JP4743773B2 - Edge detection method, apparatus, and program - Google Patents

Description

  The present invention relates to an edge detection method, apparatus, and program for detecting an edge corresponding to an object from two-dimensional image data acquired for the object.

  For example, in a three-dimensional measurement device such as a surface measurement type three-dimensional digitizer, two-dimensional image data is acquired for an object, and the three-dimensional data of the object is calculated based on the two-dimensional image data. In some applications, the two-dimensional image data is subjected to image processing, and an edge corresponding to the shape of an object included in the two-dimensional image may be detected. Here, when the object is a three-dimensional object, a portion deviating from the in-focus position is inevitably generated, and thus the two-dimensional image includes blur. Therefore, the edge may not be detected with high accuracy.

Conventionally, in such a case, a method has been adopted in which one predetermined analytical function is applied to an edge detection location for fitting, and the edge position is detected by obtaining the peak position of the fitting function. Further, in Patent Document 1, a differential image is generated by differentiating two-dimensional image data, and the peak position of the approximate curve of the differential curve is detected as an edge for a portion where the differential curve is steep, and the differential curve is moderated. A technique for detecting the center of gravity of the differential curve as an edge is disclosed for such a point.
JP-A-9-2574226

  The degree of image blur varies depending on the distance from the in-focus position. That is, depending on the three-dimensional shape of the object, the degree of blurring of the image at the edge portion varies. However, in the conventional method, fitting is performed by applying a certain analysis function regardless of the degree of image blur. Accordingly, there is a problem that the fitting accuracy is deteriorated, and as a result, the edge detection accuracy is lowered. This problem cannot be sufficiently solved even by the technique of Patent Document 2 that uses the approximate curve and the center of gravity separately. That is, after all, the method of Patent Document 2 depends on the centroid calculation when image blurring occurs, and the edge cannot be detected with high accuracy by such centroid calculation.

  The present invention has been made in view of the above problems, and provides an edge detection method, apparatus, and program capable of accurately detecting an edge even when an image blur occurs at an edge detection location. For the purpose.

According to a first aspect of the present invention, there is provided an edge detection method comprising: obtaining two-dimensional image data about the object by photographing the object using a predetermined optical system; and the object from the photographing position. distance information to, the steps of acquiring have you in various places within the image based on the two-dimensional image data, in accordance with the distance information, it is applied to each of the edge detection portion for detecting edges existing in the image A step of determining an analytic function; and a step of detecting an edge by applying the determined analytic function to each of the edge detection locations, and the step of detecting the edge comprises the two-dimensional image data Differential processing by differential processing, and a step of fitting an analytical function applied to each of the edge detection locations to the differential image , Characterized in that it comprises the steps of detecting the vertex of the fitted analytical function as an edge.

According to this method, distance information from the photographing position to the object is acquired based on a two-dimensional image of the object, and an analysis function to be applied to each of the edge detection locations is determined according to the distance information. Therefore, an optimal analysis function corresponding to the three-dimensional shape of each edge detection location is applied. For example, edge detection can be performed with an analytic function that takes into account the degree of deviation from the in-focus position at each edge detection location. Therefore, the edge detection accuracy can be improved.

In addition, the step of detecting the edge includes a step of creating a differential image by differentiating the two-dimensional image data, and a step of fitting an analysis function applied to each of the edge detection locations to the differential image; And the step of detecting the apex of the fitted analytical function as an edge, so that the peak value (edge) can be easily obtained based on the amount of change in the differential image, and the edge detection process can be simplified.

In this case, the pre-Symbol analysis function, we are desirable to include a step of deriving, based on the pre-designed value of the optical system (claim 2). The optimal analysis function corresponding to the distance information to the object can be derived from the design value of the optical system. According to this method, the step of deriving in advance the analysis function corresponding to the distance information based on the design value is executed, so that the analysis function to be applied to each edge detection location can be determined quickly and accurately. .

Also, the pre-Symbol analysis function, we are desirable to include a step of deriving from advance the actual result of the sample taken with the optical system (claim 3). According to this method, an analysis function based on the actual measurement result can be prepared according to the distance information, so that the analysis function to be applied to each edge detection location can be determined quickly and accurately. .

  In any one of the above methods, it is preferable that the analysis function includes a step of forming a table according to the distance information. Preferably, the method further includes a step of obtaining a derivation function for deriving the analytic function in accordance with the distance information. According to these methods, a table of analytic functions according to distance information and a derivation function for deriving analytic functions according to distance information are prepared, facilitating determination processing of analytic functions to be applied to edge detection locations. can do.

Further, in determining the pre-Symbol analysis function, in addition to the distance information, predetermined lens information about the photographing lens can be made to be utilized (claim 6). According to this method, the analysis function is determined using predetermined lens information. That is, by referring to the distance information and the lens information for the optical system used when acquiring the two-dimensional image data, the relationship between the amount of deviation from the in-focus position and the degree of image blur can be grasped more accurately. It becomes like this. Therefore, it becomes possible to estimate the shape of the object represented in the two-dimensional image due to the degree of blur, and to accurately determine the analysis function to be applied to the edge detection location by matching the shape of the object.

  In this case, focal length information and / or lens F value information of the photographing lens can be used as the lens information. According to this method, information on the degree of blur from the in-focus position is obtained from the combination of distance information, focal length information, or a combination of distance information and lens F value information, or from three information of distance information, focal length information, and lens F value information. Be grasped. Since the focal length information and the lens F value information are parameters that greatly affect the degree of blur from the in-focus position, the degree of blur can be accurately grasped by using these information.

  In any one of the above methods, it is preferable that the method includes a step of deriving in advance information on a degree of blurring of an image accompanying a shift from a focus position based on the design value of the optical system based on the distance information and the lens information. (Claim 8). Alternatively, the method may include a step of deriving information related to the degree of blurring of the image accompanying the deviation from the in-focus position from an actual measurement result obtained by performing sample photographing using the optical system in advance in association with the distance information and the lens information. Desirable (Claim 9). According to these methods, information on the degree of blurring of the image accompanying the deviation from the in-focus position is grasped in advance based on the design value of the optical system and the actual measurement result, so that the analysis function to be applied to each edge detection point is determined. Can be performed quickly and accurately.

In any one of the above methods, it is preferable that the analysis function includes a step of creating a table in association with the distance information and the lens information. Or
Preferably, the method further comprises a step of obtaining a derivation function for deriving the analytic function in accordance with the distance information and the lens information. According to these methods, the analysis function table associated with the distance information and the lens information and the derivation function for deriving the analysis function according to the distance information are prepared. Function determination processing can be facilitated.

Furthermore, in any of the above methods, it is desirable to use at least a Gaussian function and / or a quadratic function as the analytic function ( claim 12 ). According to this method, for example, when the degree of blur at the edge detection portion is relatively large, fitting is performed using a Gaussian function as an analysis function, and when the degree of blur is relatively small, fitting is performed using a quadratic function as an analysis function. Can be used properly.

An edge detection apparatus according to a thirteenth aspect of the present invention includes an imaging unit that acquires two-dimensional image data about an object, and distance information from the position of photographing by the imaging unit to the object. to a distance calculation means for calculating the various locations in the image rather based, an edge candidate specifying means for specifying a candidate to become possible edge defining an edge existing in the two-dimensional image, the analysis function applied to each of the edge candidate Analysis function deriving means for deriving each according to the distance information, edge detection means for detecting the edge by applying the derived analysis function to each of the edge candidates, and differentiating the two-dimensional image data Differential image generation means for creating a differential image by processing, wherein the edge detection means is an analytic function derived for each of the edge candidates The fitting processing on the differential image, and detects the apex of the fitting has been analyzed functions as an edge.

  According to this apparatus, the two-dimensional image data of the object is acquired by the imaging unit, and the distance information of the image is acquired by the distance calculation unit based on the two-dimensional image data. Then, for each of the edge candidates specified by the edge candidate specifying means, an analysis function to be applied according to the distance information is determined by the analysis function deriving means. Therefore, it is possible to apply an optimal analysis function according to the location where each edge candidate exists, and the edge detection accuracy can be improved.

Further, the image processing apparatus includes a differential image generation unit that generates a differential image by performing differential processing on the two-dimensional image data, and the edge detection unit fits an analysis function derived for each of the edge candidates to the differential image. Then, since the vertex of the fitted analytical function is detected as an edge, the peak value (edge) can be easily obtained based on the change amount of the differential image, and the arithmetic processing in the edge detecting means can be simplified.

In the above apparatus, the photographing lens used in the imaging means includes lens information obtaining means for obtaining at least focal length information and / or lens F value information at the time of photographing, and the analysis function deriving means includes in addition to the distance information, it is desirable to derive the analysis function by referring to the focal length information and / or the lens F value information of said photographing lens (claim 14). According to this configuration, the analysis function is derived by referring to the focal length information and / or the lens F value information in addition to the distance information. The focal length information and the lens F value information are parameters that have a large influence on the degree of blurring of the image in the imaging unit that captures two-dimensional image data. The degree of blur can be accurately grasped. Therefore, it is possible to derive the analytic function more accurately by matching the shape of the edge candidate location.

In the above apparatus, the analysis function storage means for holding the analysis function in a table format corresponding to the distance information is provided, and the analysis function derivation means includes the distance information of the edge candidate and the table of the analysis function storage means. It is desirable to derive an analysis function to be applied to each of the edge candidates by performing collation ( claim 15 ). According to this configuration, since the analysis function table corresponding to the distance information is held in the analysis function storage means, the analysis function deriving means only performs a process of collating the distance information of each edge candidate with the table. An analytic function can be derived. Therefore, it is possible to facilitate analysis function determination processing applied to the edge detection location.

The apparatus further includes a derivation function storage unit that holds a derivation function for deriving the analytic function according to the distance information, and the analytic function derivation unit uses the distance information of the edge candidate as the derivation function. by giving, it is desirable to derive an analysis function applied to each of the edge candidate (claim 16). According to this configuration, since the derivation function storing unit holds the derivation function for deriving the analytic function according to the distance information, the analytic function derivation unit simply applies the distance information of each edge candidate to the derivation function. An analytic function can be derived. Therefore, it is possible to facilitate analysis function determination processing applied to the edge detection location.

Further, the analysis function storage means for holding the analysis function in a table format according to the distance information and the focal length information and / or lens F value information, the analysis function derivation means includes the edge candidate It is desirable to derive an analysis function to be applied to each of the edge candidates by comparing the distance information and the acquired focal length information and / or lens F value information with a table of the analysis function storage means. ( Claim 17 ). According to this configuration, since the analysis function storage unit stores the analysis function table corresponding to the distance information, the focal length information, and / or the lens F value information, the analysis function deriving unit is provided with each edge. An analysis function that takes into account the degree of blur can be derived only by performing processing for collating candidate distance information with the table. Therefore, it is possible to facilitate analysis function determination processing applied to the edge detection location.

In addition, a derivation function storage unit that holds a derivation function for deriving the analytic function according to the distance information and the focal length information and / or lens F value information, by giving distance information of the edge candidates and acquired the focal length information and / or the lens F value information on the derivation function, it is desirable to derive an analysis function applied to each of the edge candidate (claim 18 ). According to this configuration, the derivation function storage means holds the derivation function for deriving the analysis function in accordance with the distance information, the focal length information and / or the lens F value information. By simply applying the distance information of the edge candidate to the derivation function, an analysis function in consideration of the degree of blur can be derived. Therefore, it is possible to facilitate analysis function determination processing applied to the edge detection location.

An edge detection program according to claim 19 of the present invention reads a two-dimensional image data of an object photographed using a predetermined optical system into a computer capable of data processing of the two-dimensional image data; Based on the two-dimensional image data, obtaining distance information from the shooting position to the object at various points in the image based on the two-dimensional image data; and candidates for determining edges existing in the two-dimensional image; Identifying an edge candidate, a step of deriving an analytic function to be applied to each of the edge candidates according to the distance information, and applying the derived analytic function to each of the edge candidates, respectively. wherein detecting the edge, a Te, as the step of detecting the edge, differentiating the two-dimensional image data Execution and creating a differential image by management, comprising the steps of fitting processing in the differentiated image analysis function applied to each of the edge detection points, and detecting the vertices of the fitted analytical functions as an edge It is characterized by making it.

  According to this program, the two-dimensional image data stored in the predetermined storage means is read by receiving a transfer directly from the imaging means or from the imaging means, and distance information is acquired from the two-dimensional image data. Then, it is possible to cause the computer to perform processing for determining the analysis function applied to each of the edge detection locations in accordance with the distance information. As a result, an optimal analysis function corresponding to the three-dimensional shape of each edge detection location is applied. For example, edge detection can be performed with an analytic function that takes into account the degree of deviation from the in-focus position at each edge detection location. Therefore, the edge detection accuracy can be improved.

  According to the edge detection method, apparatus, and program according to the present invention, even when image blur having a different blur condition occurs in an edge detection location, an analysis function suitable for each edge detection location is applied and the accuracy is improved. Edges can be detected. Therefore, it is possible to accurately grasp the edge of the object as compared with the case where a certain analysis function is applied as in the conventional method. For example, when performing various measurements based on two-dimensional image data, it is accurate. Measurement data and analysis data can be acquired.

Hereinafter, an embodiment of the present invention will be described with reference to the drawings.
FIG. 1 is a block diagram schematically illustrating an embodiment of a hardware configuration of an edge detection apparatus S according to the present invention, and FIG. 2 is a block diagram illustrating an electrical configuration of the edge detection apparatus S. The edge detection device S is a non-contact type three-dimensional that obtains a three-dimensional shape of the target object 10 from two-dimensional image data acquired for the target object 10 (illustrated here is an automobile door portion). In addition to the shape measuring device, the device can detect an edge corresponding to the object 10. The edge detection device S is configured to include an imaging device 20 (imaging means) that acquires two-dimensional image data of the object 10 and a personal computer 30 that performs predetermined analysis calculation processing on the two-dimensional image data. .

  The imaging device 20 acquires two-dimensional image data of the object 10 using a method called a light cutting method. As shown in FIG. 2, the imaging device 20 includes a light projecting unit 21, a light receiving unit 22, an optical system driving unit 23, an image sensor 24 including a CCD (Charge Coupled Device) area sensor, a timing generator (TG) 25, An output processing circuit 26, a lens information output unit 27, a data memory 28, a measurement control unit 29, and an I / F 201 are included. As shown in FIG. 1, the imaging device 20 has a substantially rectangular parallelepiped housing shape, and is provided with a light projecting window for the light projecting unit 21 and a light receiving window for the light receiving unit 22. The light projecting unit 21 is provided at a position above the light receiving unit 22 by a predetermined distance according to the baseline length.

  The light projecting unit 21 includes a laser diode serving as a light source, a light projecting optical system, a rotating mirror, and the like, and emits slit light that is a laser beam spreading in the horizontal direction. The slit light is planar light that spreads in a horizontal direction at a predetermined radiation angle (fan shape) and has a width in the vertical direction. The light projecting unit 21 scans the object 10 with the slit light under the control of the measurement control unit 29.

  The light receiving unit 22 includes a plurality of photographing lenses constituting a light receiving optical system, a diaphragm, a lens moving mechanism for focusing and zooming, and the like, and the slit light is reflected by the surface of the object 10. Part of the light is incident. Incidentally, the personal computer 30 obtains the distance to the reflection point based on the principle of triangulation based on the projection angle of the laser beam, the reception angle of the reflected light, and the baseline length between the projection point and the reception point. Based on such distance information, three-dimensional data of the object 10 is acquired.

  The optical system driving unit 23 includes a stepping motor and the like, and generates a driving force for driving the photographing lens of the light receiving unit 22 for focus singing and zooming under the control of the measurement control unit 29.

  The imaging device 24 generates two-dimensional image data for the target object 10 by photoelectrically converting the optical image of the target object 10 formed by the light receiving unit 22. As the imaging element 24, a plurality of photoelectric conversion elements composed of, for example, photodiodes are two-dimensionally arranged in a matrix, and the light receiving surfaces of the photoelectric conversion elements have different spectral characteristics, for example, R (red), G A Bayer array CCD color area sensor in which (green) and B (blue) color filters are arranged in a ratio of 1: 2: 1 can be used.

  The timing generator 25 generates timing pulses for controlling photographing operations (charge accumulation based on exposure, reading of accumulated charges, etc.) by the image sensor 24. For example, the timing generator 25 generates a vertical transfer pulse, a horizontal transfer pulse, a charge sweeping pulse, and the like based on the imaging control signal supplied from the measurement control unit 29 and supplies it to the image sensor 24.

  The output processing circuit 26 performs predetermined signal processing on the image signal (analog signal group received by each pixel of the CCD area sensor) output from the image sensor 24, and then converts it into a digital signal and outputs it. The output processing circuit 26 includes a CDS circuit (correlated double sampling circuit) for reducing reset noise included in the analog image signal, an AGC circuit for correcting the level of the analog image signal, and the analog image signal as a 14-bit digital image, for example. An A / D conversion circuit that converts the signal (image data) is provided.

  The lens information output unit 27 (lens information acquisition means) acquires and outputs at least focal length information and lens F value information at the time of shooting for the shooting lens used in the light receiving unit 22. The focal length information and the lens F value information are parameters that greatly affect the degree of blurring of the image due to the deviation from the in-focus position in the captured two-dimensional image. By using these pieces of information, the degree of blur according to the distance from the in-focus position can be accurately grasped, and the analysis function can be derived more accurately by matching the shape of the location where the edge candidate exists.

  The data memory 28 includes a RAM (Random Access Memory) or the like, and temporarily stores various data. For example, the data memory 28 temporarily stores two-dimensional image data about the object 10 output from the output processing circuit 26, focal length information output from the lens information output unit 27, and lens F value information.

  The measurement control unit 29 includes a CPU (Central Processing Unit) and the like, and controls the operation of each unit of the imaging device 20. Specifically, the measurement control unit 29 scans the slit light by the light projecting unit 21, drives the photographing lens by the optical system driving unit 23, generates a timing pulse by the timing generator 25, and outputs a signal from the output processing circuit 26. The processing operation and the data recording operation for the data memory 28 are controlled.

  The I / F 201 is an interface for enabling data communication with the personal computer 30. Two-dimensional image data and the like of the target object 10 temporarily stored in the data memory 28 is transmitted to the personal computer 30 via the I / F 201.

The personal computer 30 reads the two-dimensional image data about the object 10, obtains distance information about the object 10, the process of specifying edge candidates that are candidates for determining edges existing in the two-dimensional image, the edge A process of deriving an analytic function to be applied to each candidate according to the distance information, a process of detecting an edge by applying the derived analytic function to each of the edge candidates, and the like are performed. As shown in FIG. 2, the personal computer 30 includes an I / F 301 that is an interface that enables data communication with the imaging device 20, an operation unit 31 that includes a keyboard and a mouse, a display unit 32 that includes a liquid crystal display, and the like. A ROM (Read Only Memory) 33 for storing various control programs, a RAM 34 for temporarily storing various data and storing preset setting data, and a CPU 35 for performing various arithmetic operations are provided.

  FIG. 3 is a functional block diagram showing a functional configuration of the RAM 34 and the CPU 35. The CPU 35 executes a control program stored in the ROM 33 to thereby execute a data buffer 351, a three-dimensional data calculation unit 352 (distance calculation unit), a filter processing unit 353, an edge candidate pixel detection unit 354 (edge candidate identification unit), An analysis function deriving unit 355 (analysis function deriving unit), a differential processing unit 356 (differential image generating unit), a fitting processing unit 357, an edge value detection unit 358 (edge detection unit), and a processing control unit 359 function. . The RAM 34 includes an analysis function storage unit 341 and a derivation function storage unit 342.

  The data buffer 351 temporarily stores the two-dimensional image data and lens information read from the imaging device 20 via the I / F 301.

  The three-dimensional data calculation unit 352 reads the two-dimensional image data, and projects the projection angle of the laser light emitted from the projection unit 21 of the imaging device 20, the reception angle of the reflected light received by the light receiving unit 22, and the projection Based on the baseline length between the point and the light receiving point, the distance to the reflection point on the object 10 is obtained by the principle of triangulation. The projection angle is obtained from the rotation angle of the rotating mirror of the projection unit 21. The light receiving angle is calculated from the image position of the reflected light detected on the light receiving surface of the image sensor 24. That is, the three-dimensional data calculation unit 352 acquires distance information to the target object 10 in units of pixels of the image sensor 24 or predetermined pixel group units (locations in the image). By acquiring this distance information, it is possible to determine the amount of deviation from the in-focus position of each pixel. Furthermore, based on the distance information, three-dimensional data (for example, polygon mesh data) of the object 10 is obtained.

  The filter processing unit 353 performs preprocessing such as gradation conversion processing and noise removal processing on the two-dimensional image data, and then performs filter processing for extracting edges on the two-dimensional image data. The edge is a portion where the brightness is rapidly changed in the image, and corresponds to the contour of the object 10 or the boundary of the uneven shape. Note that the outline of a planar character, pattern, or the like drawn on the surface of the object 10 is also an edge. Since fitting to which an analytical function is applied is not necessary for such a planar edge, this embodiment Then, explanation is omitted. As a filter for edge extraction, for example, a differential filter, a Sobel filter, a gradient filter, a Laptian filter, or the like can be used.

  The edge candidate pixel detection unit 354 detects address information of pixels in which the filter value of the filter processing executed by the filter processing unit 353 exceeds a predetermined threshold value. Such pixel existence points serve as candidate points (edge candidates) that determine edges existing in the two-dimensional image.

  The analysis function deriving unit 355 derives an analysis function to be applied to each of the edge candidate pixels according to the distance information obtained by the three-dimensional data calculation unit 352. When the object 10 is a three-dimensional object, the image is blurred due to, for example, the occurrence of a portion deviating from the in-focus position in the outline. For the portion where the image blur occurs, fitting is performed by applying an analysis function in a fitting processing unit 357 described later, and an edge is detected. The analytic function deriving unit 355 derives the analytic function to be applied to each edge candidate according to the degree of image blur.

  The degree of image blur varies depending on the distance from the in-focus position. Therefore, the degree of blur of the image can be grasped by obtaining the distance from the in-focus position using the distance information obtained for the pixel specified as the edge candidate. The degree of change in image blur varies depending on the focal length of the photographing lens (the optical system of the light receiving unit 22) and the lens F value at the time of photographing. Therefore, the analysis function deriving unit 355 derives an analysis function suitable for each edge candidate with reference to the distance information of the edge candidate pixels and the focal length information and / or lens F value information of the photographing lens. This makes it possible to accurately derive the analytical function by matching the shape of the location where the edge candidate exists. In deriving the analytic function, the analytic function deriving unit 355 refers to the analytic function storage unit 341 or the derived function storage unit 342 of the RAM 34 described later.

  The differential processing unit 356 reads the two-dimensional image data from the data buffer 351 and generates a differential image by differentiating the two-dimensional image data. This differential image is for obtaining a peak of a luminance change caused by an edge included in the two-dimensional image.

  The fitting processing unit 357 performs processing for fitting the analytic function derived by the analytic function deriving unit 355 to the differential image for each of the edge candidates. For example, when the “quadratic function” is derived by the analytic function deriving unit 355 for a certain edge candidate, the quadratic function is obtained by using an approximation method such as the least square method for the differential value of the pixel output for the edge candidate. And the parameters of the fitting function are obtained.

  The edge value detection unit 358 obtains the vertex of the fitting function obtained by the fitting processing unit 357, and outputs this as an edge value. That is, the vertex of the fitting function is a change peak of the pixel output, and such a pixel position can be treated as an edge on the image.

  The processing control unit 359 receives the operation signal given from the user via the operation unit 31 and performs the above operation, the three-dimensional data calculation unit 352, the filter processing unit 353, the edge candidate pixel detection unit 354, the analysis function derivation unit. 355, overall control for operating the differential processing unit 356, the fitting processing unit 357, and the edge value detection unit 358 in a timely manner.

  In the analysis function storage unit 341 of the RAM 34, the analysis function used in the fitting processing unit 357 includes distance information (distance information from the in-focus position) with respect to the object 10 and lens information (focal length information and lens of the photographing lens). (F value information) and stored as a table. The analysis function deriving unit 355 includes distance information from the in-focus position based on the calculation data of the three-dimensional data calculation unit 352, the focal length information of the photographing lens and the lens F value output from the lens information output unit 27 (FIG. 2). The information is compared with the table as reference information, and an analysis function matching the reference information is derived from the table. The table may be a table using only the distance information as a variable, a table using distance information and focal length information as variables, or a table using distance information and lens F value information as variables.

  In the derivation function storage unit 342, the analysis function used in the fitting processing unit 357 includes distance information (d) from the in-focus position, lens information (focal length information (f) of the photographing lens, and lens F value information ( FNo)) is derived and derived functions F (d, f, FNo) are stored. The analysis function deriving unit 355 includes distance information (d) based on the calculation data by the three-dimensional data calculation unit 352, focal length information (f) of the photographing lens output from the lens information output unit 27, and lens F value information (FNo. ) Is applied to the derived function F (d, f, FNo) to derive the analytic function. The derivation function F is a derivation function F (d) having only the distance information as a variable, a derivation function F (d, f) having distance information and focal length information as variables, or distance information and lens F value information. May be a derivation function F (d, FNo).

  Both the analysis function storage unit 341 and the derivation function storage unit 342 may be provided, but only one of them may be provided. By preparing the analytic function storage unit 341 and / or the derived function storage unit 342, the analytic function deriving unit 355 can easily derive the derived function in accordance with the distance information and the lens information. It is possible to facilitate the determination process of the analytic function to be applied.

  Next, a specific example of an edge detection method using the edge detection device S configured as described above will be described. FIG. 4 is a process flowchart showing an edge detection process flow. Prior to actually performing the edge detection, a step of deriving an analytic function in advance according to the characteristics of the optical system of the light receiving unit 22 is performed. Here, as an analysis function derivation example, a method of deriving an analysis function from a design value of the optical system of the light receiving unit 22 (step # 1) and a method of deriving from an actual measurement result obtained by taking a sample using the optical system in advance (step # 1). Step # 2) is illustrated. It suffices to perform either one of Step # 1 and Step # 2.

  The above step # 1 can be executed, for example, by inputting a design value of the optical system as a parameter to an appropriate optical simulator or the like. That is, the shape (blurring degree) of the two-dimensional image when the distance deviates from the in-focus position can be calculated from the lens design value of the photographing lens. Therefore, an analysis function corresponding to the distance information can be derived based on the simulation result. By combining focal length information and lens F value information with this, an analysis function suitable for each condition can be obtained. According to this method, the analysis function can be quickly derived based on the design value of the optical system.

  Further, step # 2 can be executed by actually imaging the calibration chart CH as shown in FIG. The calibration chart CH is a flat chart in which a circular silver mark CH1 is printed on the black ground CH2. Such a calibration chart CH is arranged in front of the imaging device 20, and photographing is performed a plurality of times while gradually changing the distance between the two (while gradually shifting from the in-focus position). From this photographing result, the peak shape (blurring condition) of the silver mark CH1 is analyzed for each distance, and an analysis function corresponding to each distance is derived. In addition, when adding the focal length information and / or lens F value information of the photographic lens, this operation is performed by changing the focal length and / or the F value, and the peak shape of the silver mark CH1 is similarly analyzed. The analytical function for each condition is derived. According to this method, an accurate analytical function can be derived based on the actual measurement result.

  As the analysis function, various fitting functions can be used. For example, a quadratic function as shown in FIG. 6A and a Gaussian function as shown in FIG. 6B are preferably used. it can. That is, a quadratic function corresponding to a relatively sharp peak shape and a Gaussian function corresponding to a relatively gentle peak shape can be used. In this case, for example, fitting can be performed using a Gaussian function when the degree of blur at the edge detection portion is relatively large, and fitting using a quadratic function when the degree of blur is relatively small. .

  If the analytic function is derived in this way, a step (step # 3) of making this a table according to the distance information (+ focal length information and / or lens F value information), or the analytic function as the distance information A step (step # 4) of converting to a derivation function derived according to (+ focal length information and / or lens F value information) is executed.

  7 to 9 are tables in a table format showing an example of the table created in step # 3. FIG. 7 shows an example of a table to which analysis functions are assigned according to distance information with respect to the object 10. Note that the distance from the in-focus position is obtained from the difference between the in-focus distance and the distance information to the object 10 obtained by the three-dimensional data calculation unit 352. FIG. 8 shows an example of a table to which analysis functions are respectively assigned according to distance information on the object 10 and focal lengths (tele, middle, wide). Further, FIG. 9 shows an example of a table to which analysis functions are respectively assigned according to distance information with respect to the object 10, a focal length, and an F value. Such a table is stored in advance in the analysis function storage unit 341 of the RAM 34 (step # 5).

  When executing Step # 4, a derived function F (d) for deriving the derived analytic function in accordance with distance information (d) from the in-focus position, distance information (d), and focal length information (f) Derivation function F (d, FNo) derived according to the above, derivation function F (d, FNo) derived according to the distance information (d) and lens F value information (FNo), or all of these A derived function F (d, f, FNo) to be derived is obtained. These derived functions are stored in advance in the derived function storage unit 342 of the RAM 34 (step # 5). The above is the preparation step executed prior to the actual edge detection.

  Next, the object 10 is imaged by the imaging device 20, and two-dimensional image data of the object 10 is acquired (step # 6). The imaging here is, for example, imaging by a light cutting method involving light projection and reception of slit light as described above, and the acquired two-dimensional image data is data that can calculate three-dimensional data about the object 10. Subsequently, the personal computer 30 obtains distance information to the object 10 in units of pixels (pixel groups) of the image sensor 24 using the principle of triangulation (step # 7). As a result, the distance from the in-focus position is also determined in units of pixels. Thereafter, an edge detection process described in detail later is executed (step # 8), and edge data is acquired.

  FIG. 10 is a plan view showing an example of the two-dimensional image data 40 of the object 10 acquired in step # 6. Here, the door part of a motor vehicle is illustrated as the target object 10, and in this case, a substantially rectangular outline shape is an edge. When imaging the target object 10, the imaging apparatus 20 performs AF control, for example, at the position of the center portion of the target object 10 to determine the focal distance and perform imaging.

  In the two-dimensional image data 40 acquired in this way, attention is paid to the first edge portion 11 surrounded by a square denoted by reference numeral 11 and the second edge portion 12 surrounded by a square denoted by reference numeral 12. While the first edge portion 11 is substantially at the same height as the door center portion selected as the AF point, the object 10 has a shape such that the second edge portion 12 is curved rearward. In this case, since the degree of deviation from the in-focus position of the second edge portion 12 is large, the degree of blurring of the image is relatively large. This blur condition depends not only on the distance from the in-focus position but also on the focal length of the light receiving optical system and the lens F value.

  FIG. 11A is a graph showing the differential value of the pixel output around the first edge portion 11, and FIG. 11B is a graph showing the differential value of the pixel output around the second edge portion 12. Since the first edge portion 11 has a short distance from the in-focus position, the peak curve P1 is sharp, indicating that there is relatively little blurring of the image. On the other hand, since the second edge portion 12 has a long distance from the in-focus position, the peak curve P2 is a gentle curve, indicating that the image is relatively blurred.

  When the peak value (edge value) is obtained by applying the same analysis function to the peak curves P1 and P2 of the first edge portion 11 and the second edge portion 12 having different degrees of blur as described above, the fitting accuracy is improved. As a result, the edge detection accuracy decreases. Therefore, in this embodiment, the analysis function used for fitting is changed according to the degree of blurring of the image in the portion where the edge is detected. For example, a quadratic function as shown in FIG. 6A is applied to an edge detection location where a relatively sharp peak curve P1 is detected as shown in FIG. As shown in FIG. 6B, a Gaussian function as shown in FIG. 6B is applied to an edge detection location where a peak curve P2 having a relatively gentle curve is detected. This eliminates the problem of reduced edge detection accuracy due to the difference in distance from the in-focus position.

  Subsequently, the edge detection process executed in the personal computer 30 in the edge detection process step of Step # 8 will be described. FIG. 12 is a flowchart showing an edge detection process flow in the personal computer 30.

  When processing is started, two-dimensional image data is read from the data buffer 351 (see FIG. 3) under the control of the processing control unit 359, and the two-dimensional image is detected by the filter processing unit 353 and the edge candidate pixel detection unit 354. A process for detecting an edge candidate pixel that is a candidate for determining an edge existing in is performed (step S11). Details of the processing will be described with reference to FIG. 15. In short, processing for extracting pixels corresponding to a portion where the amount of change in pixel output is large (portion where the brightness is suddenly changed in the image) as edge candidate pixels. Do.

  FIG. 13 is a schematic diagram for explaining processing from identification of edge candidate pixels to edge detection. The two-dimensional image data 40 is composed of pixels 41 arranged in a matrix. In FIG. 13, the shading of each pixel 41 represents the magnitude of the pixel output. In this case, for example, a portion where the pixel output changes in the left-right direction is extracted as the edge candidate pixel 41E. Then, for example, the change degree of the pixel output of the edge candidate pixel 41E is fitted with an analysis function, and the change peak (vertex) is detected as the edge value 51. A line segment connecting such edge values 51 is an edge 52. The edge candidate pixel detection process in step S11 is a process that is the beginning of a series of edge detection processes as described above, and edge candidate pixels are detected over the entire acquired two-dimensional image.

  Next, the analysis function deriving unit 355 selects which analysis function is to be applied to each detected edge candidate pixel to perform fitting (step S12). The detailed processing contents will be described with reference to FIG. 16, but a process of assigning an analysis function corresponding to the distance from the in-focus position or the like for each edge candidate pixel is executed.

  After that, the fitting processing unit 357 performs an analysis function fitting process for performing an approximation process such as a least square method on each edge candidate pixel using each selected analysis function (step S13). FIG. 14 is an explanatory diagram showing an example of such a fitting process. First, as shown in FIG. 14A, the differential value of the edge candidate pixel is obtained based on the differential image obtained by differentiating the two-dimensional image data by the differential processing unit 356. Then, the fitting function f is obtained by performing approximate calculation so that the selected analytical function fits the differential value.

  Thereafter, the edge value detection unit 358 performs an operation for obtaining the vertex of the fitting function f (step S14). This vertex is treated as an edge value 51 in one edge candidate pixel. Then, it is confirmed whether or not the edge detection process has been executed for all the edge candidate pixels detected in step S11 by the process control unit 359 (step S15), and if not completed (NO in step S15), step Returning to S13, the process is repeated. On the other hand, when the edge detection process for all edge candidate pixels is completed (YES in step S15), the process control unit 359 ends the process.

  FIG. 15 is a flowchart showing details of the edge candidate pixel detection process in step S11 shown in FIG. First, the two-dimensional image data is read from the data buffer 351 by the filter processing unit 353, and a filter process for extracting edge candidates included in the two-dimensional image is executed (step S21).

  Next, the edge candidate pixel detection unit 354 detects edge candidate pixels based on the image data after the filter processing (step S22). Specifically, address information of a pixel whose filter value after the filter processing exceeds a predetermined threshold is detected as an edge candidate pixel. As shown in FIG. 13, the edge candidate pixels are detected as a group of pixels 41 (edge candidate pixels 41E) existing in the vicinity of the edge. Then, it is confirmed whether or not the edge candidate pixel detection processing has been performed for all the pixels (step S23), and if not completed (NO in step S23), the processing returns to step S22 and the processing is repeated. On the other hand, when the calculation for all the pixels is completed (YES in step S23), the process proceeds to the next analysis function selection process (step S12). At this time, numbering (E1, E2,... En) is performed on each detected edge candidate pixel.

  FIG. 16 is a flowchart showing details of the analysis function selection processing in step S12 shown in FIG. In this step, first, the analysis function deriving unit 355 acquires distance information of the plurality of edge candidate pixels E1 to En detected as edge candidates from the three-dimensional data calculation unit 352 (step S31). Further, lens information (focal length information, lens F value information) of the optical system of the light receiving unit 22 is acquired from the data buffer 351 (step S32). Further, a differential image (differential value data of pixel output) created in the differential processing unit 356 is acquired (step S33).

  Subsequently, the counter is set to the edge candidate E = E1 (step S34), and one of the plurality of detected edge candidate pixels E1 to En (the first edge candidate pixel E1 in the processing order) is detected. It is targeted for processing. Then, a distance value from the in-focus position is obtained for the pixels constituting the edge candidate E based on the distance information acquired in step S31 (step S35). The analysis function deriving unit 355 performs collation processing on the analysis function storage unit 341 or the derived function storage unit 342 of the RAM 34 using the distance value and the focal length information and / or lens F value information as reference information (step S36). . For example, a process of collating with a table as shown in FIG. 9 is performed. Then, an analysis function corresponding to the distance value, focal length, and lens F value is selected, and an analysis function to be applied to the edge candidate pixel E1 is determined (step S37).

  Thereafter, the analysis function deriving unit 355 confirms whether or not the selection of the analysis function has been completed for all the edge candidate pixels, that is, whether E = En (the edge candidate pixel En whose processing order is final). S38). If not completed (NO in step S38), the counter is incremented as E = E + 1 (step S39), and the edge candidate pixel with the next processing order (the second edge candidate pixel E2 with the processing order) is processed. As a target, the processing from step S35 onward is repeated. On the other hand, when the processing for all the edge candidate pixels is completed (YES in step S38), the process proceeds to the next analysis function fitting process (step S13).

  The embodiment of the edge detection apparatus S according to the present invention has been described above, but the present invention is not limited to this, and for example, modified embodiments such as the following [1] to [4] can be taken. .

[1] In the above-described embodiment, an example in which the imaging device 20 obtains two-dimensional image data (three-dimensional data) of an object by scanning and irradiating slit light. Instead, a pattern light projection method may be employed in which a slit mask or the like is disposed in front of a light source such as a halogen lamp and pattern light is projected onto the object.

[2] In the above embodiment, an example in which the focal length and the lens F value are referred to as the lens information of the optical system of the light receiving unit 22 has been described. However, for example, the exit pupil position, the aperture value, the peripheral light amount state, and the like May also be referred to.

[3] In the above-described embodiment, the example in which the fitting process is performed after the analysis function to be applied to each edge candidate is determined to be either a quadratic function or a Gaussian function is shown. Instead, it is possible to perform fitting processing on one edge candidate using both a quadratic function and a Gaussian function, evaluate both approximation levels, and select a fitting result with a higher approximation level.

[4] In the above-described embodiment, the example in which the edge detection device S is configured by the imaging device 20 and the personal computer 30 has been described. However, the imaging device 20 is provided with a three-dimensional data calculation function, an edge detection function, and the like. The edge detection device S may be configured by 20 units. Furthermore, it can also be provided as a program for executing the edge detection method performed by the edge detection device S described above. Such a program can be recorded on a computer-readable recording medium such as a flexible disk, a CD-ROM, a ROM, a RAM, and a memory card attached to the computer and provided as a program product. Alternatively, the program can be provided by being recorded on a recording medium such as a hard disk built in the computer. A program can also be provided by downloading via a network.

It is a block diagram which shows simply one Embodiment of the hardware constitutions of the edge detection apparatus S which concerns on this invention. 2 is a block diagram showing an electrical configuration of an edge detection device S. FIG. 2 is a functional block diagram showing a functional configuration of a personal computer 30. FIG. It is a process flowchart which shows an edge detection process flow. It is a top view which shows an example of the chart for calibration. It is explanatory drawing which shows an example of an analytic function, Comprising: (a) has shown the quadratic function, (b) has shown the Gaussian function, respectively. It is a table format figure which shows an example of the table which derives | leads-out an analysis function. It is a table format figure which shows an example of the table which derives | leads-out an analysis function. It is a table format figure which shows an example of the table which derives | leads-out an analysis function. It is a top view which shows an example of the two-dimensional image data of a target object. (A), (b) is a graph which shows the differential value of the pixel output around an edge. 4 is a flowchart showing an edge detection processing flow in the personal computer 30. It is a schematic diagram for demonstrating the process from the identification of an edge candidate pixel to edge detection. (A), (b) is explanatory drawing which shows an example of a fitting process. It is a flowchart which shows the detail of the edge candidate pixel detection process of step S11 shown in FIG. It is a flowchart which shows the detail of the analysis function selection process of step S12 shown in FIG.

10 object 20 imaging device (imaging means)
DESCRIPTION OF SYMBOLS 21 Light projection part 22 Light reception part 23 Optical system drive part 24 Image pick-up element 25 Timing generator 26 Output processing circuit 27 Lens information output part (lens information acquisition means)
28 Data memory 29 Measurement control unit 30 Personal computer 34 RAM
341 Analysis function storage unit (analysis function storage means)
342 Derived function storage unit (derived function storage means)
35 CPU
351 Data buffer 352 Three-dimensional data calculation unit (distance calculation means)
353 Filter processing unit 354 Edge candidate pixel detection unit (edge candidate specifying means)
355 Analysis function deriving unit (analysis function deriving means)
356 Differential processing unit (differential image generating means)
357 Fitting processing unit 358 Edge value detection unit (edge detection means)
359 Processing control unit S Edge detection device

Claims (19)

Obtaining two-dimensional image data about the object by photographing the object using a predetermined optical system ;
Obtaining distance information from the shooting position to the object at various points in the image based on the two-dimensional image data;
Determining an analytic function to be applied to each of the edge detection locations for detecting edges present in the image according to the distance information;
Applying each of the determined analysis functions to each of the edge detection locations to detect edges , and
Detecting the edge comprises:
Creating a differential image by differentiating the two-dimensional image data;
Fitting an analysis function to be applied to each of the edge detection points to the differential image;
Detecting an apex of the fitted analytic function as an edge, and an edge detection method. Edge detection method according to claim 1, the pre-Symbol analysis function, characterized in that it comprises a step of deriving, based on the pre-designed value of the optical system. The pre-Symbol analysis function, edge detection method according to claim 1, characterized in that it comprises a step of deriving from advance the actual result of the sample taken with the optical system.   The edge detection method according to claim 1, further comprising a step of tabulating the analysis function according to the distance information.   The edge detection method according to claim 1, further comprising a step of obtaining a derivation function for deriving the analytic function according to the distance information. In determining the pre-Symbol analysis function, in addition to the distance information, the edge detecting method according to claim 1, characterized in that predetermined lens information about the photographing lens is utilized.   The edge detection method according to claim 6, wherein focal length information and / or lens F value information of the photographing lens is used as the lens information.   8. The method according to claim 6, further comprising: deriving information on a degree of blurring of an image accompanying a deviation from a focus position based on the design value of the optical system based on the distance information and the lens information. The edge detection method described in 1.   Including a step of deriving information on the degree of blurring of the image accompanying the deviation from the in-focus position in association with the distance information and the lens information from an actual measurement result obtained by performing sample photographing using the optical system in advance. The edge detection method according to claim 6 or 7.   The edge detection method according to claim 6, further comprising a step of making the analysis function into a table in association with the distance information and the lens information.   The edge detection method according to claim 6, further comprising a step of obtaining a derivation function for deriving the analytic function in accordance with the distance information and the lens information. As the analysis function, the edge detection method according to any one of claims 1 to 11, characterized in that at least the Gaussian function and / or secondary function is used. Imaging means for acquiring two-dimensional image data about the object;
The distance information from the position of photographing by the image pickup means to said object, a distance calculation means for calculating the various locations within the two-dimensional image data based rather in the image,
Edge candidate specifying means for specifying an edge candidate as a candidate for determining an edge existing in the two-dimensional image;
An analytic function deriving unit for deriving an analytic function to be applied to each of the edge candidates according to the distance information;
Edge detection means for detecting an edge by applying the derived analytic function to each of the edge candidates ;
Differential image generation means for creating a differential image by differential processing the two-dimensional image data,
The edge detection unit is characterized in that an analytic function derived for each of the edge candidates is fitted to the differential image and a vertex of the fitted analytic function is detected as an edge. About the taking lens used for the image taking means, it is provided with lens information obtaining means for obtaining at least focal length information and / or lens F value information at the time of photographing,
14. The edge detection according to claim 13 , wherein the analytic function deriving unit derives an analytic function by referring to focal length information and / or lens F value information of the photographing lens in addition to the distance information. apparatus. An analysis function storage means for holding the analysis function in a table format according to the distance information;
The analyzing function deriving means and collates the table of the distance information and the analyzing function storing means of the edge candidates, to claim 13, wherein the deriving an analysis function applied to each of the edge candidate The edge detection apparatus as described. Derivation function storage means for holding a derivation function for deriving the analytic function according to the distance information;
The edge detection apparatus according to claim 13 , wherein the analytic function deriving unit derives an analytic function to be applied to each of the edge candidates by giving distance information of the edge candidate to the derived function. Analysis function storage means for holding the analysis function in a table format according to the distance information and the focal length information and / or lens F value information;
The analysis function deriving unit collates the distance information of the edge candidate, the acquired focal length information and / or lens F value information, and the table of the analysis function storage unit, thereby each of the edge candidates. The edge detection apparatus according to claim 14 , wherein an analytic function to be applied to is derived. Derivation function storage means for holding a derivation function for deriving the analytic function according to the distance information and the focal length information and / or lens F value information;
The analytic function deriving unit derives an analytic function to be applied to each of the edge candidates by giving the distance information of the edge candidate and the acquired focal length information and / or lens F value information to the derived function. The edge detection apparatus according to claim 14 , wherein: To a computer capable of data processing for 2D image data,
Reading two-dimensional image data about an object photographed using a predetermined optical system ;
Based on the two-dimensional image data, obtaining distance information from the shooting position to the object at various points in the image based on the two-dimensional image data ;
Identifying edge candidates that are candidates for determining edges present in the two-dimensional image;
Deriving an analysis function to be applied to each of the edge candidates according to the distance information;
Applying each of the derived analytic functions to each of the edge candidates to detect edges ; and
As the step of detecting the edge,
Creating a differential image by differentiating the two-dimensional image data;
Fitting an analysis function to be applied to each of the edge detection points to the differential image;
Detecting an apex of the fitted analytical function as an edge; and executing an edge detection program.

Images