【作者】 葛玉峰；

【导师】 郑加强； 周宏平；

【作者基本信息】 南京林业大学 ， 机械设计及理论， 2003， 硕士

【摘要】 目前我国森林病虫害形势严峻，而其主要的防治手段-化学防治效率是相当低下的。因此研究基于实时视觉传感技术的农药精确对靶施用技术迫在眉睫，以大大减低林木病虫害防治过程中农药的使用成本，提高农药使用效率，同时又能保护森林生态环境，保护操作者的人身安全。 本文从相似理论入手，建立了室内模拟农药精确对靶施用系统。在实验室内开展一系列的试验和分析，对主要的问题和技术难点作了较为深入的研究。并以此为基础探索在户外进行农药精确对靶施用系统的可行性和效果。 本文设计建立了由CCD、图像采集卡和PC组成的图像采集和由喷雾台架、喷头、电磁阀、继电器组成的智能喷雾两套硬件子系统。同时用VC++和汇编语言编制了上位机和下位机的控制程序，实现如下功能：(1)模拟树木图像的实时采集，并从视频流中提取单帧进行图像处理，提出一种利用相对色彩因子实现绿色树木和背景分离的算法；(2)对分割的结果进行测量，应用图像统计学的方法获得了目标图像的一系列形状位置特征参数，为进一步的树木图像理解和模式识别做准备；(3)对摄像机的成像模型做了剖析，提出了一种利用成像模型和特征参数来求取树木深度信息的算法；(4)初步分析了智能喷雾决策系统，同时提出了一种用微型喷头控制特定区域实现农药精确喷洒的控制方法；(5)实现上、下位机的信息通信，将决策结果转化为喷雾指令，控制喷雾执行系统。 本文的研究为在户外真实环境中开展基于机器视觉的农药精确对靶施用与植保机械设计提供了理论和实践依据。更多还原

【Abstract】 Presently the infestation of forest plant diseases and insects in China causes severe problems while the chemical treatment, as the major solution, is of very low efficiency. So, developing "smart" chemical application methods based on the real-time vision sensing technology is urgent and necessary, which bears the promises of low cost and high efficiency of chemical application in diseases and insects treatment and moreover, will protect the forestry ecological environment, shield the operators from the dangerous concoction.In the paper, an indoor-simulation smart chemical application system was developed based on the Theory of Similarity. A series of experiments and analysis, which covered the main puzzles and technical difficulties, were carried out to predict the feasibility and reliability of the system for out-door purposes.Two hardware sub-systems were established in the paper. One was an image capturing sub-system composed of a CCD camera, an image grabber and PC. The other was a smart spray execution sub-system composed of spraying table, nozzles, solenoid-valves and relays. Two sets of software were written in VC++ and assembly language respectively to realize the functions as follows: (1) Real-time capturing, extracting and processing the key frame of tree image from video stream, segmenting the green target from its background by a classifier called relative color indices; (2) Measuring the segmented image by statistic method to obtain morphological and location features of the target, which provides info for intelligent understanding and pattern recognition; (3) Analyzing different camera models, putting forward a tree depth info acquisition algorithm by combining camera model and tree location features; (4) Making an introductory explain for decision making system, putting forward a decision making algorithm by using specific mini-nozzles to control specific zones; (5) Realizing data exchange between PC and SCM, converting decision making results to spraying instructions to control the spray execution system..The research in the paper will provide valuable experiences both theoretically and practically for the outdoor smart chemical applications based on machine vision and the design of plant protection machineries.更多还原