The goal of object detection is to percept static/dynamic objects, which potentially are obstacles to the autonomous vehicle.
Object detection consists of multiple steps. At first, the ground plane is optionally segmented and filtered from the point cloud.
One variant is the point-wise segmentation.
Another variant is to generate a 2.5D Occupancy-Grid and to extract coherent objects through a connected-component algorithm. Either way, the last step is to fit a bounding-box for every object.