| Soil organic matter（SOM）is an important factor affecting crop yield and quality,and an important indicator of soil quality evaluation.Mastering the changes of SOM content is helpful to promote soil testing and formulated fertilization.Traditional chemical analysis method of SOM content is time-consuming,laborious and expensive.Compared with chemical methods,spectral detection technology has been widely used in the field of SOM content detection due to its rapid detection and simple operation.Previous studies have shown that the diffuse reflectance spectral information of some wavelengths in the visible-near-infrared band can be used to evaluate the SOM content.The research content of this paper is based on the spectral detection technology,and mainly aims at the problems of high instrument cost and redundant spectral information collected in the current common portable spectrometers.Using the detection method that combines photoelectric sensors and specific wavelengths of LED light sources,this paper developed a soil organic matter content detection device based on visible-near-infrared diffuse reflectance spectroscopy,which could realize the rapid detection of SOM content and promote soil testing and formula fertilization.The research content and results are as follows:（1）Collect the spectrum information of soil samples,and use four characteristic wavelength extraction algorithms to screen out the characteristic wavelengths of soil organic matter.Soil samples were divided into two parts for spectral collection and organic matter content determination respectively.After detecting abnormal samples with Monte Carlo cross validation（MCCV）,this paper used successive projection algorithm（SPA）,threshold optimization window competitive adaptive reweighted sampling（TO-WCARS）,TO-WCARS-SPA and random frog（RF）to select characteristic wavelengths of SOM.Then the partial least squares regression（PLSR）model of SOM content was established.Compared the accuracy and regression coefficient of the regression models established by different characteristic wavelengths,and combined with the wavelength of the narrow-band LED light source,three characteristic wavelengths of SOM were determined,which were 590nm,780nm and 940nm.（2）Design the hardware system of SOM content diffuse reflectance spectrum detection device based on STM32.The hardware of the detection device consists of a probe and a controller,mainly including:control module,light source constant current drive module,photoelectric conversion module,data storage module,human-computer interaction module and power module;according to the principle of diffuse reflectance spectrum detection,based on the combination of LED light source with specific wavelength and photoelectric sensor detection method,the detection light path and probe structure were designed;with STM32microcontroller as the core,according to the modular design method,combined with the requirements of the external interface position,the PCB and controller structure were designed;based on 3D printing technology,the prototypes of probes and controller were made.（3）Based on C language,the software system of SOM content diffuse reflectance spectrum detection device was developed.The software system was divided into five tasks based on the u COSIII operating system,including initialization task,touch screen task,data acquisition task,display task and data storage task.The AD conversion function of the STM32microcontroller was used to realize the data acquisition task;the SD card was managed based on the Fat Fs file system,and then the data storage task was designed;based on the em Win graphics system,the display task interface was designed to drive the screen.Finally,this paper realized the whole working process of the detection device from hardware initialization,light source control,spectral information acquisition to data storage and human-computer interaction.（4）The parameters of the detection device were optimized,the prediction model of SOM content was established,and the accuracy of the transplanted SOM content prediction model was verified.Adjust the LED drive current through the light source stability test to reduce the influence of temperature on the luminous intensity when led was lit multiple times in a short time interval;the pinion and rack mechanism was used to realize the straight line movement of the photoelectric sensor,and the optimal distance was determined by the cross validation model determination coefficient of the spectral data from the photoelectric sensor at different distances to the soil surface.The prediction model of SOM content was established,transplanted into the detection device,and the prediction accuracy was verified.The results showed that the correlation coefficient between the predicted value of SOM content and the measured value was 0.731,and the root mean square error was 2.642 g·kg-1,basically meeting the requirements of the determination of organic matter content. |
PDF Full Text Request