查詢結果分析
來源資料
頁籤選單縮合
題 名 | 以類神經網路技術作青椒形狀及大小之分級 |
---|---|
作 者 | 謝志誠; 林勁助; | 書刊名 | 農業機械學刊 |
卷 期 | 3:3 1994.09[民83.09] |
頁 次 | 頁15-27 |
分類號 | 434.251 |
關鍵詞 | 類神經網路; 農產品; 分級; 青椒; Neural networks; Agricultural products; Grading; Pimento; |
語 文 | 中文(Chinese) |
中文摘要 | 分級是農產品運銷過程中相當重要的工作,而分級後品質均一的產品,更是提高產品市場形象與支持銷售行情的必要條件。利用機械或光電感測技術來擷取農產品的重量、形狀、顏色、表面瘀傷與瑕疵等物性作為依據,以輔助解決農產品的分級選別作業,已成為農業自動化中的一個重要工作。 本文以青椒為對象,首先利用數位影響處理技術從人工分級後的各等級產品的影像中萃取出周長、投影面積、長度、寬度、投影面積對周長平方比與寬度對長度比等可作為判斷青椒等級的分級準則,再利用多層前饋式類神經網路來模擬農產品的分級作業,並採用反向傳遞演算法學習各項分級準則與等級判斷結果間的關係,以量化了解各項分級準則對等級判斷結果的影響程度。利用所建立的類神經網路模型,以購自市場已分級的樣本測試後,與人工分級結果比對有84%的相符性,顯示類神經網路在青椒分級上的可應用性。 |
英文摘要 | Grading is a very important stage in the agricultural products marketing process. With the aids of mechanical or photo-electrical sensor technologies, many researchers took some agricultural products' physical properties as criterions for automatic grading. These physical properties consist of weight, size, colour, bruises, and surface defects. Based on digital image processing techniques, the shape and size grading of pimento by using neural network is proposed in this article. In order to judge the grade of pimento, some parameters are extracted from two-dimensional image of pimento. These extracted parameters consist of length, width, perimeter, projected area, projected area to squared perimeter ratio, and width to length ratio. The extracted parameters and grade of pimento judged by experienced workers are used as input and desired patterns to estimate the connection weights of neural network. In this paper, a three-layer feed-forward neural network and back-propagation learning algorithm are suggested for the shape and size grading of pimento. The degree of agreement of judgement by the generated neural networks model is about 84% compared with human inspection. The applicability of the neural networks to judge the grade of pimento is confirmed in this report. |
本系統中英文摘要資訊取自各篇刊載內容。