Design And Implementation Of Rice Variety Recognition System Based On Machine Vision

Rice plays an important role in agricultural production in China.The quality of rice seeds directly affects the yield of rice.The purity of seed varieties is one of the important indicators for the quality of seed.Traditional detection of seed categories is a time-consuming and labor-intensive task.In order to improve the efficiency of detection,this paper proposes a method based on machine vision for rice variety identification.The main research content includes:(1)Construction of an image acquisition system.In the four directions of the closed frame,halogen light was used to provide a light environment for the experiment,and a digital camera was used to collect pictures of the rice seeds.The collected pictures are directly transmitted to the computer for processing and analysis,and the results show that the images are clear and the texture quality is good to meet the experimental requirements.(2)Image processing.The acquired image will be noisy,and Gaussian filter are used to improve the quality of the image to remove unwanted information.Then,the image is segmented by using the local adaptive binary method,and the obtained image is masked with the original image to eliminate the background of the original image.(3)Rice feature extraction and modeling.The geometrical features,color features and texture features of rice were calculated,and the extracted features were reduced by principal component analysis.The extracted features are placed into a multi-classification support vector machine and a neural network to model and compare the optimal models.After comparison,it is found that the accuracy of the neural network is higher than that of the support vector machine,and the classification of the six varieties of rice can reach about 96%.(4)Developed a rice variety identification software system,which can be used to process static rice samples in combination with image acquisition equipment,and use ANN to identify rice varieties.Experiments have shown that it is feasible to identify rice varieties using machine vision without damage.