Research On Flower Image Classification And Recognition Model Based On Deep Learning

China is a large flower-growing country with abundant species resources,a long cultivation history,and a large number of relevant employees.Among them,flower classification is an important basic work in the field of botany research and flower industry production,so flower taxonomy is a basic research with long-term significance.But their classification process needs a lot of manpower and material resources.With the development of imaging technology such as mobile phone devices,people collect,recognize and distinguish flower images,laying the foundation for intelligent classification and recognition of flowers.In recent years,deep learning has emerged from machine learning and has become a vigorously developed new discipline.Its development has greatly promoted the progress of science and technology,and at the same time has continuously changed our understanding of the world and led us into the era of artificial intelligence.At present,deep learning has been widely used in the fields of image recognition,speech recognition and other fields.The research on the recognition and classification of flower images has always been a hotspot of deep learning.It can not only help non-professionals understand and recognize flowers,and further classify them accurately,but also reduce the time and effort required by plant experts to classify flowers,help them design botanical robots,and realize automatic garden management.For the classification of flower images,this paper discusses the basic theory and related technologies of deep learning,and applies some of these techniques to the task of flower classification to improve the accuracy of flower image classification and recognition.The main work of this article is as follows:(1)Introduce the research background and significance of the classification and recognition of flower images,and explain the domestic and foreign research trends.(2)Introduce the relevant knowledge of deep learning and some related algorithms of flower image classification and recognition.(3)Because the traditional methods of flower image classification and recognition often require human intervention,are dependent,and often have features that are designed to be accidental,they cannot achieve accurate classification of flower images,so a variety of convolutional neural networks are cited to perform flower image features.Learn and compare to find a convolutional neural network model with a good classification and recognition rate.(4)Due to the problems of small flower parts,complicated background,and background interference in flower images,the attention mechanism is cited on the basis of the convolutional neural network model,and the local features of the intermediate convolutional layer and the global features of the fully connected layer are carried out Fusion,construct attention features,realize the enhancement of the flower part in the flower image,and suppress the unrelated background and other parts.(5)Since the flower images still have the problems of similarity and intra-class difference,a flower image classification model A-LDCNN based on deep learning is proposed.This model introduces the idea of linear discriminant analysis LDA on the basis of convolutional neural network and attention mechanism,constructs a loss function based on discriminant distance measurement,and then combines the Softmax loss function to construct a new loss function LDloss to participate in network training,So that the extracted features have the largest inter-class distance and the smallest intra-class distance,to achieve accurate classification of flower images under natural conditions.