Real-Time Detection of Non-Stationary Objects Using Intensity Data in Automotive LiDAR SLAM

Abstract

This article aims at demonstrating the feasibility of modern deep learning techniques for the real-time detection of non-stationary objects in point clouds obtained from 3-D light detecting and ranging (LiDAR) sensors. The motion segmentation task is considered in the application context of automotive Simultaneous Localization and Mapping (SLAM), where we often need to distinguish between the static parts of the environment with respect to which we localize the vehicle, and non-stationary objects that should not be included in the map for localization. Non-stationary objects do not provide repeatable readouts, because they can be in motion, like vehicles and pedestrians, or because they do not have a rigid, stable surface, like trees and lawns. The proposed approach exploits images synthesized from the received intensity data yielded by the modern LiDARs along with the usual range measurements. We demonstrate that non-stationary objects can be detected using neural network models trained with 2-D grayscale images in the supervised or unsupervised training process. This concept makes it possible to alleviate the lack of large datasets of 3-D laser scans with point-wise annotations for non-stationary objects. The point clouds are filtered using the corresponding intensity images with labeled pixels. Finally, we demonstrate that the detection of non-stationary objects using our approach improves the localization results and map consistency in a laser-based SLAM system.

Application of Laser Systems for Detection and Ranging in the Modern Road Transportation and Maritime Sector

“…The tests on the KITTI pedestrian detection dataset show that the proposed approach outperforms the one using only camera imagery [54] CNN-based classification of objects using camera and lidar data from autonomous vehicles, where point cloud lidar data are upsampled and converted into the pixel-level depth feature map, which is then fused with the RGB images and fed to the deep CNN Results obtained on the public dataset support the effectiveness and efficiency of the data fusion and object classification strategies, where the proposed approach outperforms the approach using only RGB or depth data [55] Real-time detection of non-stationary (moving) objects based on the CNN using intensity data in automotive lidar SLAM…”

“…Test data are collected by some of the known commercial lidar systems combined with optical cameras, or using KITTI datasets. There are various applications of CNNs, where some studies use them only for a specific step, such as vehicle detection [51,52], pedestrian detection [53], and object classification [54][55][56][57][58], while some try to use them for the whole process. For example, in [59], the authors proposed an end-to-end (E2E) self-driving algorithm utilizing a CNN that provided the vehicle speed and angle as outputs based on the input camera and 2D lidar data.…”

Adopting the YOLOv4 Architecture for Low-Latency Multispectral Pedestrian Detection in Autonomous Driving

“…The first will focus on utilizing data streams from both modalities when the images are not co-aligned, making a practical solution when two separate cameras are combined in a single sensory setup. The other direction of future research will investigate the detection of small objects in images and the addition of preprocessed LiDAR data [41] for better detection of pedestrians from long distances.…”