NDT point cloud registration algorithm and device based on GPU, and electronic equipment

Abstract

The embodiment of the invention provides an NDT point cloud registration algorithm and device based on a GPU, and electronic equipment, and belongs to the technical field of point cloud registration computers, and the algorithm comprises the steps of extracting target point cloud data, extracting source point cloud data, calculating a Gaussian approximation constant, giving an initial guess pose, transforming points in a source point cloud, calculating a Jacobian matrix and a Hessian matrix, judging whether points in the source point cloud after pose transformation are effective points, calculating a gradient vector of each effective point and a Hessian matrix H, respectively adding the gradient vector and the Hessian matrix H, and adding the sum to Hapos; a Jacobi method is used to realize a singular value decomposition equation to solve a pose increment and a norm thereof, the pose increment is normalized, a line search algorithm is used to update the pose increment, a source point cloud is transformed according to new pose transformation, the solving process of a gradient vector, a Hessian matrix and the pose increment is repeated, and an optimal transformation pose is obtained through iteration. Through the processing scheme provided by the invention, the problem of high delay of the algorithm at the CPU end is reduced, and the working efficiency of the GPU is improved.

Description

technical field [0001] The present application relates to the field of point cloud registration computer technology, and in particular to a GPU-based NDT point cloud registration algorithm, device and electronic equipment. Background technique [0002] The positioning task of autonomous driving is often completed by a combination of GNSS (Global Navigation Satellite System, Global Navigation Satellite System), IMU (Inertial Measurement Unit, Inertial Measurement Unit) and RTK (Real-Time Kinematic, carrier phase difference technology), but these positioning Methods are often constrained by some real-world conditions at runtime. For example, in canyons and tall buildings, the positioning accuracy of GNSS may drop from 1-3 meters to 10-50 meters; after continuous integration of IMU, the positioning accuracy will deteriorate due to the accumulation of positioning errors; RTK is not used in most areas, etc. Wait. Therefore, in today's autonomous driving positioning tasks, posit...

AI Technical Summary

Problems solved by technology

[0004] The current NDT algorithms are all calculated by the CPU. However, the CPU cannot guarantee the low latency of the algorithm when performing large-scale computing tasks. The reason is that the CPU has limited computing units; this problem is more obvious on the vehicle computing platform, because The computing power of the CPU is further limited by the number of CPUs

Images