Research And Implement Of Gold Monkey Recognition Based On Convolutional Neural Network

Golden monkey has significant scientific study and ornamental with beautiful appearance and very high evolutionary position.Golden monkey is the First-Grade State Protection animal.Individual identification is one of the most basic elements in the process of scientific research on golden monkeys.In recent years,the researchers mainly use manual marking and "photo physiognomy" for individual identification of monkey.Real-time monitoring of golden monkey with digital imaging technology can obtain rich image data.Therefore,Shooting a large number of golden monkey images are handled and analysed quickly.However,the use of manual processing can not meet the demand.The researchers eager to have a quick processing and analysis of golden monkey image data to reduce the heavy burden of manual processing and analysis.The Convolutional Neural Network(CNN)is used to solve the problem of large-scale golden monkey face recognition.This method presents a new solution for individual recognition of golden monkeys.This thesis mainly elaborates the research significance and development status of golden monkey face recognition.The basic theory of CNN is summarized.The facial features of golden monkey under the wild environment is analyzed.The typical CNN model is studied.The use of Le Net,Alex Net and other shallow network identification accuracy is low and the use of Res Net and other deep network training for a long time in the process of golden monkey face recognition.Therefore,an improved CNN model is proposed to solve the above two problems.Using this model to identify the golden monkey face and then determine the individual of golden monkey.The golden monkey face detection and identification software was developed to automatically identify the golden monkey based on the improved CNN model.(1)Based on the analysis of the facial features of golden monkey,An improved CNN model for the recognition of golden monkey face is proposed in this thesis.The non-linear relationship between the features of golden monkey face is obtained by cross-layer connection.in this model.New residual blocks are added to improve the feature expression of the model and reduce the number of layers of the network to reduce the complexity of the network.The improved CNN model has eight convolution layers.Three residual blocks are constructed with the feature maps of different convolution layers.The convolution layer is added at some of the cross-layer connections of the network structure and increases convolution cores in the partial convolution.Experiments show that the model have good feature expression ability and improve the accuracy of golden monkey face recognition by reducing the depth of the network.(2)Based on the research on the recognition method of golden monkey face,this thesis designed the detection and recognition software of golden monkey face based on Qt.The software uses C++ language to develop detection and recognition of golden monkey.In the Qt environment combined with Open CV function to deal with the image.To achieve the detection function of the loading image,save the image and image binarization.The trained models,test images and labels are used to identify the individual faces of golden monkeys with the classification function.On the basis of CGM-23 and CGM-26 were created in the laboratory,the results of software test showed that 90% of the golden monkey face area could be detected.The improved CNN model was used to identify the golden monkey face and achieved good recognition effect.