Researches On Applications Of Barrier Image Segmentation Algorithms For Security Robots

In recent years,the intelligent field has attracted the attention of people from all walks of life,and the intelligent robot as the intelligent high-end equipment has become the focus of major companies and people.Because the application of intelligent robot is very broad,and the security field is the major companies and government agencies to solve the problem of public safety in the hot area,so the combination of intelligent robot security concept of security robot naturally become an important direction of robot research.Robot navigation is an important link in the research of robot defense,and obstacle detection and recognition is the key technology of robot navigation.The research content of this thesis is obstacle detection and recognition based on image segmentation,which is realized and applied by security robot.In order to achieve the research and practice objectives,the main work of this paper is described as follows:(1)from the traditional image processing method is studied,first introduces security robot application obstacles based on OTSU threshold method for image segmentation of obstacle avoidance,and then introduces the principle of the gaussian filter,and on the basis of gaussian filter adaptive method for image segmentation,image processing and comparing with OTSU threshold method,and the conclusion is that the gaussian filter adaptive image segmentation method of robot obstacle avoidance in traditional image processing method is a kind of good method.(2)Based on the deep learning method,the image segmentation method based on pix2 pix neural network is studied,and the model is trained and analyzed,so as to find out the shortcomings and limitations of the image segmentation method based on pix2 pix neural network.Through lucubrating GAN network,pix2 pix neural network is optimized,and through comparison with deep learning image segmentation algorithm,an image segmentation method that can make up for the shortcomings and limitations of pix2 pix neural network is found.(3)In the practice of the security robot project,a simple image labeling application program is implemented for the preprocessing of image labeling.In order to solve the compatibility problem of security robot platform,a method of TensorFlow model to Caffe model was found.In the process of putting the model into practice,obstacle ranging is needed.Therefore,an interface of obstacle ranging is provided to the security robot project team.Through the above obstacle detection and recognition research and implementation,and integrated into the security robot project to practice the application.The experimental results show the feasibility and practicability of the research method of obstacle detection and recognition in the actual scene.