查詢結果分析
來源資料
頁籤選單縮合
題 名 | 胴裂米以類神經網路檢測之研究=Inspection of Cracked Rice Kernels by Using Neural Networks |
---|---|
作 者 | 萬一怒; 楊智超; 盛中德; 何榮祥; | 書刊名 | 農業工程學報 |
卷 期 | 44:1 1998.03[民87.03] |
頁 次 | 頁67-80 |
分類號 | 434.111 |
關鍵詞 | 胴裂米; 類神經網路; 稻米品質; 機械視覺; 影像處理; Cracked kernel; Neural network; Rice quality; Machine vision; Image processing; |
語 文 | 中文(Chinese) |
中文摘要 | 本研究探討以類神經網路配合影像處理技術檢測胴裂米,研究中共使用了米粒灰 階直方圖法和區塊法等輸入模組,結果顯示以區塊法平均灰階差值的 13 區塊輸入模組及其 所用的三層式類神經網路表現較好, 其學習可達到 97% 以上的正確率,胴裂粒測試平均正 確率在 88% 以上,完整粒測試平均正確率 97% 以上,研究顯示有學習歸納能力的類神經網 路,可應用於胴裂米之檢測。 |
英文摘要 | In this research, computer neural networks combining with image processing techniques are applied for cracked rice kernel inspection. Histogram and block methods are used for neural network training and testing. According to the tests, block method of block-gray-level-difference with 13 blocks and its associated 3 layers neural network model have the better performance in detecting cracked kernel. The correct ratio of neural network in learning is above 97%, the correct ratio of neural work in testing cracked kernel is above 88% and in testing sound kernel is above 97%. The study shows that neural netowrk with learning and inducting ability is proper for the application of cracked rice kernel detection. |
本系統中英文摘要資訊取自各篇刊載內容。