Apparatus and method for classification biological species based on deep learning using images
The present invention relates to an apparatus and method for classification of biological species based on deep learning using images, which construct a biological species classification model through image learning using deep learning for specific traits, which are unique morphological characterist...
Saved in:
Main Authors | , |
---|---|
Format | Patent |
Language | English Korean |
Published |
05.01.2023
|
Subjects | |
Online Access | Get full text |
Cover
Summary: | The present invention relates to an apparatus and method for classification of biological species based on deep learning using images, which construct a biological species classification model through image learning using deep learning for specific traits, which are unique morphological characteristics of biological species, and allow prediction of biological species using only images for specific traits. The present invention collects image data to construct the biological species classification model, corrects the collected images, constructs a species classification model for classifying biological species by learning the corrected images using a MobileNet-based convolutional neural network (CNN) algorithm, confirms the species classification model by verifying the model through k-fold cross validation (k = 3) to confirm the accuracy of the constructed species classification model, and predicts biological species by processing images subject to classification in real time in the field using the confirmed species classification model. According to the present invention, even non-majors without specialized knowledge of animals and plants can easily and conveniently classify biological species.
생물학적 종(생물종)(Biological Species)에 대해 고유한 형태적 특징인 특정 형질을 딥러닝(Deep Learning)을 이용한 이미지 학습을 통해 생물종 분류 모델을 구축하고, 특정 형질에 대한 이미지만을 이용하여 생물학적 종을 예측할 수 있도록 한 이미지를 이용한 딥러닝 기반의 생물종 분류장치 및 방법에 관한 것으로서, 생물종 분류 모델을 구축하기 위한 이미지 데이터를 수집하고, 수집한 이미지를 보정하며, 보정된 이미지를 Mobilenet 기반의 Convolutional Neural Network(CNN) 알고리즘을 이용하여 이미지를 학습하여 생물종 분류를 위한 종분류 모델을 구축하고, 구축된 종분류 모델의 정확도를 확인하기 위해 k-fold cross validation(k = 3)의 교차검증 방법으로 모델을 검증하여 종 분류 모델을 확정하고, 확정한 종분류 모델로 현장에서 실시간으로 분류 대상 이미지를 처리하여 생물종을 예측한다. |
---|---|
Bibliography: | Application Number: KR20210085150 |