Apparatus for discriminating bad cocoons based on neural network model

The present invention relates to an apparatus for discriminating bad cocoon based on a neural network model, capable of rapidly and accurately discriminating the generation of a defect by analyzing a cocoon image based on a neural network model. The apparatus for discriminating bad cocoon based on a...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: LEE SOO JANG, KIM BALGEUM, LIM JONGGUK, LEE AH YEONG
Format: Patent
Sprache:eng ; kor
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The present invention relates to an apparatus for discriminating bad cocoon based on a neural network model, capable of rapidly and accurately discriminating the generation of a defect by analyzing a cocoon image based on a neural network model. The apparatus for discriminating bad cocoon based on a neural network model includes: a conveyor unit transferring a rotatory cocooning frame on which multiple n×m cocoons are hung in a two-dimensional matrix form and transferring the rotatory cocooning frame again after temporarily stopping the transfer of the rotatory cocooning frame for a predetermined time when the rotatory cocooning frame reaches a predetermined image obtaining position; an LED spot illumination unit including m LED light sources arranged in a row and radiating LED light by row unit of the rotatory cocooning frame while moving the m LED light sources at fixed intervals on the upper part of the image obtaining position in the conveyor unit; an image obtaining unit positioned in the lower part of the image obtaining position in the conveyor unit and repeatedly repeating and performing a motion of obtaining m cocoon images by row unit at the same time by photographing and analyzing the rotatory cocooning frame, n times; and a defect discriminating unit including the neural network model in which a correlation between a defective state and the cocoon image is learned in advance. The defect discriminating unit also confirms and reports whether the bad cocoon is generated based on an output of the neural network model after inputting n×m cocoon images in the neural network model in order. 본 발명의 신경망 모델 기반의 불량 누에고치 판별장치에 관한 것으로, 이는 n×m개의 다수의 누에고치가 2차원 매트릭스 형태로 달려있는 회전섶을 이송하되, 회전섶이 기 설정된 영상 획득 위치에 도달하면 회전섶 이송을 기 설정 시간 동안 일시 중지시킨 후 다시 이송시키는 컨베이어부; 일렬 배치되는 m개의 LED 광원을 구비하고, 상기 m개의 LED 광원을 상기 컨베이어부내 영상 획득 위치의 상부 상에서 일정 간격으로 이동시키면서 회전섶의 열 단위로 LED 광을 조사하는 LED 스팟 조명부; 상기 컨베이어부내 영상 획득 위치의 하부에 위치되어, 상기 회전섶을 촬영 및 분석하여 m개의 누에고치 이미지를 열 단위로 동시 획득하는 동작을 n개 반복 수행하는 영상 획득부; 및 누에고치 이미지와 불량 상태간 상관관계가 사전 학습된 신경망 모델을 구비하고, 상기 신경망 모델에 n×m개의 누에고치 이미지를 순차 입력한 후, 상기 신경망 모델의 출력 기반으로 불량 누에고치 발생 여부를 확인 및 통보하는 불량 판별부를 포함할 수 있다.