A Target Detection And Tracking System Based On Global Vision

The technology of environment perception,navigation positioning and target tracking based on visual sensor is a hot topic in the field of robotics at present,at the same time,the target detection algorithm based on neural network has achieved good results,the combination of these two technologies provides a more extensive application prospect for mobile robot visual navigation.In this paper,visual sensors are used to construct the global environment,and realizes the target location by using the deep neural network model,the omnidirectional mobile platform is selected as an execution structure,the theoretical foundation and related realization method of the motion control are studied.Finally,the initial requirement of the mobile robot localization and autonomous navigation are achieved.The main research work of this paper are as follows:(1)Research on the method of constructing global environment based on image stitching.Firstly,the visual navigation system is built,the data collected by four cameras are transmitted to the host computer through network communication,then the feature extraction and matching,image registration,waveform correction and exposure compensation are obtained,and the global visual information is acquired through image stitching,and the camera is calibrated to realize the transformation of pixel coordinate to world coordinate,then transform the original image into a bird's eye view through the inverse perspective transformation.(2)A target detection method based on image processing and depth learning is studied.Firstly,the method of edge detection based on Canny operator is introduced.In view of the shortcomings of traditional method in complex environment and light changing scene,a target detection method based on deep neural network is proposed.Using the original image data set to adjust the parameters of the model,then the trained model is applied to the actual environment,and the global location of the target is completed,which improve the location positioning accuracy effectively,and the features obtained are more natural and more versatile.(3)A global path planning and path following algorithm based on known environment is designed and implemented.Research on grid map construction and path planning method in global environment based on *A algorithm.On this basis,the path following control process is designed.The deviation of the position and pose information of the mobile robot is used as the input of the controller,and the decision is made to the next step according to the deviation relation.A preliminary experimental test was carried out on the built environment,and the correlation analysis of the experimental results is also introduced.