基于多光谱结构光立体视觉的大田棉花高精度定位与成熟度分级

项目名称： 基于多光谱结构光立体视觉的大田棉花高精度定位与成熟度分级

项目编号： No.61503146

项目类型： 青年科学基金项目

立项/批准年度： 2016

项目学科： 自动化技术、计算机技术

项目作者： 徐胜勇

作者单位： 华中农业大学

项目金额： 18万元

中文摘要： 采棉机器人是智能农机装备的重要组成部分。由于作业环境的复杂性，当前的采棉机器人视觉系统对棉花的识别、定位和成熟度判定的精确度较低，成为制约智能采棉技术发展的重要因素。本课题面向智能采棉需求，探索大田环境中棉花的空间定位和成熟度判定问题的解决思路和方法。主要研究内容包括：（1）测试棉田冠层的多光谱反射特性，使用测定频段的结构光投射于棉田产生显著测量点，以此研究高效的相机标定、图像分割和双目立体视觉方法，恢复棉花表面的三维信息。（2）分析棉花冠层图像中粘连与分裂现象的发生规律，提出图像和空间特征双约束的处理策略；利用棉花的表面积和图像特征，建立描述棉花成熟度的特征描述和分级模型，以精确判定棉花的成熟等级。通过本课题的研究，以期推动智能采棉机器人的发展和应用，并对其他农业采收机器人的相关研究提供借鉴和参考。

中文关键词： 采棉机器人；立体视觉；成熟度分级；结构光；多光谱反射特性

英文摘要： Cotton-harvesting robot plays an important role in the intelligent robot farm machinery and equipment. Because of the complexity of the operating environment, the current robotvision system in recognize location and maturity judgement is less accurate, which has become an constraint for the development of smart cotton-picking technology. For the need of smart cotton-harvesting the ideas and methods for spatial orientation and maturity determination of cotton in fields will be explored. The main contents are as follows: (1) The multispectral reflectance characteristics of cotton canopy will be tested. According to the result, the structured light with specific frequency spectrum will be projected on the cotton fields and many significant measurement points have been produced, then the efficient camera calibration, image segmentation and binocular stereo vision methods will be researched for recovering three-dimensional information of cotton surface. (2) Laws of division and adhesions in cotton canopy image are analyzed. We will propose a strategy with the constraints underboth of the image and spatial characteristics of cotton. Then the characterization and classification model of cotton maturitywill be established on the base of the characteristics of surface area, volume and image contour, to judging the maturity of the cottonprecisely. This project may hopefully promotes the development of intelligent cotton-picking robot effectively. And it will have an important benchmarking and reference value to the research of other kinds of intelligence picking robot.

英文关键词： cotton-picking robot;stereo vision;maturity classification;structured light;multispectral reflectance characteristics