Image analysis – Applications – Manufacturing or product inspection
Reexamination Certificate
1998-05-01
2001-01-16
Mehta, Bhavesh (Department: 2721)
Image analysis
Applications
Manufacturing or product inspection
C382S190000, C382S216000, C117S201000
Reexamination Certificate
active
06175644
ABSTRACT:
FIELD OF THE INVENTION
This invention relates to machine vision for industrial processes, and more particularly relates to machine vision techniques for analyzing feature information about an object based on an acquired image of the object.
BACKGROUND OF THE INVENTION
Machine vision systems are increasingly employed to replace human vision in a wide range of industrial processes such as manufacturing operations. A machine vision system typically provides automated, computer-based image acquisition and analysis capabilities that can be employed for tasks such as, e.g., measurement and inspection of parts or materials, or monitor of the state of a manufacturing process. For such tasks, a machine vision system typically is configured with a camera for acquiring an image of a process environment or an object of interest, e.g., a part being produced, and further is configured with processing functionality to process the acquired image and produce information about the process or object. Frequently, in part measurement or process monitoring tasks, the object image acquisition and analysis functions are tailored to extract and analyze specific process and/or object features that are of interest for the selected task and given manufacturing process.
For many manufacturing environments, reliable machine vision analysis of images of the manufacturing process is made difficult by variable and sometimes unpredictable conditions in the environment. For example, changes in the surface characteristics of a part under analysis can cause unexpected fluctuations in part features as they appear in images of the part, leading to potentially faulty conclusions being made about the part based on the images. Similarly, changes in lighting conditions of a manufacturing process or in the reflectivity of parts involved in the process can cause shifts in features as they appear in images of the process, leading to potentially faulty conclusions being made about the process status based on the images. Optical and/or electrical noise generated by the manufacturing process systems can further exacerbate fluctuations in imaged features of a part or process and in addition can introduce spurious, invalid feature data.
But optimally, machine vision analysis of a manufacturing process or part being processed is based on imaged features of interest for which there is associated a high degree of confidence in their validity. Inconclusive and unreliable analyses can otherwise result. This is especially true for complicated manufacturing processes characterized by a high degree of unpredictability in conditions of the process which are to be analyzed. Ideally then, machine vision image analysis and feature extraction is robust with respect to variability, noise, unpredictability, and other such characteristics of a process to be monitored.
Conventional machine vision filtering and thresholding techniques, while typically capable of compensating for predictable, generally uncomplicated image feature fluctuation and noise, cannot typically accommodate unpredictable and complex image feature shifts and noise introduction. In addition, typical filtering operations can affect “useful,” i.e., meaningful, data as well as extraneous, stray noise data. If the extraneous noise data is substantial, its removal can distort features of interest to an extent at which they lose meaning. As a result, noise removal operations preferably must be combined with complicated feature analysis operations to maintain integrity of the data.
Machine vision image acquisition and analysis is further complicated for a range many manufacturing processes in which full-view image acquisition of a part or process environment cannot be accomplished. Specifically, for complicated three-dimensional object shapes, and more specifically for complicated opaque objects, and for manufacturing tooling configurations, one or more process or object regions may obscure other such regions from the line-of-sight view of an image acquisition camera's position. As a consequence, it may not be possible from a single camera position to simultaneously view related processes or object features such as circumferential points of a complete cross-sectional object profile. In other words, unlike that of substantially two-dimensional objects or a two-dimensional object face, related features of a complicated and opaque three-dimensional object or process environment are not guaranteed to be together fully exposed for simultaneous image acquisition. Instead, only a portion of the object and a subset of related object features are likely to be fully exposed to a single image acquisition camera angle.
The complications of this scenario are compounded in many applications where the location of a single image acquisition camera is limited by the manufacturing environment; e.g., where an optimum camera location cannot be accommodated. For example, in a scenario where an orthogonal, top-down view of a three-dimensional object may be known to encompass a complete set of object features, such vertical location of the camera may not be practical. For a large class of manufacturing applications, accommodation can be made for only a single, oblique camera location that results in a acquisition of only a perspective view of an object; and for some camera angles this can result in a large fraction of related object features being obscured in the object image.
A further complication is added for machine vision applications in which an object to be viewed is moving, e.g., rotating, during the process being monitored. In this case, a different subset of related features, e.g., a different portion of an object's cross-sectional profile or shape, is in view of the camera's line-of-sight at any given time. Traditional vision system techniques, developed typically for orthogonal image acquisition and analysis of substantially two-dimensional object surfaces, are generally ineffective at addressing these combinations of complicated object configurations, object movement, and manufacturing constraints. The unpredictability, variability, and noise conditions often characteristic of such scenarios further are not effectively addressed by traditional vision system techniques, resulting in potentially questionable validity of machine vision image analysis of such scenarios.
SUMMARY OF THE INVENTION
The invention provides the ability to validate detected features in acquired images to thereby enhance the integrity of any analysis carried out on the detected and validated features. In a machine vision method provided by the invention for detecting features of an object in a selected feature plane of the object, a sequence of images of the object is acquired, each image in the sequence corresponding to a distinct orientation of the object about a selected object axis. Images in the image sequence are inspected for feature points of the selected feature plane, as-projected into the images, in a first feature detection location. Images in the image sequence are also inspected for feature points of the selected feature plane, as-projected into the images, in a second feature detection location. The second feature detection location is configured at an image position at which a feature point detected in the first feature detection location in a first inspected image is expected to appear in a second inspected image. Valid object feature points are identified as being those feature points which are detected in both the first feature detection location in a first inspected image and in the second feature detection location in a second inspected image of the image sequence.
The invention provides a wide range of adaptations of this technique; for example, images in the image sequence can be inspected for feature points of the selected feature plane, as projected into the images, in a number, n, of feature detection locations. Here the n feature detection locations are configured with respect to each other at image positions based on an expected shift in image location of a given feature point
Jacobson Lowell D.
Ruzhitsky Vladimir N.
Scola Joseph R.
Cognex Corporation
Lober Theresa A.
Mehta Bhavesh
LandOfFree
Machine vision system for object feature analysis and... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Machine vision system for object feature analysis and..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Machine vision system for object feature analysis and... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2552083