Learning-Based Human Segmentation and Velocity Estimation Using Automatic Labeled LiDAR Sequence for Training

Abstract

In this paper, we propose an automatic labeled sequential data generation pipeline for human segmentation and velocity estimation with point clouds. Considering the impact of deep neural networks, state-of-the-art network architectures have been proposed for human recognition using point clouds captured by Light Detection and Ranging (LiDAR). However, almost all conventional datasets are either a collection of single LiDAR scanning with label information or sequential LiDAR scanning without label information. This limitation has disturbed the progress of research to date. Therefore, we have developed an automatic labeled sequential data generation pipeline, in which we can control any parameter or data generation environment with pixel-wise and per-frame ground truth segmentation and pixel-wise velocity information for human recognition. Our approach uses a precise human model and reproduces a precise motion to generate realistic artificial data. We present more than 7K sequences, where each sequence consists of 32 frames generated by the proposed pipeline. With the proposed sequence generator, we confirm that human segmentation performance is improved when using the sequential data compared to when using the data from a single LiDAR scan. We also evaluate our data by comparing with data generated under different conditions. In addition, we estimate pedestrian velocity with LiDAR by only utilizing data generated by the proposed pipeline.

Novel CE-CBCE feature extraction method for object classification using a low-density LiDAR point cloud

“…This is due to density changes with the distance from the LiDAR [ 34 ]. Especially in a very sparse LiDAR data recognition performance drastically decrease as distance between human and LiDAR increases, due to the number of points being inversely proportional to the square of the distance between a human to LiDAR [ 35 ].…”

Human Segmentation with Dynamic LiDAR Data

“…Due to the difficulty of annotation, most of the dynamic point cloud datasets are generated with simulation. Kim et al developed a pipeline for single frame [19] [20] and sequence of point cloud data generation [7]. SYNTHIA [8] dataset provided many road scene videos generated by a game engine.…”

“…Unfortunately, as far as we know, there exists so far no point cloud dataset that provides both segmentation and scene flow annotation, due to the difficulty of labeling. Alternatively, we generate sequential LiDAR data with pixel-wise segmentation information and pixel-wise velocity information for training and evaluation, using the pipeline proposed in [7].…”

“…release of several synthetic and real datasets [7], [8], [9] for dynamic point cloud is accelerating some related researches. The point-wise method [10] and voxel-based method [11], which were already proved effective in static point cloud domain, have been extended to dynamic point cloud domain.…”