Research On Key Technologies Of Inland River Navigation And Target Tracking Based On Image

Nowadays,science and technology are developing rapidly,people begin to be keen on the research of intelligent transportation system.As an important member of surface traffic safety system,unmanned boat has been applied more and more in real life.It has considerable research value in theory and broad prospect for development.Among them,computer vision which is the eyes of unmanned boat navigation is also a crucial part.At present,the development speed of computer vision is accelerating,and its application has been widely popularized in traffic safety,industrial testing,biomedicine,military,aerospace and other fields.However,there are many challenges and technical difficulties in vision that need to be solved.In this paper,the intelligent navigation of unmanned boat was taken as the research object.Based on the capture of environmental information in autonomous driving technology,the technology of inland river line detection and midline detection,feature extraction and tracking in forward ship detection was studied by using visual sensors.In this paper,the detection of inland river lines was studied,and the morphology was used.The experimental results of the Sobel operator and Canny operator were compared,and finally Canny operator with better effect was adopted.Then the improved probabilistic Hough transform was used.Finally,the channel line and midline are detected and the angle of the unmanned boat’s deviation from the ideal channel line(midline)was sent to the control terminal to adjust the course of the unmanned boat.In order to realize the function of navigation and obstacle avoidance,the target needs to be identified and tracked.Extracting target feature is the key technology of target recognition.Here,the feature description was studied.In this paper,entropy was used to describe the complexity of image edges.Firstly,moving object edge detection was carried out.First-order difference and second-order difference were made for the coordinates of the edge contour of the moving target,and then the information entropy of the first-order difference and second-order difference was calculated according to the frequency of the combined values of first-order difference and second-order difference coordinates.Based on the feature that the first-order difference and second-order difference of adjacent point coordinates contain the direction,shape information and changing trend of the edge.A new method was proposed to describe the complexity of image edges by calculating the first-order and second-order differential information entropy of edge point coordinate.Kalman filter has been widely used in moving target tracking.In this paper,velocity,position and information entropy were adopted as three state variables of Kalman filtering to track ship targets.It has strong prediction ability for tracking target’s follow-up state and achieves good tracking effect.