External parameter estimation method and system for vehicle-mounted look-around multi-camera system

Abstract

The invention discloses an external parameter estimation method for a vehicle-mounted look-around multi-camera system. The method comprises the following steps: acquiring environment pictures througha vehicle-mounted look-around multi-camera system which moves along with a vehicle plane and has calibrated internal parameters; detecting feature points and feature lines in the pictures acquired byeach camera, matching the feature points of the adjacent frames, and calculating a translation vector of each camera; extracting a vertical feature line perpendicular to the real ground in each frameof picture, and calculating a normal vector perpendicular to the vertical feature line and emitted from one camera corresponding to the current picture; classifying the pictures collected by the cameras; randomly selecting a frame of straight driving picture, and calculating an initial value of a rotation matrix of a corresponding camera and a vehicle body coordinate system by using a normal vector and a translation vector corresponding to the straight driving picture; optimizing initial values. According to the invention, the problem that external parameters of a camera system are difficult to calibrate under the extreme condition that no (few) view field overlapping area exists among cameras is solved, the field is saved, and the design and manufacturing cost of a marker is saved.

Description

technical field [0001] The invention belongs to the technical field of multi-camera systems, and in particular relates to a method for estimating external parameters of a vehicle-mounted surround-view multi-camera system and a system thereof. Background technique [0002] As one of the most potential technologies in the world today, unmanned driving means that the car senses the surrounding environment and completes navigation tasks through its own sensors without the need for artificial intelligence. PricewaterhouseCoopers predicts that the popularization of unmanned driving technology will reduce overall traffic accidents by 90%; KPMG Automotive Research Center predicts that unmanned driving technology will drive productivity and energy efficiency to be improved, and new businesses will emerge. model. [0003] Self-driving cars are often equipped with sensors such as cameras, inertial measurement units (IMUs), lidar, and global positioning systems (GPS). Among them, the ...

AI Technical Summary

Problems solved by technology

However, for large vehicles (such as heavy trucks with trailers), the number of cameras on the body is limited under the premise of taking into account the lack of blind spots and cost considerations

There are two problems at this point: (1) There will be a large distance between cameras; (2) There is no (or only a small) field of view overlap between some cameras

These actual conditions make it very difficult to calibrate the external parameters of the car surround view multi-camera system

Simply copying the above-mentioned calibration strategy for small vehicle surround-view multi-eye cameras will require a lot of space

Images