UAV-Borne LiDAR Crop Point Cloud Enhancement Using Grasshopper Optimization and Point Cloud Up-Sampling Network

Jian Chen,Shubo Wang,Zichao Zhang,Kai Zhang,Yu Han

doi:10.3390/rs12193208

Jian Chen, Shubo Wang + Show 3 more

Open Access

https://doi.org/10.3390/rs12193208

Copy DOI

Journal: Remote sensing	Publication Date: Oct 1, 2020
Citations: 9	License type: CC BY 4.0

Affiliation: China Agricultural University

Abstract

Because of low accuracy and density of crop point clouds obtained by the Unmanned Aerial Vehicle (UAV)-borne Light Detection and Ranging (LiDAR) scanning system of UAV, an integrated navigation and positioning optimization method based on the grasshopper optimization algorithm (GOA) and a point cloud density enhancement method were proposed. Firstly, a global positioning system (GPS)/inertial navigation system (INS) integrated navigation and positioning information fusion method based on a Kalman filter was constructed. Then, the GOA was employed to find the optimal solution by iterating the system noise variance matrix Q and measurement noise variance matrix R of Kalman filter. By feeding the optimal solution into the Kalman filter, the error variances of longitude were reduced to 0.00046 from 0.0091, and the error variances of latitude were reduced to 0.00034 from 0.0047. Based on the integrated navigation, an UAV-borne LiDAR scanning system was built for obtaining the crop point. During offline processing, the crop point cloud was filtered and transformed into WGS-84, the density clustering algorithm improved by the particle swarm optimization (PSO) algorithm was employed to the clustering segment. After the clustering segment, the pre-trained Point Cloud Up-Sampling Network (PU-net) was used for density enhancement of point cloud data and to carry out three-dimensional reconstruction. The features of the crop point cloud were kept under the processing of reconstruction model; meanwhile, the density of the crop point cloud was quadrupled.

Highlights

Light Detection and Ranging (LiDAR) is an active sensing technology that can quickly acquire spatial information of a target or environment [1,2]
In order to verify the stability of the Unmanned Aerial Vehicle (UAV)-borne LiDAR scanning system, the experiments were carried out in integrated navigation testing, point cloud density clustering, and point cloud density enhancement
Considering that particle swarm optimization (PSO) converges faster and takes less time in iterations, the PSO-based density-based spatial clustering of applications with noise (DBSCAN) algorithm was employed for crop point cloud clustering segmentation

Summary

Introduction

Light Detection and Ranging (LiDAR) is an active sensing technology that can quickly acquire spatial information of a target or environment [1,2]. The Unmanned Aerial Vehicle (UAV)-borne LiDAR scanning system is one of the most common platforms for crop point cloud obtaining. The crop point clouds captured by UAV-borne LiDAR scanning system are sparse and unordered [5]. The navigation and positioning accuracy of the UAV-borne LiDAR scanning system influences the accuracy of the point cloud data greatly. Density is an important indicator to measure the quality of point cloud data. Higher point cloud density represents richer information of the target or environment [6]. Ensuring the accuracy and density characteristics of the UAV-borne LiDAR point cloud data while maintaining the cost is crucial for the continuous and efficient execution of LiDAR detection tasks

Results

Discussion

Conclusion