Omnibearing obstacle detection method based on multi-sensor fusion

Abstract

The invention relates to an omnibearing obstacle detection method based on multi-sensor fusion. The method comprises the following steps: firstly, collecting images and laser point cloud data in different scenes through a laser radar and a camera; carrying out aerial projection on the laser point cloud data, carrying out feature extraction after two-dimensional gridding segmentation, and obtaininga target candidate box in an aerial view; obtaining an image region candidate box by using a one-stage target detection network model; and then fusing the target area candidate box and the image areacandidate box in the aerial view by using spatial registration; designing a segmentation sub-network of an encoder decoder structure in a three-dimensional space point cloud classification branch toclassify each point cloud to obtain an accurate category of an obstacle target in a three-dimensional space; and the three-dimensional candidate box position regression branch calculating the coordinate deviation and loss value of the predicted target and the marked target of the corresponding category, and outputting the deviated predicted obstacle position information to obtain more accurate position information of the three-dimensional space obstacle.

Description

technical field [0001] The invention relates to the technical field of unmanned vehicles, in particular to an omnidirectional obstacle detection method based on multi-sensor fusion. Background technique [0002] An unmanned vehicle is a highly intelligent device integrating environmental perception, dynamic decision-making and planning, behavior control and execution. Among them, environmental perception technology is an important basic guarantee, and the detection of dynamic obstacle targets is an important research content in environmental perception, which is of great significance to the autonomous navigation of unmanned vehicles. Sensors such as lidar, millimeter-wave radar, ultrasonic, and RGB cameras provide the necessary data for unmanned vehicles, but different types of sensors have advantages and disadvantages, and a single sensor cannot fully cover the functionality and safety of unmanned driving. For example, compared with RGB cameras, lidar has poor resolution, ...

AI Technical Summary

Problems solved by technology

[0004] The technical problem to be solved by the present invention is: how to use the information provided by each sensor to detect obstacles, and overcome the low robustness and The disadvantage of a single amount of information

Images