| Modern trams have a significant role in alleviating urban congestion and improving travel efficiency.At present,the operation safety of most trams mainly depends on the subjective judgment of the driver on the road condition,which contains certain hidden danger.Therefore,intelligent obstacle detection system is needed to improve the safety performance of trams.The most commonly used sensor for detecting obstacles is machine vision or radar.However,the performance of machine vision is greatly affected by weather and light,and radar is difficult to obtain the target’s color and texture characteristics.Considering the shortcomings of a single sensor,this thesis uses a millimeter wave radar and machine vision information fusion method to detect obstacles in front of a tram.Based on this,the related content of radar target screening,machine vision target detection and information fusion are studied.The research content is of great significance to enhance the active obstacle avoidance performance of trams and improving the safety of trams.The main contributions of this article are as follows:(1)Screen radar targets and design tracking algorithms.The complex background of the radar working environment and the interference of the flicker noise are taken into account.This section first completes the primary selection of effective targets based on certain rules,and then uses Sage-Husa adaptive Kalman filter track the selected effective targets,which improves the accuracy and reliability of the millimeter wave radar detection.(2)Set up a fusion model to integrate radar data into image data according to the temporal relationship.The method and process of joint calibration of two sensors are described in detail.The radar target point information is successfully mapped into the corresponding region of interest in the image.The problem of inconsistent radar and camera acquisition frequency is solved by buffering data and time stamping.(3)Preliminary recognition of target based on information fusion of radar and machine vision.Different from the common practice of using machine vision to identify targets based on radar detection,this thesis uses radar and machine vision to detect obstacles independently.Finally,we analyze the two detection results and distinguish the target types.(4)Designed and implemented an obstacle detection system in the tram operating environment.The orbital area based on multi threshold track region segmentation method is used to reduce the range of target screening,so as to achieve accurate target positioning.The detection system is established and the real-time data of multi-sensor is processed by multithreading to ensure the real-time detection of the system. |
