农机作业信息的数据挖掘方法研究

发布时间：2018-05-29 02:17

本文选题：精准农业 + 数据挖掘　；参考：《哈尔滨工业大学》2017年硕士论文

【摘要】：伴随着物联网技术的兴起和应用,基于位置服务的数据交换方式不断出现,越来越多的领域开始将重心放在数据本身,目前精准农业领域正是这样一个以信息技术为基础的领域。利用传感器检测和数据挖掘技术,人们能够准确、及时地控制农业耕作,实现生产效益最大化。本文所研究的数据挖掘方法着手于地理信息系统中的农机作业信息,对相关耕作指标进行精准测算,同时对耕作质量进行监测评估,使农机能够尽快地调整耕作状态,从而减小不合理耕作现象对于农业资源的浪费。首先,本文对现有农机管理系统中的设备耕作轨迹进行数据预处理,利用高斯克吕格投影变换将原有椭球坐标在数值和方向上尽量减少失真地投影成直角坐标;建立完善的农机轨迹数据集,利用轨迹点的深度信息获取采集轨迹坐标的耕作路段,利用距离和时间间隔将轨迹坐标做分段处理,利用段内线性插值方式弥补了实际设备采样不足的缺陷,提高了算法性能。其次,针对农机平台运行过程中的耕作面积及相关计算指标,本文提出了基于采集轨迹坐标的面积测算方法,能够在移动端精确测算农机设备当天耕作面积,同时利用等效轨迹乘以幅宽面积数值可以有效衡量轨迹重耕率,提出一种基于网格的面积测算算法覆盖计算耕作区域面积,从而获得耕作漏耕率。基于耕作轨迹的时间序列数据随着时间的不断推进,其体量会变得十分巨大,本文采用改进的网格聚类算法,将这些轨迹通过聚类形成地块,利用边缘检测算法提取地块边缘点和几何中心点,并用其代替原始耕作记录。采用基于特定数据索引的地图叠加算法将指定区域不同时间段图层叠加分析,进而检测历史重耕问题,提出采用区域交集方法快速计算历史重耕面积。最后,本文提出一种基于支持向量机分类的耕作质量评估方法,通过比较和筛选确定影响农机耕作质量的特征信息,将这些特征信息作为农机耕作质量的参考输入序列最小化支持向量机训练模型中,经过参数调优处理最终得到适用于当前数据平台的质量评估模型,为耕作提供客观的评估标准。
[Abstract]:With the rise and application of the Internet of things technology, data exchange based on location services is emerging, and more fields begin to focus on the data itself. At present, precision agriculture is such an area based on information technology. By using sensor detection and data mining technology, people can accurately and timely control agricultural tillage and maximize the benefit of production. The data mining method studied in this paper is based on the information of agricultural machinery operation in GIS, and measures the relevant tillage index accurately. At the same time, it can monitor and evaluate the tillage quality so that the agricultural machinery can adjust the tillage status as soon as possible. In order to reduce the unreasonable farming phenomenon to the waste of agricultural resources. Firstly, this paper preprocesses the track of equipment tillage in the existing agricultural machinery management system, and uses the Gao Si Kruger projection transformation to reduce the original ellipsoid coordinates to Cartesian coordinates in the numerical and direction as far as possible. A perfect track data set of agricultural machinery is established, and the depth information of the locus points is used to obtain the cultivation section where the track coordinates are collected, and the trajectory coordinates are segmented by distance and time intervals. Intra-segment linear interpolation is used to make up for the shortage of sampling in practical equipment, and the performance of the algorithm is improved. Secondly, aiming at the tillage area and related calculation indexes in the running process of agricultural machinery platform, this paper puts forward an area measurement method based on the collection track coordinate, which can accurately measure the farming area of agricultural machinery and equipment on the moving end. At the same time, the equivalent track multiplied by the width area value can be used to measure the track retillage rate effectively. A grid-based area calculation algorithm is proposed to calculate the area of the tillage area, so as to obtain the tillage leakage rate. The volume of time series data based on tillage trajectory will become very large with the development of time. In this paper, the improved grid clustering algorithm is used to form the plots by clustering these tracks. Edge detection algorithm is used to extract the edge points and geometric center points and replace the original tillage records with them. The map overlay algorithm based on specific data index is used to analyze the layer overlay of different time periods of the designated area, and then the problem of historical retillage is detected, and the regional intersection method is proposed to calculate the area of historical retillage quickly. Finally, a method of farming quality evaluation based on support vector machine classification is proposed in this paper. The characteristic information that affects farming quality is determined by comparison and screening. The characteristic information is used as the reference input sequence of agricultural machinery tillage quality minimization support vector machine training model. After parameter optimization processing, the quality evaluation model suitable for current data platform is obtained. Provide objective assessment criteria for farming.
【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】：S22;TP311.13

【参考文献】