Inspection apparatus, control method, and program
An inspection apparatus ( 100 ) detects an inspection object ( 90 ) from first image data ( 10 ) in which the inspection object ( 90 ) is included. The inspection apparatus ( 100 ) generates second image data ( 20 ) by performing a geometric transform on the first image data ( 10 ) in such a way that a view of the detected inspection object ( 90 ) becomes a view satisfying a predetermined reference. In an inference phase, the inspection apparatus ( 100 ) detects, by using an identification model for detecting an abnormality of the inspection object ( 90 ), an abnormality of the inspection object ( 90 ) included in the second image data ( 20 ). Further, in a learning phase, the inspection apparatus ( 100 ) learns, by using the second image data ( 20 ), an identification model for detecting an abnormality of the inspection object ( 90 ).
1. An inspection apparatus comprising:
a memory storing instructions; and
a processor configured to execute the instructions to:
detect an inspection object from first image data in which the inspection object is included;
generate second image data by performing a geometric transform on the first image data in such a way that a view of the detected inspection object satisfies a predetermined reference; and
detect, by using an identification model, an abnormality of the inspection object included in the second image data, wherein
the identification model is learned by using image data in which learning inspection objects having views satisfying the predetermined reference are included, and
in the geometric transform, a degree of divergence between a second parameter representing a second view of the inspection object on the second image data and the reference becomes smaller than a degree of divergence between a first parameter representing a first view of the inspection object on the first image data and the reference.
2. The inspection apparatus according to claim 1 , wherein
the second image data is generated by
computing the first parameter representing the first view of the inspection object on the first image data;
acquiring a reference parameter representing the predetermined reference; and
determining, based on the first parameter and the reference parameter, the geometric transform to apply to the first image data in such a way that the degree of divergence between the second parameter representing the second view of the inspection object on the second image data and the reference parameter is reduced.
3. The inspection apparatus according to claim 2 , wherein
the first parameter includes any one of a position where a line representing the inspection object and any one edge of the first image data intersect each other, a size of an angle formed by the line representing the inspection object and the any one edge of the first image data, a length of a normal line drawn from an origin to the line representing the inspection object, and an angle of the normal line.
4. The inspection apparatus according to claim 1 , wherein
the inspection object is detected from the first image data by using teaching data in which an image area representing the inspection object on the first image data is specified.
5. The inspection apparatus according to claim 1 , wherein the processor is configured to execute the instructions to further:
display an image area representing the inspection object detected from the first image data; and
accept an input for correcting the image area representing the inspection object.
6. The inspection apparatus according to claim 1 , wherein the processor is configured to execute the instructions to further:
determine an area on the first image data equivalent to an abnormal part of the inspection object detected from the second image data and
display a predetermined display superimposed on the determined area.
7. The inspection apparatus according to claim 1 , wherein
the geometric transform uses the second parameter as the reference.
8. An inspection apparatus comprising:
a memory configured to store instructions; and
a processor configured to execute the instructions including:
detect an inspection object from first image data in which the inspection object is included;
generate second image data by performing a geometric transform on the first image data in such a way that a view of the detected inspection object satisfies a predetermined reference; and
learn, by using the second image data, an identification model for detecting an abnormality of the inspection object, wherein
in the geometric transform, a degree of divergence between a second parameter representing a second view of the inspection object on the second image data and the reference becomes smaller than a degree of divergence between a first parameter representing a first view of the inspection object on the first image data and the reference.
9. The inspection apparatus according to claim 8 , wherein
the second image data is generated by
computing the first parameter representing the first view of the inspection object on the first image data;
acquiring a reference parameter representing the predetermined reference; and
determining, based on the first parameter and the reference parameter, the geometric transform to be applied to apply to the first image data in such a way that the degree of divergence between the second parameter representing the second view of the inspection object on the second image data and the reference parameter is reduced.
10. The inspection apparatus according to claim 9 , wherein
the first parameter includes any one of a position where a line representing the inspection object and any one edge of the first image data intersect each other, a size of an angle formed by the line representing the inspection object and the any one edge of the first image data, a length of a normal line drawn from an origin to the line representing the inspection object, and an angle of the normal line.
11. The inspection apparatus according to claim 8 , wherein
the inspection object is detected from the first image data by using teaching data in which an image area representing the inspection object on the first image data is specified.
12. The inspection apparatus according to claim 8 , wherein the processor is configured to execute the instructions to further:
display indicating an image area representing the inspection object detected from the first image data; and
accept an input for correcting the image area representing the inspection object.
13. The inspection apparatus according to claim 8 , wherein the processor is configured to execute the instructions to further:
determine an area on the first image data equivalent to an abnormal part of the inspection object detected from the second image data and
display a predetermined display superimposed on the determined area.
14. A control method comprising:
detecting, by a processor, an inspection object from first image data in which the inspection object is included;
generating, by the processor, second image data by performing a geometric transform on the first image data in such a way that a view of the detected inspection object satisfies a predetermined reference; and
detecting, by the processor using an identification model, an abnormality of the inspection object included in the second image data, wherein
the identification model is learned by using image data in which learning inspection objects having views satisfying the predetermined reference are included, and
in the geometric transform, a degree of divergence between a second parameter representing a second view of the inspection object on the second image data and the reference becomes smaller than a degree of divergence between a first parameter representing a first view of the inspection object on the first image data and the reference.
15. The control method according to claim 14 , wherein further comprising:
generating the second image data comprises:
computing the first parameter representing the first view of the inspection object on the first image data;
acquiring a reference parameter representing the predetermined reference; and
determining, based on the first parameter and the reference parameter, the geometric transform to apply to the first image data in such a way that the degree of divergence between the second parameter representing the second view of the inspection object on the second image data and the reference parameter is reduced.
16. The control method according to claim 15 , wherein
the first parameter includes any one of a position where a line representing the inspection object and any one edge of the first image data intersect each other, a size of an angle formed by the line representing the inspection object and the any one edge of the first image data, a length of a normal line drawn from an origin to the line representing the inspection object, and an angle of the normal line.
17. The control method according to claim 14 , wherein
the inspection object is detected from the first image data by using teaching data in which an image area representing the inspection object on the first image data is specified.
18. The control method according to claim 14 , further comprising:
displaying, by the processor, an image area representing the inspection object detected from the first image data; and
accepting, by the processor, an input for correcting the image area representing the inspection object.
19. The control method according to claim 14 , further comprising:
determining, by the processor, an area on the first image data equivalent to an abnormal part of the inspection object detected from the second image data; and
displaying, by the processor, a predetermined display superimposed on the determined area.
20. A non-transitory computer readable medium storing a program executable by a computer to perform the control method according to claim 14 .