数媒基于深度学习的水果识别系统设计与实现

北京理工大学珠海学院2020届本科生毕业设计Fruit Recognition Based on Deep LearningAbstractWith the increase of image data,the task of image recognition becomes more andmore difficult,and the traditional manual method is no longer competent.Comparedwith the current deep learning technology and artificial feature extraction technology,it can be clearly found that deep learning technology can extract the hidden informationin the image data more accurately and quickly,and has a higher feature extractionability for the image,so we propose a fruit recognition system based on deep learningalgorithms has great practical significance.The order of the algorithm structure of this paper introduces the development ofdeep learning algorithm convolutional neural network(CNN)in detail.It mainlyintroduces several basic hierarchical structures of convolutional neural networks:connections and differences between these different structures:convolutional layer,excitation layer,pooling layer(downsampling layer)and output layer.The typical deeplearning convolutional neural networks Alexnet,VGGNet,ResNet and their mainnetwork structures are introduced,and the advantages and disadvantages of deeplearning algorithm convolutional neural networks are summarized.The experimentalpart of this article is to first complete the installation of the experimental environment,sample preparation and sample preprocessing followed by the training of the deeplearning network model,and the prediction accuracy of the test data set after the trainingmodel is obtained,and finally a fruit recognition visualization interface map Shows thetype of each fruit.Keywords:Fruit recognition;deep learning;CNN;Feature extraction