查詢結果分析
來源資料
頁籤選單縮合
題名 | 育苗箱種苗影像特性與背景雜訊之去除 |
---|---|
作者 | 林達德; 黃俊誠; | 書刊名 | 農業工程學報 |
卷期 | 39:2 1993.06[民82.06] |
頁次 | 頁65-77+ |
分類號 | 434.251 |
關鍵詞 | 育苗箱; 背景雜訊; 種苗; 影像; 機器視覺系統; 影像處理; 數位遮罩; Machine vision system; Seedling; Image processing; Digital mask; |
語文 | 中文(Chinese) |
中文摘要 | 本研究以育苗箱中之西瓜與甘藍種苗原始影像為基礎,經人工判斷建立其標準影像,求取兩者之差異影像以探討影像雜訊空車分佈特性。隨後依雜訊之特性設計數位遮罩並建立量化指標,以評估各遮罩對種苗影像的處理效率,從而強化種苗缺株判斷機器視覺系統之功能。研究結果顯示苗格中種苗亮點之出現機率以苗格中央部分最高而向四周遞減,而西瓜種苗葉片比甘藍種苗葉片較為集中於苗格中央,此特性印證了機器視覺系統以常態分佈或三角分佈加權處理有效像素的方法。研究中以十二種遮罩對西瓜與甘藍種苗影像進行背景雜訊處理與比較分析,各遮罩對西瓜種苗影像之雜訊處理效率ε平均值在60.6%至78.1%間,對甘藍種苗影像之處理效率在65.5%到79.3%之間。同一遮罩對於甘藍種苗之處理效率變異性一般較西瓜種苗為大。而由量化指標SNR值、遮罩處理效率ε值與運算時間等進行評估,得以較客觀地選擇最佳之水位遮罩應用於種苗缺株判斷機器視覺系統之影像先期處理。 |
英文摘要 | Digital images of watermelon or cabbage seedlings in a nursery tray were acquired and artificially modified into standard images. The difference image of the original image and the corresponding standard image were generated to analyze the characteristics of image background noise. Digital masks were subsequently designed based on the characteristics of the background noise. Quantitative noise elimination indexes were also established to evaluate the effect of image processing through digital masking. The cumulative frequency of the pixels representing seedling leaves appeared to be highest at the center of the tray cells. The plots of the cumulative pixel frequency also indicated that cabbage seeding leaves were more widely spread in a tray cell. This image characteristics has confirmed the use of normal distribution and triangular distribution weighing methods in the machine vision system previously developed. Twelve digital masks were evaluated for their performance in removing the image background noise. Their noise elimination efficiency ε ranged from 60.6% to 78.1% for watermelon seedling image and ranged from 65.6% to 79.3% for cabbage seedling image. The variability of the efficiency of the masking processing appeared to be greater for the processing of cabbage seedling image. On the basis of the calculated SNR value, elimination efficiency ε value and the time required for the masking process, optimum masks were selected objectively for image preprocessing in the machine vision system for seedling detection. |
本系統之摘要資訊系依該期刊論文摘要之資訊為主。