基于SIFT与SVM的钉螺数字图像识别  

The Oncomelania Digital Image Identification Based on SIFT & SVM

作　　者：严深海[1] 黄贤通[1] 

机构地区：[1]赣南师范学院数学与计算机科学学院,江西赣州341000

出　　处：《赣南师范学院学报》2011年第6期58-61,共4页Journal of Gannan Teachers' College(Social Science(2))

基　　金：国家科技支撑计划重点项目(2009BAI78B02)

摘　　要：监测与控制钉螺生长区域,对有效控制血吸虫病疫情的传播和蔓延作用十分重要.采用数字图像识别技术,可以高效地实现对钉螺生长区域的监控.面对杂乱、图像部分遮挡、光照变化等复杂背景下的钉螺数字图像识别,提出了解决方案:首先,提取钉螺数字图像的SIFT特征向量集;然后,应用K均值聚类获取更具代表性的特征子集,作为分类器的训练样本集;最后,通过支持向量机方法获得分类器,用于识别钉螺数字图像.实验结果表明,提出的方法能够获得较好的识别效果.It plays a very important role in the control of schistosomiasis transmission and spread monitoring and controlling the oncomelania＇s growth areas. The oncomelania＇s growth areas can be effectively monitored and controlled by applying the image recognition technology. The high cluttered background,the part occlusion, the illumination and so on seriously affect the identification of the oncomelania digital image. The paper presents an approach to implement the oncomelania digital image recognition under above back- ground based on scale invariant feature tranformation （SIFT） features and support vector machine （SVM）. First it extracts the images＇ SIFT features set and then clusters it by K means method to obtain the more representative subset as the classifier＇ s training set. At last, it get the SVM classifier and use it to identify the oncomelania image. Experimental results show that proposed method can achieve good recognition performances.

关 键 词：通用对象类识别 尺度不变特征变换 聚类 支持向量机 钉螺 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]