Research And Implementation Of Plant Leaf Deficiency Detection Based On Deep Network

With the increasing demand for the quality of agricultural products,in order to meet people's needs,the application of advanced scientific and technological means in agriculture has become more and more important.Deep learning has been widely used in image recognition.The difference between traditional image recognition methods and in-depth learning is that in-depth learning,there is no need for complex image preprocessing,and the features of images are obtained by learning and induction through network,and no need for manual design features.Convolutional neural network model is one of the most widely used deep learning models in image recognition.For plant image recognition,the leaves,flowers and seeds of plants are relatively stable,and when using a computer to process images,plants of these three states can be used as data sets.Plant leaves are more suitable for two-dimensional image processing,and can be collected most of the year,so at this stage,the image of the leaves is mainly used to identify plants.Aiming at plant image recognition,computer is used to process plant leaves.Lack of different nutrients in plants can lead to different symptoms,such as discoloration,deformation,growth retardation,etc.These symptoms will reduce the yield and quality of crops.Using computer vision technology to monitor the deficiency information in real-time during plant growth can grasp the growth status of crops in time.In this paper,the problem of plant leaf element deficiency is studied based on deep neural network.The network nodes and layers are designed.The detection algorithm of plant element deficiency is given.The detection information system of plant leaf element deficiency is realized.The system realizes the detection of cucumber leaf element deficiency information.On the basis of APP application framework,the deep neural network algorithm is applied to detect plant leaf element deficiency.The specific work of this paper is as follows:1.Preprocessed the images in cucumber leaf data sets,including enhanced image contrast,flipped image and partial blackening;2.Analyzing VGGNet,ResNet,GoogLeNet,MobileNet and AlexNet networks to detect the deficiency of cucumber leaves,and selecting the best network.3.To improve the network,four kinds of network depths of 2,3,4 and 5 layers convolution operation are designed and compared with two kinds of convolution cores of 5×5 and 3×3 hybrid cross-combinations.Sampling layer is constructed by maximum pooling and average pooling.Full connection layer is designed as three-layer structure to construct network structure.4.Based on the constructed convolutional neural network,a mobile terminal application system for detecting plant leaf deficiency was designed and implemented.The functions of the system are:detecting the deficiency information of cucumber leaves in the input pictures,providing symptoms and coping methods of cucumber leaf deficiency.The experimental results show that the network detection accuracy based on this paper can reach 95.31%.The method can detect the deficiency information of cucumber leaves,and the detection system is simple and easy to use.