JP5515177B2 - Shape recognition device - Google Patents

Description

  The present invention relates to a shape recognition device for recognizing an outer surface shape of an object from three-dimensional data representing the outer surface shape of the object.

  There exists a thing of patent document 1 as a shape recognition apparatus comprised as mentioned above. This shape recognition device includes a three-dimensional measurement control unit that generates three-dimensional measurement data of the entire surface shape of a pressed product by performing fringe analysis on a photographed image of the surface of the pressed product such as a vehicle body or a door panel. A strain evaluation device is also provided.

  The shape recognition apparatus includes an approximate curve applying unit, a curvature deriving unit, a uniform range determining unit, and an approximate curve deriving unit. The approximate curve applying means is a part of a circle corresponding to the curve of a plurality of first data groups along the length direction of the cross section among the two-dimensional data generated from the three-dimensional measurement data in the length direction of the cross section. A plurality of circles are made to correspond to the curve by associating the sections. The curvature deriving unit obtains the curvature of each circle and generates a first approximate curve. Next, the uniform range determining means determines a uniform range from the first approximate curve, and the approximate curve deriving means derives a second approximate curve having a certain curvature. This curvature is the reciprocal of the radius of the circle.

  Next, the distortion evaluation device extracts the distortion data extraction means using the data whose deviation from the second approximate curve is equal to or larger than the set allowable range as distortion data. In other words, in Patent Document 1, the shape of each part of the curve is regarded as a set of arcs of a circle having a predetermined radius, a second approximate curve is generated based on the set of arcs of the circle, and the curvature is based on the second approximate curve. It is determined that there is distortion in a region where the deviation is from.

  Considering the surface strain detection processing of press-processed products such as vehicle bodies and door panels, for example, comparison with 3D data such as CAD data (for example, CAD data of molds) indicating the ideal surface shape of the processed product It is possible to do. However, in a press-processed product, the shape is usually deformed in a direction in which the shape slightly swells immediately after processing due to the spring back, and detection of surface distortion by comparing with CAD data is relative to CAD data. It may be difficult in practical use, including the accuracy of positioning.

  Therefore, as described in Patent Document 1, the curvature of each part of the surface of the measurement target is obtained from the two-dimensional cross-sectional data when the three-dimensional data created by photographing the surface of the measurement target is cut in a predetermined direction. It becomes effective to determine that there is distortion in a region where the curvature deviates from the comparison between the curvature of each part and the curvature of each part.

  In other words, even if the car body or door panel is distorted so as to be curved as a whole compared to the designed curve, there is almost no visual sense of distortion, and there are some depressions and protrusions. In some cases, even if the amount of depressions or protrusions is small, it often feels visually distorted. With respect to such a problem, in the apparatus described in Patent Document 1, the uniform curvature corresponds to the shape of the base to be measured, and the curvature of each part is compared based on the shape of the base. As a result, a good aspect has been revealed that local distortion can be detected.

  However, as described in Patent Document 1, when sample points are set at short intervals in the length direction of two-dimensional cross-sectional data and a circle is approximated for each sample point, a large number of sample points are linear. In the case of a line-up shape, the radius of the circle becomes a very large value, and the processing load tends to increase.

  In addition, when the shape curve drawn by the sample point is convex upward or downward, the center coordinates of the circle are greatly different. For example, in the vicinity of the inflection point of the shape curve, the shape curve is When the center coordinates of the circle that approximates one part of the shape curve are set based on the inflection point, and the center coordinates of the circle that approximates the other part of the curve based on the inflection point Since the position of the switch is greatly changed, a large numerical value has to be handled inevitably, which tends to be a burden of processing, and there is room for improvement.

  When there is a portion whose shape changes with a large curvature in a relatively narrow region on the surface of the measurement object, it is often impossible to obtain an accurate curvature due to the positional relationship between the sample points. This can be explained because the longer the interval between the sample points, the more likely it is to introduce an error. In order to eliminate this inconvenience, the interval between the sample points may be shortened. However, if the interval between the sample points is shortened, the amount of data increases, and the processing time for approximating the outline curve becomes longer, which is not realistic.

  Accordingly, it is conceivable to obtain the curvature of the three-dimensional data by generating the two-dimensional data excluding the influence of the large curvature after excluding the region including the curvature exceeding the predetermined value in the outer surface shape represented by the three-dimensional data.

  For example, when there is a part whose shape changes with a large curvature in a relatively narrow area on the surface of the measurement object, such as a seating surface provided on a door panel, an accurate curvature can be obtained depending on the positional relationship of the sample points. There are many things that cannot be requested. For this reason, as in the above-described prior art, for a region having a curvature exceeding a predetermined value, the curvature corresponding to the original shape is acquired by excluding data.

  However, in the setting of simply excluding a curvature greater than a predetermined value, it is not possible to appropriately exclude only the area that is actually desired to be excluded. This tends to be calculated as a relatively gentle curvature around the part where the shape changes with a large curvature. For this reason, it is conceivable to set the predetermined value (threshold value) to a low value. However, if the threshold value is set to a low value, parts that should not be excluded may be excluded. Therefore, the threshold value has to be set high, and all the regions to be excluded cannot be appropriately excluded.

JP 2007-127610 A

  In view of the above problems, an object of the present invention is to provide a shape recognition device that can easily perform processing for approximating a shape curve represented by cross-sectional data of a two-dimensional structure.

  The feature configuration of the shape recognition apparatus according to the present invention for achieving the above object is to recognize the outer surface shape of the object from the three-dimensional data representing the outer surface shape of the object. Curvature exclusion region extraction means for extracting a region where the curvature exceeds a preset first predetermined value, and the amount of change in curvature of the outer surface shape used for the extraction of the region is calculated, and the amount of change is preset A change amount exclusion region extraction unit for extracting a region exceeding the second predetermined value, and both the region extracted by the curvature exclusion region extraction unit and the region extracted by the change amount exclusion region extraction unit are set as exclusion regions. Exclusion region setting means for performing the processing, detailed shape data reflecting in detail the shape of the predetermined setting region in the three-dimensional data, and the shape of the predetermined setting region being rougher than the detailed shape data And generating the respective corrected shape data by excluding the data included in the exclusion region from the respective shape data, and based on the curvature of the respective corrected shape data, And a shape evaluation means for extracting a strain value.

  With such a characteristic configuration, the curvature exclusion region extraction unit sets a region having a curvature exceeding the first predetermined value as the curvature exclusion region, and the change amount exclusion region extraction unit in the adjacent region adjacent to the curvature exclusion region Can be extracted as a change amount exclusion region including a change amount of curvature exceeding the second predetermined value. As a result, it is possible to easily distinguish a region where the curvature greatly changes from a region where the curvature does not change greatly by the curvature exclusion region extraction unit, and the amount of change in curvature changes suddenly in the adjacent region adjacent to the curvature exclusion region. It is possible to accurately classify the sudden change portion to be performed by the change amount exclusion region extraction means. Since the exclusion area setting means sets the logical sum of the two areas thus divided as the exclusion area, the amount of change in the curvature changes abruptly in an area adjacent to the area where the curvature changes greatly and where the curvature does not change significantly. It is possible to set the exclusion region in consideration of the sudden change portion. For this reason, even if it is the outer surface shape of the object which has a design which has a shape whose curvature changes suddenly, it is possible to appropriately perform the distortion recognition processing of the outer surface shape of the object after excluding the design. Therefore, it is possible to improve the distortion recognition performance.

It is a perspective view which shows the whole three-dimensional measuring system. It is a block circuit diagram of a three-dimensional measurement system. It is a block circuit diagram of a shape recognition device It is a flowchart of an evaluation process routine. It is a perspective view which shows a measuring object and a coordinate system. It is a side view of a measuring object. It is a perspective view which shows the conversion form in a data conversion process part. It is a figure which shows typically about the setting of an exclusion area | region. It is a figure which shows the size of two types of area | regions when reflecting the height value of the point group of the cut-out three-dimensional data. It is a figure which shows the surface shape of the three-dimensional data acquired by cutting out. It is a figure which shows typically the process at the time of setting an exclusion area | region. It is a figure which shows the process at the time of acquiring a distortion value from the difference in curvature.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
〔System configuration〕
As shown in FIG. 1 and FIG. 2, an image is taken with an articulated robot hand H having a photographing unit V having a digital camera at the tip, a robot hand controller A that controls the robot hand H, and a photographing unit V. A three-dimensional measurement system including an image acquisition unit B that acquires image data of a measurement target T, and a shape recognition device C that acquires image data from the image acquisition unit B, extracts the distortion of the outer shape of the object, and outputs the distortion. Is configured. The shape recognition apparatus C having such a configuration has a function of recognizing the outer surface shape of the object from the three-dimensional data representing the outer surface shape of the object.

  The photographing unit V includes a pattern projection device 1 that projects a lattice pattern onto the surface of the measurement target T via the projection lens 1L, and a digital camera that photographs the surface of the measurement target T via the photographing lens 2L. 2 are provided. The robot hand H includes a plurality of arm units 3 and an electric motor 4 that drives each arm unit 3 independently. The robot hand controller A executes a preset program so that the imaging unit V The operation of photographing the measurement target T from different directions is performed, and the entire surface of the measurement target T is photographed. The image acquisition unit B determines the shape of the measurement target T from the lattice pattern of the image data photographed by the photographing unit V and generates a three-dimensional image.

  The robot hand controller A includes arm control means 6 for controlling the arm unit 3 in accordance with the control program 5. In the image acquisition unit B, the image data acquisition means 7 acquires image data from the camera 2, and the three-dimensional image generation means 8 generates a three-dimensional image from this image data. As a technique for generating a three-dimensional image from image data, a known one is used as disclosed in Japanese Patent Application Laid-Open No. 2002-257528, Japanese Patent Application Laid-Open No. 2004-317495, and the like. A three-dimensional image that is a point group in which the surface shape of the object T is represented in a three-dimensional coordinate system is generated.

  The shape recognition device C is constituted by a general-purpose computer to which a display d, a keyboard k, a mouse m, and the like are connected. The shape recognition device C extracts the amount of distortion on the surface of the measurement target T from the three-dimensional image acquired from the image acquisition unit B and performs evaluation. The evaluation result is output to the display d.

[Control configuration]
As shown in FIGS. 2 and 3, the shape recognition device C includes a data acquisition unit 11, a data conversion unit 12, a noise removal unit 13, a curvature exclusion region extraction unit 14, and a change amount exclusion region extraction unit 15. , Exclusion region setting means 16, base shape data generation means 17, detailed shape data generation means 18, curvature acquisition means 19, shape evaluation means 20, evaluation image generation means 21, and measurement target shape acquisition means 22 It has.

  These data acquisition means 11, data conversion means 12, noise removal means 13, curvature exclusion area extraction means 14, change amount exclusion area extraction means 15, exclusion area setting means 16, base shape data generation means 17, detailed shape data generation means 18, the curvature acquisition means 19, the shape evaluation means 20, the evaluation image generation means 21, and the measurement target shape acquisition means 22 are configured by software developed on a storage of a general-purpose computer that functions as the shape recognition device C. However, these may be configured by a combination of software and hardware such as logic, or may be configured only by hardware such as logic.

[Control form]
FIG. 4 shows a flowchart of control based on such a control configuration, and the function of the control configuration will be described together with the control flow shown in the flowchart.

  The data acquisition unit 11 acquires the 3D image generated by the 3D image generation unit 8 of the image acquisition unit B via the communication cable (Step # 01). In the shape recognition apparatus C, not only the 3D image generated from the image captured by the imaging unit V is processed, but 3D data having a data structure such as an STL file is processed. It is also possible. That is, the base shape and detailed shape data can be calculated directly from the point group corresponding to the outer surface to be measured by the base shape data generating means and the detailed shape data.

  5 and 6 show a door panel as a measurement object T (an example of an object), the front-rear direction in a posture in which the door panel is provided on the vehicle body is the X-axis direction, the vertical direction is the Y-axis direction, The width direction is shown as Z. A seat surface Ta on which a door handle (not shown) is mounted and an opening Tb are formed in the panel portion of the door panel.

  The data conversion means 12 sets a point group (a collection of points indicating the Z coordinate value / height value) indicating the surface shape of the measurement target T from the acquired three-dimensional image on a lattice point on the XY plane with a set pitch. Three-dimensional data is generated by converting to a point cloud (step # 02). That is, in the three-dimensional coordinate system of the three-dimensional image acquired by the data acquisition unit 11, the front-rear direction and the vertical direction of the door panel do not necessarily coincide with the XY axis direction, and the height value indicating the outer shape (Z direction) The pitch of the point cloud indicating the value at (1) cannot be said to be an appropriate value for processing by the shape recognition apparatus C.

  For this reason, as shown in FIG. 7, the data conversion means 12 sets the front / rear direction of the door panel as the measurement target T to the X-axis direction, sets the vertical direction to the Y-axis direction, and sets the width direction (door Is set in the Z-axis direction, and a three-dimensional data value is generated from the three-dimensional image data value. Note that the point group in the three-dimensional data generated by the data conversion unit 12 is a virtual point existing on the surface of the measurement target T, and indicates a point whose position is specified by three-dimensional coordinates.

  The three-dimensional data generated in this way has a data structure in which the height value is expressed as a Z coordinate with reference to an XY plane with a set pitch of less than several millimeters. When performing this conversion, it is necessary to obtain a height value in an area where the point cloud (shown as a three-dimensional image value in the figure) obtained by the data obtaining unit 11 does not exist. As in the bilinear method and the nearest neighbor method, height values reflecting the height values of adjacent point groups are newly created by interpolation processing. By performing these processes, three-dimensional data having a simple data structure whose height value is indicated in the Z direction with reference to the XY plane is generated, and as a result, the data amount is reduced. Further, the data conversion means 12 performs processing for removing this data including data for generating a polygon, such as an STL file.

  The noise removing unit 13 removes noise included in the three-dimensional data (Step # 03). In this removal, a process of taking a moving average of a plurality of coordinates (height values) indicating the surface shape for each predetermined region is performed. By this process, when the height value at the target coordinate is a value far from the surrounding height value, it is converted to a value reflecting the surrounding height value, and as a result, noise is removed. Note that, as shown in Patent Document 1, the noise removal processing form by the noise removing unit 13 is one of a large number of point groups indicating a surface shape as a specific point, and the coordinate value (height value) of the specific point. ) And the coordinate value (height value) of a point adjacent thereto, and a smoothing process may be performed based on the difference in height value. .

  The curvature exclusion area extraction unit 14 includes a curvature acquisition module and an area setting module, and performs a curvature exclusion area setting process (step # 04). The curvature acquisition module acquires the three-dimensional data from which noise has been removed and supplies it to the curvature acquisition means 19, whereby the measurement object calculated from the point group of the two-dimensional data generated from the three-dimensional data by the curvature acquisition means 19 A curvature indicating the shape of T is acquired. An example of the curvature obtained in this way is shown in FIG. The region setting module extracts a region where the curvature of the outer surface shape represented by the three-dimensional data exceeds a preset first predetermined value (the radius of the curve is small), and this region and a region surrounded by this region are X- A curvature exclusion region whose position is represented in the Y coordinate system is set (the processing form of the curvature acquisition means 18 will be described later). The curvature exclusion region set in this way is shown in FIG.

  In this process, the process of calculating the curvature of the outer shape of the measurement target T by the curvature acquisition means 19 is essential, and it is desirable to calculate the curvature for each narrow region as much as possible.

  In this embodiment, the curvature exclusion region is set based on the curvature, but not only the curvature, but if there is a part protruding beyond the set amount from the average surface, or a region recessed beyond the set amount, The processing mode of the region setting module may be set so that a portion including these or a portion surrounded by these is a curvature exclusion region.

  The change amount exclusion area extraction unit 15 includes a curvature acquisition module, a feature extraction module, and an area setting module, and performs change amount exclusion area setting processing (step # 05). Since the curvature acquisition module is the same as that described above, the description thereof is omitted (see FIG. 8A). The feature extraction module extracts features from the curvature acquired by the curvature acquisition module. The curvature acquired by the curvature acquisition module is the curvature used for extraction of the region exceeding the first predetermined value. The curvature feature corresponds to the amount of change in curvature. For this reason, the feature extraction module is configured using a high-pass filter. The amount of change in curvature obtained by filtering with the high-pass filter is shown in FIG. As a result, it is possible to accurately distinguish a region (abrupt change portion) where the amount of change in curvature is large and a region where the amount of change in curvature is small. The region setting module extracts a region where the amount of change in curvature of the outer surface shape exceeds a second predetermined value set in advance, and the position of this region or a part surrounded by this region is represented in the XY coordinate system. The change amount exclusion region is set (the processing mode of the curvature acquisition unit 18 will be described later). FIG. 8D shows a binarized amount of change in curvature.

  Also, in this embodiment, the change amount exclusion region is set based on the change amount of the curvature, but not only the change amount of the curvature, but also the portion protruding beyond the set amount from the average surface and the set amount are set. If there is a region that is recessed beyond it, the processing mode of the region setting module may be set so that the change amount of the region including these or the region surrounded by these regions is set as the change amount exclusion region.

  The exclusion area setting means 16 has an area setting module and performs exclusion area setting processing (step # 06). The exclusion area setting means 16 sets both the area extracted by the curvature exclusion area extraction means 14 and the area extracted by the change amount exclusion area extraction means 15 as exclusion areas. Therefore, both the curvature exclusion area and the change amount exclusion area are set as exclusion areas by the exclusion area setting means 16. The excluded area set in this way is shown in FIG.

  In the shape recognition device C of the present invention, as will be described later, the base shape data generation means 17 and the detailed shape data generation means 18 cut out three-dimensional data from the three-dimensional data from which noise has been removed, and point groups of this three-dimensional data The surface shape data having a two-dimensional data structure reflecting the surface shape of the region cut out from the image is generated, and the surface shape data is given to the curvature acquisition means 19. The curvature acquisition means 19 calculates a plurality of curvatures from the given point group of the surface shape data, and feeds back to the base shape data generation means 17 and the detailed shape data generation means 18, respectively. Is generated.

  However, when there is a part whose shape changes with a large curvature in a relatively narrow region such as the seating surface Ta of the measurement target T, the position of the point cloud and the actual curved surface of the measurement target T Depending on the relationship, it is often impossible to obtain an accurate curvature. When such an area is included in the area to be evaluated, the accuracy of the evaluation is reduced. Therefore, in order to exclude an area that causes a decrease in accuracy when the outer shape of the measurement target T is evaluated, The setting means 16 sets an exclusion area.

  The base shape data generation unit 17 and the detailed shape data generation unit 18 include a three-dimensional data extraction module, an exclusion processing module, a two-dimensional data conversion module, and a curvature acquisition module.

  The detailed shape data generation means 18 cuts out three-dimensional data having a set width from the three-dimensional data from which noise has been removed, and single surface shape data having a two-dimensional data structure in which the surface shape of the cut-out three-dimensional data is reflected. And detailed shape data represented by a plurality of curvatures acquired from the surface shape data is generated (step # 07).

  Specifically, the three-dimensional data cutout module cuts out three-dimensional data in a region having a set width in the Y-axis direction, and as shown in FIG. 9, the first set interval D1 (about 30 mm. A process of reflecting the point group included in the range of (an example of one setting range) in the Z coordinate (height value) is performed. In this process, the first setting interval D1 is set as the size of the region reflecting the Z coordinates of a plurality of point groups. The region is moved in the X-axis direction at the pitch of the point group, and the region is moved. A Z coordinate (height value) is generated at a pitch equal to the point group pitch in the X-axis direction by processing such as averaging (# 07 step <a>). Next, when there is a point P (height value) included in the exclusion region in the three-dimensional data, the exclusion processing module excludes the point P (<b> in step # 07).

  Thereafter, the two-dimensional data conversion module performs coordinate conversion to convert the three-dimensional data into two-dimensional data to obtain surface shape data (<c> in step # 07). Then, the curvature acquisition module gives the surface shape data to the curvature acquisition means 19 so as to acquire a plurality of curvatures corresponding to the shape of the surface shape data from the curvature acquisition means 19 and to obtain detailed shape data consisting of a plurality of curvatures. (<D> in step # 07).

  Similarly to the detailed shape data generation unit 18, the base shape data generation unit 17 cuts out three-dimensional data having a set width from the three-dimensional data from which noise has been removed, and two-dimensional data in which the surface shape of the cut-out three-dimensional data is reflected. A single surface shape data to be a structure is generated, and base shape data represented by a plurality of curvatures acquired from the surface shape data is generated (step # 08).

  As a specific process, the three-dimensional data cutout module cuts out three-dimensional data of an area having a set width in the Y-axis direction, and a second set interval D2 (first set interval in the X-axis direction as shown in FIG. A process of reflecting a point group included in the range of about 2 to 10 times D1 (an example of the second setting range) on the Z coordinate (height value) is performed. In this process, the second set interval D2 is set as the size of the region reflecting the Z coordinates of a plurality of point groups, and the region is moved while moving the region of this size in the X-axis direction at the point group pitch. A Z coordinate (height value) is generated at a pitch equal to the point group pitch in the X-axis direction by processing such as averaging (# 08 step <a>). Next, when there is a point P (height value) included in the exclusion region in the three-dimensional data, the exclusion processing module excludes the point P (<b> in step # 08).

  Thereafter, the two-dimensional data conversion module performs coordinate conversion to convert the three-dimensional data into two-dimensional data to obtain surface shape data (<c> in step # 08). Then, the curvature acquisition module gives the surface shape data to the curvature acquisition means 19 so as to acquire a plurality of curvatures from the curvature acquisition means 19 and generate base shape data composed of a plurality of curvatures (in step # 08). <D>).

  As described above, in the process of removing the point P included in the excluded region in the steps # 07 and # 08, the process of setting the height value to “0” is not performed, but the height value (point P / point group) is not performed. ) Does not exist.

  As shown in FIGS. 5 and 6, when the surface shape of the measurement target T is curved, the height value indicated by the point group of the cut out three-dimensional data also shows an arc shape. Therefore, when Y1, Y2,..., Yn are given as the coordinate values in the Y-axis direction for the cut out three-dimensional data, the height values corresponding to Y1, Y2,. The height is different from each other in the direction.

  For this reason, as the coordinate conversion in the processing of step (c) of step # 07 and step (c) of step # 08, the normal line for each of a plurality of points P in the Y axis direction and the reference parallel to the Z axis are used. Surface shape data of a two-dimensional data structure is generated by coordinate transformation in which the image is rotated on the corresponding curved center by an angle θ between the axis and projected on the same plane. The center of curvature is calculated for each point P from the shape indicated by the distribution of the point group (height value) in the Y-axis direction indicating the surface shape in the three-dimensional data.

  The curvature acquisition means 19 has a quadratic curve application module and a curvature calculation module. In the process of (d) of step # 07 and (d) of step # 08, surface shape data is given, and a plurality of curvatures corresponding to the shape curve represented by the point group of the surface shape data are generated. In this curvature acquisition means 19, a quadratic curve application module applies a quadratic curve corresponding to a shape curve represented by a point group, and a curvature calculation module calculates a curvature from this quadratic curve, and the calculation result is obtained. This is fed back to the detailed shape data generation means 18 and the base shape data generation means 17.

  The quadratic curve application module of the curvature acquisition means 19 uses a point group corresponding to the shape curve of the surface shape data as the sample point P as shown in FIGS. 11A and 11B, and the plurality of sample points P. Among them, three or more sample points P are set. Then, a quadratic curve passing through the vicinity of the point group having the vertex near the center is applied as the approximate curve F, and the curvature is calculated based on the curvature at the center of the sample point P of the applied approximate curve. Since this approximate curve F is a quadratic curve, a shape passing through a position that matches or approximates a plurality of sample points P is set by increasing or decreasing the coefficient value. The quadratic curve can be understood as a function that can be quadratic differentiated.

  When an approximate curve F composed of a quadratic curve is set, if the shape curve represented by the sample point P is convex downward, the coefficient of the approximate curve F is positive “+”, and is convex upward. The coefficient becomes negative “−”. Further, the closer the shape curve represented by the sample point is to a straight line, the closer the coefficient of the approximate curve F is to “0”. Therefore, even at the inflection point where the shape curve represented by the sample point P changes from convex upward to convex downward, it is only necessary to switch the coefficient value from positive to negative. A curve F can be set. In this curvature acquisition means 19, a quadratic curve is used as the approximate curve F. However, for example, a higher-order function curve higher than a cubic function, a sine curve, a hyperbola, or the like may be used.

  In order to obtain the curvature by the curvature calculation module, it is also possible to obtain directly from the degree of curvature at the sample point center portion (near the apex) of the approximate curve. As one specific example, a process of converting the degree of curve of the approximate curve into a numerical value and storing it as a table and acquiring a curvature value from the table based on the degree of curve can be considered. Further, as a specific processing form for obtaining the curvature, the curvature calculation module sets the center line Q, the distance from the center sample point P in the direction of the center line Q, and the direction orthogonal to the center line Q. Assuming that the center point S is a position where the distance to the approximate curve F in FIG. 1 coincides with a half of the distance to the center sample point P, the radius of the center point S that is inscribed in the vertex is defined as the center point S. An operation may be performed in which the radius of curvature r is set and the inverse of the radius r is set as the curvature. Thus, the setting of the center line Q and the setting of the center point S in the process of obtaining the curvature are not limited to those shown in FIG. 11, for example, setting the center line Q to the middle of two sample points P, It is also possible to set an arbitrary one suitable for processing.

  From such processing, the base shape data is composed of a plurality of curvatures, and the detailed shape data is also composed of a plurality of curvatures.

  In this embodiment, the curvature is excluded by an instruction from the curvature exclusion region extraction unit 14, the change amount exclusion region extraction unit 15, the exclusion region setting unit 16, the base shape data generation unit 17, and the detailed shape data generation unit 18. The acquisition means 19 calculates the curvature. Instead, the curvature exclusion area extraction means 14, the change amount exclusion area extraction means 15, the exclusion area setting means 16, the base shape data generation means 17, and the details The shape data generation unit 18 may include a module that performs the same processing as that of the curvature acquisition unit 19, and the module may perform processing for calculating the curvature.

  When generating the detailed shape data and the base shape data, in order to evaluate the shape of the entire surface of the measurement target T, the three-dimensional data cutout module shifts the cutout area in the Y-axis direction, while # 07 and # 08. Step processing is repeated (step # 09). In particular, when shifting the object to be cut out in the Y-axis direction, the processing mode is adopted so that the three-dimensional data cut out first and the three-dimensional data cut out after that do not overlap in the Y-axis direction. The processing mode may be set so as to partially overlap in the Y-axis direction.

  The shape evaluation means 20 compares the curvatures in the same region in the X-axis direction among the plurality of curvatures of the base shape data and the detailed shape data, and acquires a distortion value (Step # 10). This distortion value is a simple process of calculating the difference in curvature value, but as shown in FIG. 12, the difference from the curvature of the detailed shape data is extracted as a distortion value with reference to the curvature of the base shape data.

[Different processing forms of shape evaluation means]
In particular, in the present invention, when acquiring the strain value, two types of external curves (not shown) assumed from a plurality of curvatures of the base shape data and the detailed shape data are generated by calculation, and the base shape data is obtained. The processing form of the shape evaluation means 20 may be set so as to obtain the distortion value from the difference (offset amount) between the corresponding outer shape curve and the outer shape curve corresponding to the detailed shape data. By adopting this processing mode, the unevenness amount of the detailed shape data based on the base shape data is acquired with a value close to the actual size.

  The evaluation image generation unit 21 acquires shape image data indicating an outline indicating the outer edge of the measurement target T from the measurement target shape acquisition unit 22, and sets an area including the same distortion value with respect to the internal region of the shape image data. Then, for each area, the hue / density corresponding to the distortion value is applied to the area, and this image is displayed on the display d (step # 11).

  As a result of the processing of the evaluation image generating means 21, an image of the shape of the measurement target T is displayed on the display d, and distortion in the protruding direction and depressions with respect to the internal region of the image of this shape. By displaying the distortion of the direction in different hues, and by changing the density and hue according to the amount of distortion, it is possible to visually grasp the area where the distortion exists, the direction of the distortion, and the degree of distortion To do.

[Outline of Embodiment]
As described above, according to the present invention, three-dimensional data is generated from photographing data having a three-dimensional data structure acquired by the photographing unit V, and after removing noise, the shape of a predetermined setting region in the three-dimensional data is detailed. The detailed shape data including the reflected curvature group is generated, and the base shape data including the curvature group reflecting the shape of the predetermined setting region of the three-dimensional data more roughly than the detailed shape data is generated. Thereafter, the data included in the exclusion region is excluded from the respective shape data (ie, the detailed shape data and the base shape data) to generate the respective corrected shape data (ie, the corrected detailed shape data and the corrected base shape data). . The distortion value of the outer surface shape is extracted based on the curvature of each of the generated corrected shape data (corrected detailed shape data and corrected base shape data). That is, the distortion value is obtained by comparing the curvature of the same region in the modified detailed shape data and the modified base shape data. As a result, for example, even when a concave distortion exists in a part of the narrow region of the outer surface shape that is a gentle convex shape as a whole, the concave distortion can be accurately detected.

  In addition, when there is distortion in the measurement target T, an image having the same shape as the measurement target T is displayed on the display d, a distortion distribution area is displayed inside the image, and distortion is performed for each area. Since the paint of the hue / density corresponding to is performed, the position where the distortion exists visually can be grasped, and at the same time, the direction of the distortion (convex or concave) and the degree of the distortion can be visually grasped.

[Another embodiment]
In the present invention, the processing form is set so that the detailed shape data and the base shape data are generated directly from the acquired three-dimensional data, whether it is three-dimensional data acquired by photographing or three-dimensional data having an STL file structure. May be set. As a specific example, the detailed shape data generation means 18 acquires the coordinates of a point group having a set width in a predetermined direction among the coordinates of the point group of the three-dimensional data indicating the outer surface shape of the measurement target T. A point group for each first set interval D1 is acquired from the group, an approximate plane is defined from the peripheral point group of this point group by the method of least squares, and the origin 0 of this approximate plane is set.

  Next, a map (two-dimensional data) in which the coordinates of the peripheral point group are reflected on a plane including the Z ′ axis (for example, the Z′-X ′ plane) with the normal direction of the approximate plane as the Z ′ axis. Is generated. By applying an approximate curve to this map and calculating the curvature of the approximate curve, the curvature of the detailed shape data with the origin 0 as the inspection point is obtained. Similarly, the base shape data generating means 16 obtains the curvature of the base shape data with the origin 0 as the inspection point from the point group for each second set interval D2. By this principle, it is possible to obtain the curvature of the detailed shape data and the base shape data based on a plurality of origins 0, and by obtaining and displaying the distortion value from the curvature of the detailed shape data and the base shape data by the same processing as described above. The distortion can be grasped.

  In this alternative embodiment, the process of excluding the excluded area is not described, but the excluded area is set in the three-dimensional data acquired by the base shape data generating unit 17 or the detailed shape data generating unit 18 and excluded from the processing target. If the acquired 3D data may contain noise, the processing form is set so that noise is removed after the base shape data generation unit 17 or the detailed shape data generation unit 18 acquires the noise. May be.

  INDUSTRIAL APPLICABILITY The present invention can be used for evaluation of three-dimensional data for manufacturing a mold and three-dimensional data for machining with a machine tool.

C: Shape recognition device 14: Curvature exclusion region extraction means 15: Change amount exclusion region extraction means 16: Exclusion region setting means 20: Shape evaluation means

Claims (1)

A shape recognition device for recognizing an outer shape of an object from three-dimensional data representing the outer shape of the object,
Curvature-exclusion region extraction means for extracting a region where the curvature of the outer surface shape represented by the three-dimensional data exceeds a preset first predetermined value;
A change amount exclusion region extraction means for calculating a change amount of the curvature of the outer surface shape used for the extraction of the region and extracting a region in which the change amount exceeds a preset second predetermined value;
An exclusion region setting unit that sets both the region extracted by the curvature exclusion region extraction unit and the region extracted by the change amount exclusion region extraction unit as an exclusion region;
The detailed shape data reflecting the shape of the predetermined setting area in the three-dimensional data in detail and the base shape data reflecting the shape of the predetermined setting area more roughly than the detailed shape data are obtained, and the respective shapes Shape evaluation means for generating each corrected shape data by excluding data included in the exclusion region from the data, and extracting the distortion value of the outer surface shape based on the curvature of the respective corrected shape data;
A shape recognition device comprising:

Images