基于支持向量机的多光谱番茄磷营养诊断  

Design and Experiment on Joint Sliding Mode Control System of Fruit Harvesting Robot

作　　者：梁喜凤 周涛 王斌锐 

机构地区：中国计量学院机电工程学院,杭州310018

出　　处：《农业机械学报》2016年第3期1-7,共7页Transactions of the Chinese Society for Agricultural Machinery

基　　金：“十二五”国家科技支撑计划项目（2014BAD08B03）资助项目、江苏大学高级人才基金资助项目（13JDG077）、江苏省博士后基金资助项目（1402076B）和江苏高校优势学科建设工程资助项目（苏政办发［２０１４］３７号）

摘　　要：为提高番茄磷营养水平检测精度,针对目前基于光谱分析的作物磷营养水平检测精度较低以及磷的光谱反射率受叶绿素和花青素影响的问题,提出了结合番茄样本光谱特征和生理特征融合的番茄磷营养水平诊断策略。以自行培育的25%,50%,75%,100%,150%五个梯度水平的磷营养胁迫水培番茄样本为研究对象,分别利用光谱分析仪和叶绿素仪采集不同磷营养水平番茄叶片的光谱数据和SPAD值并对叶片花青素含量进行测定,提取各样本在不同波长下的光谱反射率和生理特征（SPAD值和花青素含量）作为番茄磷营养诊断的特征变量,基于最小二乘支持向量机建立诊断模型,通过改进粒子群优化算法获取支持向量机的最优参数。将120个番茄样本随机分为校验集和测试集分别进行试验,结果表明,采用该文的建模方法结合番茄样本光谱特征和生理特征能够建立精度较高的番茄磷营养水平预测模型,高于对比的其他方法,其相关系数和均方根误差分别为0.9611和0.461,具有较好的诊断作用,为番茄磷素的快速检测提供了新的思路。The control system is one of essential constitute parts of the fruit harvesting robots, which affects the work efficiency and quality of the robots. In generally, fruit harvesting robots are nonlinear spatial open chains with multi-input and multi-output and serious coupling. It is difficult t.o set up accuracy mathematical model for the joint control system because there are many uncertain factors during working, such as joint friction, assembly errors, etc. Sliding mode control method has more advantages to solve nonlinear problems with strong robustness, anti-disturbance, etc. To improve the control performance of the fruit harvesting robot and solve the vibration of the control system, the sliding mode control （SMC） method based on genetic algorithm （GA） optimization was proposed. In this method, the parameters of SMC algorithm were optimized by GA through adjusting the parameters of switching function and exponential approach law in real time. Simulation and experimental platforms of a joint control system were designed and manufactured based on STM32 microcontroller, AS5045 position feedback module and CAN bus communication module. Moreover, experiments on position tracking and joint control system response were conducted in unloading and loading conditions respectively. The results show that the response velocity of position tracking was increased greatly by adopting GA. The vibration amplitude and time of the control system caused by external disturbance and load changes were reduced significantly. The experiment results also showed that the response time during position tracking of joint No. 6 in the unloading condition is 0.3 s shorter than that in the loading condition. However, the position tracking accuracy and maximum deviation are not affected greatly by adding load. In addition, the controller response time in actual experiments increases 0.5 s comparing with that of simulation experiment owing to ignoring some uncertain factors, such as joint friction, assembly errors, inertia force, etc.

关 键 词：番茄 磷 光谱 营养诊断 支持向量回归 

分 类 号：S22[农业科学—农业机械化工程] TP241.3[农业科学—农业工程]