查詢結果分析
來源資料
頁籤選單縮合
題名 | 應用於種苗移植作業之機器視覺系統= |
---|---|
作者 | 賴天明; 林達德; |
期刊 | 農業工程學報 |
出版日期 | 19921200 |
卷期 | 38:4 1992.12[民81.12] |
頁次 | 頁91-110 |
分類號 | 434.251 |
語文 | chi |
關鍵詞 | 移植; 視覺系統; 種苗; 機器; |
中文摘要 | 本研究之目的為建立種苗移植作業之機器視覺系統,以偵測育苗箱中種苗缺株情形。此機器視覺系統以OCULUS 150影像處理卡和照明系統擷取二元化的種苗影像。所得影像經過四種影像處理方法--常態分佈法、三角分佈法、判斷區域設定方法和數位遮罩處理,以得到代表各苗格內種苗之有效像素。此有效像素可作為偵測苗格是否缺株之指標,根據苗格內之有效像素數目,再利用變異數判斷法或統計圖判斷法判斷出育苗箱中之缺株苗格。機器視覺系統完成後以西瓜和甘藍種苗進行偵測試驗。實驗結果顯示影響判斷準確率的因素為育苗箱背景、種苗葉片面積、苗格中生長位置和照明之均勻程度。以黃土栽培的種苗在相同的灰度分界值下,缺株偵測之結果較差,最小的判斷正確率為79%,而對於深色泥土或黑色介質栽培的種苗,四種處理方法和兩種缺株判斷方法可百分之百偵測出缺株苗格和非缺株苗格。單一影像處理時間與處理方法有關,利用區域設定方法以286個人電腦處理時,所需的處理時間約為7秒鐘。此機器視覺系統之設計可適用於各類育苗箱,處理所得缺株苗格座標,可提供未來發展之種苗移植機構定位之用。 |
英文摘要 | A machine vision system was developed in this research for inspecting seedlings in nursing trays. Binary images of different types of nursing tray were acquire with an OCULUS 150 image processing car and a lighting setup. These images were further processed with four filtering methods, normal distribution method, triangular distribution method, AOI setting method and masking method to obtain effective pixels representing seedling images in tray cells. The number of effective pixels was used as a criterion for determining the existence of seeding in a tray cell. Two classification algorithms, variance method and histogram method, were developed to detect the absence of seedling based on the effective pixel criterion. Nursing trays planted with watermelon seedlings and cabbage seedlings were tested and inspected with the machine vision system developed. The major factors which may affect the accuracy of seedling tray inspection were: leaf area, seedling position in the tray, soil background and uniformity of lighting. The worst case of inspection accuracy was 79% for seedlings planted on yellow soil due to similar grey levels of the leaves and soil background. For watermelon seedlings and cabbage seedlings planted on dark soil or medium, all four filtering method and two classification algorithms have proved to be successful, and the existence or absence of seedling in a tray cell could be accurately identified without error. Individual image of seeding nursing tray could be processing in 7 seconds using an AT 286 personal computer. The machine vision system was designed to be flexible for a change in tray configuration. The detected coordinates of missing seedlings in a nursing tray can be used for positioning operation of transplant mechanisms to be developed. |
本系統之摘要資訊系依該期刊論文摘要之資訊為主。