机载非接触式近红外土壤墒情检测系统研制

Development of an airborne non-contact near-infrared soil moisture detection system

针对探针式土壤水分传感器插入土壤后因反馈点固定而需大量布点、成本高、破坏耕层等问题,该研究提出一种基于法布里-珀罗干涉近红外传感器的非接触式土壤墒情在线检测系统。系统硬件部分由机载自动检测装置、电气控制箱和北斗双天线实时差分定位系统(Real Time Kinematic,RTK)组成。整套系统样机的试制包括:传感器的选型和模块设计封装、升降检测装置设计、传感器避障与采样点北斗定位、土壤含水量预测建模、软件中的二次开发和系统与润禾2ZBA-2型移栽机的集成等。田间试验结果表明:当移栽机以0.3m/s速度行进时,土壤水分传感器参比校准后进行土壤水分的测定,5 s内工控机上实时显示水分含量值,水分含量预测值与实测值的相对误差范围为0.18%~14.46%,平均相对误差7.77%,所测水分值结合北斗RTK系统测得的定位坐标生成土壤表层含水率分布图,为后续喷灌、滴灌等变量灌溉提供参考依据。

A non-contact and online detection system was developed for the soil moisture using Fabry Perot interference near-infrared chip. A large number of points were reduced for the cost-saving and non-destruction of the cultivated layer after the probe of the soil moisture sensor was inserted into the soil. The hardware of the system consisted of an airborne automatic detection device, an electrical control box, and Beidou dual antenna real-time differential positioning. The specific sensors were then selected to detect the soil water content. The packaging of the module was designed to protect the internal structure of the sensor from damage. A lifting detection device was used to control the soil moisture sensor onto the soil surface for the detection. The obstacle avoidance of the device with the ultrasonic sensor was installed to realize the automatic positioning of the measured height. The spectral data in the range of 1 750-2 150 mm was collected from the soil samples with different water content and types. A partial least squares(PLS) prediction model was established for the water content of the soil surface. The determination coefficients of all prediction models were above 0.9. There was a high curve fitting degree of the overall soil model, where the R2 reached 0.933 4. Although there was slightly lower than that of the single soil prediction model, there was a high universality of the overall model suitable for the prediction of most soil water content. Furthermore, the secondary development of the sensor data acquisition was realized to embed the prediction model of soil water content into the original near-infrared sensing chip system. As such, the measured value of soil water content and the corresponding spectral data curve were more intuitively displayed on the industrial computer in real time, when measuring soil samples. The measured distance information by the ultrasonic sensor was transformed into the voltage signals using the cooperated Xinjie PLC and analog module. The closed-loop stepping motor was utilized to adjust the height of the soil moisture sensor in real time, according to the PLC feedback signals, thereby realizing the coordination between the positioning system and the detection device.Correspondingly, a trial prototype was fabricated to integrate the near-infrared detection system of soil moisture and the automatic self-propelled transplanter. The field test results show that the soil moisture sensor after online calibration was dropped onto the soil surface with the lifting detection mechanism for measurement when the transplanter moved at the inspection speed of 0.3 m/s. After that, the moisture content value rose within 5 s, after the real-time display of the moisture content on the industrial computer. The measured water content of surface soil was combined with the positioning of the Beidou RTK system, thereby calculating the soil moisture content under the accurate longitude and latitude. The distribution map of soil moisture content was generated on the measured plot. Subsequently, the sampling points after the test were sampled, pretreated, dried, and calculated to obtain the actual moisture content of the sampling points, where the relative error was calculated between the measured and actual moisture content. Consequently, the measured soil moisture content in the test was basically consistent with the actual one, where the relative error of continuous detection was less than 10%. The distribution map of soil surface water content can be expected to directly match the early warning level of soil moisture for the visualization of the abstract information. Therefore, the soil moisture information of the field can be detected to accurately display in real time. The finding can provide a strong reference and practical significance for the variable irrigation, such as the sprinkler and drip irrigation, thereby optimizing the regional water and soil adaptation in the spatiotemporal pattern of water productivity. Intelligent agricultural machinery can also be created to fully realize the “border inspection and side management” of field management.