JP2007298376A - Method and device for determining boundary position, program for making computer function as boundary position determination device, and recording medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To accurately detect the position of an edge of a panel formed on a glass substrate. <P>SOLUTION: A process executed by an edge position detection device includes: a step S710 for executing an initial setting based on an instruction input from the outside to the device; a step S720 for extracting an image corresponding to a target region from an image representing the whole of the glass substrate; a step S730 for generating image data generated by emphasizing the edge of the extracted image; a step S740 for generating one-dimensional projection accumulation data by integrating date representing the image with the edge emphasized in accordance with the direction of a projection accumulation direction; a step S750 for determining a target region for calculating a statistics value; and a step S760 for calculating the statistics value based on the determined region, and calculating a threshold value based on the statistics value; and a step S770 for determining the position of a panel edge by using the threshold value. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a technique for determining the position of a boundary, and more specifically to a technique for determining the position of a boundary included in an image acquired by photographing a subject.

  The inspection process in the manufacturing process plays an important role as a quality control method for manufacturing a high-quality thin panel. In recent years, various defect inspection apparatuses to which image processing technology is applied have been developed in order to stabilize quality and reduce costs by reducing visual inspectors. In general, image processing requires a longer processing time as the processing area is larger. For this reason, the processing time is shortened by inspecting the image using only the necessary area.

  On the other hand, as a method for producing a thin panel, it is common to produce a plurality of panels from a single glass substrate. In such a defect inspection of a glass substrate, an image acquired by imaging with a camera is used. In this defect inspection, as a proposal for shortening the image inspection processing time, there is a method of inspecting only the inside of each panel and not inspecting other portions.

  In order to inspect only the inside of the panel, it is necessary to determine the boundary of the panel region in the image of the glass substrate, that is, the edge position of the panel.

  As a method of determining the edge position, there is a method of using design data of the edge position. Specifically, the position of the edge in the image is calculated from the position of the edge with respect to the edge of the glass substrate and the positional relationship between the camera and the glass substrate.

  However, this method generally causes an error of about 10 mm due to positioning errors between the camera and the glass substrate. Therefore, it is necessary to determine the edge position by image processing.

  As a method for determining an edge position by image processing, for example, Japanese Patent Laid-Open No. 5-266194 (Patent Document 1) discloses a technique capable of obtaining an end point of a straight line portion of a contour line without receiving interference due to noise. ing. According to this technique, first, an image is gray-projected in the direction of an edge to obtain one-dimensional projection accumulated data. Next, the gradient magnitude is obtained from the one-dimensional projection accumulated data, and a position where the gradient magnitude reaches a peak is detected as an edge position.

  Moreover, as a technique for detecting a connection portion of a plurality of parts displayed in an image, for example, Japanese Patent Application Laid-Open No. 2005-284725 (Patent Document 2) discloses the surface of a material to be joined such as an extrusion die material or a plate material to be joined together. Even if there is a disturbance factor such as dirt or scratches, a technique that can accurately detect the joining line is disclosed.

  As a technique for detecting an edge, for example, Japanese Patent Application Laid-Open No. 2004-272313 (Patent Document 3) discloses a technique for quantifying an image with poor edge quality and detecting the edge with high accuracy. .

Further, regarding the extraction of the edge of a subject, for example, Japanese Patent Laid-Open No. 2005-167802 (Patent Document 4) reads an original image by an area sensor, extracts the edge of the original by image processing from image data, and A technique for detecting a document size based on a contour and a position is disclosed.
Japanese Patent Laid-Open No. 5-266194 JP 2005-284725 A JP 2004-272313 A JP 2005-167802 A

  Consider a case in which the panel edge of a panel on a glass substrate is determined using the first technique disclosed in Japanese Patent Laid-Open No. 5-266194. A wiring pattern region exists around the panel. The outside of the wiring pattern region is a glass substrate.

  When a glass substrate having such a configuration is imaged by a camera, the panel edge, the edge of the wiring pattern, and the edge of the glass substrate are output on the image data from the inside of the panel to the outside.

  According to the first technique, by integrating the pixel values of the original image along a plurality of directions, one-dimensional projection accumulation data is obtained, and a peak value of the projection gradient data is obtained for each of the plurality of the plurality of the plurality of directions. Among these directions, the peak value in the largest direction is set as the edge position.

  According to this technique, the position of the pattern edge is selected and the position of the panel edge may not be selected, and there is a problem that the edge cannot be determined with high accuracy.

  Japanese Patent Laid-Open No. 5-266194 discloses a second technique for obtaining a two-dimensional gradient distribution of an original image before obtaining one-dimensional projection accumulation data.

  As a method for obtaining the two-dimensional gradient distribution, there is a method using image data after performing the Sobel filter processing. Also in this case, as in the method according to the first technique described above, the position of the pattern edge is selected and the position of the panel edge is determined even if the peak position of the gradient magnitude is obtained from the one-dimensional projection accumulated data. May not be selected. As a result, the panel edge may not be determined with high accuracy.

  The present invention has been made to solve the above-described problems, and a first object of the present invention is to provide a position determination device that can accurately determine the position of a boundary. The second object is to provide a boundary position determination device capable of improving the efficiency of image processing inspection of an inspection object.

  A third object is to provide a method capable of accurately determining the position of the boundary. A fourth object is to provide a boundary position determination method capable of improving the efficiency of image processing inspection of an inspection object.

  A fifth object is to provide a program for causing a computer to function as an apparatus that can accurately determine the position of a boundary. A sixth object is to provide a program for causing a computer to function as a boundary position determination device capable of improving the efficiency of image processing inspection of an inspection object.

  A seventh object is to provide a recording medium storing a program for causing a computer to function as an apparatus that can accurately determine the position of a boundary. The eighth object is to provide a recording medium storing a program for causing a computer to function as a boundary position determination device capable of improving the efficiency of image processing inspection of an inspection object.

  In order to solve the above problems, according to one aspect of the present invention, a boundary position determining apparatus is provided. This apparatus includes storage means for storing original image data representing an original image acquired by photographing a subject. The original image data is represented according to a two-dimensional coordinate axis. When the original image has a plurality of partial images, the apparatus includes first generation means for generating enhanced image data representing an image in which the boundary of each partial image is enhanced based on the original image data, and the boundary is A plurality of brightness values of each partial area corresponding to each partial image after being emphasized are integrated in the direction of the boundary, and an average value is calculated, thereby representing a plurality of values expressed in accordance with a one-dimensional coordinate axis of the two-dimensional coordinate axes. A first generation unit for generating one-dimensional data, a first determination unit for determining a range of data to be determined for determining a boundary position based on the design data of the subject, and a data range A calculation means for calculating a threshold value for a luminance value by calculating a statistic using a plurality of included one-dimensional data, and a luminance value exceeding the threshold based on the plurality of one-dimensional data included in the data range Vs. To comprising specifying means for specifying a 1-dimensional data, a coordinate value on a one-dimensional coordinate axis of the one-dimensional data specified by the specifying means, a second determination means for determining a position of the boundary.

  Preferably, the first determining means includes means for determining a data range based on a range determined by the original image data and a range specified by the design data.

  Preferably, the means for determining the data range calculates one-dimensional data of the boundary as the calculated position data based on the design data, and 1 of the boundary as the actual position data based on the original image data. Dimension data is calculated, and the data range is determined based on the difference between the calculated position data and the actual position data.

  Preferably, the subject includes a first part to be inspected and a second part not to be inspected. The original image includes a first image corresponding to the first portion and a second image corresponding to the second portion. The first generation means generates data representing an image in which the boundary between the first image and the second image is emphasized as the emphasized image data.

  Preferably, the subject is a glass substrate. The first part is a panel region processed into a glass substrate. The first generation means generates data representing an image in which a portion where an image corresponding to the panel area and an image corresponding to an area other than the panel area are in contact is emphasized.

  Preferably, the calculation means calculates an average value and a standard deviation for a plurality of one-dimensional data, and calculates a threshold value based on the average value and the standard deviation.

  Preferably, the specifying means specifies a plurality of one-dimensional data corresponding to a luminance value exceeding a threshold, and for each of the specified one-dimensional data, a difference between the upper limit value of the data range and the one-dimensional data. And means for determining one-dimensional data that minimizes each calculated difference as one-dimensional data corresponding to a luminance value exceeding a threshold value.

  Preferably, the original image includes a rectangular area. The boundary is formed by the outline of the rectangular area. The first generation means generates data representing an image in which the boundary formed by the contour is emphasized as the emphasized image data.

  Preferably, the second generation means integrates the luminance values of the pixels corresponding to the rectangular area after the boundary is emphasized in one direction in the rectangular area, and calculates an average value, thereby calculating a plurality of values. Generate one-dimensional data. The first determining means determines the data range so that the range defined by the one-dimensional data for the rectangular area includes the data range.

  Preferably, the statistic includes an average value and a standard deviation. The calculating means calculates the threshold value using each formula defined in advance according to each value of the plurality of standard deviations. Each formula varies depending on the value of each standard deviation.

  According to another aspect of the invention, a method is provided for a computer to determine the location of a boundary included in an image. The computer includes a storage device that stores data and a processing device that executes a predetermined process using the data. In this method, the processing device includes a step of reading original image data representing an original image acquired by photographing a subject from a storage device. The original image data is represented according to a two-dimensional coordinate axis. In this method, when the original image has a plurality of partial images, the processing device generates enhanced image data representing an image in which the boundary of each partial image is emphasized based on the original image data; Is obtained by integrating the luminance values of the partial areas corresponding to the partial images after the boundary is emphasized in the direction of the boundary and calculating an average value, thereby calculating a one-dimensional coordinate axis out of the two-dimensional coordinate axes. Generating a plurality of one-dimensional data represented according to the following: a processing device determining a range of data for determining a boundary position based on subject design data; and a processing device, Calculating a threshold value for a luminance value by calculating a statistic using a plurality of one-dimensional data included in the data range; and a processing device including a plurality of primary elements included in the data range. Identifying one-dimensional data corresponding to a luminance value exceeding a threshold value based on the data, and determining a coordinate value on the one-dimensional coordinate axis of the identified one-dimensional data as a boundary position With.

  According to still another aspect of the present invention, a program for causing a computer to function as an apparatus for determining the position of a boundary included in an image is provided. The computer includes a storage device that stores data and a processing device that executes a predetermined process using the data. The program causes the processing device to execute a step of reading original image data representing an original image acquired by photographing a subject from the storage device. The original image data is represented according to a two-dimensional coordinate axis. When the original image has a plurality of partial images, the program generates, on the basis of the original image data, enhanced image data representing an image in which the boundary of each partial image is emphasized, and the boundary is enhanced. Then, the luminance values of the partial areas corresponding to the partial images after being added are integrated in the direction of the boundary, and an average value is calculated, whereby a plurality of 1's represented according to the one-dimensional coordinate axis out of the two-dimensional coordinate axes. Using a step of generating dimensional data, a step of determining a range of data to be determined for determining the position of the boundary based on the design data of the subject, and a plurality of one-dimensional data included in the range of data By calculating a statistic, a step of calculating a threshold value for the luminance value and corresponding to a luminance value exceeding the threshold value based on a plurality of one-dimensional data included in the data range Identifying a dimension data, a coordinate value on a one-dimensional coordinate axis of the one-dimensional data is identified, and a step of determining a position of the boundary.

  According to still another aspect of the present invention, a computer-readable recording medium storing the above program is provided.

  According to the boundary position determining apparatus according to the present invention, the position of the boundary (for example, the edge portion of the panel) can be determined with high accuracy. According to the method of the present invention, the computer can accurately determine the position of the boundary. According to the program according to the present invention, the computer functions as an apparatus that can accurately determine the position of the boundary. When a recording medium according to the present invention is mounted on a computer and a program stored in the recording medium is executed, the computer that executes the program functions as an apparatus that can accurately determine the position of the boundary. be able to. As a result, the efficiency of the image processing inspection of the inspection object can be improved.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. In the following description, the same parts are denoted by the same reference numerals. Their names and functions are also the same. Therefore, detailed description thereof will not be repeated.

  With reference to FIG. 1, an inspection system 10 according to an embodiment of the present invention will be described. FIG. 1 is a block diagram showing a hardware structure of an inspection system 10 having a function of determining an edge position. The inspection system 10 is electrically connected to the light 110, the edge position detection device 100 electrically connected to the light 110, the camera 120 electrically connected to the edge position detection device 100, and the edge position detection device 100. And a connected display 130. The edge position detection apparatus 100 detects the position of a boundary (edge) between a panel region formed on the glass substrate and a region other than the panel region. Specifically, as described later, the edge position detection device 100 detects the position of the boundary as a coordinate value on a coordinate axis defined in advance.

  A glass substrate 140 that is an object to be inspected is carried into the inspection system 10 and mounted at a predetermined position. Two thin panels 141 and 142 are formed on the glass substrate 140, respectively.

  Based on a signal from the edge position detection device 100, the light 110 irradiates the glass substrate 140 mounted at the position with light having a preset intensity. The irradiated light is, for example, general white light.

  The camera 110 receives and captures the reflected light from the glass substrate 140 and sends it to the edge position detection device 100 as image data. The camera 120 may be an area sensor camera or a line sensor camera. The imaging part of the camera 120 has a photoelectric conversion mechanism such as a charge coupled device (CCD) or a complementary metal oxide semiconductor (CMOS).

  The image size described in the present embodiment is, for example, horizontal 1000 pixels and vertical 1200 pixels. However, the size of the image is not limited to the above, and may be other sizes (for example, horizontal 500 pixels, vertical 600 pixels).

  In addition, the luminance value of the pixel of the image described in this embodiment is a value in the range of “0” to “255”. In the following description, when the luminance value is “0”, the corresponding pixel corresponds to black. When the luminance value is “255”, the corresponding pixel corresponds to white. Note that the luminance value of the image pixel used in the present embodiment is not limited to a value included in the range of “0” to “255”, and may be a value included in another range. .

  With reference to FIG. 2, the image of the glass substrate 140 image | photographed with the camera 120 is demonstrated. FIG. 2 is a diagram illustrating an image 240 of the glass substrate 140. The image 240 includes panel edges 211, 212, 213, 214 formed at the boundary between the panel 210 and the glass substrate, and a wiring pattern region 215 formed inside the glass substrate 140 so as to surround the panel 210.

  Similarly, for the panel 220, an image representing a wiring pattern area formed between the glass substrate 140 and the panel 220 and a panel edge surrounding the panel 220 in a rectangular shape is generated. Furthermore, a wiring pattern edge (for example, a wiring pattern edge 216) is generated between the wiring pattern region 215 and the glass substrate 140. A region 206 is an image corresponding to a region to be inspected in the image 240.

  Therefore, the region 206 will be described with reference to FIG. FIG. 3 is a diagram showing an image 300 of the region 206.

  The image 300 shows an image edge 201 corresponding to the edge of the area 206, a wiring pattern edge 216, a panel edge 211, and noise 230. The accumulation direction shown in FIG. 3A represents the direction in which the luminance values included in each region are accumulated. This direction is defined in advance with respect to the glass substrate 140. For example, when the glass substrate 140 is rectangular, a direction parallel to any one side is defined as the accumulation direction.

  FIG. 3B is a diagram illustrating values calculated by integrating the luminance values corresponding to the image 300 in the accumulation direction. Specifically, for the region 310, the luminance value in each region is expressed as at least 160 or more when the X coordinate value is between 0 and a. For the region 320, the luminance value is expressed as 180 when the X coordinate value is between a and b. Further, regarding the region 330, the luminance value is represented as 0 in the range where the X coordinate value is b or more.

  With reference to FIG. 4, the image produced | generated by the filter process is demonstrated. FIG. 4 is a diagram illustrating an edge-enhanced image 400 generated by executing the Sobel filter process on the image corresponding to the region 206 in the image 240.

  As shown in FIG. 4A, the edge enhancement image 400 corresponds to an image 410 corresponding to a part of the panel area, an image 420 corresponding to a part of the wiring pattern area, and a part of the glass substrate. Image 430. The edge enhanced image 400 further includes an image 211 corresponding to the panel edge, an image 216 corresponding to the wiring pattern edge, and noise 230. An accumulation direction 450 represents a direction in which luminance values of pixels corresponding to the respective images are integrated.

  FIG. 4B is a diagram showing a result (integrated value) obtained by integrating luminance values according to the accumulation direction 450. For example, the integrated value for the image 410 corresponding to the normal area in the panel area is “0”. The integrated value of the noise 230 is “20”. The integrated value for the image 211 corresponding to the panel edge is “40”. The integrated value for the image 420 corresponding to a part of the wiring pattern region is “0”. The integrated value for the image 216 corresponding to the wiring pattern edge is “127”. The integrated value for the image 430 corresponding to the glass substrate is “0”.

  With reference to FIG. 5, the edge position detection apparatus 100 which comprises the test | inspection system 10 is demonstrated. FIG. 5 is a block diagram showing a configuration of each function realized by the edge position detection apparatus 100. The edge position detection apparatus 100 includes a design data input unit 510, an image input unit 520, a temporary storage unit 530, a storage unit 540, and a processing unit 550. The processing unit 550 includes a writing unit 551, an initial setting unit 552, a coordinate acquisition unit 553, an extraction unit 554, an image generation unit 555, an integration unit 556, a data range determination unit 557, and a threshold calculation unit 558. A position determination unit 559 and a display control unit 560.

  Design data input unit 510 accepts data input representing design specifications of panels 141 and 142 formed on glass substrate 140. This data includes, for example, the size information of the glass substrate 140, the size information of the panels 141 and 142, or data defining the positional relationship between the panels 141 and 142 and the glass substrate 140.

  The image input unit 520 receives input of image data output from the camera 120. Temporary storage unit 530 temporarily (volatilely) holds data received by design data input unit 510 or image input unit 520 or data generated by processing unit 550. Temporary storage unit 530 is realized as a RAM (Random Access Memory) or a frame buffer, for example. Note that the temporary storage unit 530 is not limited to being physically implemented as a single memory, and may be configured by being divided into two or more memories.

  The writing unit 551 reads data held in the temporary storage unit 530 and writes the data in an area secured in the storage unit 540. The storage unit 540 is a storage device that can hold data in a nonvolatile manner. The initial setting unit 552 is an initial unit for performing an inspection process to be performed on the glass substrate 140 based on data stored in the storage unit 540 and an external instruction input to the edge position detection device 100. The set value is read from the storage unit 540 and stored in an area secured in the temporary storage unit 530.

  Storage unit 540 stores image data generated by photographing glass substrate 140 (for example, data representing image 240 shown in FIG. 2). The storage unit 540 is realized as a flash memory or a hard disk device, for example.

  The coordinate acquisition unit 553 acquires the coordinates of the panel edges 211, 212, 213, and 214 based on the image data of the glass substrate 140 stored in the storage unit 540. In addition, from the viewpoint of promoting the inspection of image processing, the coordinates of any specific panel edge (for example, the panel edge 211) may be acquired instead of the mode of acquiring the coordinates of each panel edge.

  The extraction unit 554 extracts the image of the region 206 (that is, the partial image divided by the boundary) from the image 240 based on the data stored in the storage unit 540. Specifically, first, the storage unit 540 stores coordinate values for specifying the area 206 set by the initial setting unit 552. The extraction unit 554 refers to each coordinate value, determines an area corresponding to the coordinate value from the image 240, and extracts data included in the determined area. The extraction unit 554 stores the extracted data in the temporary storage unit 530 (extracted image data 531).

  The image generation unit 555 executes processing for enhancing the contour of the image based on the image data stored in the temporary storage unit 530. Specifically, the image generation unit 555 uses the extracted image data 531 to use filter processing, which is a general image conversion method in image processing, and based on parameters input in advance as initial setting values. The filter process is executed. The filter processing to be executed is, for example, a 3 × 3 Sobel filter that is a general edge extraction filter. The image generation unit 555 stores the image data generated by executing the filter processing in the temporary storage unit 530 (image data 532 with edge enhancement).

  The accumulating unit 556 accumulates the luminance values of the pixels constituting the image in one predetermined direction using the data stored in the temporary storage unit 530. Specifically, the integration unit 556 integrates the luminance values of the pixels having the same coordinate value with respect to the X coordinate defined in the image 240 in a predefined accumulation direction (for example, the negative direction of the Y axis). Further, the integrating unit 556 calculates an average value of luminance values for pixels having the same X coordinate value by dividing the integrated value by the number of integrated pixels. The accumulating unit 556 writes the primary projection data 533 thus calculated in the temporary storage unit 530.

  The data range determination unit 557 determines the range of data that is a target for determining the position of the edge 211 using the primary projection data 533 stored in the temporary storage unit 530. When the data range determining unit 557 determines the data range, the data range determining unit 557 stores the data range in the area secured in the temporary storage unit 530 as data 534 representing the data range.

  The threshold value calculation unit 558 calculates a threshold value by executing a predetermined statistic calculation process for the range defined by the data 534 representing the data range, using the primary projection data 533. The threshold value according to the present embodiment is a luminance value used for determination to determine the position of the edge 211. When the threshold value calculation unit 558 determines the threshold value, the threshold value calculation unit 558 writes the threshold value in the temporary storage unit 530 (threshold value 535).

  The position determination unit 559 determines the position (X coordinate value) of the panel edge 211 based on the primary projection data 533 and the threshold value 535. For example, the position determination unit 559 specifies the primary projection data 533 for the luminance value exceeding the threshold 535, and determines the X coordinate value of the specified primary projection data 533 as the position of the panel edge 211. The position determination unit 559 writes the position data determined in this way into the temporary storage unit 530 (data 536 representing the position of the edge).

  Display control unit 560 reads the data stored in temporary storage unit 530 based on the display instruction input via setting data input unit 510, and sends the data to the display unit (for example, display 130). The captured image is displayed.

  With reference to FIG. 6, the data structure of the edge position detection apparatus 100 is demonstrated. FIG. 6 is a diagram conceptually showing one mode of data storage in storage unit 540. Storage unit 540 includes memory area 602 to memory area 630 for storing data.

  Image data acquired by photographing the glass substrate 140 on which the panels 141 and 142 are formed is recorded in the memory area 602. An offset value (Poff) input in advance for determining the data range is stored in the memory area 604. Coordinate values input in advance to define the area 206 as an object for detecting the edge position are input to the memory areas 606 and 608, respectively. For example, the coordinate value Plt at the upper left corner of the area 206 is stored in the memory area 606 as (800, 1150). Similarly, the coordinate value Pbr (X, Y) representing the lower right corner of the area 206 is stored in the memory area 608 as (1000, 700).

  A data number Nd defined in advance as the number of data to be integrated for calculating the statistic is stored in the memory area 610. A constant “a” used in the formula for determining the threshold value is stored in the memory area 612. The projection accumulation direction Dp determined for integrating the luminance values of the pixels in one predetermined direction is stored in the memory area 614 as, for example, “data indicating the negative direction of the Y coordinate axis”.

  The parameter values used for executing the Sobel filter process are stored in the memory area 616. Design data determined by panel design for forming panels 141 and 142 on glass substrate 140 is stored in memory area 618. The design data includes, for example, the vertical and horizontal sizes (for example, the length) of the glass substrate 140, the vertical and horizontal sizes (for example, the number of pixels) of the panels 141 and 142, the distance between the glass substrate 140 and the panels 141 and 142, This includes the interval between the panels 141 and 142.

  The X coordinate value (Xd) of the panel edge (for example, panel edge 211) calculated based on the design data is stored in the memory area 620. Before the X coordinate value is calculated, for example, data representing “NULL” is stored. The same applies to other regions. The X coordinate value (Xr) of the panel edge (for example, the panel edge 211) calculated using the image data acquired by photographing with the camera 120 is stored in the memory area 622. The X coordinate value (Xe) of the maximum value of the panel edge (for example, the panel edge 211) calculated from the calculation by the data range determining unit 557 is stored in the memory area 624. Similarly, the minimum value (Xs) of the panel edge is stored in the memory area 626. Data stored in memory areas 624 and 626 corresponds to data 534 representing the data range shown in FIG. 5, for example. When the data range determining unit 557 does not execute the above-described processing, the memory areas 624 and 626 store information (NULL) indicating that no data exists in each area.

  The threshold value Vth is stored in the memory area 628. A program used to realize the above functions of the edge position detection apparatus 100 is stored in the memory area 630. When the program is executed by an arithmetic processing unit such as a processor, each function constituting the processing unit 550 shown in FIG. 5 is realized. The program is configured, for example, by combining modules that realize each function.

  A control structure of the edge position detection apparatus 100 will be described with reference to FIG. FIG. 7 is a flowchart showing a procedure of processing executed by processing unit 550. When the edge position detection apparatus 100 is realized by a computer system, a processor included in the computer system functions as the processing unit 550.

  In step S710, processing unit 550 performs data reading and other initial settings based on an instruction input from the outside to edge position detection device 100. Specifically, the processing unit 550 reads data stored in the storage unit 540 as the initial setting unit 552 and writes the data in the temporary storage unit 530. The data written by the initial setting is, for example, the actual panel edge 211 calculated based on the X coordinate value Xs (211) of the panel edge 211 included in the image 240 and the data of the image 240, calculated by the design data. Among the errors between the X coordinate value Xr (211) and the coordinate value Xs (211), the maximum error Xme (211), the coordinate value specifying the region 206, the offset Poff, the number of data Nd, the constant a, Projection accumulated direction Dp.

  In step S720, processing unit 550 extracts an image corresponding to region 206 from image 240 representing the entire glass substrate 140. Specifically, the processing unit 550 reads out the image data stored in the storage unit 540 as the extraction unit 554, refers to the upper left coordinate value Plt and the lower right coordinate value Prb, and selects a rectangular region to be extracted. The data included in the area is specified and stored in another area secured in the temporary storage unit 530 (extracted image data 531). As an example, in the image 300 shown in FIG. 3A, the area 206 is a rectangular area, and the coordinate value Plt (X, Y) at the upper left end of the area 206 is (800, 1150). The coordinate value Prb (X, Y) at the lower right end of the area 206 is (1000, 700). Therefore, the luminance value of the pixel corresponding to the area included in the area specified by these coordinate values is stored in the temporary storage unit 530.

  In step S730, processing unit 550 generates image data in which the edge of the extracted image is emphasized. Specifically, the processing unit 550, as the image generation unit 555, executes a Sobel filter process defined in advance with reference to the extracted image data 531, thereby enhancing the contour included in the region 206. Generated image data (image data 532). The image generated by the Sobel filter process is, for example, an image 400 shown in FIG.

  In step S740, processing unit 550 integrates the data representing the edge-enhanced image according to the direction of projection accumulation direction Dp to generate one-dimensional projection accumulation data. Specifically, the processing unit 550 reads out the edge-enhanced image data 532 as the integration unit 556 from the temporary storage unit 530, and calculates the pixels having the same X coordinate value along the negative direction of the Y coordinate axis. The luminance value is integrated. Referring to the coordinate values (memory areas 606 and 608) shown in FIG. 6, the size of the image in the Y-axis direction is 450 pixels (= 1150-700). Therefore, 450 luminance values are integrated. In the example of FIG. 4B, the above integration is executed along the accumulation direction 440.

  In step S750, processing unit 550 determines a region for which a statistic is to be calculated. Specifically, the processing unit 550 refers to the primary projection data 533 as the data range determination unit 557 and derives data 534 representing the data range. In the present embodiment, each luminance value is projected according to the Y-axis direction. Therefore, the range is designated by an X coordinate range (Xs to Xe).

  The coordinate value Xr (211) corresponding to the actual panel edge 211 included in the image 240 needs not to be included in the region for calculating the statistic from the viewpoint of accurately determining the position of the edge. However, in reality, due to positioning errors between the camera 120 and the glass substrate 140, the X coordinate value Xs (211) of the panel edge 211 in the image 240 calculated based on the design data, and the image 240 are included. The actual panel edge 211 may not match the X coordinate value Xr (211). In this case, at the stage where step S750 is executed, the value of Xr (211) is unknown. Therefore, the maximum error X coordinate value (Xe) that defines the range of data used for calculating the statistic is subtracted from the X coordinate value Xs (211) from the maximum error Sme (211) and the offset Poff. Thus, the actual panel edge 211 included in the image 240 can be set so as not to be included in the region for calculating the statistics. Similarly, the X coordinate value (Xs) of the lower limit value of the area is calculated by subtracting the data number Nd from the upper limit value Xe.

For example, the calculation using the values shown in FIG. 4B is performed as follows.
Xe = Xs (211) -Xme (211) -Poff
= 950-20-10
= 920.

  Here, “10” is used as the offset value (Poff), but other values may be used. The offset value is determined as follows, for example.

  First, when Poff = 0, Xe = 950-20-0 = 930. This value coincides with the end of the area for calculating the statistic when the actual panel edge 211 has a maximum error shift. Therefore, the offset value may be at least “1” or more.

  In the present embodiment, assuming that there is a gradient as a luminance value that is not noise near the actual panel edge, the offset value is set so that the average value and variance of the luminance values are more stable as non-defective products. “10” is set. In the present embodiment, pixels are used as the unit of the offset value.

  Further, the X coordinate value Xs of the smaller region for calculating the statistic is obtained by subtracting the number of data Nd for calculating the statistic from the upper limit value Xe.

Xs = Xe-Nd + 1
= 920-50 + 1
= 871.

  Here, the larger the value of the data number Nd, the longer the processing time, but the edge detection accuracy is improved. On the other hand, the smaller this value is, the shorter the processing time is, but the processing result is more susceptible to local noise, and the accuracy is reduced. Therefore, the number of data Nd is determined in consideration of the trade-off relationship between processing time and detection accuracy. This determination is made based on, for example, the results of actual processing (processing time and detection accuracy).

  As a result of the above calculation, the region for calculating the statistic in the present embodiment is determined as 871 to 920 for the X coordinate.

  Taking the case of determining the other panel edges 212, 213, 214 into consideration, the above contents will be generally described as follows. First, from the “coordinate value of the panel edge in the image calculated using the design data” to the inside of the panel, “the panel value in the image calculated from the actual coordinate value of the panel edge in the image and the design data The number of pixels corresponding to the “maximum error from the edge coordinate value” is shifted. Next, the position shifted by the number of pixels corresponding to “offset, which is a displacement amount for determining a region for calculating a statistic”, is a boundary on the panel edge side of “region for calculating a statistic”. Furthermore, the position where the number of pixels corresponding to the number of data (Nd) −1 for calculating the statistic is shifted from the “boundary on the panel edge side of the region for calculating the statistic” is the other boundary.

  Referring to FIG. 7 again, in step S760, processing unit 550 calculates a statistic based on the region determined in step S750, and further calculates threshold Vth based on the statistic. The statistic includes, for example, an average value (Vave) and a standard deviation (Vsig) of one-dimensional projection accumulation data included in the region. Specifically, the processing unit 550 derives the threshold value 535 as the threshold value calculation unit 558 with reference to the data 534 representing the data range.

  For example, in the example shown in FIG. 4B, Nd (= 50) one-dimensional values included in the upper limit X coordinate value Xe of the area from the lower limit X coordinate value Xs of the area for calculating the statistic. As for each value of the projection accumulation data, the value of eight data is “20”, and the other values are “0”. In this case, the average value Vave is “3.2”, and the standard deviation Vsig is about “7.4”. Therefore, as shown below, the threshold value Vth is calculated using Equation 1.

Vth = Vave + a × Vsig (1)
= 3.2 + 3.0 × 7.4
= 25.4.

  Here, the value of “3” in 3σ used in the field of quality control is used as the value of the threshold determination constant a.

  Referring to FIG. 7 again, in step S770, processing unit 550 determines the position of panel edge 211 using threshold value Vth. Specifically, the processing unit 550 refers to the threshold value 535 as the position determination unit 559 and derives data 536 representing the position of the edge. For example, from the upper limit value Xe (= 920) on the panel edge 211 side of the glue to calculate the statistic, the threshold value Vth (in the positive direction of the X axis) from the upper limit value Xe (= 920) in FIG. In the example, the coordinates that exceed 25.4) first are specified. In the present embodiment, as shown in FIG. 4B, the X coordinate value exceeding the threshold Vth (25.4) is “940”. As a result, this value coincides with the X coordinate value Xs (211) of the panel edge 211.

  In step S760, (Equation 1) is used as an equation for calculating the threshold value Vth in the present embodiment, but other equations may be used. For example, as described below, the edge position can be determined with higher accuracy by changing the equation for calculating the threshold value Vth according to the value of the standard deviation Vsig, that is, according to the degree of noise.

  As a first case, when almost no noise is included in the image (when the relationship of “Vsig ≦ 1.0” is established), Expression (2) may be used.

Vth = Vave + a (Equation 2)
However, a = 9.0.

  Here, “9.0” is used as the value of threshold determination constant “a”, which is larger than “6” in 6σ used in the field of quality control. A value larger than this value may be used.

  As a second case, when there is some noise in the image (for example, when the relationship of “1.0 <Vsig ≦ 5.0” is established), the expression (3) may be used.

Vth = Vave + a × Vsig (Formula 3)
However, a = 6.0.

  Here, “6” in 6σ used in the field of quality control is used as the value of the threshold determination constant a.

  As a third case, when there is a lot of noise (for example, when the relationship of “5.0 <Vsig” is established), Expression (4) may be used.

Vth = Vave + a × Vsig (Formula 4)
However, a = 3.0.

  By setting the equations and constants in this way, when there is almost no noise (corresponding to the first case), it is possible to avoid erroneous detection of a certain local noise as an edge position. it can. When the noise is large (corresponding to the third case), the difference between the edge strength of the panel edge 211 and the strong noise edge strength is small by reducing the constant a for determining the threshold value Vth. In this case, erroneous detection of the position of the panel edge 211 can be avoided.

  Here, another technique for determining the panel edge of the glass substrate 800 will be described with reference to FIGS. FIG. 8 is a diagram illustrating the appearance of the glass substrate 800. The glass substrate 800 includes two panels 810 and 820.

  A wiring pattern region 815 is formed around the panel 810. Panel edges 811, 812, 813, and 814 exist between the panel 810 and the wiring pattern region 815. The outside of the wiring pattern region 815 is a region where a panel is not formed on the glass substrate 800. The panel 820 is also formed with regions similar to the above-described regions for the panel 810.

  A region 806 is a target for explaining the technology. As is apparent from FIG. 8, the region 806 includes a part of the panel 810, a part of the wiring pattern region 815, and a part of the glass substrate 800.

  FIG. 9 is a diagram illustrating an image 900 acquired when the glass substrate 800 is photographed with a camera. An imaging area 850 for the glass substrate 800 shown in FIG. 8 corresponds to the area 950 in the image 900. The image 900 includes images 910 and 920 of the panel 810, an image 911 of the panel edge 811, and a wiring pattern edge image 916 that is a boundary between the wiring pattern region 815 and the glass substrate 800. An image 906 occupying a part of the image 900 is an image of the area 806 occupying a part of the glass substrate 800.

  Next, the case where the calculation process is executed in the accumulation direction shown in FIG. 10 will be described. FIG. 10 is a diagram for explaining the image 906 shown in FIG. As shown in FIG. 10A, the image 906 includes an image of the panel, a panel edge 1020, a wiring pattern edge 1030, and an end portion 1010 of the image from the panel toward the glass substrate.

  For example, when the technique disclosed in Japanese Patent Laid-Open No. 5-266194 is used, one-dimensional projection accumulation data is obtained by integrating pixel values of an original image along a plurality of directions. The peak value of the projection gradient data is obtained for each. The peak value in the largest direction among the plurality of directions is set as the edge position.

  Specifically, in FIG. 10A, the edge position is clearly the direction in which the peak value of the projection gradient data in the accumulation direction 1040 is the largest. In this case, the one-dimensional projection accumulation data when the pixel value is projected in the accumulation direction 1040 is shown in FIG. Here, the luminance value on the vertical axis of the one-dimensional projection accumulation data shown in FIG. 10B represents the average value of the luminance values. When the peak position of the gradient magnitude is obtained from the one-dimensional projection accumulation data shown in FIG. 10B, the position of the wiring pattern edge 1030 is selected, and the position of the panel edge 1020 is not selected.

  FIG. 11 is a diagram for explaining still another technique. Japanese Patent Laid-Open No. 5-266194 also discloses a technique other than the above technique. The technique is to obtain a two-dimensional gradient distribution of an original image before obtaining one-dimensional projection accumulation data.

  Here, FIG. 11A shows an image obtained by performing Sobel filter processing as a method for obtaining a two-dimensional gradient distribution on the image 906 corresponding to the region 806. In this case as well, the edge position is the projected gradient in the accumulation direction 1140 shown in FIG. 11A, as in the case where the method of obtaining the two-dimensional gradient distribution described with reference to FIG. 10 is applied. It is clear that the peak value of the data is in the largest direction.

  FIG. 11B shows one-dimensional projection accumulation data when the two-dimensional gradient distribution shown in FIG. 11A is projected in the accumulation direction 1140. The luminance value on the vertical axis of the one-dimensional projection accumulation data shown in FIG. 11B represents the average value of the luminance values. Even when the peak position of the gradient magnitude is obtained from the one-dimensional projection accumulation data shown in FIG. 11B, the position of the pattern edge 1130 is selected, and the position of the panel edge 1120 is not selected.

  As described above, according to the conventional technique, the panel edge that is the boundary between the panel and the wiring pattern region, for example, the panel edge 1020 in FIG. 10 and the panel edge 1120 in FIG.

  However, according to the inspection system 10 according to the embodiment of the present invention described with reference to FIGS. 1 to 7, the position of the edge is determined accurately and in a short time based on the following technical features. be able to.

  First, as described in step S750, an area for calculating a statistic is set inside a panel which is a rectangular area, and a threshold value Vth is determined using the statistic of a partial area inside the panel. . In this way, since the periphery of the panel area (that is, the wiring pattern area) is not a target for the calculation of the statistic, the boundary between the inside of the panel and the outside of the panel that is the wiring pattern area is determined with high accuracy.

  Further, as described in step S770, since the position closest to the area for calculating the statistic is determined as the edge position among the positions exceeding the threshold value Vth, the boundary between the inside of the panel and the wiring pattern area is determined. Determined with high accuracy.

  Further, as described in step S710, the coordinates Xs (211) of the panel edge 211 in the image calculated from the design data are used to determine the target region. Accordingly, the region 206 including the panel edge 211 can be cut out (extracted) from the image 240 with an appropriate size without waste, and the time required to determine the position of the edge can be shortened.

  Further, as described in step S710, the maximum error Xme (211) between the coordinates Xr (211) in the image of the panel edge 211 and the coordinates Xs (211) of the panel edge 211 in the image calculated from the design data. ) And are used. As described in step S750, the statistic is calculated by setting a coordinate value that is at least Xme (211) pixels away from the coordinate Xs (211) as the limit of the area for calculating the statistic. The coordinate Xr (211) in the image of the panel edge 211 is removed from the region for the purpose. Thereby, the statistic of the noise component not including the panel edge 211 is calculated. As a result, the threshold value Vth can be determined appropriately, and the edge position can be determined with high accuracy.

  As described above in detail, the edge position detection apparatus 100 according to the embodiment of the present invention is realized as a hardware configuration by combining circuit elements that execute each process, but may also be configured in other aspects. Can do. For example, it is also realized by a mode in which software and hardware cooperate by causing a CPU (Central Processing Unit) or other processor to execute a program for executing each process.

  A computer system 1200 that implements the edge position detection apparatus 100 will be described with reference to FIG. FIG. 12 is a block diagram showing a hardware configuration of computer system 1200.

  The computer system 1200 includes, as main components of hardware, a CPU 1210, a mouse 1220 and a keyboard 1230 that accept input of instructions from a user of the computer system 1200, and processing that is defined in advance by input data or programs. A RAM 1240 for storing temporarily generated data, a hard disk 1250 capable of storing a large amount of data, a CD-ROM (Compact Disk-Read Only Memory) driving device 1260, a monitor 1280, a communication IF (Interface) 1290. Each component described above is connected by a data bus. A CD-ROM 1262 is mounted on the CD-ROM drive 1260.

  Processing in the computer system 1200 that functions as the edge position detection apparatus 100 is realized by the hardware and software executed by the CPU 1210. Such software may be pre-stored in the RAM 1240 or the hard disk 1250, or may be stored in a CD-ROM 1262 or other recording medium and distributed as a program product, and the CD-ROM drive 1260 or other reading device may store the software. In some cases, the data may be read from the recording medium and temporarily stored in the hard disk 1250. The software is read from the RAM 1240 or the hard disk 1250 and executed by the CPU 1210. The hardware itself of the computer system 1200 shown in FIG. 12 is general. Therefore, it can be said that the essential part of the present invention is software stored in the RAM 1240, the hard disk 1250, the CD-ROM 1262, and other recording media. Since the operation of each hardware of computer system 1200 is well known, detailed description will not be repeated.

  Note that the scope to which the technical idea included in the present embodiment described in detail above is applied is not limited to the detection of the boundary (edge) as described above. For example, the present invention can be similarly applied to the case where the position of the outer frame of the rectangular area in the image representing the rectangular area including the boundary is detected as the boundary.

  Further, in order to simplify the description of the present embodiment, the present embodiment has been described in the case where an image acquired by photographing a subject is a rectangle. However, the present invention can be similarly applied to a case where the image is not a rectangle, for example, a polygon. For example, only one side constituting the polygon is set as a position detection target, the above-described processing is executed, and the processing is repeated for each side. By doing so, it is possible to determine the position of the boundary between the region having the shape and the other region even for a polygonal image having a shape other than the rectangle.

  The embodiment disclosed this time should be considered as illustrative in all points and not restrictive. The scope of the present invention is defined by the terms of the claims, rather than the description above, and is intended to include any modifications within the scope and meaning equivalent to the terms of the claims.

  The present invention is used for, for example, an image inspection device, an image conversion device, an image recognition device, a pattern matching device, a positioning device using an image, an autonomous robot with an image recognition function, an industrial robot with an image recognition function, and the like.

It is a figure showing the structure of the test | inspection system 10 which concerns on embodiment of this invention. It is a figure showing the image 240 of the glass substrate 140. FIG. FIG. 6 is a diagram illustrating an image of an area 206. It is a figure showing the edge emphasis picture 400 generated by performing Sobel filter processing. 3 is a block diagram illustrating a configuration of each function realized by the edge position detection device 100. FIG. FIG. 11 is a diagram conceptually illustrating one aspect of data storage in a storage unit 540. 5 is a flowchart showing a procedure of processing executed by processing unit 550 of edge position detection apparatus 100. It is a figure showing the external appearance of the glass substrate. It is a figure showing the image 900 acquired when the glass substrate 800 is image | photographed with the camera. It is a figure for demonstrating the image 906. FIG. It is a figure showing the image which performed the Sobel filter process with respect to the image 906. FIG. 2 is a block diagram illustrating a hardware configuration of a computer system 1200. FIG.

Explanation of symbols

  10 inspection system, 140 glass substrate, 141, 142 panel, 206 area, 210, 220, 240, 300 image, 211, 212, 213, 214, 811, 812, 813, 814 panel edge, 215, 815 wiring pattern area, 216 Wiring pattern edge.

Claims (13)

Storage means for storing original image data representing an original image acquired by photographing a subject, wherein the original image data is represented according to a two-dimensional coordinate axis;
First generation means for generating, when the original image has a plurality of partial images, based on the original image data, emphasized image data representing an image in which a boundary between the partial images is emphasized;
The luminance value of each partial area corresponding to each partial image after the boundary is emphasized is integrated in the direction of the boundary, and an average value is calculated, so that one-dimensional of the two-dimensional coordinate axes Second generation means for generating a plurality of one-dimensional data represented according to the coordinate axes of:
First determining means for determining a range of data to be determined for determining the position of the boundary based on the design data of the subject;
Calculating means for calculating a threshold value for a luminance value by calculating a statistic using the plurality of one-dimensional data included in the range of the data;
Specifying means for specifying the one-dimensional data corresponding to a luminance value exceeding the threshold based on the plurality of one-dimensional data included in the data range;
A boundary position determining apparatus comprising: a second determining unit that determines a coordinate value on the one-dimensional coordinate axis of the one-dimensional data specified by the specifying unit as the position of the boundary.   2. The boundary according to claim 1, wherein the first determination unit includes a unit that determines a range of the data based on a range determined by the original image data and a range specified by the design data. Positioning device. The means for determining the range of the data is:
Based on the design data, as the calculated position data, calculate the one-dimensional data of the boundary,
Based on the original image data, the one-dimensional data of the boundary is calculated as actual position data,
The boundary position determination apparatus according to claim 2, wherein a range of the data is determined based on a difference between the calculated position data and the actual position data. The subject includes a first part to be inspected and a second part not to be inspected,
The original image includes a first image corresponding to the first portion and a second image corresponding to the second portion;
2. The boundary position determination according to claim 1, wherein the first generation unit generates data representing an image in which a boundary between the first image and the second image is enhanced as the enhanced image data. apparatus. The subject is a glass substrate;
The first portion is a panel region processed into the glass substrate,
5. The boundary according to claim 4, wherein the first generation unit generates data representing an image in which a portion where an image corresponding to the panel area and an image corresponding to an area other than the panel area are in contact is emphasized. Positioning device.   2. The boundary position determination according to claim 1, wherein the calculation unit calculates an average value and a standard deviation for the plurality of one-dimensional data, and calculates the threshold based on the average value and the standard deviation. apparatus. The specifying means is:
Means for identifying a plurality of the one-dimensional data corresponding to luminance values exceeding the threshold;
Means for calculating the difference between the upper limit of the range of the data and the one-dimensional data for each of the identified one-dimensional data;
2. The boundary position determining apparatus according to claim 1, further comprising: means for determining, as the one-dimensional data corresponding to a luminance value exceeding the threshold value, one-dimensional data in which each calculated difference is minimum. The original image includes a rectangular region;
The boundary is formed by the outline of the rectangular region;
The boundary position determination device according to claim 1, wherein the first generation unit generates data representing an image in which the boundary formed by the contour is emphasized as the emphasized image data. The second generation means integrates each luminance value of each pixel corresponding to the rectangular area after the boundary is emphasized in one direction in the rectangular area, and calculates an average value, thereby calculating the average value. Generate multiple one-dimensional data,
The boundary position according to claim 8, wherein the first determination unit determines the range of the data such that the range of the data is included in a range defined by the one-dimensional data for the rectangular region. Decision device. The statistic includes an average value and a standard deviation,
The said calculation means calculates the said threshold value using each formula prescribed | regulated beforehand according to each value of several standard deviation, Each said formula changes with the value of each said standard deviation, The claim | item 2. Boundary positioning device. A method for determining a position of a boundary included in an image by a computer, the computer comprising a storage device for storing data and a processing device for executing a predetermined process using the data, The method is
The processing device comprises a step of reading original image data representing an original image acquired by photographing a subject from the storage device, and the original image data is represented according to a two-dimensional coordinate axis,
When the original image has a plurality of partial images, the processing device generates, based on the original image data, enhanced image data representing an image in which a boundary between the partial images is enhanced;
The processing device adds the luminance values of the partial areas corresponding to the partial images after the boundary is emphasized in the direction of the boundary to calculate an average value, thereby calculating the two-dimensional coordinate axis. Generating a plurality of one-dimensional data represented according to a one-dimensional coordinate axis of
The processing device determining a range of data to be determined for determining the position of the boundary based on the design data of the subject;
Calculating a threshold value for a luminance value by calculating a statistic using the plurality of one-dimensional data included in the data range;
The processing device identifying the one-dimensional data corresponding to a luminance value exceeding the threshold based on the plurality of one-dimensional data included in the range of the data;
A method for determining a position of the boundary, comprising: determining, as the boundary position, a coordinate value on the one-dimensional coordinate axis of the identified one-dimensional data. A program for causing a computer to function as a device for determining a position of a boundary included in an image, the computer including a storage device for storing data and a processing device for executing a predetermined process using the data The program is stored in the processing device.
A step of reading original image data representing an original image acquired by photographing a subject from the storage device is executed, and the original image data is represented according to a two-dimensional coordinate axis;
When the original image has a plurality of partial images, based on the original image data, generating enhanced image data representing an image in which the boundary of each partial image is enhanced; and
The luminance value of each partial area corresponding to each partial image after the boundary is emphasized is integrated in the direction of the boundary, and an average value is calculated, so that one-dimensional of the two-dimensional coordinate axes Generating a plurality of one-dimensional data represented according to the coordinate axes of:
Determining a range of data to be determined for determining the position of the boundary based on the design data of the subject;
Calculating a threshold value for a luminance value by calculating a statistic using the plurality of one-dimensional data included in the data range;
Identifying the one-dimensional data corresponding to a luminance value exceeding the threshold based on the plurality of one-dimensional data included in the data range;
And determining a coordinate value on the one-dimensional coordinate axis of the identified one-dimensional data as the position of the boundary.   A recording medium storing the program according to claim 12.

Images