| CPC G06T 17/00 (2013.01) [G06V 10/764 (2022.01)] | 6 Claims |
|
1. A method for generating point cloud data to cause a processor to execute a process, the method comprising executing on the processor the steps of:
generating two-dimensional color image data by capturing an image of a target object, the target object including a plurality of sub-target objects;
classifying a plurality of pixels that form the color image data into three or more categories to generate classified image data containing information on the plurality of pixels, the three or more categories corresponding to indexes that are determined by identifying each of the plurality of sub-target objects using a semantic segmentation method;
determining kernels associated with the three or more categories, the kernels having sizes according to sizes of the plurality of sub-target objects formed of pixels that belong to the three or more categories;
generating first data on a point cloud formed of points each identified in a three-dimensional coordinate system by capturing an image of the target object using an infrared stereo camera;
projecting the first data onto a plane to generate second data on a point cloud formed of points each corresponding to a point in the point cloud indicated by the first data and each identified in a two-dimensional coordinate system;
associating a point of the point cloud contained in the second data with a pixel in the classified image data to classify the points of the point cloud contained in the first data into the three or more categories via the points of the point cloud contained in the second data; and
generating, from the first data, third data on a point cloud formed of points each identified in the three-dimensional coordinate system by filtering information associated with the points in the first data, the filtering using the kernels according to the categories of the points contained in the first data.
|