Research And Implementation Of Individual Identification Of Leopard Based On Deep Learning Model

Leopard,rare and endangered species,the national primary protection of wildlife.Leopard was common in Chinese provinces,but as a result of the population of leopard is too small and isolated,the excessive hunting by humans and the destruction of habitats by human actions and so on,the number of leopard in China has declined sharply.The population monitoring of leopard is one of the important tasks for its effective protection.The accurate recognition of leopard is the basis for effective monitoring,management and protection of leopard.At present,it is widely used to recognition the individual of leopard by the shape difference recognition technology based on the image data of leopard.There are two main problems in this method:(1)Accuracy is affected because of the pattern will change with the movement;(2)Increasing labor costs with the increase of the image data of leopard.To solve these problems,based on the fact that the deep learning method has a significant classification effect in the field of image recognition,in this thesis,the depth characteristics that exists leopard image data are tried to dig up and the individual of leopard recognition technology is studied in the following aspects:(1)Preprocessing of leopard image data.Data preprocessing work is an important prerequisite for the follow-up recognition method.For the application scenarios of the individual recognition,specific data processing method is required;(2)Research on leopard individual recognition model.Considering the variety of leopard image data,this thesis studies how to effectively build individual recognition model of leopard based on deep learning and ensure the accuracy of recognition;(3)Study on individual recognition model of leopard.Aiming at the less and unbalanced of leopard image data,this thesis studies on the method to ensure more individual of leopard are identified.Through research on the above-mentioned problems,in this thesis,an individual recognition model based on deep learning is proposed and popularized.Compared with the traditional methods of individual recognition methods,the main contributions of the thesis are as follows:(1)In this thesis,the deep learning method is first applied to the study of the individual recognition of leopard,and the individual recognition method of leopard based on the improved Cifar-10 deep learning model is put forward;(2)Improving the individual recognition rate of leopard.Through the use of three leopard individuals image data sets from the Shanxi Wocheng Institute of Ecology and Environment in 2010-2017 to the improved deep learning model training and validation,this identification method achieves 99.3% accuracy;(3)The model is more versatile.The model is verified by using the image data sets of 12 leopard individuals,and its recognition accuracy is as high as 93.3%.The data sets of 12 leopard individuals are used to verify the prediction ability of the known individual,with the accuracy rate above 86.6%.