Research On Fusion Sensing Algorithm Of Intelligent Ship Navigation Environment Based On Image Recognition

With the increasing globalization of the shipping economy and the increasing busyness of various routes,the problems of serious environmental pollution,high labor costs and insufficient safety related to water transportation have become increasingly prominent and have attracted widespread attention from all walks of life.In order to solve the above problems,in the context of continuous technological advancement,the concept of intelligent ships came into being.Navigation environment state perception technology is an indispensable content in the development of intelligent ships.Whether it is the realization of autonomous navigation or the improvement of navigation situation awareness capabilities,it depends on the ship’s navigation environment perception system.In intelligent navigation,ship collision avoidance and decision control have higher technical requirements for environmental perception than other application scenarios.This paper researches the intelligent ship environment perception problem,builds a scenario-based ship environment perception platform,and carries out research on target detection,tracking and fusion algorithms.The main research contents are as follows.First,an intelligent ship navigation environment awareness platform was designed,and the design ideas and implementation methods of the intelligent ship navigation environment awareness platform were clarified,including equipment selection,system architecture,equipment parameter calibration,perception system software design and perception data processing.The construction of the platform is the foundation of data collection and algorithm verification.Secondly,a ship target detection algorithm based on YOLO（You Only Look Once,YOLO）v3 is proposed.Based on the traditional YOLOv3,the loss function is optimized,the ship target detection algorithm is improved,the identification of different ship types and the real-time detection of surface ship targets are realized,and the effectiveness of the target detection algorithm is verified by self-made data sets.The tests have performed well.Third,a ship tracking algorithm based on Deep Sort is proposed.By collecting ship images,making a data set to train a deep neural network,extracting ship depth features,optimizing model parameters,and obtaining a ship target tracking model.The deduction and verification were carried out in conjunction with the built ship environment perception platform.The results show that the tracking stability of the target when it is blocked is effectively improved,and the algorithm has good tracking accuracy.Finally,A fusion algorithm that fuses the ship’s automatic identification system,radar and image tracking data is proposed.Correlate AIS,radar and image perception data in space and time,fuse the perception information to get more reliable environmental information.Based on the perception platform to collect data and conduct experimental verification,the results show that the fusion algorithm effectively improves the accuracy of environmental perception compared to a single perception method.The research results in this paper can be used as a theoretical basis for the environmental sensing system required for autonomous navigation of intelligent ships,and have a good application prospect in ship assistance driving systems and ship traffic service systems.