Method And System For Map Construction

What is this patent about?

’609 is related to the field of map construction , particularly the automated generation of lane-level roadmaps from raw image data. Traditional mapmaking is labor-intensive and expensive, requiring a balance between accuracy and cost. This patent addresses the problem of efficiently creating accurate maps suitable for advanced navigation and autonomous driving systems.

The underlying idea behind ’609 is to automate the process of extracting lane geometry from road images by first creating a simplified representation of the road using polylines , then partitioning these polylines into segments. These segments are then used to construct a general outline of the road and, subsequently, to identify the individual outlines of each lane within that road.

The claims of ’609 focus on a method, a system, and a non-transitory computer readable storage medium for constructing a map. The key steps involve constructing an outline that encompasses multiple lanes on a road, using segments derived from polylines that represent the road's shape. These segments are created at the start, end, and turning points of the polylines. The individual outlines of each lane are then identified based on these segments and the overall outline.

In practice, the system receives raw image data, possibly from a vehicle's camera and GPS. This data is visualized as a dot plot, which is then converted into a polyline plot. The polylines are segmented, and a lane geometry construction circuit builds a general outline around all lanes. A lane identification circuit then uses the segments and the general outline to determine the boundaries of each lane. This process may involve smoothing the lane boundaries using techniques like B-spline transformation and generating waypoints for navigation.

The invention differentiates itself by automating the lane extraction process. Instead of manual digitization or relying solely on GPS data, it uses image processing to identify lane markings and boundaries. The grid-based approach for map extraction allows for efficient retrieval of lane segment data based on the vehicle's location. This automated approach reduces the cost and effort associated with creating and maintaining accurate lane-level roadmaps, making it suitable for real-time navigation and autonomous driving applications.