基于最小矩形法的柑桔横径实时检测方法研究

农产品的大小是农产品分级的重要特征之一．本文以柑桔为研究对象，对利用计算机图像处理技术检测柑桔横径的方法进行了探索．研究了柑桔图像的平滑和边缘提取方法，比较了利用Sobel算子和跟踪虫（tracking bug)法所得到的柑桔边缘，发现跟踪虫法寻找边界的速度比用Sobel算子快，且不需要进行细化处理．为了适应实际生产中柑桔方向的随机性和外形的不规则性的要求，使柑桔尺寸检测的方法有更好的适应性，设计了一种利用柑桔的最小外接矩形（MER）求最大横径的方法，并在柑桔的外形尺寸检测中进行了验证，实际最大横径Dr与预测最大横径的相关性为0．9982，利用此方法估测果径的最大误差在1．30mm以内，最大相对误差为2.64％，平均相对误差为1.35％，GB／T12947-91鲜柑桔分级标准中均以5mm为分等标准差，故MER法的精度能满足实际生产中柑桔分级精度的要求．本研究结果不仅为进一步研究开发基于图像处理技术的柑桔品质检测系统打下了基础，而且可用于对其他农产品进行检测．

Real-time Size Inspection of Citrus with Minimum Enclosing Rectangle Method

The size of agricultural products is one of the most important features in classi-fication. To precisely detect the size of citrus at real time, the method of image smooth and edgeextraction were studied. The image edge was obtained by means of Sobel operator and track-ing bug method were compared. The result indicated that tracking bug method not only couldextract the edge much faster, and but also no thinning treatment was needed. In order to meetthe requirement of real-time detection, a new way to estimate the maximum diameter by findingthe Minimum Enclosing Rectangle (MER) around the object was designed, and the correlationcoefficient of real maximum diameter versus detected one reached 0.9962. The maximum relativeerror was 2.64%, and the average relative error was 1.35%. These results not only lay a foundationfor further developing citrus quality detection system using image-processing technology, but alsocould be used to detect the size of other objects.