| Seed industry is a strategic and basic core industry,which is a basic guarantee for promoting the long-term stable development of Chinese agriculture and national food security.High-quality crop seeds are of great significance to increase and stabilize yield,resist insect diseases and reduce agricultural pollution.《Rules for agricultural seed testing》defines purity as the degree to which a variety is typically consistent in terms of characteristic characteristics,expressed as the percentage of the number of seeds of the variety to the number of seeds of the sample of the crop to be tested.Many breeding experts face the risk of seed mixing while cultivating high quality crop seeds,so the purity identification of crop seeds plays an important role in cultivating high quality crop seeds and improving agricultural production.At the same time,moisture content of crop seeds is a very important indicator in seed storage.Too high or too low moisture content will affect seed activity and thus affect its storage.Therefore,moisture content detection of seeds is a very important part in the process of seed preservation.In order to realize the automatic detection of wheat seed purity and moisture content,this project developed a wheat seed purity/moisture content detection system based on hyperspectral imaging technology,combined with current hardware and software technology,and embedded the purity/moisture content detection model into the system to verify the feasibility of the system.In the purity test,the mean spectrum of wheat seeds is taken as the input feature,the wheat seeds that meet the demand and the impurity seeds are taken as the output feature of positive and negative samples,respectively.In the water content test,the mean spectrum is taken as the input feature and the water content is taken as the output feature.The main work of the project is as follows:1.Aiming at functional integration and automatic adjustment of operating parameters,the overall functional framework of the system is established,and the hardware selection and design are completed.The project uses the STM32 series chip as the control core,and uses I/O port output PWM wave and timer to complete the adjustment of the motor speed and direction.STC89C52 chip and RS232 bus are combined to adjust the brightness of light source,remote control of light source brightness is realized.The RS485 bus is used to achieve the transmission of environmental information(temperature and humidity),and the first USART serial port of STM32 is used to achieve data interaction with the host computer.In addition,the power module and drive circuit are designed or added,and the above hardware circuit modules are integrated into the circuit schematic and PCB design and board marking,to achieve the integration of the circuit.Through the selection or design of the above hardware circuit,the final realizations are as follows: 1)To control the operating speed of the motor to make it consistent with the exposure time of the camera;2)To read ambient temperature and humidity information to minimize interference caused by thermal noise and moisture in the air;3)To control the automatic adjustment of light source brightness,so that the camera can fully obtain sample information.2.For the high precision detection of wheat seed purity,a seed purity hyperspectral detection model based on Synthetic Minority Over Sampling Technique(SMOTE),Uninformative Variable Elimination(UVE),and Support Vector Machines(SVM)is proposed to solve the problem of sample imbalance caused by difficult acquisition of impurity samples in practical application scenarios.Firstly,the SMOTE algorithm was used to expand the minority(impurities)samples of wheat seeds to improve the imbalance of the samples.Then,UVE was used to select high-dimensional hyperspectral features to reduce the risk of degradation of model robustness due to band redundancy.Finally,the SVM model was constructed as a classifier to further improve the purity detection performance under the condition of small samples.Experiments based on five types of wheat seeds showed that: the proposed method significantly improved the accuracy of seed purity detection under the condition of insufficient number of negative samples.The average accuracy rate,precision rate and negative sample detection rate reached 95.98%,94.94% and 89.32%,respectively,which were 3.89%,7.18% and 12.42% higher than the traditional methods.Based on hyperspectral images of wheat seeds and partial least squares regression algorithm for modeling,the purity and water content models were embedded into the system to verify the feasibility of the system.3.On the basis of the friendly operation interface,complete system functions,and convenient operation,the software system of human-computer interaction interface and back-end control is developed.The system software takes Windows operating system as the operating platform,uses Python for logical code implementation,uses Py Qt for graphical interface rendering,uses QSS for interface rendering,builds a database on the Windows platform,and combines Py Qt and Matplotlib libraries to achieve accurate positioning of wheat seed purity test results.According to the actual application requirements,the following functions are established for the system detection software: 1)Display of temperature and humidity information,light source brightness control interface,motor speed adjustment interface,etc.;2)The contour detection algorithm based on the calibrator can automatically configure the optimal speed of the motor;3)Hyperspectral image acquisition and black and white plate image correction of test samples;4)The front-end and back-end implement of wheat seed and moisture content detection and analysis;5)Database management of personnel and data.During the development of the system,a method was designed and implemented to automatically open the number of threads when the number of wheat samples was uncertain,and to accurately locate the result of wheat seed purity detection. |
PDF Full Text Request