JP4836086B2 - 3D shape detector - Google Patents

Description

  The present invention relates to a three-dimensional shape detection apparatus that detects a three-dimensional shape to be detected.

  In general, in order to detect a detection target three-dimensionally, images of the detection target are taken from two or more viewpoints separated by a certain distance, and distance calculation is performed on common feature points such as edge portions of the detection target, for example. The stereo measurement principle to be performed is widely used. However, it is difficult to associate the feature points when the viewpoints for photographing the detection target are separated. Therefore, the conventional apparatus uses a moving stereo method, that is, a method of moving the camera along a predetermined trajectory and associating the feature points of each continuously captured image in a daisy chain. However, feature points can be easily associated (for example, Non-Patent Document 1 and Patent Documents 1 and 2).

  In another conventional apparatus, in addition to the moving stereo method described above, the accuracy of the three-dimensional coordinates is improved by evaluating the accuracy of the three-dimensional coordinates obtained from two still images. That is, the accuracy of the three-dimensional coordinates is improved by evaluating the accuracy of the three-dimensional coordinates using the vertical parallax between successive still images and selecting a set of still images in which the vertical parallax is a predetermined value or less. . In addition, by evaluating the accuracy of the three-dimensional coordinates using the disappearance ratio of the orientation points for obtaining the three-dimensional coordinates, and selecting a set of still images so that the disappearance ratio of the orientation points is less than a predetermined value, The accuracy of three-dimensional coordinates is increased (for example, Patent Document 3).

IEICE Transactions '86 / 11 Vol. J69-D No. 11 P1631-P1638 JP 7-78252 A JP-A-9-35061 JP 2005-332177 A

  By the way, automating the operation of a string-like flexible object (hereinafter referred to as a cable) is an important issue in the field of factory automation (FA). That is, in order to grasp a cable by a robot arm, insert it into a connector, search a route for working while avoiding the cable, and perform an operation for determining the overlapping cable up and down, it is necessary to detect the three-dimensional shape of the cable. However, since the conventional three-dimensional shape detection apparatus as described above detects a three-dimensional shape using feature points of the detection target itself such as an edge portion, for example, there are few feature points in the detection target itself such as a cable. It is difficult to detect a three-dimensional shape.

  The present invention has been made in order to solve the above-described problems. The object of the present invention is to extract feature points more reliably even in a detection target with few feature points such as cables, and to more reliably form a three-dimensional shape. The object is to provide a three-dimensional shape detection device capable of detection.

The three-dimensional shape detection apparatus according to the present invention includes an imaging unit that is displaced from a first position to a second position and that images a detection target from a plurality of positions including the first and second positions, and a predetermined shape for the detection target. Irradiating means for irradiating the pattern light and adding a plurality of bright spots to the detection target, and acquiring a plurality of pieces of image information from the photographing means, and extracting a plurality of bright spots as feature points from the image information. And the feature points of the image information photographed at the first and second positions using the feature points of the image information photographed between the second positions and the three-dimensional detection target based on the associated feature points. Processing means for detecting a shape, the irradiating means includes a light source that generates detection light, and a pattern light generation unit that irradiates the detection target as pattern light composed of a plurality of slit lights. , Shooting means A trajectory candidate generating unit that generates a plurality of trajectory candidates to be displaced, a slit light extending direction detecting unit that detects the extending direction of the slit light, and an intersection angle between the extending direction of the trajectory candidate and the extending direction of the slit light. And a detection operation control unit that determines a combination of the extension direction of the trajectory candidate and the extension direction of the slit light so that the intersection angle approaches a right angle.

  According to the three-dimensional shape detection apparatus of the present invention, the irradiation means adds a plurality of bright spots to the detection target, and the processing means extracts the plurality of bright spots from the image information as feature points. The feature points can be extracted more reliably even with a small number of detection targets, and the three-dimensional shape can be detected more reliably using the moving stereo method.

The best mode for carrying out the present invention will be described below with reference to the drawings.
Embodiment 1 FIG.
FIG. 1 is a block diagram showing a three-dimensional shape detection apparatus according to Embodiment 1 of the present invention. FIG. 2 is a perspective view showing the camera 1 and the pattern light irradiation device 4 of FIG. In the figure, the three-dimensional shape detection apparatus includes a camera 1, a viewpoint moving mechanism 2, a drive controller 3, a pattern light irradiation device 4, an irradiation controller 5, and a processing unit 6. In this embodiment, the photographing means is composed of the camera 1, and the irradiation means is composed of the pattern light irradiation device 4 and the irradiation controller 5.

  As shown in FIG. 1, the camera 1 is for taking an image of a cable 10 that is a detection target. The camera 1 is attached to a viewpoint moving mechanism 2 such as a robot arm or an electric stage. As shown in FIG. 2, the viewpoint moving mechanism 2 displaces the camera 1 from the first position 2a to the second position 2b along a predetermined locus. The camera 1 images the cable 10 from a plurality of positions including the first and second positions 2a and 2b. Returning to FIG. 1, the drive controller 3 controls the operation of the viewpoint movement mechanism 2 based on the drive control command 62 a from the processing means 6.

  As shown in FIG. 2, the pattern light irradiation device 4 includes a laser light generator 40 and a diffraction grating 41. The diffraction grating 41 is rotatably disposed between the laser light generator 40 and the cable 10, and the pattern light composed of a plurality of linear slit lights 42 parallel to the laser light of the laser light generator 40. 43. The pattern light 43 is applied to the cable 10 to add a plurality of bright spots 10 a to the cable 10. Returning to FIG. 1, the irradiation controller 5 controls the start and end of irradiation of the pattern light 43 based on the irradiation control command 62 b from the processing means 6. Further, the irradiation controller 5 controls the extending direction 42a (see FIG. 2) of the slit light 42 by rotating the diffraction grating 41 based on the irradiation control command 62b. In this embodiment, the light source is constituted by the laser light generator 40, and the pattern light generation unit is constituted by the irradiation controller 5 and the diffraction grating 41.

  The processing means 6 is electrically connected to the camera 1, the drive controller 3, and the irradiation controller 5. The processing means 6 controls the displacement of the camera 1 by the viewpoint moving mechanism 2 and the irradiation of the pattern light 43 by the pattern light irradiation device 4 and the irradiation controller 5. The processing unit 6 acquires a plurality of pieces of image information 1a generated by the camera 1, and detects the three-dimensional shape of the cable 10 based on the image information 1a. The processing unit 6 is a computer including a storage unit (RAM and ROM) that stores information such as a program and a processing unit (CPU) that performs arithmetic processing based on the information in the storage unit.

  Specifically, the processing means 6 includes a trajectory candidate generation unit 60, a slit light extending direction detection unit 61, a detection operation control unit 62, an image acquisition unit 63, a feature point extraction unit 64, an association processing unit 65, A tracking stability verification unit 66, a three-dimensional coordinate calculation unit 67, a shape characteristic model storage unit 68, and a three-dimensional shape restoration unit 69 are included.

  For example, external environment information 60b indicating the ambient environment of the camera 1, physical interference, and the like is input to the processing unit 6 from the outside. The trajectory candidate generation unit 60 generates a plurality of trajectory candidates 60a for displacing the camera 1, such as a trajectory for avoiding an obstacle. The trajectory may be two-dimensional or three-dimensional depending on the configuration of the viewpoint moving mechanism 2 such as a robot arm or an electric stage. The slit light extending direction detector 61 detects the extending direction 42a of the slit light 42 by detecting the rotational position of the diffraction grating 41 (see FIG. 2) based on a signal from a sensor (not shown).

  The detection operation control unit 62 is based on the trajectory candidate 60a from the trajectory candidate generation unit 60 and the extending direction information 61a from the slit light extending direction detecting unit 61, and a reliability evaluation index indicating the reliability of the three-dimensional shape detection. Ask for. Specifically, a reliability evaluation index is obtained based on the intersection angle between the extending direction 42a of the slit light 42 and the extending direction 2c of the locus candidate 60a (see FIG. 2). That is, the reliability evaluation index can be represented by the inner product of the vector of the slit light 42 and the vector of the locus candidate 60a. Further, the detection operation control unit 62 determines that the trajectory candidate 60a and the diffraction grating are set so that the intersection angle between the trajectory candidate 60a and the extending directions 2c and 42a of the slit light 42 approaches a right angle based on the obtained reliability evaluation index. The combination with the rotational position 41 is determined. This is because the detection sensitivity of the bright spot 10a is higher as the crossing angle is closer to a right angle. Furthermore, the detection operation control unit 62 sends a drive control command 62a and an irradiation control command 62b to the drive controller 3 and the irradiation controller 5 based on the selection result of the locus candidate 60a and the rotation position of the diffraction grating 41, respectively. input.

  The camera 1 takes an image of the cable 10 irradiated with the pattern light 43 while being displaced along a predetermined locus (see FIG. 2). The image acquisition unit 63 acquires a plurality of pieces of image information 1 a from the camera 1. Note that a digital video camera or the like is used as the camera 1, and a capture or a dedicated port is used as the image acquisition unit 63 depending on the camera interface type such as NTSC, USB, or IEEE1394.

  The feature point extraction unit 64 performs feature extraction processing on the plurality of pieces of image information 1a acquired by the image acquisition unit 63, and extracts feature points. Here, the feature points extracted by the feature point extraction unit 64 include, in addition to the bright spots 10a added by the pattern light 43, interpolation points for interpolating between the extracted bright spots 10a, and a cable 10 such as printing. And its own features.

  Characteristic portions of the cable 10 itself such as printing can be extracted by using a corner detection filter such as a SUSAN operator and a Harris operator, or an edge detection filter such as Sobel and Canny.

  The bright spot 10a can be extracted from a difference image between a cable image in a state where the pattern light 43 is not irradiated and a current image in a state where the pattern light 43 is irradiated. In the case of a relatively simple background, the background and the cable can be separated by using, for example, a frequency band filter. At this time, the mid-range component intensity image from which the high-frequency component (fine noise component) and the low-frequency component (gradual brightness change component such as gradation due to illumination variation) are removed is an image obtained by removing the cable 10.

  In addition, the interpolation point can be created by extracting a point having a high entropy Exy of the direction code (OC: Orientation Code) described below, other than the characteristic portion of the cable 10 itself or the region already extracted as the bright point 10a. It becomes. First, the direction code (OC) is a value obtained by quantizing the maximum gradient direction of brightness change of each pixel with an integer value, and is robust to the influence of contrast fluctuation, brightness change, highlight and shadow. Further, the direction code has a characteristic that it is difficult to change even when the luminance value on the image changes between consecutive frames due to the influence of a change in illumination conditions or the like. The direction code Cxy in the pixel (x, y) is defined by the following expression 1.

  Here, θxy represents the lightness gradient direction, Δθ represents the quantization number, and | ΔIx | and | ΔIy | represent differential values in the x direction and the y direction, respectively. Γ is a threshold value that suppresses the influence of the noise level. When the frequency of appearance of each direction code in the local region of the region (size Me × Me) centered on the pixel (x, y) is hxy (C) (C = 0, 1,..., N), the direction The code entropy Exy is defined by the following equation 2. Pxy (C) (C = 0, 1,..., N) represents a relative frequency related to the direction code excluding the defective code number hxy (N). When the interpolation point using the direction code entropy Exy is printed on the cable, the printed portion is preferentially extracted as the feature point.

  This direction code entropy Exy is used to determine the reliability of the extracted feature points in addition to the creation of interpolation points. That is, the feature point extraction unit 64 preferentially validates a feature point having a high direction code entropy Exy from a plurality of extracted feature points, and invalidates a feature point having a low direction code entropy Exy. In other words, the feature point extraction unit 64 validates feature points whose rank of direction code entropy Exy is higher than a predetermined rank, and invalidates feature points having a low rank. This feature selection has robustness against changes in density gradient values and the effects of illumination fluctuations, and has a feature that is less dependent on edges in one direction.

The association processing unit 65 performs feature point tracking using the feature points of the image information 1a photographed between the first and second positions 2a and 2b, and image information photographed at the first and second positions 2a and 2b. The feature points 1a are associated with each other. Specifically, the association processing unit 65 associates feature points by directional code matching (OCM) using the above-described difference in directional code values. That is, for the j-th feature point in the i-th image, the direction code Cj, i of the template for correspondence search (size Mo × Mo) and the direction code Ci + 1 in the next viewpoint image are used for each pixel. The point at which the average residual absolute value D obtained by dividing the sum of the residual absolute values d (Cj, i, Ci + 1) obtained by the above is divided by the number of pixels Mo ² is the corresponding point.

  The tracking stability verification unit 66 obtains a stability index indicating the stability of the feature point tracking by the association processing unit 65, that is, the stability of the association of the feature points. Here, it is assumed that the feature points 66a to 66c corresponding to each other are displaced as shown in FIG. It is well known that epipolar restraint works between corresponding points between each photographing position. That is, if the extraction accuracy of the feature points 66 a to 66 c is high, the feature points 66 a to 66 c are arranged on the epipolar line 66 d corresponding to the trajectory of the camera 1. Therefore, the stability of the feature point tracking can be evaluated by evaluating the deviation angles θ1 and θ2 from the epipolar line 66d. Further, for example, when the distances D1 and D2 (see FIG. 2) between the photographing positions of the cable 10 are equally spaced, the distances X1 and X2 (see FIG. 3) between the feature points 66a to 66c on the image plane are also used. Evenly spaced. That is, since there is a correlation between the distance relationship between the shooting positions of the cable 10 and the distance relationship assumed between the feature points 66a to 66c, the distances D1 and D2 between the shooting positions of the cable 10 and the feature points 66a to 66c. The stability of the feature point tracking can be evaluated by evaluating the deviation of the relationship between the distances X1 and X2. In other words, the tracking stability verification unit 66 has at least one of deviation angles θ1 and θ2 from the epipolar line 66d and the deviation between the distance relationship between the photographing positions of the cable 10 and the distance relationship assumed between the feature points 66a to 66c. Is used to determine the stability index, and the stability index is determined based on a criterion based on a predetermined value, and the corresponding feature points 66a to 66c are invalidated.

  Returning to FIG. 1, the three-dimensional coordinate calculation unit 67 obtains the three-dimensional coordinates of the feature points based on the stereo measurement principle. The shape characteristic model storage unit 68 stores a shape characteristic model 68a of the cable 10 in advance. The shape characteristic model 68a is information input from the outside, and indicates that the approximate three-dimensional shape of the cable 10 is a string. The three-dimensional shape restoration unit 69 restores the three-dimensional shape of the cable 10 from the three-dimensional coordinates calculated by the three-dimensional coordinate calculation unit 67 by spline interpolation or the like. Here, as described above, the feature point is selected by the tracking stability verification unit 66 or the like, but the feature point whose position is shifted may remain. When the three-dimensional shape is restored in a state that includes feature points whose positions are shifted, the restored three-dimensional shape becomes distorted. The three-dimensional shape restoration unit 69 corrects the restored three-dimensional shape based on the shape characteristic model 68a, and thereby eliminates a portion where excessive deformation occurs as compared with the shape characteristic model 68a as an error portion. The three-dimensional shape restoration unit 69 outputs the corrected three-dimensional shape to the outside as the three-dimensional shape data 69a.

  Next, the operation of the three-dimensional shape detection apparatus will be described. FIG. 4 is a flowchart showing a three-dimensional shape detection operation by the three-dimensional shape detection apparatus of FIG. In the figure, when ambient environment information 60b is input to the processing means 6 from the outside, a plurality of trajectory candidates 60a are generated by the trajectory candidate generation unit 60 (step S1), and the rotational position of the diffraction grating 41 is set to the slit light beam. The direction detection unit 61 detects the position (step S2), and the detection operation control unit 62 determines the locus of the camera 1 and the rotational position of the diffraction grating 41 (step S3).

  Next, an irradiation control command 62b is input from the detection operation control unit 62 to the irradiation controller 5, the diffraction grating 41 is rotated by a predetermined angle, and irradiation of the pattern light 43 is started (step S4). . When the irradiation of the pattern light 43 is started, a drive control command 62a is input from the detection operation control unit 62 to the drive controller 3, and the camera 1 is moved to the initial position, that is, the first position 2a (step S5).

  When the camera 1 is moved to the first position 2a, image information 1a photographed at the first position 2a is acquired by the image acquisition unit 63 (step S6), and the bright spot 10a, the characteristic portion of the cable 10 itself, and interpolation are performed. A point is extracted from the image information 1a by the feature point extraction unit 64 as a feature point (step S7). When feature points are extracted, feature point association is performed by the association processing unit 65 (step S8), and tracking stability evaluation is performed by the tracking stability verification unit 66 (step S9). Note that when the camera 1 is located at the first position 2a, there is only one image information 1a, so that the feature point association and the tracking stability evaluation are not substantially performed.

  Next, it is determined by the detection operation control unit 62 whether or not the cable 10 has been photographed by the camera 1 at all photographing positions (step S10). At this time, if it is determined that there is a position at which shooting is not performed, a drive control command 62a is input again from the detection operation control unit 62 to the drive controller 3, and the camera 1 is moved to the next shooting position. (Step S11). Thereafter, until it is determined that shooting at all shooting positions has been completed, acquisition of image information 1a, feature point extraction, feature point association, tracking stability evaluation, shooting end determination, and camera 1 movement ( Steps S6 to 11) are repeated. At this time, the feature point selection using the direction code entropy Exy is performed by the feature point extraction unit 64, so that a feature point suitable for tracking is preferentially selected. Further, the feature point is correlated by the direction code collation by the association processing unit 65, so that the feature point is accurately tracked. Furthermore, the tracking stability evaluation based on the stability index is performed by the tracking stability verification unit 66, so that a feature point suitable for three-dimensional coordinate calculation is selected.

  Next, when it is determined that the photographing at all photographing positions is completed, the irradiation of the pattern light 43 is finished (step S12), and the three-dimensional coordinate calculation of the feature point based on the stereo measurement principle is performed. 67 (step S13), a three-dimensional shape restoration process based on the calculated three-dimensional coordinates is performed by the three-dimensional shape restoration unit 69 (step S14). When the three-dimensional shape restoration processing is completed, the shape characteristic model 68a of the cable 10 stored in the shape characteristic model storage unit 68 is acquired by the three-dimensional shape restoration unit 69, and the three-dimensional shape based on the shape characteristic model 68a is acquired. Correction processing is performed (step S15).

  In such a three-dimensional shape detection apparatus, the pattern light irradiation device 4 adds a plurality of bright spots 10a to the cable 10, and the processing means 6 extracts the plurality of bright spots 10a as feature points from the image information 1a. The feature points can be more reliably extracted even with the cable 10 having a small number of feature points, and the three-dimensional shape can be detected more reliably using the moving stereo method.

  In addition, the feature point extraction unit 64 preferentially validates the feature point having a high direction code entropy from among the plurality of extracted feature points, and invalidates the feature point having the low direction code entropy. The selected feature points can be preferentially selected, and the detection accuracy of the three-dimensional shape can be improved.

  Furthermore, since the association processing unit 65 associates feature points by direction code collation, the feature points can be tracked with high accuracy, and the detection accuracy of the three-dimensional shape can be improved.

  Furthermore, the tracking stability verification unit 66 includes deviation angles θ1 and θ2 from the epipolar line 66d and the deviation between the distance relationship between the photographing positions of the cable 10 and the distance relationship assumed between the feature points 66a to 66c. A stability index is obtained based on at least one, the stability index is determined based on a determination criterion based on a predetermined value, and the corresponding feature points 66a to 66c are invalidated. As a result, feature points suitable for three-dimensional coordinate calculation can be selected more reliably, and the detection accuracy of the three-dimensional shape can be improved.

  In addition, since the three-dimensional shape restoration unit 69 corrects the restored three-dimensional shape based on the shape characteristic model 68a, even if a feature point having a misalignment remains at the time of three-dimensional shape restoration, the feature point The deformation of the three-dimensional shape can be eliminated, and the detection accuracy of the three-dimensional shape can be improved.

  Further, the detection operation control unit 62 obtains an intersection angle between the extension direction of the slit light 42 and the extension direction of the locus candidate 60a, and the locus candidate 60a and the diffraction grating 41 so that the intersection angle approaches a right angle. The combination with the rotation position is determined. Thereby, the extraction accuracy of the bright spot 10a can be increased, and the detection accuracy of the three-dimensional shape can be improved.

  In the first embodiment, the pattern light 43 applied to the detection target has been described as having a striped shape composed of a plurality of linear slit lights 42 parallel to each other. However, for example, a plurality of concentric circles, meshes, etc. The shape of the pattern light may be any shape spreading in two dimensions.

Embodiment 2. FIG.
FIG. 5 is a block diagram showing a three-dimensional shape detection apparatus according to Embodiment 2 of the present invention. FIG. 6 is an explanatory diagram showing the pattern light 43 generated by the laser light generator 200 and the irradiation controller 210 of FIG. In the first embodiment, it has been described that the irradiation means includes the pattern light irradiation device 4 (laser light generator 40 and diffraction grating 41) and the irradiation controller 5 (see FIG. 2). The second irradiation means includes a laser light generator 200 and an irradiation controller 210. In this embodiment, the light source is configured by the laser light generator 200, and the pattern light generation unit is configured by the irradiation controller 210.

  The laser light generator 200 generates slit light 42 extending linearly as shown in FIG. The irradiation controller 210 scans the slit light 42 in a direction 42 b orthogonal to the extending direction 42 a of the slit light 42. The irradiation controller 210 turns the slit light 42 on and off when scanning the slit light 42. In FIG. 6, a circle shown at one end of the slit light 42 indicates ON / OFF of the slit light 42. That is, the second embodiment is configured such that the pattern light 43 is generated by scanning the slit light 42 while being turned on and off. In other words, the interval 42c of the slit light 42 can be arbitrarily controlled by controlling ON / OFF of the slit light 42.

  Here, the number of bright spots 10a (see FIG. 2) can be increased as the interval 42c of the slit light 42 is reduced, and the detection accuracy of the three-dimensional shape can be increased. However, the resolution of the image information 1a generated by the camera 1 is limited. Therefore, if the camera 1 and the cable 10 are far apart, the interval 42c of the slit light 42 may be crushed on the image information 1a, and the bright spot 10a may not be extracted.

  The distance information 221 from the outside indicating the separation distance between the camera 1 and the cable 10 is input to the processing means 220 of the second embodiment. The detection operation control unit 222 inputs an irradiation control command 222b to the irradiation controller 210 based on the input distance information 221, and controls the interval 42c of the slit light 42. Specifically, if the distance between the camera 1 and the cable 10 is large, the interval 42c of the slit light 42 is widened, and if the distance is small, the interval 42c of the slit light 42 is narrowed. The slit light 61 detects the extending direction 42 a of the slit light 42 by detecting the rotational position of the laser light generator 200. In addition, the detection operation control unit 320 generates a laser beam so that the extension direction 42a of the slit light 42 and the extension direction of the locus candidate 60a are close to orthogonal by inputting the irradiation control command 62b to the irradiation controller 210. The container 200 is rotated. Other configurations are the same as those in the first embodiment.

  According to such a three-dimensional shape detection apparatus, the detection operation control unit 222 controls the interval 42c of the slit light 42 based on the distance information 221, so that the interval 42c of the slit light 42 can be set optimally. The detection accuracy of the three-dimensional shape can be further improved.

Embodiment 3 FIG.
FIG. 7 is a block diagram showing a three-dimensional shape detection apparatus according to Embodiment 3 of the present invention. When the intervals 42c (see FIG. 6) of the slit light 42 are equal, the bright spots 10a (see FIG. 2) are added to the cable 10 at equal intervals, so that the feature points between the image information 1a photographed at different positions In some cases, the accuracy of the association becomes low. That is, a local pattern is formed in the pattern light 42 by changing the interval between the bright spots 10a and the size of the bright spots 10a. By using this local pattern, the accuracy of associating feature points between the image information 1a can be improved. After associating feature points, the association processing unit 310 of the processing unit 300 inputs correspondence result information 310 a indicating the association result to the detection operation control unit 320. Specifically, the correspondence result information 310a indicates the distribution state of the effective feature points.

  The detection operation control unit 320 controls at least one of the interval 42c and the width of the slit light 42 by inputting an irradiation control command 320b to the irradiation controller 210 based on the correspondence result information 310a. That is, the above-mentioned local pattern is formed in a region where the feature point association accuracy is low. In FIG. 6, the interval 42d for turning the slit light 42 ON and OFF is shown apart, but the width of the slit light 42 can be controlled by making this interval 42d adjacent. Other configurations are the same as those of the second embodiment.

  In such a three-dimensional shape detection apparatus, the detection operation control unit 320 controls at least one of the interval 42c and the width of the slit light 42 based on the correspondence result information 310a, so that the feature point association accuracy is low. A local pattern can be formed, and the detection accuracy of the three-dimensional shape can be further improved.

It is a block diagram which shows the three-dimensional shape detection apparatus by Embodiment 1 of this invention. It is a perspective view which shows the camera and pattern light irradiation apparatus of FIG. It is explanatory drawing which shows the displacement of the feature point extracted by the feature point extraction part of FIG. It is a flowchart which shows the three-dimensional shape detection operation by the three-dimensional shape detection apparatus of FIG. It is a block diagram which shows the three-dimensional shape detection apparatus by Embodiment 2 of this invention. It is explanatory drawing which shows the pattern light produced | generated by the laser beam generator and irradiation controller of FIG. It is a block diagram which shows the three-dimensional shape detection apparatus by Embodiment 3 of this invention.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 Camera (imaging means), 1a Image information, 2a 1st position, 2b 2nd position, 3 Drive controller, 4 Pattern light irradiation apparatus (irradiation means), 5,210 Irradiation controller (pattern light generation part), 6 , 220, 300 Processing means, 10 cable (detection target), 10a bright spot, 40, 200 laser light generator (light source), 41 diffraction grating (pattern light generation unit), 42 slit light, 43 pattern light, 60 locus candidate Generation unit, 60a Trajectory candidate, 61 Slit light extension direction detection unit, 62, 222, 320 Detection operation control unit, 64 Feature point extraction unit, 65, 310 Association processing unit, 66 Tracking stability verification unit, 67 Three-dimensional coordinate calculation Unit, 68 shape characteristic model storage unit, 68a shape characteristic model, 69 three-dimensional shape restoration unit.

Claims (7)

An imaging unit that is displaced from the first position to the second position and that images the detection target from a plurality of positions including the first and second positions;
Irradiating means for irradiating a pattern light of a predetermined shape to the detection target to add a plurality of bright spots to the detection target;
A plurality of image information is acquired from the photographing means, the plurality of bright spots are extracted as feature points from the image information, and the feature points of the image information photographed between the first and second positions are used. Processing means for associating feature points of the image information photographed at the first and second positions with each other and detecting a three-dimensional shape of the detection target based on the associated feature points ,
The irradiation means includes
A light source that generates detection light;
A pattern light generation unit that irradiates the detection target with the detection light as the pattern light including a plurality of slit lights;
Including
The processing means includes
A trajectory candidate generator for generating a plurality of trajectory candidates for displacing the photographing means;
A slit light extending direction detector for detecting the extending direction of the slit light;
The intersection angle between the extension direction of the locus candidate and the extension direction of the slit light is obtained, and a combination of the extension direction of the locus candidate and the extension direction of the slit light so that the intersection angle approaches a right angle. Detecting operation control unit to determine
3-dimensional shape detection device which comprises a. The processing means includes
The feature point extracting unit that extracts the feature point from the image information and preferentially validates the feature point having a high direction code entropy among the plurality of extracted feature points. The three-dimensional shape detection apparatus according to 1. The processing means includes
The three-dimensional shape detection apparatus according to claim 1, further comprising an association processing unit that performs the association of the feature points by direction code matching. The processing means includes
Associating the feature points using at least one of a deviation angle from the epipolar line of the feature points and a deviation between a distance relationship between the imaging positions of the detection target and a corresponding distance relationship between the corresponding feature points A tracking stability verification unit that obtains a stability index indicating the stability of the image, determines the stability index based on a determination criterion based on a predetermined value, and invalidates the corresponding feature point. The three-dimensional shape detection apparatus according to any one of claims 1 to 3. The processing means includes
A three-dimensional coordinate calculation unit for obtaining three-dimensional coordinates of the feature points based on a stereo measurement principle;
A shape characteristic model storage unit that stores in advance the shape characteristic model of the detection target;
A three-dimensional shape restoring unit that restores the three-dimensional shape of the detection target based on the three-dimensional coordinates and corrects the three-dimensional shape based on the shape characteristic model. The three-dimensional shape detection apparatus according to claim 1. The irradiation means includes
A light source that generates detection light;
A pattern light generation unit that irradiates the detection target as the pattern light composed of a plurality of slit lights with the detection light,
The processing means includes
A detection operation control unit configured to input an irradiation control command to the pattern light generation unit based on distance information from the outside indicating a separation distance between the imaging unit and the detection target, and to control an interval of the slit light. The three-dimensional shape detection apparatus according to any one of claims 1 to 5, wherein The irradiation means includes
A light source that generates detection light;
A pattern light generation unit that irradiates the detection target as the pattern light composed of a plurality of slit lights with the detection light,
The processing means includes
2. A detection operation control unit configured to input an irradiation control command to the pattern light generation unit based on a result of the association of the feature points, and to control at least one of an interval and a width of the slit light. The three-dimensional shape detection apparatus according to any one of claims 1 to 6 .

Images