Abstract: This invention provides a system and method for finding patterns in images that incorporates neural net classifiers. A pattern finding tool is coupled with a classifier that can be run before or after the tool to have labeled pattern results with sub-pixel accuracy. In the case of a pattern finding tool that can detect multiple templates, its performance is improved when a neural net classifier informs the pattern finding tool to work only on a subset of the originally trained templates. Similarly, in the case of a pattern finding tool that initially detects a pattern, a neural network classifier can then determine whether it has found the correct pattern. The neural network can also reconstruct/clean-up an imaged shape, and/or to eliminate pixels less relevant to the shape of interest, therefore reducing the search time, as well significantly increasing the chance of lock on the correct shapes.

Abstract: This invention provides a system and method for finding multiple line features in an image. Two related steps are used to identify line features. First, the process computes x and y-components of the gradient field at each image location, projects the gradient field over a plurality subregions, and detects a plurality of gradient extrema, yielding a plurality of edge points with position and gradient. Next, the process iteratively chooses two edge points, fits a model line to them, and if edge point gradients are consistent with the model, computes the full set of inlier points whose position and gradient are consistent with that model. The candidate line with greatest inlier count is retained and the set of remaining outlier points is derived. The process then repeatedly applies the line fitting operation on this and subsequent outlier sets to find a plurality of line results. The process can be exhaustive RANSAC-based.

Abstract: This invention provides a system and method for runtime determination (self-diagnosis) of camera miscalibration (accuracy), typically related to camera extrinsics, based on historical statistics of runtime alignment scores for objects acquired in the scene, which are defined based on matching of observed and expected image data of trained object models. This arrangement avoids a need to cease runtime operation of the vision system and/or stop the production line that is served by the vision system to diagnose if the system's camera(s) remain calibrated. Under the assumption that objects or features inspected by the vision system over time are substantially the same, the vision system accumulates statistics of part alignment results and stores intermediate results to be used as indicator of current system accuracy. For multi-camera vision systems, cross validation is illustratively employed to identify individual problematic cameras.

Abstract: This invention provides a system and method for finding multiple line features in an image. Two related steps are used to identify line features. First, the process computes x and y-components of the gradient field at each image location, projects the gradient field over a plurality subregions, and detects a plurality of gradient extrema, yielding a plurality of edge points with position and gradient. Next, the process iteratively chooses two edge points, fits a model line to them, and if edge point gradients are consistent with the model, computes the full set of inlier points whose position and gradient are consistent with that model. The candidate line with greatest inlier count is retained and the set of remaining outlier points is derived. The process then repeatedly applies the line fitting operation on this and subsequent outlier sets to find a plurality of line results. The process can be exhaustive RANSAC-based.

Abstract: This invention provides a system and method for finding patterns in images that incorporates neural net classifiers. A pattern finding tool is coupled with a classifier that can be run before or after the tool to have labeled pattern results with sub-pixel accuracy. In the case of a pattern finding tool that can detect multiple templates, its performance is improved when a neural net classifier informs the pattern finding tool to work only on a subset of the originally trained templates. Similarly, in the case of a pattern finding tool that initially detects a pattern, a neural network classifier can then determine whether it has found the correct pattern. The neural network can also reconstruct/clean-up an imaged shape, and/or to eliminate pixels less relevant to the shape of interest, therefore reducing the search time, as well significantly increasing the chance of lock on the correct shapes.

Abstract: This invention provides a system and method for selecting the correct profile from a range of peaks generated by analyzing a surface with multiple exposure levels applied at discrete intervals. The cloud of peak information is resolved by comparison to a model profile into a best candidate to represent an accurate representation of the object profile. Illustratively, a displacement sensor projects a line of illumination on the surface and receives reflected light at a sensor assembly at a set exposure level. A processor varies the exposure level setting in a plurality of discrete increments, and stores an image of the reflected light for each of the increments. A determination process combines the stored images and aligns the combined images with respect to a model image. Points from the combined images are selected based upon closeness to the model image to provide a candidate profile of the surface.

Abstract: This invention provides a system and method for calibration of a 3D vision system using a multi-layer 3D calibration target that removes the requirement of accurate pre-calibration of the target. The system and method acquires images of the multi-layer 3D calibration target at different spatial locations and at different times, and computes the orientation difference of the 3D calibration target between the two acquisitions. The technique can be used to perform vision-based single-plane orientation repeatability inspection and monitoring. By applying this technique to an assembly working plane, vision-based assembly working plane orientation repeatability, inspection and monitoring can occur. Combined with a moving robot end effector, this technique provides vision-based robot end-effector orientation repeatability inspection and monitoring. Vision-guided adjustment of two planes to achieve parallelism can be achieved.

Abstract: This invention provides a system and method for finding multiple line features in an image. Two related steps are used to identify line features. First, the process computes x and y-components of the gradient field at each image location, projects the gradient field over a plurality subregions, and detects a plurality of gradient extrema, yielding a plurality of edge points with position and gradient. Next, the process iteratively chooses two edge points, fits a model line to them, and if edge point gradients are consistent with the model, computes the full set of inlier points whose position and gradient are consistent with that model. The candidate line with greatest inlier count is retained and the set of remaining outlier points is derived. The process then repeatedly applies the line fitting operation on this and subsequent outlier sets to find a plurality of line results. The process can be exhaustive RANSAC-based.

Abstract: This invention provides a system and method for selecting the correct profile from a range of peaks generated by analyzing a surface with multiple exposure levels applied at discrete intervals. The cloud of peak information is resolved by comparison to a model profile into a best candidate to represent an accurate representation of the object profile. Illustratively, a displacement sensor projects a line of illumination on the surface and receives reflected light at a sensor assembly at a set exposure level. A processor varies the exposure level setting in a plurality of discrete increments, and stores an image of the reflected light for each of the increments. A determination process combines the stored images and aligns the combined images with respect to a model image. Points from the combined images are selected based upon closeness to the model image to provide a candidate profile of the surface.

Abstract: This invention provides a system and method for finding multiple line features in an image. Two related steps are used to identify line features. First, the process computes x and y-components of the gradient field at each image location, projects the gradient field over a plurality subregions, and detects a plurality of gradient extrema, yielding a plurality of edge points with position and gradient. Next, the process iteratively chooses two edge points, fits a model line to them, and if edge point gradients are consistent with the model, computes the full set of inlier points whose position and gradient are consistent with that model. The candidate line with greatest inlier count is retained and the set of remaining outlier points is derived. The process then repeatedly applies the line fitting operation on this and subsequent outlier sets to find a plurality of line results. The process can be exhaustive RANSAC-based.

Abstract: This invention provides a system and method for selecting the correct profile from a range of peaks generated by analyzing a surface with multiple exposure levels applied at discrete intervals. The cloud of peak information is resolved by comparison to a model profile into a best candidate to represent an accurate representation of the object profile. Illustratively, a displacement sensor projects a line of illumination on the surface and receives reflected light at a sensor assembly at a set exposure level. A processor varies the exposure level setting in a plurality of discrete increments, and stores an image of the reflected light for each of the increments. A determination process combines the stored images and aligns the combined images with respect to a model image. Points from the combined images are selected based upon closeness to the model image to provide a candidate profile of the surface.

Abstract: This invention provides a system and method for finding patterns in images that incorporates neural net classifiers. A pattern finding tool is coupled with a classifier that can be run before or after the tool to have labeled pattern results with sub-pixel accuracy. In the case of a pattern finding tool that can detect multiple templates, its performance is improved when a neural net classifier informs the pattern finding tool to work only on a subset of the originally trained templates. Similarly, in the case of a pattern finding tool that initially detects a pattern, a neural network classifier can then determine whether it has found the correct pattern. The neural network can also reconstruct/clean-up an imaged shape, and/or to eliminate pixels less relevant to the shape of interest, therefore reducing the search time, as well significantly increasing the chance of lock on the correct shapes.

Abstract: This invention provides a system and method for finding multiple line features in an image. Two related steps are used to identify line features. First, the process computes x and y-components of the gradient field at each image location, projects the gradient field over a plurality subregions, and detects a plurality of gradient extrema, yielding a plurality of edge points with position and gradient. Next, the process iteratively chooses two edge points, fits a model line to them, and if edge point gradients are consistent with the model, computes the full set of inlier points whose position and gradient are consistent with that model. The candidate line with greatest inlier count is retained and the set of remaining outlier points is derived. The process then repeatedly applies the line fitting operation on this and subsequent outlier sets to find a plurality of line results. The process can be exhaustive RANSAC-based.

Abstract: This invention provides a system and method for aligning a first work piece with an underlying second work piece in the presence of occlusion by the first work piece of critical alignment features of the second work piece. The vision system, which guides the motion of a manipulator holding the first work piece and a motion stage holding the second work piece, learns secondary alignment features at least one of the first and second work pieces. Using these secondary features, the vision system determines alignment between the work pieces and guides the manipulator and the motion stage to achieve alignment as the first work piece engages the second work piece. The secondary features are used to define a course alignment. Deterministic movements of the manipulator and/or motion stage are used to learn the relationship between the secondary and primary features. Secondary features are used to direct alignment.

Abstract: This invention provides a calibration target with a calibration pattern on at least one surface. The relationship of locations of calibration features on the pattern are determined for the calibration target and stored for use during a calibration procedure by a calibrating vision system. Knowledge of the calibration target's feature relationships allow the calibrating vision to image the calibration target in a single pose and rediscover each of the calibration features in a predetermined coordinate space. The calibrating vision can then transform the relationships between features from the stored data into the calibrating vision system's local coordinate space. The locations can be encoded in a barcode that is applied to the target, provided in a separate encoded element, or obtained from an electronic data source. The target can include encoded information within the pattern defining a location of adjacent calibration features with respect to the overall geometry of the target.

Abstract: This invention provides a system and method for selecting the correct profile from a range of peaks generated by analyzing a surface with multiple exposure levels applied at discrete intervals. The cloud of peak information is resolved by comparison to a model profile into a best candidate to represent an accurate representation of the object profile. Illustratively, a displacement sensor projects a line of illumination on the surface and receives reflected light at a sensor assembly at a set exposure level. A processor varies the exposure level setting in a plurality of discrete increments, and stores an image of the reflected light for each of the increments. A determination process combines the stored images and aligns the combined images with respect to a model image. Points from the combined images are selected based upon closeness to the model image to provide a candidate profile of the surface.

Abstract: This invention provides a system and method for finding multiple line features in an image. Two related steps are used to identify line features. First, the process computes x and y-components of the gradient field at each image location, projects the gradient field over a plurality subregions, and detects a plurality of gradient extrema, yielding a plurality of edge points with position and gradient. Next, the process iteratively chooses two edge points, fits a model line to them, and if edge point gradients are consistent with the model, computes the full set of inlier points whose position and gradient are consistent with that model. The candidate line with greatest inlier count is retained and the set of remaining outlier points is derived. The process then repeatedly applies the line fitting operation on this and subsequent outlier sets to find a plurality of line results. The process can be exhaustive RANSAC-based.

Abstract: This invention provides a system and method for finding multiple line features in an image. Two related steps are used to identify line features. First, the process computes x and y-components of the gradient field at each image location, projects the gradient field over a plurality subregions, and detects a plurality of gradient extrema, yielding a plurality of edge points with position and gradient. Next, the process iteratively chooses two edge points, fits a model line to them, and if edge point gradients are consistent with the model, computes the full set of inlier points whose position and gradient are consistent with that model. The candidate line with greatest inlier count is retained and the set of remaining outlier points is derived. The process then repeatedly applies the line fitting operation on this and subsequent outlier sets to find a plurality of line results. The process can be exhaustive RANSAC-based.

Abstract: This invention provides a system and method to validate the accuracy of camera calibration in a single or multiple-camera embodiment, utilizing either 2D cameras or 3D imaging sensors. It relies upon an initial calibration process that generates and stores camera calibration parameters and residual statistics based upon images of a first calibration object. A subsequent validation process (a) acquires images of the first calibration object or a second calibration object having a known pattern and dimensions; (b) extracts features of the images of the first calibration object or the second calibration object; (c) predicts positions expected of features of the first calibration object or the second calibration object using the camera calibration parameters; and (d) computes a set of discrepancies between positions of the extracted features and the predicted positions of the features.

Abstract: This invention provides a system and method for determining the pose of shapes that are known to a vision system that undergo both affine transformation and deformation. The object image with fiducial is acquired. The fiducial has affine parameters, including degrees of freedom (DOFs), search ranges and search step sizes, and control points with associated DOFs and step sizes. Each 2D affine parameter's search range and the distortion control points' DOFs are sampled and all combinations are obtained. The coarsely specified fiducial is transformed for each combination and a match metric is computed for the transformed fiducial, generating a score surface. Peaks are computed on this surface, as potential candidates, which are refined until a match metric is maximized. The refined representation exceeding a predetermined score is returned as potential shapes in the scene. Alternately the candidate with the best score can be used as a training fiducial.