[ NODAR PRODUCTS ]
GridDetect
Real-time object detection and tracking layer
Real-Time 3D Pointcloud Processing for Accurate Object Tracking
GridDetect processes Hammerhead’s high-resolution 3D pointcloud data to provide a real-time bird’s-eye view (BEV) of the environment, as well as object lists with velocity and direction.
GridDetect BEV Views
Using particle filters, clusters of points above the groundplane (objects) are tracked and a realtime occupancy map is produced.
Features
Detect and track objects up to 250m in realtime
6-250m BEV and object list with relative speed and heading
Groundplane detection with tunable parameters, such as max obstruction height and max grade
Fast, reliable, deterministic signal processing approach does not require training
Small object detection - 2cm @10m or 19cm @200m (30˚ FOV, 90% TP)
Robust performance in high-vibration, low-light and inclement weather
Preserves obstacles during occlusion
Adapts to multiple sensor resolution
Differentiate between static and dynamic obstacles
Enables trajectory prediction
Benefits
Reliable object detection with no prior training
BEV and object list output simplify integration
High-reliability and long-range uniquely enable ALKS L3 ADAS
Appropriate for safety and security critical applications
Reduce false positives
GridDetect Architecture
GridDetect In Action
NODAR Viewer realtime output above: RGB (upper-left), Depthmap (lower-left), pointcloud with bounding boxes (center), BEV (right)