Inductively coupled plasma (ICP) algorithm-based point cloud file registration method and device and storage medium

Abstract

The invention discloses a point cloud file registration method based on an ICP algorithm, and the method comprises the steps: obtaining a point cloud file set, and starting a plurality of processes; dividing the point cloud file set into a plurality of point cloud segments, and pre-distributing the point cloud segments to corresponding processes in sequence; creating a shared memory containing the first variable and the second variable, and initializing the shared memory; according to the first variable, idle GPUs or CPUs are scheduled in real time in each process to execute an ICP algorithm, target point clouds in each process are registered through the ICP algorithm, and a registered point cloud file is obtained; wherein when any process completes registration of all distributed target point clouds in advance, the point cloud segment with the maximum number of unregistered target point clouds is searched through the second variable, and the unregistered target point clouds in the point cloud segment are dynamically distributed to the idle process for registration. Therefore, the parallelization degree of registration of the series of point cloud files, the utilization rate of computing resources and the efficiency of parallel registration of the point cloud files can be improved.

Description

technical field [0001] The present invention relates to the technical field of signal processing, in particular to an ICP algorithm-based point cloud file registration method, device and computer-readable storage medium. Background technique [0002] At present, the methods for ICP (Iterative Closest Point) registration of point cloud files mainly include: serial point cloud file serial ICP registration algorithm and ICP parallel algorithm. Serial ICP registration algorithm for a series of point cloud files: Point cloud library (PCL for short) provides a step-by-step matching scheme for two or two point cloud files. Unified into the coordinate system of the source point cloud. ICP parallel algorithm: It mainly studies the ICP parallel acceleration between two point cloud files. The idea is to use GPU or multiple processor cores (CPU cores) to parallelize certain steps in the ICP algorithm. [0003] However, the above two registration algorithms use CPU+GPU or multiple CPU ...

AI Technical Summary

Problems solved by technology

[0004] 1. The degree of parallelization is insufficient. The existing technology focuses on the parallel registration between two point cloud files. It mainly accelerates some steps of the ICP algorithm. It is a level of parallelism, and the parallel level is low. For cloud file set registration, its degree of parallelization is insufficient

[0005] 2. The computing resource utilization rate of the computer is low. The existing technology mainly uses the CPU+GPU mode or multiple CPU core modes for parallel acceleration, which fails to make full use of the computing resources of the computer.

During CPU serial execution, the computer only starts one CPU core for calculation, and the rest of the CPU cores and GPU are in an idle state; after the serial execution of the CPU part is completed, the computer starts the GPU to accelerate the parallel part, and all CPU cores in the computer enter the idle state Or waiting state, that is, during the entire registration process, the CPU and GPU are asynchronous, and the computer only has one CPU core or GPU running at the same time, which fails to fully use the computing resources of the computer

(2) The existing multiple CPU core parallel mode is to parallelize the part of the ICP algorithm that is suitable for multi-core parallelism, and the rest of the algorithm still maintains the CPU serial operation

Similarly, when the CPU is executing serially, only one CPU core of the computer is calculating at the same time, and the remaining CPU cores and GPU are idle; when multiple CPU cores are executing in parallel, the GPU of the computer is idle, which is also not fully utilized. computer computing resources

Images