Summary of Application of Parallel Computing in lidar, GIS, and RS

I. Current Work Progress

1, Application in Lidar

Recently, the main task is to read articles on the Application of Parallel Computing in lidar technology. Currently, it is rare to introduce parallel computing into the field of Lidar Data Processing. I have searched for three articles: a grid-based Lidar Data Processing Platform Architecture and a three tier.
Architecture for lidar interpolation and Analysis (a three-tier architecture for interpolation and Analysis of LIDAR data) and parallel
Algorithm for Linear Feature Detection from Airborne Lidar Data (LIDAR Linear Feature Detection Based on Parallel Algorithms).

The system structure of a grid-based Lidar Data Processing Platform analyzes the process of Lidar Data Processing, and proposes a system of LIDAR data processing platform, data processing tasks are reasonably divided and allocated to grid nodes of various distributions. Parallel and collaborative computing of each node improves the computing speed. The architecture mentioned in this article is not complex. The difficulty lies in the fact that the algorithm for processing data in lidar is complex. It is impractical to implement data processing through Algorithm Programming; how to rationally divide data processing tasks in a PC group is also to be discussed.

A three tier architecture for lidar interpolation and
Analysis proposes a three-layer architecture that uses the portal, workflow, and grid technologies to coordinate various distributed resources for the interpolation and Analysis of LIDAR data, it boasts superior performance in processing dense laser points.

Parallel Algorithm for Linear Feature Detection from airborne
The Application of Parallel Computing in Lidar Data is mainly to convert the original aerial LIDAR data into a rule grid by using the nearest adjacent interpolation method, so as to perform parallel operations on the PC group.

From the above three documents, the main idea of parallel computing and processing of LIDAR on a PC group is to rationally divide the laser points, considering the number of sub-nodes, load capacity, and communication between nodes, the split data and metadata are allocated to the sub-nodes. After calculation, the results and metadata are returned to the master node together, integrate it to get the final result.

In addition, for applications of parallel computing and grid computing in lidar, the traditional Serial Algorithm of LIDAR data processing needs to be changed to a parallel algorithm for processing on PC clusters.

2. Application in Remote Sensing Image Processing

There are three features in remote sensing image processing: 1) the data volume is extremely large, and sequential processing cannot meet the requirements of massive Real-time Data Operations; 2) each data point is computed in the same way (except for the boundary). When a region is divided into several small regions, the problem can be resolved to several small-scale subproblems related to small regions. 3) the interaction between variables is local, that is, in the calculation of each data point, only the values at the neighboring points within a small distance are needed. Among them, the first feature facilitates the objective needs of remote sensing image processing for Parallel Processing Technology. The last two features indicate that remote sensing digital image processing is inherently parallel and a distributed parallel image processing system is established, it facilitates real-time processing of massive digital image data.

Therefore, there are various discussions and applications in the field of remote sensing image processing at home and abroad, such as the Key Technology Research on rapid evaluation of flood disasters proposed by Chi Tianhe. it highlights the rapid processing of large data volumes of remote sensing images in parallel computing, and emphasizes the importance of real-time disaster assessment. For example, "high-precision parallel monitoring classification of remote sensing images", this paper studies the parallel geometric correction algorithm in the distributed storage environment. Application of PVM-based network parallel computing in remote sensing image processing. the parallel computing algorithm is used in Linux to test DEM visualization in the digital Qingjiang River basin using OpenGL, as well as remote sensing image stretching parallel computing described in the Application of Distributed Parallel Computing Technology in remote sensing data processing.

The common feature of the above articles is to transform the algorithms of different branches in traditional remote sensing image processing in parallel so that massive data can be processed quickly and in real time on the cluster. Data Division in Distributed Parallel remote sensing image processing discusses how to reasonably divide computing tasks on parallel machines; the Model Research and Experiment of the distributed parallel remote sensing image processing system based on registration service is to explore how to establish a fast gridded and parallel Remote Sensing Processing Technology for concurrent operations on large databases. problem.

3, Application of Parallel Computing and Grid Technology in GIS

Similar to the Application of Parallel Computing in remote sensing, parallel GIS is also widely used, such as parallel conversion of spatial data formats in GIS, and construction of Water Resource Management Information System Based on Parallel Computing and GIS technology, and buffer analysis in the grid environment.

Ii. Work conception and obstacles

I have consulted many documents on the Application of Parallel Computing in lidar, GIS, and remote sensing, and it is not difficult to find some universal rules: parallel computing is mainly used for fast and efficient processing of massive data, the computing time complexity will also be converted into space complexity. The steps that will form high-density computing operations in the processing process will be transformed in parallel by the existing complete algorithms, the obtained parallel processing algorithm runs on a PC group. Finally, this type of parallel algorithm is encapsulated into grid services using a Grid container to form a functional component with inherent rules of the high-performance parallel computing grid.

Currently, the main research areas are as follows:

1. Divide the LIDAR data into several subnodes, calculate the DSM or Dem, and return the data to the master node for integration to obtain the grid DSM (DEM ).

2. Parallel Computing of remote sensing image processing, such as geometric correction and linear stretching.

3. Parallel GIS.

Theoretically, as long as the massive data can be reasonably divided into multiple layers or images, the target algorithm can be transformed into a parallel algorithm. However, the current difficulties are as follows:

1,
I don't know about PC Clusters and how they run.

2,
There are a lot of materials on the Internet, but we have never seen GIS or LIDAR products that are actually computed on PC clusters. It is difficult to use research articles in practice.