process.

The control unit 11 first performs the imaging step (S1) by causing the image acquisition unit 16 to function, and causes the TOF camera 4 to capture the distance image and the received light intensity image. Next, the first reference setting step (S2) is performed by causing the reference setting unit 17 to function, and a set of reference points is set. Next, the first estimation step (S3) is performed by causing the estimation unit 18 to function, and the rotation angle parameter θx is calculated. Next, the second reference setting step (S4) is performed by causing the reference setting unit 17 to function, and a set of reference points is reset. Next, the second estimating step (S5) is performed by causing the estimating unit 18 to function, and the rotation angle parameter θy is calculated. Next, the third reference setting step (S6) is performed by causing the reference setting unit 17 to function, and a set of reference points is reset. Next, the third estimation step (S7) is performed by causing the estimation unit 18 to function, and the rotation angle parameter θz is calculated. The rotation angle parameter theta _x, theta _y, each value convergence theta _z performed to determine whether or not to (stable) (S8), the rotation angle parameter θ _x, θ _y, θ _z converges (stable) These steps (S2 to S7) are repeated until When the rotation angle parameters θ _x , θ _y , θ _z converge (stable), the rotation angle parameters θ _x , θ _y , θ _z are stored in the storage unit 15 (S9).

As shown here, by repeating the previous steps (S2 to S7) until the rotation angle parameters θ _x , θ _y , θ _z converge (stabilize), the rotation angle parameters around each device axis can be accurately obtained. Can be derived.

Note that the second reference setting step and the third reference setting step are not necessarily required, and a set of reference points set in the previous reference setting step may be used as they are. Further, the order of the rotation angle parameters θ _x , θ _y , and θ _z estimated in the first estimation step, the second estimation step, and the third estimation step may be switched. Further, instead of repeating each step until the rotation angle parameters θ _x , θ _y , and θ _z converge (stable), each step may be performed once.

  FIG. 4 is a flowchart illustrating an example of the imaging step.

  The imaging step is a step of generating a distance image by associating a distance with each pixel. In this flowchart, a plurality of distance images are acquired from the TOF camera 4, and an average distance image is generated by averaging the distances associated with each pixel among the distance images. Note that the imaging step may use an average distance image or a single image as it is. Further, not the simple distance but the coordinate position on the device axis x, y, z may be associated with each pixel.

  Specifically, the image acquisition unit 16 first outputs a control signal to the TOF camera 4 and receives a distance image from the TOF camera 4 (S11). The image acquisition unit 16 repeats this step until a plurality of (for example, 10) distance images are acquired. Next, the image acquisition unit 16 averages the distances associated with the pixels belonging to the plurality of distance images among the distance images, and generates an average distance image (S12). Next, the image acquisition unit 16 stores the generated average distance image in the storage unit 15 (S13).

  FIG. 5 is a flowchart illustrating an example of the reference setting step.

  The reference setting step is a step of setting two reference points on the average distance image stored in the storage unit 15. In this flowchart, the platform end 1C shown in FIG. 1 is extracted from the average distance image stored in the storage unit 15, and two reference points corresponding to a line on the platform 1A parallel to the platform end 1C are defined as 1 Two reference points having a positional relationship parallel to two spatial axes are set. Since the platform end 1C has a large step (for example, a height of about 1500 mm) between the adjacent line 1B, the platform end 1C is a distance feature point and is suitable as a structure extracted from a distance image.

  In setting the reference point, in addition to extracting a structure in the real space from the distance image, a structure in the real space (for example, a light reception intensity feature point such as a marker or a line tape) is extracted from the light reception intensity image, and the distance image is extracted. A corresponding reference point may be set, or a distance image or a received light intensity image may be presented to the worker, and a reference point corresponding to a structure designated by the worker may be set in the distance image.

  Specifically, the reference setting unit 17 reads a distance (z-axis coordinate in the illustrated example) associated with each pixel belonging to the average distance image stored in the storage unit 15, and sets adjacent pixels to each other. Compare (S21, S22). If the difference in distance between adjacent pixels exceeds a threshold value (in the example shown, the line at the end of the platform and a step height of 1500 mm is used as a threshold value), it is estimated that it is a stepped portion in real space. For this reason, the reference setting unit 17 determines a pixel whose distance difference between adjacent pixels exceeds a threshold value or a pixel in the vicinity thereof (for example, one or two pixels closer to the platform than the pixel). The line is stored as a line on the platform 1A parallel to the platform end 1C (S23). The reference setting unit 17 estimates whether all the pixels match a line on the platform 1A parallel to the platform end 1C.

Next, the reference setting unit 17 sets a set of reference points (x ₁ , y ₁ , z ₁ ), (x ₂ , y ₂ , z ₂ ) that coincide with a line on the platform 1A parallel to the platform end 1C. Set and store in the storage unit 15 (S24). The reference setting unit 17 sets a set of reference points (x ₁ , y ₁ , z ₁ ), (x ₂ , y ₂ , z ₂ ) by a predetermined number (for example, 10 sets).

The reference point (x ₁ , y ₁ , z ₁ ) and the reference point (x ₂ , y ₂ , z ₂ ) are selected as two points separated from each other by a certain distance (for example, 190 mm to 210 mm) or more. It is preferable. In addition, this flowchart is preferably used in the first reference setting step and the second reference setting step in the rotation angle estimation process shown in FIG.

  FIG. 6 is a flowchart illustrating another example of the reference setting step.

  In this flowchart, not a line on the platform 1A parallel to the platform end 1C but a reference point that coincides with a line on the platform 1A orthogonal to the platform end 1C is set.

  More specifically, the reference setting unit 17 reads the distance associated with each pixel belonging to the average distance image stored in the storage unit 15 and compares adjacent pixels (S31, S32). First, a line parallel to the platform end 1C is estimated. Therefore, the reference setting unit 17 stores a pixel whose distance difference between adjacent pixels exceeds a threshold value or a pixel in the vicinity thereof as a line on the platform 1A parallel to the platform end 1C (S33). The reference setting unit 17 estimates whether all the pixels match a line on the platform 1A parallel to the platform end 1C.

Next, the reference setting unit 17 sets a reference point (x ₁ , y ₁ , z ₁ ) located on a line on the platform 1A parallel to the platform end 1C and stores it in the storage unit 15 (S34).

Next, the reference setting unit 17 estimates a line that passes through the reference point (x ₁ , y ₁ , z ₁ ) and is orthogonal to the platform end 1C. Therefore, the distance (z-axis coordinate in the illustrated example) associated with each pixel is compared with the reference point (x ₁ , y ₁ , z ₁ ) (S35, S36). The reference setting unit 17 sets the distance read from each pixel within a certain range with respect to the reference point (x ₁ , y ₁ , z ₁ ) (in the illustrated example, the difference between the z-axis coordinates is within a range of 30 mm before and after). A certain pixel is estimated to coincide with a line on the platform 1A that passes through the reference point (x ₁ , y ₁ , z ₁ ) and is orthogonal to the platform end 1C, and is stored in the storage unit 15. The reference setting unit 17 estimates whether all the pixels coincide with a line on the platform 1A orthogonal to the platform end 1C.

Next, the reference setting unit 17 passes through the reference point (x ₁ , y ₁ , z ₁ ), and matches the reference point (x ₂ , y ₂ , z ₂₎ that matches the line on the platform 1A orthogonal to the platform end 1C. ) Is set (S34). In this case, it is preferable to select a point separated from the reference point (x ₁ , y ₁ , z ₁ ) by a certain distance (for example, 190 mm to 210 mm) as the reference point (x ₂ , y ₂ , z ₂ ). is there. The reference setting unit 17 sets a set of reference points (x ₁ , y ₁ , z ₁ ), (x ₂ , y ₂ , z ₂ ) by a predetermined number (for example, 10 sets).

  This flowchart is preferably employed in the third reference setting step in the rotation angle estimation process shown in FIG. It should be noted that whether to adopt the flowchart shown in FIG. 4 or the flowchart shown in FIG. 5 at each reference setting step is preferably determined according to the device axis from which the rotation angle parameter is to be derived. is there. The rotation angle parameter can be derived with higher accuracy by adopting two reference points arranged so as to be orthogonal to the apparatus axis from which the rotation angle parameter is to be derived. That is, when the rotation angle parameter is derived for the device axis x and the device axis y, it is preferable to adopt the flowchart shown in FIG. 4 for setting a reference point that coincides with a line parallel to the platform end 1C. When deriving the rotation angle parameter with respect to the device axis z, it is desirable to employ the flowchart shown in FIG. 5 for setting a reference point that coincides with a line orthogonal to the platform end 1C.

  If the rotation angle parameter around any device axis is already estimated before the reference setting step is performed, the distance (or device axis x, y) associated with each pixel is estimated. , Z) may be corrected based on the rotation angle parameter. By correcting in this way, a line on the platform 1A parallel to the platform end 1C and a line on the platform 1A orthogonal to the platform end 1C can be estimated more accurately.

  FIG. 7 is a flowchart illustrating an example of the estimation step.

The estimation step has a positional relationship of a set of reference points (x ₁ , y ₁ , z ₁ ) and (x ₂ , y ₂ , z ₂ ) set in the reference setting step for each of the device axes x, y, and z. This is a step of deriving the rotation angle parameters θx, θy, θz around each axis based on the above. Specifically, the estimation unit 18 reads a set of reference points (x ₁ , y ₁ , z ₁ ), (x ₂ , y ₂ , z ₂ ) from the storage unit 15, and uses a predetermined arithmetic expression to calculate A virtual axis in which a set of (x ₁ , y ₁ , z ₁ ), (x ₂ , y ₂ , z ₂ ) has a parallel positional relationship is virtualized, and the virtual axis can be realized around a predetermined device axis The rotation angle parameter at is derived (S41). The estimation unit 18 repeats this step for each set of reference points (x ₁ , y ₁ , z ₁ ) and (x ₂ , y ₂ , z ₂ ) (for example, 10 sets). Next, the estimation unit 18 averages the rotation angle parameters derived for each set of the reference points (x ₁ , y ₁ , z ₁ ), (x ₂ , y ₂ , z ₂ ), and calculates the average rotation angle parameter. Calculate (S42). The estimation step may use an average rotation angle parameter obtained by averaging a plurality of rotation angle parameters, or may directly use a rotation angle parameter calculated from a set of reference points.

  Here, an example of an arithmetic expression in the case of deriving the rotation angle parameter θx around the device axis x in the first estimation step shown in FIG. 3 will be described.

In the first estimation step, a set of reference points (x ₁ , y ₁ , z ₁ ), (x ₂ , y ₂ , z ₂ ) set in the first reference setting step, that is, platform end 1C Angle parameter θx about the device axis x based on a set of reference points (x ₁ , y ₁ , z ₁ ), (x ₂ , y ₂ , z ₂ ) that coincide with a line on the platform 1A parallel to Is derived.

  A transformation determinant from the device axes x, y, and z to the virtual axes Xx, Yx, and Zx obtained by rotating the device axes x, y, and z around the device axis x with the rotation angle parameter θx is expressed by the following equation. .

Here, in order to match the virtual axis Zx and the virtual axis Yx with the spatial axis Z and the spatial axis Y shown in FIG. 1A, a reference point (with respect to the plane formed by the virtual axis Xx and the virtual axis Zx) x ₁ , y ₁ , z ₁ ), (x ₂ , y ₂ , z ₂ ) satisfy the condition that the line segments are parallel. For this purpose, it is necessary to develop the above equation and establish the following equation that includes the rotation angle parameter θx as a variable.

  Therefore, an arithmetic expression for deriving the rotation angle parameter θx around the device axis x is as follows.

  Next, an example of an arithmetic expression for deriving the rotation angle parameter θy around the device axis y in the second estimation step shown in FIG. 3 will be described.

In the second estimation step, a set of reference points (x ₁ , y ₁ , z ₁ ), (x ₂ , y ₂ , z ₂ ) set in the second reference setting step, that is, platform end 1C Angle parameter θy about the device axis y based on a set of reference points (x ₁ , y ₁ , z ₁ ), (x ₂ , y ₂ , z ₂ ) that coincide with a line on the platform 1A parallel to Is derived.

  A transformation determinant from the device axes x, y, and z to the virtual axes Xy, Yy, and Zy obtained by rotating the device axes x, y, and z around the device axis y with the rotation angle parameter θy is expressed by the following equation. .

Here, in order to match the virtual axis Xy and the virtual axis Zy with the space axis X and the space axis Z shown in FIG. 1B, a reference point (with respect to the plane formed by the virtual axis Yy and the virtual axis Zy) x ₁ , y ₁ , z ₁ ), (x ₂ , y ₂ , z ₂ ) satisfy the condition that the line segments are parallel. For that purpose, it is necessary to develop the above equation and to establish the following equation including the rotation angle parameter θy as a variable.

  Accordingly, an arithmetic expression for deriving the rotation angle parameter θy about the device axis y is as follows.

  Next, an example of an arithmetic expression when the rotation angle parameter θz around the device axis z is derived in the third estimation step shown in FIG. 3 will be described.

In the third estimation step, a set of reference points (x ₁ , y ₁ , z ₁ ), (x ₂ , y ₂ , z ₂ ) set in the third reference setting step, that is, platform end 1C Rotation angle parameter θz around the device axis z based on a set of reference points (x ₁ , y ₁ , z ₁ ), (x ₂ , y ₂ , z ₂ ) that coincide with a line on the platform 1A orthogonal to Is derived.

  A transformation determinant from the device axes x, y, z to the virtual axes Xz, Yz, Zz obtained by rotating the device axes x, y, z around the device axis z with the rotation angle parameter θz is expressed by the following equation. .

Here, in order to match the virtual axis Xz and the virtual axis Yz with the space axis X and the space axis Y shown in FIG. 1C, a reference point ( x ₁ , y ₁ , z ₁ ), (x ₂ , y ₂ , z ₂ ) satisfy the condition that the line segments are parallel. For this purpose, it is necessary to develop the above equation and to establish the following equation including the rotation angle parameter θz as a variable.

  Therefore, an arithmetic expression for deriving the rotation angle parameter θz around the device axis z is as follows.

  As these arithmetic expressions express, the condition that the positions of the reference points on the three virtual axes obtained by rotating the three device axes around the predetermined device axes coincide on one virtual axis. By deriving a satisfied rotation angle parameter around a predetermined device axis, it is possible to accurately estimate the rotation angle at which each device axis is rotating from each space axis.

  When estimating the rotation angle parameter around each device axis, if the rotation angle parameter around any device axis has already been estimated, the position of each reference point is corrected based on the rotation angle parameter. You may keep it. By correcting in this way, the rotation angle parameter around each device axis can be estimated more accurately. For example, the rotation matrix shown by the above formula 2 can be used to correct the position of the reference point based on the average rotation angle parameter θx around the device axis x. Further, the rotation matrix shown by the above equation 5 can be used to correct the position of the reference point based on the average rotation angle parameter θy around the device axis y. Further, for the correction of the position of the reference point based on the average rotation angle parameter θz around the device axis z, the rotation matrix shown in the above equation 8 can be used.

  Finally, the average rotation angle calculated in the previous estimation step is registered in the storage unit 15 as a rotation angle parameter.

  As in the rotation angle estimation process described above, any two reference points on the distance image are extracted and set based on the pixel information, and the three spatial axes with reference to the two reference points are set. By deriving the rotation angle around each device axis that matches the three device axes and registering them as parameters, the rotation angle at which each device axis is rotating from the spatial axis is estimated based on the distance image Can do.

  In the above description, the present invention has been described using an example in which the end of the platform of the station is imaged and the reference point is set. However, the imaging device of the present invention images, for example, an automobile road and performs central separation. Extract distance feature points such as belts and sidewalks, lane markings, pedestrian crossings, braille blocks, etc., and set reference points, or receive light intensity of markers, line tapes, etc. A feature point or a distance feature point due to a step of each part may be extracted to set a reference point.

DESCRIPTION OF SYMBOLS 1 ... Station 1A ... Platform 1B ... Track 1C ... Platform end 2 ... Fall prevention fence 2A ... Sliding door 2B ... Housing 3 ... Post 4 ... TOF camera 5 ... Information processing device 6 ... Imaging device 11 ... Control unit 12 ... Throw Light unit 13: Light receiving unit 13A ... Light receiving element 14 ... Imaging unit 15 ... Storage unit 16 ... Object detection unit 17 ... Reference setting unit 18 ... Estimation unit

Claims (9)

An imaging unit for imaging a distance image obtained by imaging a target area projected on the light receiving surface;
A reference setting unit that sets two reference points on the distance image captured by the imaging unit;
The rotation angle around each device axis is set by the reference setting unit to match the three device axes orthogonal to each other with respect to the light receiving surface as to the three spatial axes orthogonal to each other based on the two reference points. An estimation unit that estimates based on a positional relationship of the two reference points on the distance image;
With
The estimation unit estimates the rotation angle around each device axis, assuming that the two reference points are in a positional relationship parallel to one of the three spatial axes on the distance image. apparatus. The reference setting unit is configured to adjust a device axis different from the device axis for which the rotation angle is to be estimated before the estimation unit is to estimate the rotation angle around each device axis. Set two reference points in a parallel positional relationship.
The imaging device according to claim 1. Based on the positional relationship corrected by the rotation angle around the device axis previously obtained by the estimation unit before the estimation unit tries to estimate the rotation angle around each device axis, the reference setting unit Set the reference point,
The imaging device according to claim 1 or 2. When estimating the rotation angle around each device axis, the estimation unit estimates the rotation angle around each device axis based on the positional relationship corrected by the rotation angle around the device axis obtained previously. ,
The imaging device according to claim 1. The estimation unit repeats derivation of the rotation angle around each device axis until the rotation angle around each device axis converges to a certain range.
The imaging device according to claim 1. The reference setting unit sets the reference point in the distance image based on a distance feature point in the distance image;
The imaging device according to claim 1. The imaging unit captures a received light intensity image obtained by capturing the same target area as the distance image together with the distance image,
The reference setting unit sets the reference point in the distance image based on a received light intensity feature point in the received light intensity image;
The imaging device according to claim 1. An imaging step of imaging a distance image obtained by imaging a target area projected on the light receiving surface of the imaging unit;
A reference setting step for setting two reference points on the distance image captured in the imaging step;
In the reference setting step, the rotation angle around each device axis is set so that the three device axes orthogonal to each other with respect to the light receiving surface match the three spatial axes orthogonal to each other with respect to the two reference points. Estimating the two reference points based on a positional relationship on the distance image, wherein the two reference points are relative to one of the three spatial axes on the distance image. An estimation step for estimating a rotation angle around each device axis, assuming that the positions are parallel to each other;
The rotation angle estimation method in which the information processing apparatus executes A program installed in an information processing apparatus,
An imaging step of imaging a distance image obtained by imaging a target area projected on the light receiving surface of the imaging unit;
A reference setting step for setting two reference points on the distance image captured in the imaging step;
In the reference setting step, the rotation angle around each device axis is set so that the three device axes orthogonal to each other with respect to the light receiving surface match the three spatial axes orthogonal to each other with respect to the two reference points. Estimating the two reference points based on a positional relationship on the distance image, wherein the two reference points are relative to one of the three spatial axes on the distance image. An estimation step for estimating a rotation angle around each device axis, assuming that the positions are parallel to each other;
Is a rotation angle estimation program that causes the information processing apparatus to execute.

Images