基于典型叶片模板自动匹配的虫损叶面积测量  

Measurement of pest-damaged area of leaf based on auto-matching of representative leaf

作　　者：钟取发[1] 周平[1] 付斌斌[1] 刘科文[1] 

机构地区：[1]浙江理工大学视觉检测研究所,杭州310018

出　　处：《农业工程学报》2010年第3期216-221,共6页Transactions of the Chinese Society of Agricultural Engineering

基　　金：国家自然科学基金资助项目(50545027);浙江省科技计划资助项目(2007C33029)

摘　　要：为了定量化评估农作物的虫害程度,提出了一种基于典型叶片模板自动匹配的叶片虫损面积测量新方法。先将叶片图像二值化并提取其外轮廓;再对提取的轮廓进行多边形近似,以多边形的顶点为端点将叶片外轮廓划分成若干子轮廓;然后采用形状上下文对完整叶片与虫损叶片之间的子轮廓进行自动配准,找出其间的相互映射关系;最后根据映射关系对虫损叶片进行重建,计算出虫损面积。对10类不同叶片的测量分析表明:该方法平均每叶片耗时0.962s,最大相对误差为8.22%,平均相对误差为4.78%。其中,形状复杂度高的叶片平均相对误差为7.48%,复杂度中等的叶片为5.99%,复杂度低的叶片为1.84%。结果表明,该方法能准确而快速地测量虫损叶面积。In order to evaluate the pest-damaged extent of crop quantitatively,the anthors proposed a novel method based on auto-matching of representative whole leaf to measure leaf pest-damaged area. Firstly,the outer contour of leaf was extracted after image binary;secondly,the contour was approximated to a polygon and segmented to many sub-contours using polygon vertexes;thirdly,the mapping relationship between the whole leaf and the pest-damaged leaf was constructed by matching their sub-contours based on the shape context;finally,the pest-damaged leaf was reconstructed by mapping their sub-contour relationship for area calculation. The experiments on ten types of different leaves showed that the average process time for one leaf was 0.952 s,the maximum relative error was 8.22% and,the average relative error was 4.78%. As to leaves with high shape complexity,the average relative error was 7.48%,and to leaves with medium and low shape complexity,those were 5.99% and 1.84%,respectively. The proposed method has proved to be an accurate and efficient method for measurement of leaf pest-damaged area.

关 键 词：图像处理 计算机视觉 测量 轮廓匹配 形状上下文 多边形近似 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]