Cognex Corporation

Simon Barker

David J Michael

Adam Wagman

Aaron S Wallack

Andrew Hoelscher

William M Silver

Drew Hoelscher

This invention applies dynamic weighting between a point-to-plane and point-to-edge metric on a per-edge basis in an acquired image using a vision system. This allows an applied ICP technique to be significantly more robust to a variety of object geometries and/or occlusions. A system and method herein provides an energy function that is minimized to generate candidate 3D poses for use in alignment of runtime 3D image data of an object with model 3D image data. Since normals are much more accurate than edges, the use of normal is desirable when possible. However, in some use cases, such as a plane, edges provide information in relative directions the normals do not. Hence the system and method defines a 'normal information matrix', which represents the directions in which sufficient information is present. Performing (e.g.) a principal component analysis (PCA) on this matrix provides a basis for the available information.

System and method for simultaneous consideration of edges and normals in image features by a vision system

Techniques include systems, computerized methods, and computer readable media for choosing placement of three-dimensional (3D) probes used for evaluating a 3D alignment pose of a runtime 3D image inside a 3D alignment system for estimating the pose of a trained 3D model image in that 3D runtime image. A plurality of features associated with a first plurality of points of interest from a 3D image are generated, wherein each feature includes data indicative of 3D properties of an associated point from the plurality of points of interest. A second plurality of points of interest are selected from among the first plurality of points of interest, based at least in part on the plurality of features associated with the first plurality of points of interest. Placements of a plurality of 3D probes are determined based at least in part on the second plurality of points of interest.

Selection of balanced-probe sites for 3-D alignment algorithms

A system and method for selecting among 3D alignment algorithms in a 3D vision system is provided. The system and method includes a 3D camera assembly to acquire at least a runtime image defined by a 3D point cloud or runtime 3D range image having features of a runtime object and a vision system processor. A training image is provided. It is defined by a 3D point cloud or 3D range image having features of a model. A selection process is operated by the vision processor. It analyzes at least one training region of the training image having the features of the model and determines a distribution of surface normals in the at least one training region. It also selects, based upon a characteristic of the distribution, at least one 3D alignment algorithm from a plurality of available 3D alignment algorithms to align the features of the model with respect to the features of the runtime object.

System and method for automatic selection of 3D alignment algorithms in a vision system

The present application discloses a probe placement module for placing probes on a virtual object depicted in an image. The probe placement module is configured to place probes on interest points of an image so that the probes can accurately represent a pattern depicted in the image. The probe placement module can be configured to place the probes so that the probes can extract balanced information on all degrees of freedom associated with the pattern's movement, which improves the accuracy of the model generated from the probes.

Probe placement for image processing

A system and method for training multiple pattern recognition and registration models commences with a first pattern model. The model is trained from multiple images. Composite models can be used to improve robustness or model small differences in appearance of a target region. Composite models combine data from noisy training images showing instances of underlying patterns to build a single model. A pattern recognition and registration model is generated that spans the entire range of appearances of the target pattern in the set of training images. The set of pattern models can be implemented as either separate instances of pattern finding models or as a pattern multi-model. The underlying models can be standard pattern finding models or pattern finding composite models, or a combination of both.

Semi-supervised method for training multiple pattern recognition and registration tool models

Systems and methods for training a machine vision system create geometric models. The disclosed methods can extract one or more corresponding features and one or more differentiating features from different sets of training images. The one or more differentiating features can be used to differentiate between the different work pieces. The disclosed methods can generate an alignment model using the corresponding features and a classification model using the one or more differentiating features.

Systems and methods for classification and alignment of highly similar or self-similar patterns

A method for training a pattern recognition algorithm including the steps of identifying the known location of the pattern that includes repeating elements within a fine resolution image, using the fine resolution image to train a model associated with the fine image, using the model to examine the fine image resolution image to generate a score space, examining the score space to identify a repeating pattern frequency, using a coarse image that is coarser than the finest image resolution image to train a model associated with the coarse image, using the model associated with the coarse image to examine the coarse image thereby generating a location error, comparing the location error to the repeating pattern frequency and determining if the coarse image resolution is suitable for locating the pattern within a fraction of one pitch of the repeating elements.

Method and apparatus for training a probe model based machine vision system

A method for training a pattern recognition algorithm for a machine vision system that uses models of a pattern to be located, the method comprising the steps of training each of a plurality of models using a different training image wherein each of the training images is a version of a single image of the pattern at a unique coarse image resolution, using the models to identify at least one robust image resolution where the image resolution is suitable for locating the pattern within an accuracy limit of the actual location of the pattern in the image and storing the at least one robust image resolution for use in subsequent pattern recognition processes.

Growing community of inventors

Sudbury, MA, United States of America

Simon Barker