Annotation-Free Curb Detection Leveraging Altitude Difference Image

TL;DR

This work tackles curb detection for autonomous driving by eliminating the need for manual annotations. It converts LiDAR point clouds into a structured Altitude Difference Image (ADI), which highlights height-difference features like curbs and is robust to lighting variations, and introduces an Automatic Curb Annotator (ACA) to generate training data automatically. A lightweight MobileOne+UNet++ network processes ADI for fast inference, followed by BEV post-processing that fits quadratic curves to deliver smooth curb delineations with substantially reduced latency. Evaluations on the KITTI 3D curb dataset show state-of-the-art accuracy with real-time performance, demonstrating the practicality of annotation-free, ADI-based curb detection for real-world deployment.

Abstract

Road curbs are considered as one of the crucial and ubiquitous traffic features, which are essential for ensuring the safety of autonomous vehicles. Current methods for detecting curbs primarily rely on camera imagery or LiDAR point clouds. Image-based methods are vulnerable to fluctuations in lighting conditions and exhibit poor robustness, while methods based on point clouds circumvent the issues associated with lighting variations. However, it is the typical case that significant processing delays are encountered due to the voluminous amount of 3D points contained in each frame of the point cloud data. Furthermore, the inherently unstructured characteristics of point clouds poses challenges for integrating the latest deep learning advancements into point cloud data applications. To address these issues, this work proposes an annotation-free curb detection method leveraging Altitude Difference Image (ADI), which effectively mitigates the aforementioned challenges. Given that methods based on deep learning generally demand extensive, manually annotated datasets, which are both expensive and labor-intensive to create, we present an Automatic Curb Annotator (ACA) module. This module utilizes a deterministic curb detection algorithm to automatically generate a vast quantity of training data. Consequently, it facilitates the training of the curb detection model without necessitating any manual annotation of data. Finally, by incorporating a post-processing module, we manage to achieve state-of-the-art results on the KITTI 3D curb dataset with considerably reduced processing delays compared to existing methods, which underscores the effectiveness of our approach in curb detection tasks.

Figures (4)

• Figure 1: Comparison between RGB image and ADI under extreme lighting conditions. The top row represents RGB images, while the bottom row represents ADIs. In the first row, the image in column (a) is in a high-exposure scenario, while the image in column (b) is in a scenario with low lighting under tree shade. It can be observed that under these extreme lighting conditions, the curb in the RGB images is difficult to discern (as indicated by the red dashed boxes). However, the curb is clearly visible in the ADIs.
• Figure 2: The overall framework of our proposed curb detection method based on ADIs with an annotation-free approach. (a): Schematic diagram of the network architecture. (b): Illustration of the process in the ACA module.
• Figure 3: The MobileOne module has two different structures during training and inference. Left: MobileOne module during training with reparameterizable branches. Right: MobileOne module during inference where the branches are reparameterized.
• Figure 4: Data flow throughout the entire process from inputting the ADI into the network to obtaining the final curb detection results.