Visual inspection apparatus, visual inspection method, image generation apparatus, and image generation method
AssigneeHitachi Astemo, Ltd.
InventorsHideaki Onozuka, Takateru Seki, Takamasa Imaizumi, Hiroyuki Hayashi
Life of this patent
Estimating the term…
Abstract
A visual inspection apparatus is configured to inspect a surface of a piston based on a captured image acquired by a camera and a learning result acquired by conducting machine learning using a plurality of defective product sample images, each of the defective product sample images being generated by combining a two-dimensional image of a defect image that is generated based on a three-dimensionally pre-generated defect model with an image of the surface of the piston.
Claims7
A visual inspection apparatus configured to inspect a surface of an inspection target, the visual inspection apparatus comprising:
- an imaging portion configured to image the surface of the inspection target;
- a storage portion configured to store a learning result acquired by conducting machine learning using a plurality of defective product sample images, each of the plurality of defective product sample images being generated by combining a two-dimensional image of a defect shape that is generated based on a three-dimensionally pre-generated defect model with an image of the surface of the inspection target;
- and
- an inspection portion configured to inspect the surface of the inspection target based on a captured image acquired by the imaging portion and the learning result,
- wherein:
- the two-dimensional image is generated in consideration of a luminance with respect to the three-dimensionally pre-generated defect model;
- the luminance with respect to the three-dimensionally pre-generated defect model is a luminance distribution in a predetermined range containing a defect and a predetermined portion around the defect in the three-dimensionally pre-generated defect model;
- and
- the luminance distribution is acquired based on:
- a physical amount regarding a preset illumination direction at coordinates of a point cloud of a surface in the predetermined range that is converted from the three-dimensionally pre-generated defect model,
- a physical amount regarding a preset imaging direction at the coordinates of the point cloud;
- and
- a physical amount regarding a normal direction to the defect in a plane containing the preset illumination direction and the preset imaging direction.
A visual inspection method for inspecting a surface of an inspection target by a computer, the visual inspection method comprising:
- imaging the surface of the inspection target;
- and
- inspecting the surface of the inspection target based on a captured image acquired in the imaging and a learning result acquired by conducting machine learning using a plurality of defective product sample images, each of the plurality of defective product sample images being generated by combining a two-dimensional image of a defect shape that is generated based on a three-dimensionally pre-generated defect model with an image of the surface of the inspection target,
- wherein:
- the two-dimensional image is generated in consideration of a luminance with respect to the three-dimensionally pre-generated defect model;
- the luminance with respect to the three-dimensionally pre-generated defect model is a luminance distribution in a predetermined range containing a defect and a predetermined portion around the defect in the three-dimensionally pre-generated defect model;
- and
- the luminance distribution is acquired based on:
- a physical amount regarding a preset illumination direction at coordinates of a point cloud of a surface in the predetermined range that is converted from the three-dimensionally pre-generated defect model,
- a physical amount regarding a preset imaging direction at the coordinates of the point cloud;
- and
- a physical amount regarding a normal direction to the defect in a plane containing the preset illumination direction and the preset imaging direction.
Dependencies are inferred from the claim text.