Please use this identifier to cite or link to this item: http://hdl.handle.net/11455/35435
標題: 類神經網路應用於糙米外觀品質檢測之研究
Appearance Quality Inspection in Brown Rice Using Artificial Neural Network
作者: YANG, CHIH - CHAU
楊智超
關鍵字: brown rice;糙米;appearance quality;neural network;外觀品質;類神經網路
出版社: 農業機械工程學系
摘要: 
本研究主要在結合類神經網路與影像處理技術,建立一套有效的糙米分類
系統。利用彩色CCD攝影機一次對30顆台農67號米取像後,首先以影像分
割將各顆米粒由整幅影像中獨立出來,再由電腦程式分別計算出各顆糙米
之外觀分類參數,包括相對投影面積、面積周長比、形狀細密度、紅色光
度平均值、綠色光度平均值、藍色光度平均值、紅色光度值標準差、綠色
色光度值標準差、藍色色光度值標準差、綠色對紅色光度分量比值、白堊
質比、短軸比、4個副短軸與主短軸差之和、4個副短軸與主短軸絕對值差
之和、左右端寬度比值及左右端寬度差等分類參數;配合類神經網路學習
歸納的能力,嘗試將糙米分為屑米、變色粒、發芽粒、褐色粒、異型粒、
碎粒、白堊質粒、畸形粒、未熟粒及完整粒等十種類別。另外,以下方打
光方式對糙米取像,並規畫出直方圖法及區塊法兩種參數擷取模式,配合
類神經網路對糙米進行胴裂檢測。糙米分類及胴裂檢測之最佳模式,乃藉
由調整網路參數而得到。

This study is to establish an effective brown rice classifying
system with combining artificial neural networks and image
processing technique.After color CCD camera grabs images of 30
"Tainung 67" brown rice at one time, image segmentation is used
to isolate each kernel from the whole image frame, the following
is to calculate the outward classifying parameter of each kernel
by computer programs. Image processing technique is operated in
coordination with learning and inductive ability in neural
networks, therefore, brown rice is classified into sound,
immature, abnormal, chalky, broken, off-type, rusty, spouted,
discolored kernel, and rice screening ten categories.Moreover,
back lighting grabs brown rice images, the result builds two
parameter models, histogram and block, which are used to detect
crack in brown rice with neural networks.The best model in brown
rice classification and fissure detection could be gotten by
adjusting network parameters.
URI: http://hdl.handle.net/11455/35435
Appears in Collections:生物產業機電工程學系

Show full item record
 

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.