Design And Implementation Of Plant Flower Recognition System

The study of plants is particularly important since plants and ecosystems have natural connection with human being.People can have a better classification,more effective protection and utilization of all kinds of plants through the identification of plant flowers.However,the traditional way of flower recognition is relied on artificial classification,which is difficult and low efficiency.In this paper,a plant flower recognition system was designed and implemented by computer and smart phone terminal.The system could identify different varieties of flowers,which made it more convenient for the flower identification,classification and appreciation of various types of people.Therefore,the work of this paper had major practical significance.In this paper,a large number of plant flower images were firstly collected,and the image data-sets of their own plant flowers recognition were constructed and transformed into the LMDB format required by caffe.Secondly,the convolutional neural network of plant flowers recognition was built and trained with the use of caffe deep learning framework.Finally,a plant flower recognition system composed of mobile and computer terminal was designed and implemented under the combination of trained convolutional neural network and Android and Web development technology.The caffe was successfully transplanted to the Android platform during the achieving of the system,meanwhile,the access problem of Java for caffe trained convolutional neural network model was solved through JNI technology.This plant flower recognition system has the following characteristics: 1)The system is consisted of two parts: the mobile end and the computer end.In comparison to the existing plant flower recognition systems which are mostly single terminal,this system is more abundantly constituted to provide more terminal selectivity.2)The mobile side of the system provides both local recognition and remote identification,which is differentfrom the existing flower recognition APP that must upload the image to the sever for identification with Internet access.And for this system,even in the absence of network,the users still can select the local recognition mode for flower identification by means of local stored convolution neural network.