Algorithm Research Of ROS-based Visual Semantic SLAM In Lowspeed Unmanned Driving

With the advent of the wave of research and application of unmanned driving,driverless technology is becoming the key to improving productivity in the future.In the high-speed field,fully autonomous driverless technology is still difficult to achieve and cannot be applied on a large scale.The current application of unmanned products is mainly concentrated in the low-speed field.Aiming at this problem,this paper studies the algorithm of visual semantic SLAM.The system extracts 3D semantic information with only the depth camera as input.In accordance with the traditional visual SLAM system,the three-dimensional scale information topology of the annotated target is realized,and the construction of the visual semantic map is realized.This paper implements a complete visual semantic SLAM system,which can record the scale information and relative position information of specific objects in the camera motion environment while recording the camera motion trajectory and historical pose,and at same time record the above information.visualization.The core tasks of the algorithm design in this paper are as follows.First,the feature points of moving objects are eliminated at the front end of the visual SLAM system,which increase robustness.Second,according to the actual operating environment of the system,the environmental word bag is extracted in a targeted manner,which increase stability.Third,the YOLOv5 algorithm is used to detect 2D targets.This paper trains the target detection model in a targeted manner,so that the system can run better in the environment.Fourth,the 3D semantic message of objects is recovered by combining the results of 2D object detection with depth information,and using point cloud filtering.This paper conducts relevant experiments on the key links of algorithm design.The experiments show that the algorithm and system designed in this paper can construct visual semantic maps more effectively.At the same time,due to the ROS operating system used,the deployment capability of the system in this paper is improved,and the practicability of the system is increased.