Optimizing a set-up stage in an automatic visual inspection process
Embodiments of the invention provide a system and method for a visual inspection process, in which images of objects on an inspection line are automatically grouped into clusters based on the appearance of the object in the images. The clustered images create a broad database of images that can be used as a reference for inspected items having different appearances, which ensures that all parts and appearances of an inspected object can be identified and inspected, thereby increasing the success of defect detection and substantially minimizing false detection of defects.
1 . A visual inspection method of a manufacturing plant, the method comprising:
detecting an object in a plurality of set-up images of an inspection line of the manufacturing plant, in each set-up image the object conforming to a criterion;
grouping the images into a plurality of separate clusters based on values of the criterion;
detecting the object in an inspection image of the inspection line;
comparing the inspection image to a cluster of the plurality of separate clusters based on a value of the criterion of the object in the inspection image; and
operating the manufacturing plant and detecting a defect on the object in the inspection image during operation of the manufacturing plant based on the comparison to increase a speed of visual inspections during production processes; and
determining that each image assigned to the cluster of the plurality of separate clusters does not show perspective distortion compared to all other images assigned to the cluster, prior to comparing the inspection image to the cluster of the plurality of separate clusters.
2 . The method of claim 1 , wherein grouping the images into the plurality of separate clusters based on values of the criterion comprises assigning a first image to a first cluster and assigning a second image to a second cluster if a difference between a value of the criterion in the first image and a value of the criterion in the second image is above a threshold.
3 . The method of claim 2 , wherein the threshold is based on the criterion and on the object.
4 . The method of claim 1 , wherein the set-up images comprise images of defect-free objects and the inspection image comprises either a defect-free or defected object.
5 . The method of claim 1 , further comprising:
determining whether a cluster of the plurality of separate clusters achieved a complete representation of the object, and if a complete representation of the object is achieved then detecting a defect on the object in the inspection image.
6 . The method of claim 5 , wherein determining whether the cluster of the plurality of separate clusters achieved a complete representation of the object comprises determining a number of images in the cluster.
7 . The method of claim 5 , wherein determining whether the cluster of the plurality of separate clusters achieved a complete representation of the object comprises comparing images assigned to the cluster to each other.
8 . The method of claim 7 , wherein determining whether the cluster of the plurality of separate clusters achieved a complete representation of the object comprises determining whether each image assigned to the cluster can be used as a distortion-less reference to all other images assigned to the cluster.
9 . The method of claim 1 , wherein if a complete representation of the object is not achieved, then the method comprises utilizing a successive image assigned to the cluster of the plurality of separate clusters to update a database of reference images.
10 . The method of claim 1 , further comprising:
displaying the cluster of the plurality of separate clusters to a user for approval.
11 . The method of claim 1 , further comprising:
retroactively detecting defects in images assigned to a cluster of the plurality of separate clusters.
12 . A system for visual inspection in a manufacturing plant, the system comprising:
a processor in communication with a display, the processor being configured to;
receive an image of an object on an inspection line of the manufacturing plant,
detect the object in the image,
assign the image to a cluster of a plurality of separate clusters based on appearance of the object in the image,
determine a status of the cluster of the plurality of separate clusters, based on a determination that the cluster achieved a complete representation of the object,
determine that an inspection algorithm should be applied on the image based on the status of the cluster, thereby achieving inspection of some objects before all clusters achieved complete representation of the object; and
determine that each image assigned to the cluster of the plurality of separate clusters does not show perspective distortion compared to all other images assigned to the cluster, prior to comparing the image to the cluster of the plurality of separate clusters;
wherein the manufacturing plant is operated and a defect on the object in the image is detected during operation of the manufacturing plant based on a comparison to increase a speed of visual inspections during production processes.
13 . The system of claim 12 , wherein the processor is further configured to determine when the cluster of the plurality of separate clusters achieved a complete representation of the object by determining when each image assigned to the cluster of the plurality of separate clusters can be used as a distortion-less reference to all other images assigned to the cluster.