Television – Special applications – Flaw detector
Reexamination Certificate
2000-04-19
2002-10-01
Le, Vu (Department: 2613)
Television
Special applications
Flaw detector
C348S092000
Reexamination Certificate
active
06459448
ABSTRACT:
FIELD OF THE INVENTION
This invention relates to a computer system for inspecting objects for manufacturing defects and more particularly to an algorithm in the computer system for inspecting electronically acquired imagery of an object for a wide variety of manufacturing defects that might be present in the object.
BACKGROUND OF THE INVENTION
While advances in computer processing power have made processing of large amounts of data possible, an automated optical inspection problem for arrays of geometric targets still exists. Automated optical inspection is the inspection of electronically acquired imagery of an object for tolerance, color, blemishes, cracks or a wide variety of manufacturing defects that might be present in the object. The automated optical inspection problem arises, among other places, in the inspection of aperture masks used in CRT-type color monitors and television sets. Problems encountered during aperture masks inspections also occur during inspections of other objects, which include inspection of an array of annular targets with periodic data. Thus, two seemly different objects, such as flat panel displays and filters for filtering bacterial particles out of a product stream of bio-engineered vaccines and chemicals have the same inspection problem. Therefore, while the following discusses inspection of aperture masks with arrays of annular targets, it should be apparent to one of ordinary skill in the art that the invention relates to inspection of all objects with arrays of geometric targets.
Aperture masks generally are comprised of thin metal sheets perforated by hundreds of thousands of tiny holes. These holes are too small to see with an unaided eye and each hole has a precise shape, or profile. The shape of each hole varies slightly and definitely according to its position across the width and/or height of the mask. The degree to which the shape of these holes can be maintained in manufacturing of the aperture mask has a direct bearing on whether the mask can be used by a manufacturer.
Inspection of aperture masks during manufacturing is a particularly demanding problem because there are vast numbers of holes, at least several hundred thousand holes, in each mask. Aperture masks are relatively inexpensive to manufacture. Nonetheless, the shape of a single hole that is out of tolerance eventually shows up as a blemish that an end user can see in a finished product. For example, the end user will see a blemish in an image that is produced by a CRT-type color monitor, which includes the defective aperture mask. Thus, automated or manual inspection of each mask has to be performed. Automated inspection by a computer leads to a formidable data reduction problem since each hole has many pixels, thereby producing billions of pixels across the length and width of each mask.
Some manufacturers use human inspectors to manually inspect each mask but do not use any special magnification method. The human inspectors hold each mask up to the light and bend it in various ways to detect an irregular hole or area in the mask. While the manufacturing of the aperture masks is automated, inspections of the masks are performed by groups of inspectors. Each person in a group may inspect a particular section of the masks and inspections of the masks are performed at much slower rates than the rates at which they are manufactured. The manual inspection process also is a relatively expensive undertaking for the manufacturer.
A current automated method uses two-dimensional video cameras to analyze data on a computer. Thereafter various mathematical operations, such as edge detection, a gradient calculation, or some other type of transform to manipulate the data, are performed on the two-dimensional representation of the data. The automated method measures the inside and outside diameter of each hole to calculate whether the inner diameter is the right range of tolerance. This method generates an unusable and impractical amount of data that may only be analyzed by a very specialized and expensive computer. Moreover, by the time images large enough to analyze by video cameras are created, a microscopic processor is usually created for each hole. In an industrial environment, where there is a lot of vibration in the manufacturing process, masks that come out of the last stage of the manufacturing process vibrate slightly. This vibration is enough to make an image useless unless the image is taken over an extremely short time period.
What is needed, therefore, is an automated method for inspecting electronically acquired imagery of an object for a wide variety of manufacturing defects that might be present in the object. The method must be unaffected by the visual effects of mechanical vibration on an image while tolerating the effects of imperfect positioning of the target object by an operator or as a result of equipment shortcomings. The method also must work in a range of different object acceptance norms and must be easily adaptable to change from one set of norms to another. Moreover, the method must be expressible as a custom or semi-custom integrated circuit chip and it must scale well to a highly parallel implementation.
SUMMARY OF THE INVENTION
The present invention relates to a system and method for inspecting electronically acquired imagery, from a one-dimensional camera, of an object for tolerance, color, blemishes, cracks or a wide variety of manufacturing defects that might be present in an object. The method to detect manufacturing defects includes an algorithm for analyzing the pattern of geometric elements in an array and detecting deviations from numerical acceptance norms, such as diameter, spacing, and symmetry, for the geometric elements. In the inventive system, a one-dimensional video camera captures a single one-dimensional “slice” of target shape of an object with every line it scans. Each slice is broken down into “segments” consisting of sets of adjacent pixels that are similar in brightness, hue, or both. The one-dimensional camera, in conjunction with processing by ancillary electronic data processing means and methods, delivers the segments to the system where they are sequenced. The system identifies every segment to determine what feature of the target shape the segment represents. Thereafter, predefined rules are used to determine if each identified segment deviates from numerical acceptance norms. The inventive method is thus used to analyze periodic arrays of any arbitrary target shape, thereby working in a range of different object acceptance norms and being easily adaptable to change from one set of norms to another.
The algorithm used in the inventive method accomplishes at least two outcomes simultaneously. It serves as a framework, for the transformation of a set of measurements that could be made on a two-dimensional image of the target area into a set of measurements made only on the one-dimensional data set. In addition, the algorithm is capable of accurately measuring a lattice constant, i.e., the spacing between the geometric elements of an array that arises from a data set only after certain manipulations have been made. The lattice constant can be measured without prior knowledge of any scan line other than the one currently being observed. Thus, it is never necessary to assemble a two-dimensional depiction of the one-dimensional video data, or of any product data arising from the manipulation of the video data. This is important in a design of computational hardware to analyze the video data stream generated by any camera, such as a one-dimensional camera or a two-dimensional camera.
Specifically, in a preferred embodiment, the algorithm is used to inspect aperture masks with arrays of annular elements The inventive method uses a pattern rule, derived from the geometry of the annular elements, and a lattice constant rule for estimating the spacing between annular elements of an array, whereby not all possibilities in the analysis work. It should be noted that other rules may be used in inspecting other objects with arrays of ge
K-G Devices Corporation
Le Vu
Morgan & Lewis & Bockius, LLP
LandOfFree
System and method for automatically inspecting arrays of... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with System and method for automatically inspecting arrays of..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and System and method for automatically inspecting arrays of... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3000501