JP4635651B2 - Pattern recognition apparatus and pattern recognition method - Google Patents

Description

  The present invention relates to a pattern recognition apparatus and a pattern recognition method for recognizing a pattern of a transparent electrode such as an ITO film provided on a transparent substrate such as a glass substrate.

In a display panel such as a liquid crystal panel, a transparent electrode such as an ITO film is formed on a transparent substrate such as a glass substrate. In mounting electronic components on the glass substrate, recognition processing for recognizing these transparent electrodes is performed (for example, patents). Reference 1). In this recognition processing, pattern recognition for identifying and detecting a recognition target from a background image is performed by performing recognition processing on an image obtained by capturing a glass substrate with a camera.
JP-A-4-58217

  However, when a transparent electrode such as an ITO film on a glass substrate is a detection target, there is a problem in that recognition is not stable due to the material characteristics of the recognition target object, and recognition errors often occur. That is, the glass substrate in the background of the ITO film appears with high brightness in the image to be recognized by coaxial illumination, and the contrast with the ITO film to be detected is low. For this reason, when there is image noise or a foreign object image in the recognized image, the correlation value in the pattern matching is lowered and a recognition error is likely to occur, and the apparatus is stopped due to the recognition error.

  SUMMARY An advantage of some aspects of the invention is that it provides a pattern recognition apparatus and a pattern recognition method that can stabilize recognition.

The pattern recognition apparatus of the present invention is a pattern recognition apparatus that recognizes a pattern of the transparent electrode by imaging a transparent substrate on which the transparent electrode is formed, and reads image data of an image captured by the camera. Based on an image reading processing unit, a read image storage unit that stores an image read by the image reading processing unit, and a first setting value and a second setting value that are set in advance and stored in the setting value storage unit A brightness correction processing unit that performs brightness correction processing on an image stored in the image storage unit, a correction image storage unit that stores a correction image on which brightness correction processing has been performed, and the correction image as a target A pattern detection processing unit that detects a pattern of the transparent electrode, and the brightness correction processing unit corresponds to the transparent electrode in the image subjected to the brightness correction processing. That than the first set value that is set based on the lower limit value of the luminances of pixels to convert the luminance value of the pixels of high brightness to the maximum tone value in the luminance value of the pixels constituting the read image, the image The minimum gradation value in the luminance value of the pixels constituting the read image is the luminance value of the pixel having a lower luminance than the second set value set based on the upper limit value of the luminance of the pixel corresponding to the transparent substrate in FIG. Execute histogram expansion processing to convert to .

The pattern recognition method of the present invention is a pattern recognition method for recognizing a pattern of the transparent electrode by imaging a transparent substrate on which the transparent electrode is formed, and reads image data of an image captured by the camera. An image reading process step, a read image storing step for storing an image read in the image reading process step, and storing in the image storage unit based on a first set value and a second set value set and stored in advance A brightness correction processing step for performing brightness correction processing on the image that has been processed, a correction image storage step for storing the corrected image on which the brightness correction processing has been performed, and a pattern of the transparent electrode for the correction image A pattern detection processing step for performing pattern detection processing to detect the brightness, and in the brightness correction processing step, the brightness correction processing target Maximum gradation in the luminance value of the pixels constituting the read image luminance value of the luminance said first high-luminance pixels than the set value that is set based on the lower limit value of the pixel corresponding to the transparent electrode in the image A luminance value of a pixel having a lower luminance than the second set value set based on the upper limit value of the luminance of the pixel corresponding to the transparent substrate in the image is converted to a value of the pixel constituting the read image A histogram expansion process for converting the luminance value into the minimum gradation value is executed .

  According to the present invention, the first setting is performed based on the lower limit value of the luminance of the pixel corresponding to the transparent electrode to be detected and the upper limit value of the luminance of the pixel corresponding to the transparent substrate in the image to be subjected to brightness correction. And a first correction in which the luminance value of the pixel having a higher luminance than the first setting value is set to be higher than the first setting value prior to the pattern detection process. Brightness that is converted to a luminance value that is greater than or equal to the value, and that converts a luminance value of a pixel that is lower in brightness than the second setting value to a luminance value that is less than or equal to the second correction value that is set to be lower than the second setting value By executing the depth correction, it is possible to eliminate the influence of the image noise in the read image to be recognized and the image of the foreign matter and perform stable recognition.

  Next, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing a configuration of a pattern recognition apparatus according to an embodiment of the present invention, FIGS. 2 and 4 are diagrams showing an operation screen of the pattern recognition apparatus according to an embodiment of the present invention, and FIG. FIG. 5 is an explanatory diagram of luminance value conversion in the brightness correction processing of the pattern recognition method of one embodiment, and FIG. 5 is a process flow diagram of the pattern recognition method of one embodiment of the present invention.

  First, the configuration of the pattern recognition apparatus will be described with reference to FIG. In FIG. 1, a transparent substrate 1 is a transparent plate-like member such as glass, and a transparent electrode 2 is formed on the transparent substrate 1 by an ITO film that is a transparent conductor. The transparent substrate 1 is positioned below the camera 3 while being held by a panel holding portion (not shown). An illuminating device 4 is disposed above the transparent substrate 1, and the transparent substrate 1 is illuminated by the camera 3 in a state where the transparent substrate 1 is illuminated from above by the half mirror 4 a disposed on the optical axis of the illuminating device 4 and the camera 3. By imaging 1, the image of the transparent substrate 1 including the transparent electrode 2 is read (see FIG. 2). And the pattern of the transparent electrode 2 is recognized by performing the pattern recognition process mentioned later with respect to the read image.

  Image data picked up by the camera 3 is sent to the processing unit 5 and is subjected to pattern recognition processing. The processing unit 5 includes processing function units such as an image reading processing unit 11, a brightness correction setting processing unit 12, a brightness correction processing unit 13, and a pattern detection processing unit 14. The processing unit 5 includes a monitor 6, an operation / input unit 7, a read image storage unit 8, a corrected image storage unit 9, and a setting value storage unit 10, and the processing unit 5 incorporates this pattern recognition device. Connected to an external control unit 15 which is a control device of an external device such as an electronic component mounting apparatus. The external control unit 15 instructs the processing unit 5 to execute pattern recognition processing, and the processing unit 5 outputs a pattern recognition processing result to the external control unit 15.

  The monitor 6 is a display panel that displays an image captured and read by the camera 3 and an operation screen for performing a setting value input operation for a brightness correction process described later. The operation / input unit 7 includes a keyboard, a mouse, operation buttons provided on the panel of the monitor 6, and the like, and performs an operation input to the panel recognition device and a setting value input operation described later. Instead of providing the operation / input unit 7 in the pattern recognition device, this function may be performed by the operation input function of the external control unit 15.

  Next, the processing function of the processing unit 5 will be described. The image reading processing unit 11 performs reading processing of an image captured by the camera 3. The image read by the image reading processing unit 11 is stored in the read image storage unit 8. The brightness correction setting processing unit 12 performs processing (setting mode) for setting a setting value used in the brightness correction processing by the brightness correction processing unit 13 described below. As will be described later, this processing is performed on an operation screen displayed on the monitor 6.

  The brightness correction processing unit 13 is stored in the read image storage unit 8 based on the first setting value and the second setting value set in advance by the brightness correction setting processing unit 12 and stored in the setting value storage unit 10. The brightness correction process described later is performed on the target image. The corrected image on which the brightness correction processing has been performed is stored in the corrected image storage unit 9. The pattern detection processing unit 14 performs a pattern detection process for detecting the pattern of the transparent electrode for the corrected image stored in the corrected image storage unit 9. The pattern detection result is output to the external control unit 15.

  Next, the content of the processing performed by the brightness correction setting processing unit 12 and the brightness correction processing unit 13 will be described in accordance with the operation screen displayed on the monitor 6. A display screen 6 a shown in FIG. 2 shows an operation screen displayed on the monitor 6 by the brightness correction setting processing unit 12. The display screen 6a is provided with an image display field 20, a histogram display field 24, set value input fields 27a and 27b, an input value increase / decrease button 28, and a brightness correction check box 29. Also, operation buttons 31 and 32 for executing brightness correction processing by the brightness correction processing unit 13 and pattern detection processing by the pattern detection processing unit 14 by manual operation, and processing by the brightness correction setting processing unit 12 (setting mode). An end button 33 is provided to end the process.

  In the image display field 20, a read image 20 a that is read by imaging the transparent substrate 1 by the camera 3 is displayed. In the histogram display field 24, a histogram 24a indicating the luminance distribution of the pixels constituting the read image 20a is displayed. In the set value input fields 27a and 27b, numerical values for setting a first set value T1 and a second set value T2 to be described later are displayed, and the input value increase / decrease button 28 can be used to increase or decrease the value. Yes. The brightness correction check box 29 is used to select whether or not to perform brightness correction, which will be described later, in the pattern recognition process (automatic processing mode), and by adding a check mark, brightness correction is performed. Execution is selected.

  By operating the setting operation button 30, the numerical values displayed in the setting value input fields 27a and 27b are taken and set as the first setting value T1 and the second setting value T2. By operating the operation button 31, brightness correction is executed based on the first set value T1 and the second set value T2. By operating the operation button 32, pattern detection processing is executed for the corrected image after the brightness correction, and by operating the end button 33, the setting mode is ended.

  Here, the read image 20a obtained by imaging the transparent substrate 1 will be described. In the read image 20a, a pattern image 22 obtained by reflecting the illumination light similarly by the transparent electrode 2 appears in the background image 21 obtained by reflecting the illumination light of the illumination device 4 by the transparent substrate 1. Yes. In the read image 20a, the pattern image 22 appears as an image element having higher brightness than the background image 21, but there is no large contrast difference between them. The background image 21 is dotted with foreign matter images 23 in which non-transparent foreign matters such as dust appear as dark portions.

If the read image 20a including such image elements is subjected to pattern matching for detecting the pattern image 22 as it is, the correlation value indicating the cross-correlation between the read image 20a and the template image is lowered, resulting in a recognition error. Cheap. That is, image noise exists in the pattern image 22 and the brightness value of the pattern image 22 varies. For this reason, when the contrast difference between the background image 21 and the pattern image 22 is small as described above, the degree to which the correlation value is reduced due to the image noise cannot be ignored. Furthermore, since the foreign object image 23 has a low luminance and a large luminance difference from the background image 21, the correlation value further decreases due to the presence of the foreign object image 23. If these factors are combined, the matching rate in pattern matching decreases, and recognition errors may occur frequently.

  In order to prevent the occurrence of such a recognition error, in the present embodiment, brightness correction is performed on the read image 20a by the method described below to eliminate the influence of the image noise and the image of foreign matter described above. Pattern matching for pattern detection is executed on the corrected image. Hereinafter, the brightness correction process will be described.

  First, the histogram 24a will be described. The histogram 24a is obtained by dividing the luminance of each pixel constituting the read image 20a into 256 gradations and displaying the luminance frequency distribution as a histogram. The gradation values of 0 to 255 correspond to the luminance values of the pixels. ing. In the histogram 24 a, a trapezoidal frequency accumulation unit 25 in which the frequencies are accumulated in the gradation values in a specific range corresponding to the pixels having the adjacent luminance values appears. That is, as shown in FIG. 4, the background image 21 and the pattern image 22 in the read image 20a appear as two frequency accumulating portions 25a and 25b that exist relatively close to each other in the high luminance range in the histogram 24a, and are transparent substrates. The foreign matter image 23 corresponding to foreign matter such as dust existing on 1 appears in a low luminance range separated from the frequency accumulation unit 25b as the frequency accumulation unit 25c in the histogram 24a.

  The histogram 24a displays setting value cursor lines 26a and 26b for setting a first setting value T1 and a second setting value T2, which will be described later, and the levels displayed in the setting value input fields 27a and 27b. By increasing / decreasing the tone value (luminance value) with the input value increase / decrease button 28, the set value cursor lines 26a, 26b move in the vertical direction in the histogram 24a. By operating the setting operation button 30, the luminance values corresponding to the setting value cursor lines 26a and 26b in the histogram 24a are set as the first setting value T1 and the second setting value T2.

  In the brightness correction process, first, for the purpose of clearly identifying the pattern image 22 in the background image 21, the two frequency accumulating units 25a and 25b existing relatively close to each other in the histogram 24a are reliably separated. A first set value T1 and a second set value T2 are set. That is, the first set value T1 is set immediately below the lower limit value of the frequency accumulation unit 25a, and the second set value T2 is set immediately above the upper limit value of the frequency accumulation unit 25b. This setting is performed by operating the input value increase / decrease button 28 and moving the set value cursor lines 26a and 26b on the display screen 6a shown in FIG. In the example shown in FIG. 3, the first set value T1 and the second set value T2 are set to the gradation values 194 and 168, respectively.

  Next, luminance values are converted based on the set first set value T1 and second set value T2. That is, the luminance value of each pixel having a luminance value equal to or higher than the first setting value T1 is converted into a first correction value (here, the maximum gradation value 255), and each luminance value equal to or lower than the second setting value T2 is converted. The luminance value of the pixel is converted to a second correction value (here, minimum gradation value 0). This conversion processing is performed using a histogram expansion processing function provided in the processing unit 5. In other words, in the histogram 24a, the range of the first setting value T1 to the second setting value T2 is expanded from the maximum gradation value (255) to the minimum gradation value (0), and accordingly. This means that the portion equal to or higher than the first set value T1 and the portion equal to or lower than the second set value T2 are compressed.

By this luminance value conversion, the histogram 24a corresponding to the read image 20a changes to a luminance distribution like the histogram 24b. In the histogram 24b, all the luminance values of the pixels belonging to the frequency accumulating unit 25a are uniformly converted into gradation values 255, and all the luminance values of the pixels belonging to the frequency accumulating units 25b and 25c are uniformly scaled. Converted to key value 0.

  FIG. 5 shows a corrected image after the brightness correction is performed in this way. That is, on the operation screen shown in FIG. 3, after setting the first set value T1 and the second set value T2 as described above, the brightness correction processing unit 13 performs brightness correction by operating the operation button 31. The process is executed, and the corrected image 20b is displayed in the image display field 20 of the display screen 6a. In this corrected image 20 b, the pattern image 22 appears in the background image 21 with a clear contrast, and the foreign object image 23 scattered in the background image 21 is converted to the same luminance value as the background image 21. It has been deleted from the image. Then, by operating the operation button 32, pattern matching between the corrected image 20b and the template image is executed by the pattern detection processing unit 14, and the pattern of the pattern image 22 in the background image 21 is detected.

  In addition, although the example which employ | adopted the maximum gradation value and the minimum gradation value is shown here as a 1st correction value and a 2nd correction value, you may set to values other than this. That is, the first correction value is a luminance value that is equal to or higher than the first set value T1, and the second correction value is a luminance value that is equal to or lower than the second set value T2, and the brightness correction image after luminance conversion and Any value may be used as long as a correlation value giving a stable matching result is obtained when pattern matching is performed with the template image.

  In other words, in the brightness correction process described above, the brightness is higher than the first set value T1 set based on the lower limit value of the brightness of the pixel corresponding to the transparent electrode 2 in the brightness correction target image. Is converted into a luminance value equal to or higher than the first correction value set higher than the first set value, and the upper limit value of the luminance of the pixel corresponding to the transparent substrate 1 in the read image 20a is converted. The luminance value of a pixel having a luminance lower than the second setting value T2 set based on the luminance value is converted to a luminance value equal to or lower than the second correction value set to a luminance lower than the second setting value T2. ing.

  In the brightness correction setting processing performed by the brightness correction setting processing unit 12, the histogram 24a indicating the luminance distribution of the pixels in the read image 20a is displayed with the settings of the first setting value and the second setting value. On the operation screen. Thus, the set value for performing the brightness correction is appropriately set while visually confirming the difference in luminance distribution between the background image 21 and the pattern image 22 to be recognized on the operation screen on the histogram 24a. It is possible to set by a simple operation.

  Next, pattern recognition processing (automatic processing mode) performed by the pattern recognition apparatus will be described with reference to each drawing along the flow of FIG. Prior to the execution of the pattern recognition process, first, the first set value and the second set value are set. This setting process is executed when a new workpiece type is targeted. A real workpiece in which the transparent electrode 2 is formed on the transparent substrate 1 is used, and the recognition target is displayed on the operation screen shown in FIG. This is performed by displaying the read image 20a.

  Then, in the histogram 24a displayed together with the read image 20a, the distribution of the frequency accumulation unit 25a corresponding to the background image 21 and the frequency accumulation unit 25b corresponding to the pattern image 22 is observed, and the frequency accumulation is performed by operating the input value increase / decrease button 28. The first setting value T1 is set to a gradation value lower than the lower limit value of the unit 25a, and the second setting value T2 is set to a gradation value higher than the upper limit value of the frequency accumulation unit 25b.

Thereafter, when execution of pattern recognition processing is instructed by the external control unit 15, the pattern recognition processing is executed by the processing unit 5. First, the transparent substrate 1 is positioned at the imaging position, and the transparent electrode 2
The recognition target part where the is formed is positioned immediately below the camera 3 and imaging is performed (ST1). Next, image data of an image captured by the camera is read (image reading processing step), and the read image is stored in the read image storage unit 8 (ST2) (read image storing step).

  Here, it is determined whether or not the brightness correction needs to be executed (whether or not the brightness correction check box 29 is checked) (ST3), and when the brightness correction is executed, the brightness correction processing unit 13 performs the above-described brightness. Execute correction. That is, the brightness correction process described above is performed on the read image 20a (FIG. 3) stored in the read image storage unit 8 based on the first set value T1 and the second set value T2 that are set and stored in advance. Perform (brightness correction processing step). Then, the corrected image 20b (FIG. 4) subjected to the brightness correction is stored in the corrected image storage unit 9 (ST4). When brightness correction is not executed, the image data in the read image storage unit 8 is transferred to the corrected image storage unit 9 (ST5).

  Then, the pattern detection processing unit 14 performs pattern detection processing of the transparent electrode 2 on the corrected image 20b stored in the correction image storage unit 9 (ST6) (pattern detection processing step), and the pattern detection result is externally controlled. It outputs with respect to the part 15, and complete | finishes the process targeting one recognition site | part.

  As described above, the pattern recognition process shown in the present embodiment is a pattern detection process in pattern recognition in which a transparent electrode 2 such as an ITO film formed on a transparent substrate 1 such as a glass substrate is detected by image recognition. In advance, the luminance value of a pixel having a higher luminance than the first set value T1 set based on the lower limit value of the luminance of the pixel of the pattern image to be detected in the image to be corrected for brightness is set to a higher luminance. The luminance value of a pixel having a lower luminance than the second set value T2 set based on the upper limit value of the luminance of the pixel of the background image is converted to a lower luminance.

  As a result, even when the pattern image 22 existing in the background image 21 with low contrast is targeted for detection in the read image 20a, the contrast between the pattern image 21 and the background image 22 can be increased. Accordingly, there is no decrease in the correlation value in pattern matching caused by image noise or foreign matter in the read image, thereby eliminating a recognition error and performing stable recognition.

  The pattern recognition apparatus and the pattern recognition method of the present invention have the effect of eliminating the influence of dust and noise in a read image and performing stable recognition, and a bright background image such as an ITO film on a glass substrate It is useful in the field of pattern recognition for detecting patterns existing in low contrast.

The block diagram which shows the structure of the pattern recognition apparatus of one embodiment of this invention The figure which shows the operation screen of the pattern recognition apparatus of one embodiment of this invention Explanatory drawing of luminance value conversion in the brightness correction process of the pattern recognition method of one embodiment of this invention The figure which shows the operation screen of the pattern recognition apparatus of one embodiment of this invention Processing flow diagram of pattern recognition method of one embodiment of this invention

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Transparent substrate 2 Transparent electrode 3 Camera 20a Reading image 20b Correction image 24a, 24b Histogram T1 1st setting value T2 2nd setting value

Claims (4)

A pattern recognition device for recognizing a pattern of the transparent electrode by capturing an image of the transparent substrate on which the transparent electrode is formed with a camera,
An image reading processing unit that reads image data of an image captured by the camera, a read image storage unit that stores an image read by the image reading processing unit, and a first preset value stored in a setting value storage unit A brightness correction processing unit that performs a brightness correction process on the image stored in the image storage unit based on the set value and the second set value, and a correction that stores the corrected image on which the brightness correction process has been performed An image storage unit, and a pattern detection processing unit that detects a pattern of the transparent electrode for the corrected image,
The brightness correction processing unit is configured to detect pixels having a luminance higher than the first set value set based on a lower limit value of the luminance of the pixel corresponding to the transparent electrode in the image subjected to the brightness correction processing. From the second set value set based on the upper limit value of the luminance of the pixel corresponding to the transparent substrate in the image, by converting the luminance value into the maximum gradation value in the luminance value of the pixel constituting the read image A pattern recognition apparatus that executes a histogram expansion process for converting a luminance value of a low-luminance pixel into a minimum gradation value in a luminance value of a pixel constituting the read image . A brightness correction setting processing unit configured to perform setting of the first setting value and the second setting value on an operation screen on which a histogram indicating a luminance distribution of pixels in the image is displayed;
The brightness correction setting processing unit sets the first setting value based on a lower limit value of a frequency accumulation unit corresponding to a pattern image obtained by reflecting reflected light by the transparent electrode in the histogram. The pattern recognition apparatus according to claim 1, wherein the second set value is set based on an upper limit value of a power accumulation unit corresponding to a background image obtained by reflecting reflected light from the transparent substrate . A pattern recognition method for recognizing a pattern of the transparent electrode by imaging a transparent substrate on which the transparent electrode is formed with a camera,
An image reading processing step for reading image data of an image captured by the camera, a read image storing step for storing an image read in the image reading processing step, a first set value and a first setting value stored in advance; A brightness correction processing step for performing brightness correction processing on the image stored in the image storage unit based on the set value of 2, and a correction image storage step for storing the correction image on which the brightness correction processing has been performed; A pattern detection processing step for performing a pattern detection process for detecting a pattern of the transparent electrode for the corrected image,
In the brightness correction processing step, pixels having higher brightness than the first set value set based on the lower limit value of the brightness of the pixel corresponding to the transparent electrode in the image subjected to brightness correction processing. From the second set value set based on the upper limit value of the luminance of the pixel corresponding to the transparent substrate in the image, by converting the luminance value into the maximum gradation value in the luminance value of the pixel constituting the read image A pattern recognition method comprising: performing a histogram expansion process for converting a luminance value of a low-luminance pixel into a minimum gradation value in a luminance value of a pixel constituting the read image . The setting of the first setting value and the second setting value is performed on an operation screen on which a histogram indicating a luminance distribution of pixels in the image is displayed,
In the histogram, the first set value is set based on a lower limit value of the frequency accumulation unit corresponding to the pattern image obtained by reflecting the reflected light by the transparent electrode, and the reflected light is reflected by the transparent substrate. 4. The pattern recognition method according to claim 3, wherein the second set value is set based on the upper limit value of the frequency accumulating unit corresponding to the background image obtained in this way.

Images