Forward train detection method based on vision and laser radar fusion

Abstract

The invention discloses a forward train detection method based on vision and millimeter-wave radar fusion, and the method comprises the steps: achieving the forward train detection through an image vision algorithm, and obtaining the precise distance of a forward train through a laser radar; possible trains in the images can be obtained through the detection algorithm, so that the system has no missed identification, train frames extracted through the detection algorithm are further classified through the image classification algorithm on the basis, error identification in the images is eliminated, and accurate forward train identification is achieved.

Description

technical field [0001] The invention belongs to the technical field of rail train driverless autonomous environment perception, and in particular relates to a forward train detection method based on fusion of vision and laser radar. Background technique [0002] In recent years, rail transit has developed rapidly, and the smart rail transit system represented by unmanned trains has been favored by more and more enterprises and scientific research institutions. At this stage, the unmanned train system mainly relies on the signal system, and the typical representative is the CBTC (communication-based automatic train control system) system. The unmanned train system based on the signal system can obtain the position of the train on the line through communication, so that the running train can obtain the distance of the vehicle ahead in real time, thereby avoiding vehicle collision. However, the communication-based driverless train system does not have specific autonomous envir...

AI Technical Summary

Problems solved by technology

Among them, the image-based forward train recognition method installs a camera on the train and recognizes the forward environment map through a machine learning algorithm to determine whether there is a train ahead. This kind of method can better realize the front train recognition, but based on The image method is difficult to achieve accurate measurement of the forward train, especially in scenes such as curves and ramps; the lidar-based method mainly installs millimeter-wave lidar or lidar on the train, actively emits radio waves through the lidar, and The distance of the forward train is obtained by the time-of-flight method, which can obtain the distance of the forward train, but it is difficult to achieve accurate identification of the forward train, so it is easy to misidentify the tunnel wall or the building in the train line as train, causing the system to falsely report

Images