农作物几何特征量测量系统设计与试验  

Design and Experiment for Crops Geometrical Feature Measuring System

作　　者：孙诚达[1] 邱威[1] 丁为民[1] 顾家冰[1] 赵三琴[1] 

机构地区：[1]南京农业大学工学院,南京210031

出　　处：《农业机械学报》2015年第12期1-10,17,共11页Transactions of the Chinese Society for Agricultural Machinery

基　　金：国家高技术研究发展计划(863计划)资助项目(2012AA101904-5);江苏省自然科学青年基金资助项目(BK20130670);江苏省产学研联合创新基金--前瞻性联合研究资助项目(BY2013050)

摘　　要：为实现复杂背景下农作物几何特征量的实时测量,设计了交互式软件系统和便携式硬件结构。软件基于Matlab图像处理技术和多个作物几何特征量的计算模型,提出了动态交互式颜色特征提取方法和逐点颜色分量差值筛选法,实现了目标作物与背景的有效分离。研究了非旋转体作物体积与投影面积的关系,提出了测量这类作物的体积等特征量的新思路;建立了单位像素实际值与物距的对应关系,避免了需要在背景中设置参照从而计算单位像素代表作物几何特征量的实际值。软件渗透了人机交互和分类设计的思想,具有较好的交互性、通用性和扩展性。硬件采用高配的PC机核心部件、快速响应的工业触摸屏、自动定焦对焦工业相机和可控的相机支架、大容量锂电池。结构紧凑（29.8 cm×19.9 cm×6 cm）、操作简便、便于手持式和分体式测量。系统经过精度测量和实时测试,能完成多种作物的几何特征量测量,一般测量误差4%~8%,自动处理时间平均1.63 s（建好颜色特征库时）,装置有较好的处理速度,有一定的实用价值。In order to measure the crop geometrical feature real-timely in the complex background, interactive software system and hardware system were designed. Based on the Matlab image process technology, crops geometrical feature calculation models and the idea of human-computer interaction and classification calculation, the software system were designed. The whole measuring process was divided into four steps, which were image acquisition, RGB color feature acquisition, crops region separation from the background and crops geometrical feature calculation. A human-computer interactive software interface was designed, which consisted of menu bar, operation button, image display window and measurement value display window, and it could control each measuring step. RGB color eigenvalues of different crops were obtained by using discrete sample point interactive selection method and region interactive selection method, then the crops region was separated from background efficiently by using screening method based on RGB color eigenvalue difference of each pixel. The software system built classified color database for some kinds of crops according to the environmental condition and crops feature, so pertinence and efficiency for extracting crops region were improved, and real-time measurement was realized. Image was rectified to enhance measurement precision. Corresponding relationships between the actual sizes represented by unit pixel and different object distances were determined previously in order to calculate the actual crops geometrical feature values in real time, so it was not needed to set reference object to calculate the actual size represented by unit pixel when the system was used. The software system built calculation models for several crops geometrical features, including volume, area, and length calculation models of revolving and non-revolving body. The crops volume and projective area regressive model was explored, which provided a new idea for calculating the volume of the crops with non-revolving featu

关 键 词：农作物几何特征量 便携式测量装置 分类式测量系统 人机交互式颜色特征提取 分量差值筛选法 

分 类 号：S126[农业科学—农业基础科学]