IMAGE PROCESSING SYSTEM, AND IMAGE PROCESSING PROGRAM

To provide an image processing system capable of improving accuracy of learning with a classifier by creating a focused image in which points of interest in the image are stably placed while efficiently reducing non-defective images by using a discrete optimization algorithm at the stage of creating...

Full description

Saved in:
Bibliographic Details
Main Authors HARA MOEKO, KASHIWABARA KEN, NAKAMURA YOSHIAKI
Format Patent
LanguageEnglish
Japanese
Published 16.06.2022
Subjects
Online AccessGet full text

Cover

Loading…
More Information
Summary:To provide an image processing system capable of improving accuracy of learning with a classifier by creating a focused image in which points of interest in the image are stably placed while efficiently reducing non-defective images by using a discrete optimization algorithm at the stage of creating a training data set for a classifier using a neural network.SOLUTION: An image processing system for classifying images into desired classes includes: a detector optimization unit that optimizes a detector that detects a point of interest in an image using a discrete optimization algorithm; a learning data creation unit for classification, which creates a focused image in which points of interest detected by the obtained optimized detector are centrally located; and a classifier learning unit that learns classifiers with a neural network using the focus image.SELECTED DRAWING: Figure 1 【課題】 ニューラルネットワークによる分類器の学習用データセットを作る段階で離散最適化アルゴリズムを用いて良品画像を効率的に減らすと共に、画像における着目点を安定して配置させた着目点画像を作成することで、分類器による学習の精度を向上させる。【解決手段】 画像を所望のクラスに分類する画像処理システムであって、画像内の着目点を検出する検出器を離散最適化アルゴリズムにより最適化する検出器最適化部と、得られた最適化済み検出器により検出された着目点が中央に配置された着目点画像を作成する分類用学習データ作成部と、着目点画像を用いてニューラルネットワークによる分類器の学習を行う分類器学習部を備えた画像処理システム。【選択図】 図1
Bibliography:Application Number: JP20200201815