Research On Object Detection Technology Of Usv Based On Information Fusion Technology

Unmanned Surface Vehicle or "USV" is a kind of ship that can realize navigation and operation through autonomous navigation or remote control,which is a mobile robot applied to the surface essentially.Among them,object detection is an important research direction in USV environment perception and object detection technology,and it is also an emerging research in all of the world.In this paper,large and medium-sized unmanned boats are used as the carrier,and the target detection research is carried out for the medium-range distance(within 300 meters)surface navigation ships.The work carried out is as follows:(1)Analyzed the important technology of USV,designed the architecture of large-andmedium-sized USV from the perspective of system engineering,and designed the autonomous navigation system and shore foundation of USV from the hardware structure system,as well as the perception module,intelligent decision module,communication module,and navigationrelated modules on the ship side;from the perspective of logical structure,it introduces the working mode and specific task execution process of USV.(2)Research on target detection technology based on millimeter wave radar and camera respectively.Firstly,an overall introduction to the surface detection technology based on millimeter wave radar in this paper is made,and the hardware selection is carried out.By comparing the object detection methods directly based on image segmentation technology,the necessity of preprocessing radar images to eliminate clutter is illustrated and the method is mainly introduced.The median filter algorithm and morphological processing in image processing technology are used to reduce The interference of the sea clutter on the radar image finally recognizes the edge of the image according to the characteristics of the binary image to realize the object detection.Then,the research of water surface object recognition method based on deep learning is introduced.It adopts the overall solution of network camera and electronic pan-tilt anti-shake.This article introduces the self-built image database method used in this article and the process of online training of YOLOv3 deep learning algorithm using Tencent cloud server.The experiment proves that the method achieves both the recognition speed and the recognition accuracy,and realizes the application of the existing deep learning algorithm to the detection of water surface targets.(3)Based on the previous research foundation,the information fusion technology based on millimeter wave radar and visual algorithm is studied and applied to the detection of mid-range distance water surface object.A decision-level information fusion method is used to generate regions of interest according to the coordinate conversion rules to improve the efficiency and accuracy of the visual recognition algorithm.According to the information fusion rules,the obstacle speed,distance,azimuth,and ship class information are output.