Design Of Near-infrared Spectrum Acquisition Device For Tree Species Identification

Wood is an important renewable resource which is widely used in all aspects of production and living.However,there are many types of wood,many of which are similar in shape and difficult to distinguish with the naked eye.With the rapid development of the national economy,The industrial field has higher requirements on the efficiency,accuracy and cost of wood identification,traditional identification methods cannot meet the needs of the development of the modern wood industry.Domestic and foreign scholars have carried out many studies on wood identification technology.The wood identification method based on near-infrared spectroscopy has become one of the most widely used identification technologies because of its fast detection speed,low cost,and no damage to samples.At this stage,most of the spectrum acquisition devices used in near-infrared spectral analysis are general-purpose,with wide spectrum coverage and high price.There are relatively few spectrum acquisition devices for the wood industry.This study developed a micro near-infrared spectrum acquisition device for the identification of wood species.Combine Principal Component Analysis(PCA)and Support Vector Machine(SVM)models,it can achieve non-destructive and rapid identification of wood.Research innovation: Select the spectral sensor selection according to the unique spectral characteristics of wood in the near-infrared region.Combined with microcontroller and industrial 485 transmission protocol,the device is miniaturized,low power consumption,fast and accurate,and meets the needs of the industrial field for online detection and identification of wood species.Main research content:(1)Comparing the research results of domestic and foreign scholars,analyse the feasibility of applying near-infrared spectroscopy technology to the field of wood species identification,and the summariz shortcomings of existing near-infrared spectroscopy acquisition devices.(2)To analyze the characteristics of wood in the near-infrared spectral region,Hamamatsu’s C14272 miniature spectral sensor was selected as the core optical element.Based on the consideration of integrated design and device cost,STMicroelectronics(ST)’s stm32f107vct6 is used as the main control chip,and the microcontroller module circuit is designed to realize the acquisition and transmission of spectral data.Using AD5541 of Analog Devices as the digital-toanalog conversion chip,an adjustable constant voltage source module circuit is designed to control the tunable filter inside the spectrum sensor.Using the ADS8320 of Texas Instruments(TI)as the analog-to-digital conversion chip,a signal acquisition module is designed to realize the acquisition of the sensor output signal.Finally,the schematic diagram and PCB diagram are drawn according to the design scheme of each module.(3)According to the principle of separation of driver and application,the device driver and application are designed based on Free RTOS real-time operating system,including adjustable constant voltage source driver design,temperature acquisition module driver design,AD conversion module driver design and communication module driver design.,and according to the requirements of wood spectrum collection,a multi-task system is established to determine the task execution logic,so as to realize the rapid and accurate collection of spectral information by the device.(4)Perform functional tests on device hardware modules.The test results are as follows: the output voltage accuracy of the adjustable constant voltage source circuit is greater than 0.68%,the peak-to-peak value is 52 mV,the temperature acquisition circuit accuracy is greater than0.05%,and the AD acquisition circuit accuracy is greater than 0.67%.Using African red sandalwood,discolored red sandalwood,rubber wood,and white oak to verify the feasibility of wood species identification,collect the near-infrared spectrum of wood samples in the1350nm～1650nm band,and use PCA principal component analysis method and SVM algorithm to analyze the collected spectral data.Modeling classification.The results showed that the four kinds of wood were identified with an accuracy of 90%.