本文應用機器視覺之理論發展一可自動檢測透明容器內異物及內容物適量與否之 演算法。我們分析透明容器內各種異物的情形，提出分區拍攝、分區檢測的原則，以簡化各 類異物檢出及內容量檢測的處理方法、縮短檢測時間。並且提出兩個可調的參數以適應不同 檢測對象的需求。本文所提出之演算法是由 C 語言構建，在 Sun/Sparc 工作站上加以模擬 實作，經由實驗結果，在測試的四種容器內無異物之情況，皆可得正確檢測結果; 另十六種 有異物之情況，則可檢測出十四種。各分區之檢測在 0.25 秒到 0.89 秒內即可完成。

　　　　　An algorithm based on machine vision theory which can inspect both the foreign matters in a transparent bottle and the nomal quantity of its content is developed. After analyzing the possible cases of foreign matters, we propose the principle of split and merge for the bottle image inspection. The inspection process is simplified and the inspection time is greatly reduced. Moreover, two adjustable parameters are designed to meet different inspection sensitivity demands. The algorithm developed in this paper is implemented in C language on a Sun/Sparc workstation. In the simulation of four without-foreign matter cases, all the inspection results are correct; while in sixteen with-foreign matter cases, fourteen cases are correctly inspected. All the inspected operations can be completed from 0.25 secoonds to 0.89 seconds.