| 摘要 |
Objects within a three-dimensional (3D) environment are detected by obtaining a point-cloud environment representation (e.g. from LIDAR) comprising a set of point locations, and converting the point-cloud to a 3D feature grid by dividing the 3D space into cells according to a grid size, determining whether the cells contain point-cloud points, and mapping any that do to a feature vector, and mapping any that do not to a zero feature vector. A detection window is also generated containing an integral number of the cells and having a set of positions which can be occupied. A detection score for each position is calculated by: voting for each cell within the window which includes at least one point, each vote being calculated using the cell feature vector (e.g. fixed-dimensional feature vector) and a weighting vector (e.g. obtained from a linear classifier), and summing the votes. The score provides a basis for determining whether each position contains an object of interest, each position with a score above a threshold being classified as containing an object of interest. The vote may comprise a scalar product of the cells feature vector and the weighting vector. A system performing the method may be provided on a vehicle. |
