WO1998034096A1

WO1998034096A1 - Method and system for detecting defects in transparent objects having spatial variations in their optical density

Info

Publication number: WO1998034096A1
Application number: PCT/US1998/001256
Authority: WO
Inventors: Robert J. Bieringer; Rex M. Kremer; Spencer D. Luster; Eric A. Roth
Original assignee: Medar Inc
Current assignee: Medar Inc
Priority date: 1997-01-31
Filing date: 1998-01-21
Publication date: 1998-08-06
Anticipated expiration: 1999-07-31

Abstract

A method and system (16) for detecting defects in transparent objects (15) having spatial variations in their optical density such as CRT panels (15) supported between at least one camera (24) and at least one light source (23) which are controlled to create a variable light profile. Each of the light sources (23) is a segmented LED array (46, 54). The light profile is changed for a 'smear' camera of the system for each scan taken. At the end of each scan, the values for the next scan are read from the table of light value and sent to an analog output circuit that controls the intensity of each segment of the LED array (46, 54) associated with the smear camera (24). In this way, the optical signals from the smear camera are kept within the dynamic range of the camera under conditions of very high gain and large negative offset during relative movement between the CRT panel (15) and the smear camera and its LED array.

Description

METHOD AND SYSTEM FOR DETECTING DEFECTS

IN TRANSPARENT OBJECTS HAVING SPATIAL

VARIATIONS IN THEIR OPTICAL DENSITY

Technical Field

This invention relates to methods and systems for detecting defects in transparent objects and, more particularly, to methods and systems for detecting defects in transparent objects which have spatial variations in their optical density.

Background Art

From the family television set to the computer terminal, the electronic display has become an indispensable way to deliver information. No other medium offers its speed, versatility and interactivity. These attributes are being used to create a wide variety of products that can provide information in any combination of text, graphics, still images or video.

The conventional cathoderay tube (CRT) remains the dominant display. In 1991, the worldwide market for CRT monitors was 37 million units of which 20 million units were high-resolution color monitors. These 37 million do not take into account the oscilloscope, radar, projection, and other special-purpose tubes that would add a few hundred thousand to the figures .

The shadow-mask tube is by far the most widespread tube for computer and high-resolution monitor displays. The shadow-mask CRT uses three electron beams deflected by one deflection coil. The beams traverse a perforated shadow or aperture mask before impinging on the selected luminescent screen material, which is usually made of transparent stripes or dots of red, blue, and green phosphors located in a matrix generated using a hexagonal unit cell. The dots are filled with the phosphors in subsequent individual slurry processes. The strips or dots are typically etched in a black, optically opaque, patterned material of a CRT panel by a photolithographic process which includes the shadow- mask. The black matrix surrounds the color phosphors for increased contrast and also reduces the precision required for effective phosphor dot placement and size. Thus, it is important to detect and to fix the defects, or to remove and clean the glass panel at this stage of manufacture, prior to adding value to an already defective product .

The arrangement of the electron optics and of the deflection system is such that the three electron beams converge on the screen after passing through the shadow-mask, each beam impinging on one color, red, blue, or green, only. An example of an arrangement is shown in Figure 1, wherein a shadow-mask 10 is positioned directly behind a phosphor dot viewing screen 12 in a three-gun color picture tube to prevent the excitation of any one color phosphor by the electron beams 14 not associated with that color.

Shadow-mask tubes use a mechanical selection of colors . The thin perforated steel or invar shadow- mask is welded onto a metallic frame suspended by supports in the tube glass faceplate. However, invariably, the holes of shadow-masks get partially or fully clogged which prevent the proper formation of the stripes or dots, thereby resulting in black matrix defects.

There are a number of different defects that are described by manufacturers of CRTs. They are described by various names that can be related to their appearance, or to their source. Some are characterized by missing, or partially missing, dots and some by changes in shape or size of the dots. While some can be related to problems caused by the shadow mask, others can occur during handling, or during the photolithographic and accompanying cleaning processes .

Defects such as missing dots, as shown in Figure 2, are relatively easy to detect. However, "subtle" defects such as "smear" or "splash" are much harder to detect . An example of smear is illustrated in Figure 3 while an example of splash is illustrated in Figure 4. Subtle defects may be defined as small changes in a relatively large number of dots.

However, as observed by inspection using light transmission, imaging and analysis, there really are two general classes of defects. One class causes local changes in light transmission (comparable to the dot size) that are relatively limited in spatial extent and exhibit moderate to high contrast . The other class of defects results from small changes in a large number of dots covering a relatively large spatial extent exhibiting low contrast. It is the latter which present the greatest challenge to conventional machine vision systems used for inspection. The detection of such subtle defects is currently being done by an operator at an operator inspection station. This station typically includes a light table on which the CRT panel is placed so that the operator can view light transmitted through the CRT panel .

Summary Of The Invention

An object of the present invention is to provide a method and system for automatically detecting defects in objects such as CRT panels in the presence of large spatial variations in the panel's optical density.

Another object of the present invention is to provide a method and system for automatically detecting distributed defects in a black matrix of an electronic display such as a CRT panel by keeping optical signals within the dynamic range of camera output under conditions of very high gain and large negative offset during relative movement between the CRT panel and a camera of the system.

Still another object of the present invention is to provide a method and system for automatically detecting defects in the presence of large changes in light transmission through different portions of the object being inspected. Preferably, an illumination profile is adjusted dynamically using open loop control.

Yet still another object of the present invention is to provide a method and system for automatically detecting distributed defects, such as "smear" and "splash" in a display panel, which could not be detected without significant increase in contrast of images, in the presence of large spatial variations in light transmission through the panel during relative movement of the panel with respect to a camera and light source .

In carrying out the above objects and other objects of the present invention, a method is provided for detecting defects in a transparent object having spatial variations in its optical density. The method includes the steps of supporting the object between at least one camera and at least one light source, and controlling the at least one light source so that the at least one light source creates a variable light profile which is transmitted through the object. The method also includes the steps of generating an image of the object during the step of controlling to obtain a signal having high frequency defect data and processing the high frequency defect data to determine the defects in real-time.

Preferably, the object is a display panel and, still preferably, is a CRT panel.

Still further in carrying out the above objects and other objects of the present invention, a system is provided for carrying out the above method steps.

The method and system of the present invention is provided for quality control rather than for quality assurance, so that the inspection is performed at manufacturing line rates, resulting in the need to inspect an entire CRT panel in less than ten seconds. This requirement suggests the use of linescan cameras and minimal digitization (such as thresholding the analogue outputs of the (CCD) cameras) . The method and system are intended to serve as an aid to visual inspec- tion and defect classification.

The invention employs a strategy of "finding the exceptions, rather than proving the rule" which means that some optical pre-processing is required to allow one to ignore non-defective regions and find only those that contain defects. The method and system detect the low contrast defects of large extent that allow one to take a broader view (something akin to concentrating on the forest rather than its trees) . In fact, these defects are often more easily observed by the naked eye than by using optical magnification. For the sake of brevity, the two classes of defects are named "local" and "distributed".

An automatic inspection station makes use of two optical and corresponding logical channels; one for each class of defects. Panels are inspected using light transmission and, due to spatial variations in the optical density of the glass resulting from panel shape, the light sources are segmented so that the illumination profiles for each channel can be adjusted both spatially and separately. Magnifications of the two channels differ by roughly an order of magnitude and they are chosen to provide a uniform signal both along a line of dots and between the dots when the B/M matrix is well aligned with the glass (relative to which the inspection optics is aligned) . Since the B/M can be rotated slightly relative to the glass, it is preferable to provide some simple electronic filtering for the local defect channel in order to assure detection of the defects by thresholding of the analogue camera output.

In the case of the distributed defect channel, since the optical contrast is extremely low for these defects, it is necessary to increase the gain and dynamic range of image acquisition optics and electronics. This requires that the illumination profile be corrected during the inspection, as the optics scan the panel. Since the detection method is more "global" in nature, these defects are detected by using electronic filtering that takes into account signal differences between scans of the panel. This combination is effective in detecting a large fraction of the distributed defects .

The above objects and other objects, features, and advantages of the present invention are readily apparent from the following detailed description of the best mode for carrying out the invention when taken in connection with the accompanying drawings .

Brief Description Of The Drawings

FIGURE 1 is a schematic diagram showing the geometry of how a shadow or aperture mask allows for electron bombardment of the red, green and blue phosphors in a working CRT;

FIGURE 2 is a view illustrating a black matrix defect of a CRT panel caused by clogged holes of the shadow-mask; FIGURE 3 is a view illustrating a black matrix or distributed defect of a CRT panel caused by a "smear" on the shadow-mask;

FIGURE 4 is a view illustrating a black matrix or distributed defect of a CRT panel caused by a splash of NH₄HF₂ on the shadow-mask;

FIGURE 5 is a schematic diagram of a manual or operator inspection station and an automatic inspection station which includes the method and system of the present invention;

FIGURE 6 is a schematic block diagram of a system for detecting defects in transparent objects having spatial variations in their optical density such as CRT panels;

FIGURE 7a is a schematic top view of an inspection window in a roller conveyor;

FIGURE 7b is a schematic top view of smear and defect LED arrays, scans for the cameras and a proximity switch, all of which are typically located in the window of Figure 7a;

FIGURE 8a is a schematic top view of the position of a CRT panel in the inspection window of Figure 7a for inspection and calibration;

FIGURE 8b is a schematic top view of smear and defect LED arrays, scans for the cameras and a proximity switch, all of which are typically located below the panel of Figure 8a; FIGURE 9 is a schematic, block diagram, data flow chart of the present invention;

FIGURE 10 is a graph illustrating a normal gray scale image for a number of pixels;

FIGURE 11 is a graph similar to the graph of

Figure 10 illustrating a selected range for negative offset enhancement; and

FIGURE 12 is a graph similar to the graphs of Figures 10 and 11 illustrating an expanded signal utilizing the invention.

Best Mode For Carrying Out The Invention

Referring now to the drawing figures, there is illustrated in Figures 5 and 6 a first or automatic inspection station, generally indicated at 16, for detecting defects in transparent objects having spatial variations in their optical densities such as electronic displays or CRT panels or parts 15. The station 16 is designed with the intent to provide an aid to visual inspection, so that the machine can do what it does well (i.e. pay attention) .

A human inspector can provide the judgment needed for defect classification at a second or manual inspection station. The automatic inspection station 16 typically reads the panel ID, inspects for defects, and provides a rough classification and a mapping of the flaws to the second station. At this second station, the human inspector can Observe the defects and classify them and then transmit the panel ID and defect classifi- ^■ l o ¬

cations to a host computer in the manufacturing plant. The two stations are preferably linked by an ethernet connection.

More particularly, the parts 15 are initially positioned on an indexing roller conveyor, generally indicated at 18. As the parts 15 enter the inspection station 16, a bar code scanner (not shown) reads a bar code affixed to each part 15 and sends it to an inspection CPU or host computer 28. The parts 15 are then loaded or indexed onto a section 20 of the roller conveyor 18 that has the rollers 22 cut short (Figures 7a and 8a) to obtain a transmissive view of the part 15. The part 15 is held in position while light sources 23 and cameras 24 (Figure 6) scan across the part 15 and do the inspection. The light sources 23 are mounted to move together and the cameras 24 are also mounted to move together synchronously with the light sources 23.

In general, once defects are detected, they are sized using connectivity analysis and their bright- ness or darkness are recorded; these parameters are reported, along with their location, to the manual inspection station via the ethernet connection. During transport between the automatic inspection station and the manual inspection station, green and red lights track with the panels indicating to the inspector when he or she can expect a panel with a defective black matrix (i.e. B/M). The manual inspection station is equipped with a PC monitor on which the flaw map can be displayed along with choices for the results of human classification. The inspector enters these into the host computer using a touch screen on the monitor, enabling that system to perform a standard Statistical Process Control analysis.

More particularly, if the part 15 appears to have flaws, an individual reject light (i.e. red) 27 (Figure 5) will be lit as the part 15 is indexed to the next location, otherwise an accept light (i.e. green) 33 will be lit. These lights 27 and 33 are repeated at each index position to allow the correct signal to follow the appropriate part. The first station 16 also sends the results of the inspection to a database for later retrieval. If enabled, the station 16 will also sound an audible signal as each possibly flawed part exits the inspection station 16.

As previously mentioned, when an operator wishes to visually inspect a part 15, he or she merely picks the part 15 off of the conveyor 18 and places it on a light table 42 at the manual inspection station

(Figure 5) where the bar code is read and the results from the prior automatic inspection are read from the database and displayed on a CRT screen 44. The operator may select which types of defects are present on the part 15 and mark it as a reject, or he or she may say that the part 15 is acceptable due to cleaning or repair. When the operator input is finished the modi- fied data is added to the database under the correct code number .

Referring specifically to Figure 6, there is illustrated schematically a machine vision system at the inspection station 16, by which the method and system of the present invention can reliably detect "subtle" or

"distributed" defects of the part 15 (see, for example, the "smear" and "splash" defects of Figures 3 and 4, respectively) . Preferably, the method and system use a variable light profile of a smear LED array 46 (Figures 7b and 8b) of the light sources 23 to detect the de- fects.

The machine vision system of Figure 6 typically includes an image digitizer/frame grabber 22 for each one of the cameras 24. Each image digitizer/frame grabber 22 samples and digitizes input images from an image source such as one of the cameras 24 along line 48 and places each input image into a frame buffer having picture elements. The image/digitizer/frame grabber 22 may be a conventional frame grabber board such as that manufactured by Matrox, Cognex, Data Translation or other frame grabbers. Alternatively, the image digitizer/frame grabber 22 may comprise a vision processor board such as made by Cognex .

Each of the picture elements may consist of an 8 -bit number representing the brightness of that spot in the image. If the cameras 24 are digital cameras, the digital camera will eliminate the need for the image digitizer/frame grabber 22 and the input image appears along a line 25.

The machine vision system also includes input/output circuits 30 to allow the system 20 to communicate with external devices such as controllers for controlling motors such as servo motors at blocks 50 and 52. The controller and servo motor for the cameras 24 are indicated by block 50 in Figure 6. The control- ler and the servo motor for the light sources 23 are indicated by block 52 in Figure 6. Typically, the cameras 24 and the light sources 23 synchronously move together and scan the part 15 as indicated by the scan lines on the smear LED array 46 and a defect LED array 54 of the light sources 23 as indicated in Figures 7b and 8b.

The cameras 24 may be image sources such as analog, digital or line scan cameras such as RS-170, CCIR, NTSC and PAL.

The machine vision system also includes a system bus 26 which may be either a PCI, an EISA, ISA or VL system bus or any other standard bus to allow inter- system communication such as with a CRT monitor 29 of the machine vision system and an image processing board 31 as will be described in greater detail hereinbelow.

The machine vision system may be programmed at a mass storage unit 32 to include custom controls for image processing and image analysis including blob analysis .

The host computer 28 of the machine vision system may be a pentium-based PC having a sufficient amount of RAM and hard disk space for computer programs for controlling the machine vision system.

There is generally indicated at 40 a block which represents one or more filters as will be de- scribed below in greater detail.

Referring again to Figures 7a and 7b, the inspection window 56 in the roller conveyor 18 and the fields of view of the defect and smear cameras, 24 are shown. The two defect cameras of the cameras 24 have a small overlap which has not been shown in Figure 7b to make it clear that there are two cameras with the fields of view laid end-to-end.

Preferably, while not shown in Figure 6, the two defect cameras are spaced apart and positioned a first distance from its respective light array 54 in a first plane. The smear camera is positioned between the defect cameras a second distance from its respective light array 46 in a second plane spaced slightly away from the first plane, the second distance being greater than the first distance.

The fields of view of the defect cameras are large enough to inspect the range of parts specified (14" and 15") . During the inspection, the LED arrays 54 and 46 and the three cameras 24 will start on one side of the inspection window 56 and scan across the inspection window 56. The side on which the optics (i.e., the cameras 24 and the light sources 23) start alternates from inspection to inspection to extend the life of the mechanical components in the system. The movement of the light arrays 46 and 54 and the cameras 24 are controlled by the pair of brushless servo motors 52 and 50, respectively, driven by synchronized control modules (i.e. within blocks 52 and 50, respectively) in order to guarantee the top and bottom sections of the optics move at the same rate .

Referring now to Figures 8a and 8b, there is illustrated where the part 15 is placed during inspec- tion and calibration. The distance calibration process uses a part 15 with three calibration distances marked so the system may determine their limits. The distance calibration process will start the movement, scanning from a home proximity switch 60 and wait for the first scan on the part 15 with each of the defect and smear cameras 24. This will allow the system to know when the data collection should be started during inspection. It will also use calibration distances 62, 63 and 64 to calibrate the pixel sizes and the camera placements for all three cameras 24. The calculation of scan calibra- tion distances 66 and 68 (between Figures 8a and 8b) is done from both the home position and the end-of-scan position on the other side of the inspection window 56 to allow the alternating scanning directions.

For gray scale light profile calibration a normal part should be used. For each defect camera, the system will step up to a position at the beginning, middle and end of the part and find the best fit for a light profile and threshold values. For the smear light profile, the system will step to each scanning position and optimize the light profile at each location. This will result in the smear light profile table that will be updated in real-time as the part 15 is inspected.

The light profile is adjusted at up to sixteen locations across the part and may have a separate set of sixteen values for each scan.

Figure 9 shows a diagram of the data flow during inspection. Any of the square boxes could be replaced by custom code that is designed to replace the supplied code's function. The data could be accessed at any of the circles for any special processes needed. The gray scale data from the cameras 24 is filtered at block 40 (Figure 6) to increase the defect discrimination. In particular, and with reference to Figure 9, the defect camera data is filtered with pixel- to-pixel filters 70 and 70' to enhance the edge detection in the pixel direction. The smear camera data is filtered with a scan-to-scan filter 74 after filtering by a pixel-to-pixel filter 72 to enhance the edge detection in the scan direction. A compare data block 76 in the diagram is a delayed past value from a previous scan that is used in the scan-to-scan filter 74.

The filtered data output goes to a camera or the image processing block 31 (Figure 6) which includes boards 78, 78' and 78'' (Figure 9) where they are thresholded and converted to hit data. Hit data consists of the pixel number where the hit occurred, the direction (rising or falling) and the threshold the signal crossed.

The two hit data streams from the defect camera are merged into a single stream at junction 80 and then the merged defect hit data stream and the hit data stream from the smear camera are used to generate hit pairs at blocks 82 and 84, respectively. Hit pairs show where the edges of possible defects are in a scan. They consist of the hit data for the two edges that define the possible defect and the thresholds crossed by and hits that are internal to the possible defect.

The hit pair data is then sent to the connectivity module block 86 (sometimes called blob analysis) for combining into flaws (blobs) . The flaw data contains the bounding rectangle, the total defective area contained within the defect, and a flag for each of the thresholds crossed by the defect.

The flaw data is then used to select the flaw type and size range of the defect at block 88 according to setup parameters entered into the system. This data is then used to decide if the part is outside of the allowed limits at block 90 and the results are sent to the previously mentioned database for use by the operator at the manual inspection station at block 92 and are used to signal the operator (by lights and buzzer) that a possibly defective part has exited the automatic inspection station 16.

The raw (gray scale) camera data comes in from the cameras 24 at a 10 MHz pixel rate. The gray scale signals have 8 -bit resolution. This gives them a range from gray scale value 0 to gray scale value 255. A standard gray scale signal uses the gray scale value 0 for the condition of no light hitting a pixel on a camera and a gray scale value of 255 for the pixel just reaching saturation due to receiving the maximum measurable amount of light. Each of the inputs from the defect cameras is a standard gray scale signal and the signal from the smear camera is a modified negative offset signal (i.e. Figure 12) to give greater discrimi- nation of the smear signal. A modified negative offset signal is an 8-bit value with range from 0 to 255, but the value for no light hitting the pixels on the array is not 0 but is a negative number. The large negative offset used for the smear camera signal requires a variable light profile on the smear camera, as provided by the present invention. The normal gray scale signal looks like the graph of Figure 10. What the negative offset does is to take a small range of the gray scale values and spread them out over the entire 255 gray scale range. In the image of Figure 11, the dotted lines show the area that would be spread over the 255 gray scale range with a negative offset that moves gray scale 107 to 0. For the negative offsets, the light intensity is also increased to set the highest gray scale value desired at gray scale value 255. For this example, the value 193 is used. This results in the highest possible discrimination over the desired range.

The increase in optical gain from the negative offset is the reason that the light profile needs to change across the part. If the signal area of interest varies outside the upper and lower limits (in this case 107 to 193) , the signal would hit the minimum or maximum value and would not display the true variation. The reasons for this happening are many and varied, they include the thickness variations of the glass across the part, the changing surface angles of the part and the shadowing of the surface by the side edges as the three most important. These can be reduced or eliminated by changing the light profile as the part is scanned.

The light profile is changed for the smear camera for each scan taken. A table is generated in the light profile calibration step that is used to supply the correct values for each scan taken. At the end of each scan, the values for the next scan are read from the table and sent to an analog output card or light controller 94 that controls the intensity of each of the sixteen segments of the LED array 46. Expanding the selected region of the signal results in a signal that looks like the signal of Figure 12. Much of the signal falls outside of the gray scale range of 0 to 255. This portion of the signal is no longer seen by the inspection system as the correct values. This is the effect that requires the variable light profile.

The signal of Figure 12 is what the signal looks like after the negative offset is applied. Any portion of the signal that is outside the 0 to 255 gray scale range would not be interpreted correctly. The variable light profile is used to bring the edges that are too low up to where they are within the 0 to 255 gray scale range .

While the best mode for carrying out the invention has been described in detail, those familiar with the art to which this invention relates will recognize various alternative designs and embodiments for practicing the invention as defined by the following claims.

Claims

What Is Claimed Is;

1. A method for detecting defects in a transparent object having spatial variations in its optical density, the method including the steps of: supporting the object between at least one camera and at least one light source; controlling the at least one light source so that the at least one light source creates a variable light profile which is transmitted through the object; generating an image of the object during the steps of controlling to obtain a signal having high frequency defect data; and processing the high frequency defect data to determine the defects in real-time.

2. The method is claimed in claim 1 wherein the object is a display panel.

3. The method as claimed in claim 2 wherein the display panel is a CRT panel whose spatial variations in optical density result from panel shape.

4. The method as claimed in claim 1 wherein the step of generating includes the step of moving the at least one camera and the at least one light source relative to the object to scan the object along a scan direction during the step of controlling.

5. A system for detecting defects in a transparent object having spatial variations in its optical density, the system comprising: a support for supporting the object; means including a first light source and a first camera spaced from the first light source for generating a first image of the object supported therebetween to obtain a first signal having a first set of high frequency defect data; a controller for controlling the first light source so that the first light source creates a variable light profile which is transmitted through the object; and means for processing the first set of high frequency defect data to determine the defects in realtime .

6. The system as claimed in claim 5 wherein the object is a display panel.

7. The system as claimed in claim 6 wherein the display panel is a CRT panel whose variations in optical density result from panel shape.

8. The system as claimed in claim 5 further comprising drive means for moving the first light source and the first camera to scan the object along a scan direction during control of the first light source.

9. The system as claimed in claim 8 wherein the support includes a conveyor for supporting the object at a position between the first light source and the first camera.

10. The system as claimed in claim 8 wherein the first camera and the first light source are moved synchronously by the drive means .

11. The system as claimed in claim 5 wherein the first light source is an LED array.

12. The system as claimed in claim 11 wherein the LED array is segmented and wherein each segment of the LED array is separately controlled by the controller.

13. The system as claimed in claim 10 wherein the drive means includes first and second motors for driving the first camera and the first light source, respectively.

14. The system as claimed in claim 13 wherein each of the motors is a servo motor.

15. The system as claimed in claim 5 further comprising a second light source and at least one second camera spaced from the second light source for generating a second image of the object supported therebetween to obtain a second signal having a second set of high frequency defect data and wherein the means for processing processes the first and second sets of high frequen- cy defect data to determine the defects in real-time, the controller also controlling the second light source.

16. The system as claimed in claim 15 further comprising drive means for moving the first and second light sources and the first and at least one second camera to scan the object along a scan direction during control of the first and second light sources.

17. The system as claimed in claim 16 wherein the support includes a conveyor for supporting the object at a position between the first light source and the first camera and between the at least one second camera and the second light source.

18. The system as claimed in claim 16 wherein the first and at least one second cameras and the first and second light sources are moved synchronously by the drive means .

19. The system as claimed in claim 15 wherein each of the first and at least one second light sources is an LED array.

20. The system as claimed in claim 18 wherein the drive means includes first and second motors for driving the first and at least one second cameras and the first and second light sources, respectively.