An Online Extrinsic Calibration Method for Object-Free Scenarios Based on the Joint Features of Millimeter-Wave Radar and Cameras
2021-11-09
Sensor fusion is critical for autonomous driving and autonomous robotics, and millimeter-wave radar–camera fusion systems have gained widespread popularity due to their complementary sensing capabilities. However, precise calibration between these two sensors is even more essential to ensure effective fusion and enhance overall system performance. Calibration encompasses both intrinsic and extrinsic alignment, with the latter being particularly crucial for achieving accurate sensor fusion. Unfortunately, many target‑based calibration methods require complex procedures and carefully designed experimental conditions, posing significant challenges for researchers. To address this issue, this paper introduces a novel approach that leverages deep learning to extract shared features from raw millimeter-wave radar data—specifically range‑Doppler‑angle information—and camera images. Rather than explicitly representing these common features, our method implicitly uses them to match identical objects across the two data sources. By treating the extracted features as a single instance, we demonstrate an online, targetless calibration method for millimeter-wave radar–camera systems. This feature‑based approach enables the estimation of the extrinsic transformation matrix; to improve calibration accuracy and robustness, RANSAC and the Levenberg–Marquardt (LM) nonlinear optimization algorithm are employed to derive the matrix. Experimental results in real-world environments validate the effectiveness and precision of the proposed method.
Millimeter-wave radar and cameras are complementary sensing modalities, widely employed in applications such as autonomous driving and robotics. Regardless of lighting or weather conditions, millimeter-wave radar can provide accurate information on distance, velocity, and angle, while cameras capture high‑resolution visual data. The integration of millimeter-wave radar and cameras enhances perception capabilities, enabling tasks like object recognition, detection, and tracking in dynamic environments. Calibration between millimeter-wave radar and cameras is critical for sensor fusion, as their sensing principles differ—particularly when determining their relative pose. A promising approach involves leveraging deep learning to extract meaningful features from raw millimeter-wave radar data and exploring the relationships between radar and image features to estimate the extrinsic transformation matrix.
Contact Us
Email:
sales@aumiwalker.com
Phone:
+86-10-63718360
Address:
Room 505, Floor 5, Building 6, Phase II, Nord Center, No. 1 Courtyard, Auto Museum East Road, Fengtai District, Beijing
Do you need an expert to recommend the right product for you?
If you have any questions about our products, please contact us promptly.
Contact Us
Phone: +86-10-63718360
Email: sales@aumiwalker.com
Address: Room 505, Floor 5, Building 6, Phase II, Nord Center, No. 1 Courtyard, East Auto Museum Road, Fengtai District, Beijing
WeChat Official Account
WeChat Mini Program
Leave a Message Online
Telephone：
E-mail：
WeCom：
My Favorites：