基于多传感器数据融合的旋耕耕深检测系统研究

Rotary Tillage Depth Detection Based on Multi-sensor Data Fusion

旋耕作业时通常用耕深衡量作业质量,适合的耕深能打破土壤板结、改善土壤结构,促进作物根系生长。目前耕深检测研究场景主要为旱田,检测方法主要为倾角传感器间接检测,而水田土质松软,作业过程中会出现机具下陷和姿态变化的问题。为此,本文设计了一种适用于不同土壤环境的耕深检测系统。为检测地形起伏及机具下陷,设计一种地面仿形机构,通过耦合仿真探究其设计参数,验证可行性;为提高单传感器间接检测稳定性,使用三点悬挂倾角检测和北斗卫星系统(Beidou system,BDS)相对高程数据融合的检测方法,并建立耕深检测模型(Tillage detection model,TDM)。提出自适应迭代扩展卡尔曼算法(Adaptive iterative extended Kalman filter,AIEKF)对传感器获取数据先滤波再融合,获得更稳定、准确的耕深。在参考传统耕深测量方法的基础上,提出了RTK-BDS高程差值的测量方法获取耕深真值,并和传统方法进行了对比。仿形机构耦合仿真试验结果表明,所设计的仿形机构检测绝对误差小于0.5 cm,仿形后土槽高程最大形变量为2.89 mm。田间试验结果表明,TDM检测模型能有效反映耕深变化,AIEKF处理后数据较KF处理后数据信噪比平均提升1.41 dB,融合得到的耕深较两种单一传感器检测的耕深MAPE平均分别提高2.30%和2.07%,融合后平均MAPE为3.95%,平均RMSE为1.08 cm。提出的RTK-BDS差值真值检测法和传统检测法最大绝对误差为2.45 cm,能较好完成耕深真值检测。

Tillage depth is usually used to measure the quality of rotary tillage.Suitable tillage depth can break soil compaction,improve soil structure and promote crop root growth.At present,the research scene of ploughing depth detection is mainly dry field,and the detection method is mainly indirect detection by inclination sensor.However,the soil of paddy field is soft,and the problems of sinking and attitude change will occur in the operation process.Therefore,a tillage depth detection system suitable for different soil environments was designed.In order to detect terrain relief and tool sag,a ground copying mechanism was designed,and its design parameters were explored through coupling simulation to verify its feasibility.In order to improve the indirect detection stability of single sensor,the detection method of three-point suspension inclination detection and relative elevation data fusion of Beidou system(BDS)were used,and a tillage detection model(TDM)was established.Adaptive iterative extended Kalman filter(AIEKF)was proposed to filter and then fuse data acquired by sensors to obtain more stable and accurate depth.On the basis of referring to the traditional tillage depth measurement method,the RTK-BDS elevation difference measurement method was proposed to obtain the true tillage depth value,and compared with the traditional method.The coupling simulation results showed that the absolute error was less than 0.5 cm and the maximum shape variable of soil trough elevation was 2.89 mm.Field experiment results showed that TDM detection model can effectively reflect the change of ploughing depth.The signal to noise ratio of AIEKF processing data was increased by 1.41 dB on average compared with that of KF processing data,and the ploughing depth data obtained by fusion was increased by 2.30%and 2.07%on average compared with the ploughing depth MAPE detected by two single sensors,respectively.After fusion,the average MAPE was 3.95%and the average RMSE was 1.08 cm.The maximum absolute error of RTK-BDS difference truth detection method and traditional detection method was 2.45 cm,which can complete the truth detection of ploughing depth well.