英特爾杯作品2010年一等獎作品摘要

　　基于智能視覺技術(shù)醫(yī)用藥劑中可見異物自動化檢測系統(tǒng) (一等獎作品)

　　Automatic Injection Impurity Detecting System Based on Intelligent visual technology

　　肖 亮 李俊杰 桑延奇

　　中南大學 Central South University)

　　摘要：針對國內(nèi)人工藥品檢測低效率、低準確率的不足，按照國家行業(yè)標準設(shè)計了一套基于智能機器視覺的可見異物自動化檢測系統(tǒng)。該系統(tǒng)主要包括機械傳動模塊、電氣控制模塊和圖像識別與處理模塊。根據(jù)異物運動連續(xù)性和噪聲運動無序性等特點，作品使用改進的二次差分算法從圖像中提取運動雜質(zhì)，然后采用基于SIFT特征的MeanShift算法對雜質(zhì)進行跟蹤，最后檢測出雜質(zhì)情況，并由此判斷產(chǎn)品質(zhì)量是否合格。測試表明，系統(tǒng)的檢測分辨率達40μm，準確率達90%以上，滿足企業(yè)要求，基本能替代人工檢測。

　　關(guān)鍵詞：機器視覺，可見異物檢測，序列圖像二次差分，異物跟蹤

　　Abstract：According to the national standards of production，our team designs an automatic impurity detecting system based on intelligent visual technology to address the low efficiency and low accuracy of domestic manual detection. The system mainly includes mechanical transmission module，electrical control system module and image processing module. According to the feature-continuity of impurities movement and discontinuity of noise movement，we propose an improved second-difference algorithm in order to extract movement impurities，then use the SIFT features and MeanShift algorithm to track impurities，finally detect the extracted impurities，and find out whether the product is qualified. Experimental results show that the system can detect impurity whose diameter is above 40 microns，and the system accuracy rate can reach 90%. It meets the requirements of most enterprise，and in most cases can replace the original manual testing.

　　Keywords：Machine vision，Visible impurity inspection，Image sequences second-difference，Impurity tracking

　　全國大學生電子設(shè)計競賽組委會

　　2011年11月9日