A Wheat Grain Variety Recognition System Based On Deep Learning

Wheat is one of the most important agricultural products in the world.High-quality breeding can improve the quality and yield of wheat,ensure food security,promote sustainable agricultural development,and help solve people’s demand for food and industrial raw materials.Wheat grain classification can screen for high-quality,highyield and high-adaptability varieties in breeding,promote the increase of planting area and yield,and reduce the impact caused by climate and environmental changes.The main disadvantage of traditional machine learning methods is that the feature extraction process requires manual participation,which will consume a lot of time and energy,and the extracted features cannot fully reflect the internal structure and laws of the data.Therefore,the deep learning method used in this paper for wheat grain recognition and classification can automatically learn features,improve accuracy and generalization ability,and can adapt to different data types and application scenarios.The main research work of this paper is as follows:(1)Wheat seed dataset construction and image pre-processing.The images of 22 commercially promoted wheat seeds with wide planting areas,such as Nongda 3416-18,Neile 288,Hengshui 6632 and Bainong 419,were captured using a body vision microscope,and three angle images were taken for each wheat seed,and the images were labeled and renamed,de-contextualized and cropped to 224 ×224 pixels,and the training set Random cropping and normalization were performed for wheat seed variety identification.(2)Convolutional neural network-based wheat seed variety recognition.A wheat seed recognition model was constructed by combining the convolutional neural network and migration learning commonly used in deep learning,and the results were compared with traditional machine learning and optimizer and learning rate parameters for finding the best.The results show that the effect of deep learning in wheat seed recognition is better than that of traditional machine learning in wheat seed recognition,with the highest recognition rate of 98.6%.Among them,the accuracy of wheat seed variety recognition was further improved by using the model after migration learning,with a recognition rate of up to 99.3%.(3)Wheat Grain Variety Identification Based on Lightweight Convolutional Neural Network.This paper builds a wheat grain variety recognition model based on a lightweight convolutional neural network,and adds a global average pooling layer(GAP)to the model structure,which greatly reduces the number of parameters and calculations of the model.Compared with the traditional convolutional neural network,it does not require excessive reliance on computing resources,and can achieve better recognition accuracy and has better practicability.(4)Design and development of mobile-oriented identification of wheat grain varieties.This thesis integrates and designs a wheat grain identification system,which can automatically identify wheat grain varieties on mobile devices.The system includes four modules: image uploading,data processing,variety identification and data storage.The design of the system has the advantages of high accuracy and low cost,and is suitable for practical applications.