Ghost targets are targets that appear at wrong locations in radar data and are caused by the presence of multiple indirect reflections between the target and the sensor. In this work, we introduce the first point based deep learning approach for ghost target detection in 3D radar point clouds. This is done by extending the PointNet network architecture by modifying its input to include radar point features beyond location and introducing skip connetions. We compare different input modalities and analyze the effects of the changes we introduced. We also propose an approach for automatic labeling of ghost targets 3D radar data using lidar as reference. The algorithm is trained and tested on real data in various driving scenarios and the tests show promising results in classifying real and ghost radar targets.
In-car activity monitoring is a key enabler of various automotive safety functions. Existing approaches are largely based on vision systems. Radar, however, can provide a low-cost, privacy-preserving alternative. To this day, such systems based on the radar are not widely researched. In our work, we introduce a novel approach that uses the Doppler signal of an ultra-wideband (UWB) radar as an input to deep neural networks for the classification of driving activities. In contrast to previous work in the domain, we focus on generalization to unseen persons and make a new radar driving activity dataset (RaDA) available to the scientific community to encourage comparison and the benchmarking of future methods.
Applications of deep learning have recently seen a surge in the field of construction. Supervised semantic segmentation of 2D or 3D data acquired from buildings requires the use of annotated data for training, validation, and testing. Although various datasets have been published targeting this application, they lack a common convention and definitions based on construction ontologies. This work presents a guideline for ontology-based semantic annotation of RGB-D and point cloud datasets for buildings. Such a contribution facilitates the use of deep learning in construction by bridging the gap between this field and computer science.The annotation guideline is available under this link https://gitlab.rhrk.uni-kl.de/kaufmann/humantech-data-annotation.
The use of cameras and radar sensors is well established in various automation and surveillance tasks. The multimodal nature of the data captured by those two sensors allows for a myriad of applications where one covers for the shortcomings of the other. While cameras can capture high resolution color data, radar can capture the depth and velocity of targets. Calibration is a necessary step before applying fusion algorithms to the data. In this work, a robust extrinsic calibration algorithm is developed for camera-radar setups. The standard geometric constraints used in calibration are extended with elevation constraints to improve the optimization. Furthermore, the method does not rely on any external measurements beyond the camera and radar data, and does not require complex targets unlike existing work. The calibration is done in 3D thus allowing for the estimation of the elevation information that is lost when using 2D radar. The results are evaluated against a sub-millimeter gro und truth system and show superior results to existing more complex algorithms. https://github.com/mahdichamseddine/CaRaCTO.
The multimodal nature of camera and radar sensor data enables various automation and surveillance tasks, where one sensor compensates for the limitations of the other sensor: cameras capture high-resolution color data, while radar measures depth and velocity of targets. Calibration is essential to fuse these data modalities effectively. This work presents a robust extrinsic calibration algorithm for camera–radar setups, extending standard geometric constraints with elevation information to enhance optimization. Unlike existing methods, this approach relies solely on camera and radar data without requiring complex targets or external measurements. The 3D calibration enables the estimation of the target elevation which is lost when using 2D radar. We evaluate our results against a sub-millimeter ground truth system, demonstrating superior performance compared to more complex algorithms. Leveraging these accurate calibration results, we subsequently employ monocular depth estimation and instance segmentation techniques to perform camera–radar data fusion, allowing 3D target and scene reconstruction. github.com/mahdichamseddine/CaRaCTO.