Vehicle pose estimation method and system based on non-overlapping vision field multi-camera system

Abstract

The invention discloses a vehicle pose estimation method based on a non-overlapping vision field multi-camera system. The vehicle pose estimation method comprises the steps: synchronously collecting environment pictures through a plurality of vehicle-mounted monocular cameras which move along with a vehicle and calibrate internal and external parameters; detecting feature points in the pictures acquired by each vehicle-mounted monocular camera, tracking the feature points, and matching the feature points of adjacent frames to obtain feature point pairs; obtaining a first optimization objectivefunction for the pose of the single camera through the orthogonal relation between the relative translation amount of the single vehicle-mounted monocular camera between two adjacent frames and the antipodal plane normal vector formed by all the direction vector pairs; converting the variables in the first optimization objective function into a multi-camera system center coordinate system througha second optimization objective function to be expressed; and performing iterative optimization on the second optimization objective function through an iterative estimation algorithm to obtain a vehicle pose. The vehicle pose estimation method can adapt to low-cost hardware and is high in accuracy.

Description

technical field [0001] The invention belongs to the technical field of multi-camera systems, and in particular relates to a vehicle pose estimation method based on a non-overlapping 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 human operation. 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] The same as the traditional manual driving operation principle, the real-time perception and positioning of the vehicle operating environment is the basis for the decision-making and control of the unmann...

AI Technical Summary

Problems solved by technology

For example, defects such as low computing efficiency, poor algorithm robustness, and poor accuracy of calculation results

The specific performance is as follows: (1) The traditional 6-point method will obtain 64 sets of solutions without using the constraints of the vehicle motion model, the computational complexity is high, and it is not suitable for real-time processing

(2) The traditional 17-point method uses too many feature points, has poor tolerance to noise and outliers, and the algorithm is not robust enough

(3) The 1-point algorithm based on the vehicle motion model and its extended 2-point algorithm, the motion model used is too ideal, and the accuracy is poor in practical applications

[0011] Although the vehicle-mounted multi-camera visual positioning method can meet the positioning requirements, there are the following problems: First, the method relies on four binocular cameras whose cost is much higher than that of ordinary cameras The data is obtained, and the pose data needs to be fused with the information of the vehicle inertial measurement unit, and the hardware cost is relatively high; the second is that the method uses 4 binocular cameras to calculate the pose separately and then perform visual fusion processing to obtain the vehicle pose. The on-board computer needs to have strong computing power to meet the real-time visual positioning requirements of the unmanned driving system, but the existing vehicle-level embedded processors are still not competent; third, the method does not mention the optimization of the visual positioning algorithm, such as Perform data processing with the existing public positioning algorithm, according to our actual evaluation and testing of various existing optimization methods (see Figure 4), its positioning accuracy and efficiency need to be improved

The vehicle pose estimation of the multi-camera system is realized with the existing public technical solutions and optimization algorithms. Due to the low matching degree of the hardware between the existing algorithm and the non-overlapping panoramic multi-camera system, the accuracy, real-time and reliability of the algorithm are caused. Insufficient, there is still a certain gap with the actual application requirements

Images