Information processing system and learning method

Abstract

An information processing system (1) is provided with a movement amount data acquisition unit (109), an image data acquisition unit (101), a difference expression extraction unit (108), and a mirror region specification unit (111). A movement amount data acquisition unit (109) acquires movement amount data of an object. An image data acquisition unit (101) acquires an image data item captured by an imaging device mounted on an object. A difference expression extraction unit (108) extracts, on the basis of two image data items imaged before and after the movement of the object, a difference expression representing feature information on the difference between the two images. A mirror surface region specifying unit (111) specifies a mirror surface region on the basis of the correlation between the difference expression and the movement amount data.

Description

technical field [0001] The present invention relates to an information processing system and a learning method. Background technique [0002] In object detection using vision sensors, objects reflected in mirrors are also detected in the same way as in other ways. Therefore, it is difficult to identify the mirror itself. Non-Patent Document 1 proposes a technique in which a predetermined pattern moving in a specific direction is captured by video, and an area where the same pattern moves in a different direction is extracted and recognized as a specular area. [0003] Non-patent literature [0004] Non-Patent Document 1: "REGION IDENTIFICATION TECHNIQUE FOR MIRROR SURFACE / TRANSPARENT OBJECT", Toshiba Review Vol.68No,10 (2013) SUMMARY OF THE INVENTION [0005] However, in the conventional technique described above, when another pattern showing a similar action exists in the imaging area, there is a possibility that the detection area of ​​the pattern is erroneously recog...

AI Technical Summary

Problems solved by technology

Therefore, it is difficult to identify the mirror itself

Images