智慧农机车载信息采集终端研制及深松作业评价方法研究

发布时间：2018-04-22 10:38

本文选题：数据采集 + 排序学习　；参考：《哈尔滨工业大学》2017年硕士论文

【摘要】：我们国家农业的体量比较大,但是实力并不强,在农业现代化方面还有很长的路要去走,在国家的第十三个五年规划(2016-2020)中,将农业的信息化和现代化作为重大的农业发展项目,现代农业同传统农业相比,不再“靠天吃饭,听天由命”,而是通过现代科技采集农田中的作物与作业信息,因地制宜地提出种植和培育方案,利用大型农机进行统一作业,从而提高单位产出,同时使得农业生产实现可持续。信息采集属于精准农业实施的第一个步骤,大型农机是实施农田作业的主体,对农业生产实现现代化起着至关重要的作用,本文首先关注农业机械在进行农田作业过程中的数据采集,主要以深松作业为背景,深松就是利用深松机械铲使得农田的深层土壤松动,提高土壤对水肥的吸收,在深松作业中主要关注农机经过的路线,深松作业情况以及农机运动状态,接着为了对作业结果进行评价,以采集到的数据为基础,提出了一种针对深松作业的评价方法。首先,本文设计了一种基于深松作业的农机信息采集终端,信息采集终端在数据采集功能的基础上也要对数据进行上传和本地备份,由功能需求设计了系统的模块化结构,设计硬件电路编写软件程序,主要工作流程是:由不同种类的传感器模块获取农机的运行轨迹信息、作业深度信息和运动状态信息,经数据提取和整合后经GPRS网络传送给远端的服务器平台,同时将数据备份到本地的存储介质中。接下来对获取到的数据实施分析,包括数据的预处理和特征提取,数据预处理包括对轨迹数据进行坐标系变换和轨迹分段,对运动状态数据进行滤波与加速度修正,从其中提取出轨迹规整度、危险驾驶指数、油耗以及深度变异系数这些能表示深松作业质量的特征值,同时进行归一化处理,用特征值构造训练样本集。最后,选取排序学习方法中的排序支持向量机对构建的训练样本集进行训练得到深松评价模型,通过对选取的不同核函数进行参数寻优,提高了在该评价体系中的模型的准确率。
[Abstract]:Our country's agricultural mass is relatively large, but its strength is not strong. There is still a long way to go in the area of agricultural modernization. In the 13th Five-Year Plan of the State, 2016-2020), the informationization and modernization of agriculture are regarded as major agricultural development projects. Compared with the traditional agriculture, modern agriculture is no longer "relying on heaven to eat and leave to fate". Instead, it collects information on crops and operations in farmland through modern science and technology, puts forward cultivation and cultivation programs according to local conditions, and makes use of large-scale agricultural machinery to carry out unified operations. In order to improve unit output, while making agricultural production to achieve sustainability. Information collection is the first step in the implementation of precision agriculture. Large-scale agricultural machinery is the main body of agricultural operations, which plays a vital role in the modernization of agricultural production. In this paper, first of all, we pay attention to the data collection of agricultural machinery in the process of farmland operation, mainly in the background of deep loosening, which is to use deep loosening machinery to loosen the deep soil of farmland, and to improve the absorption of water and fertilizer by soil. In order to evaluate the operation results, an evaluation method for deep loosening operation is put forward based on the collected data. First of all, this paper designs a kind of agricultural machinery information collection terminal based on deep loosening operation. The information collection terminal also need to upload and backup the data on the basis of the data acquisition function. The modular structure of the system is designed according to the function requirement. The hardware circuit is designed to program the software. The main work flow is to obtain the running track information, the operation depth information and the motion state information of agricultural machinery by different kinds of sensor modules. After data extraction and integration, the data is transferred to the remote server platform via the GPRS network, and the data is backed up to the local storage medium. Then the acquired data is analyzed, including data preprocessing and feature extraction. The data preprocessing includes coordinate system transformation and trajectory segmentation, filtering and acceleration correction of moving state data. The characteristic values, such as trajectory regularity, dangerous driving index, fuel consumption and depth variation coefficient, which can represent the quality of deep loosening operation, are extracted from them. At the same time, the training sample set is constructed by normalized processing. Finally, the sorting support vector machine (SVM) is selected to train the set of training samples to obtain the deep loosening evaluation model, and the parameters of the selected kernel functions are optimized. The accuracy of the model in the evaluation system is improved.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：S222;S126

【参考文献】