CNN-based Human Detection Using a 3D LiDAR onboard a UAV

Abstract

This paper addresses the problem of detecting humans in a point cloud taken with a 3D-LiDAR onboard a UAV. The potential use cases of this technology are numerous, examples include security and surveillance, disaster relief and search and rescue operations. In this paper, a CNN-based approach is proposed which is able to analyse point clouds returned by a 3D LiDAR sensor in such a way that it can detect humans. The algorithm described here consists of 3 main components: data pre-processing, post-processing, and human classification. In this paper objects were assigned to one of two classes: human and non-human. The classification was performed by projecting the 3D point cloud onto a series of 2D planes using occupancy grid mapping. This creates a set of silhouettes of the object corresponding to the top, front and side views. Classification is achieved by supervised CNNs using single-view and multi-view (3 channels) images patches. Raw DataRaw data from airborne LiDAR Data Synch. LOAM-based LiDAR Point-Cloud synchronization Ground detection RANSAC to determine the ground plane Data Alignment Align Point-cloud w/ reference coordinate axes Define ROI Use reference point to define rectangular ROI Clustering Separate Pointcloud into clusters Occ. Grid Project the Points onto 2D Grid map Process Image Convert grid to 2D image at cluster location Human classification CNN-based classifier Point cloud pre-processing Post-processing CNN-based classification

Towards Safe Robotic Agricultural Applications: Safe Navigation System Design for a Robotic Grass-Mowing Application through the Risk Management Method

“…Another sensor used for human detection in real-time applications is the 3D LiDAR, where a combination of filtering and convolutional neural network layers can detect humans in cluttered environments [19], or with a two-stage deep learning architecture, such as PointRCNN, developed by Shi et.al. [20], that decomposes the problem into bounding box proposal generation with a refining coordinate process as the last stage.…”

Intelligent Amphibious Ground-Aerial Vehicles: State of the Art Technology for Future Transportation

“…Laser generated by LiDAR has strong penetration, such that obtain information beneath shelters like trees to easily handle the three-dimensional segmentation task [68]. Furthermore, [69] analyzed the point cloud returned by a 3D LiDAR sensor with a convolutional neural network and detected humans walking on the ground. There was also a precedent for the research on airborne LiDAR-guided aircraft landing.…”

Development of Self-Synchronized Drones’ Network Using Cluster-Based Swarm Intelligence Approach

“…These algorithms are useful and great in the cases of two dimensions, however, since SDN is a three diminutions network scenario some more aspects need to be considered per in mind the high computational overhead in the three dimensions, while the drones usually have very limited and scarcity resources. The highest-degree of freedom algorithm is denoted as a connectivity-based procedure, which is principally based on the number of degrees of freedom values free to vary in the ultimate vehicle clustering statistic design and calculation [43][44].…”