Multi-sensor fusion sensing method and system for automatic driving under extreme working condition

Abstract

The invention relates to an automatic driving multi-sensor fusion sensing method and system under an extreme working condition. The method comprises the following steps: obtaining a smoke fog area in image data, obtaining corresponding laser point cloud data in a three-dimensional view cone range of the image data, projecting the laser point cloud data to an image domain corresponding to the image data, and obtaining smoke fog point cloud data of the image domain according to an intersection of the two; meanwhile, obtaining radar domain obstacle point cloud data based on the laser point cloud data. And according to the intersection of the smoke mist point cloud data of the image domain and the obstacle point cloud data of the radar domain, false obstacle fusion sensing data are obtained, and then the false obstacle fusion sensing data is combined with real obstacle data obtained by the millimeter wave radar to remove false obstacles. According to the method, a visual angle where smoke fog possibly exists is provided for the laser point cloud based on the image data, the laser point cloud data are mutually verified between the image domain and the radar domain, the laser point cloud data are combined with the millimeter wave radar, and finally high-precision automatic driving scene data are obtained.

Description

technical field [0001] The present application relates to the technical field of automatic driving decision-making and planning, in particular to a multi-sensor fusion sensing method and system for automatic driving under extreme working conditions. Background technique [0002] Current self-driving vehicles usually make decisions and plans in a three-dimensional coordinate system, which requires the environment perception system to provide accurate geometric ranging information. Therefore, lidar has become the mainstream sensor for current self-driving vehicles due to its high-precision geometric measurement characteristics. Due to the short wavelength of lidar, it is difficult to penetrate granular obstacles such as smoke and fog, so it is easy to cause false obstacles in the environmental perception system and then fail, affecting the safety of autonomous vehicles. Compared with lidar, visible light camera has rich texture information, but it is difficult to give precise ...

AI Technical Summary

Problems solved by technology

However, a unified fusion theory and an effective generalized fusion model and algorithm have not yet been established, and the problem of fault tolerance or robustness in the fusion system has not been well solved. There are still many practical problems in the design of the data fusion system.

Images