ROI-Based LiDAR Sampling Algorithm in on-Road Environment for Autonomous Driving

Abstract

As the acquisition of laser range measurements such as those from light detection and ranging (LiDAR) sensors requires a considerable amount of time, to design an effective sampling algorithm is a critical task in numerous laser range applications. The state-of-the-art adaptive methods such as two-step sampling are highly effective at handling less complex scenes such as indoor environments with a moderately low sampling rate. However, their performance is relatively low in complex on-road environments, particularly when the sampling rate of the measuring equipment is low. To address this problem, this paper proposes a region-of-interest (ROI)-based sampling algorithm in on-road environments for autonomous driving. With the aid of fast and accurate road and object detection algorithms, particularly those based on convolutional neural networks, the proposed sampling algorithm utilizes the semantic information and effectively distributes samples in the road, object, and background areas. The experimental results demonstrate that the proposed algorithm significantly reduces the mean-absolute-error in the object area by at most 52.8% compared to two-step sampling; moreover, it achieves robust reconstruction quality even at a very low sampling rate of 1%. INDEX TERMS Autonomous driving, LiDAR sampling, on-road environment, ROI-based sampling, twostage sampling.

Automotive Radar Sub-Sampling via Object Detection Networks: Leveraging Prior Signal Information

“…A further distinctive feature of our work is that we allocate to the important regions of the environment a larger sampling budget while maintaining the overall sampling budget and reconstruction complexity. In another related work [33], the acquisition of LiDAR data is guided using the Region-of-Interest information determined based on the results of image segmentation; in contrast, our work helps guide radar data acquisition utilizing the results of 2-D object detection.…”

End-to-end system for object detection from sub-sampled radar data

“…Moreover, a distinctive feature of our work is that the measurement matrix size increases for certain regions of the environment as we allocate them a larger sampling budget while maintaining the overall sampling budget and reconstruction complexity. In another related work [25], the acquisition of LiDAR data is guided using the Region-of-Interest information determined based on the results of image segmentation; our work helps guide radar data acquisition utilizing the results of 2-D object detection.…”

Content adaptive video compression for autonomous vehicle remote driving

“…Then, coding blocks are mapped and assigned ROI priorities only if they contain parts of the crucial objects. Third, the car's Lidar (Light detecting and ranging) depth maps [8] can distinguish between ROI and non-ROI coding blocks according to their distances from the vehicle. After mapping the image into non-ROI and ROI coding blocks, the encoder allocates bandwidths to both regions.…”

“…ROI is dynamically adjustable by location, size, resolution, and bitrate. The authors of [8] allocate the Lidar bitrate budget according to ROI and non-ROI, significantly improving the quality of selfdriving car depth maps. In [37], the encoder allocates the CTU bits according to their visual saliency.…”