Method, apparatus, and computer program for generating input feature map of artificial intelligence model to detect object based on LiDAR
Provided are a method, an apparatus, and a computer program for generating an input feature map of an artificial intelligence model to detect an object based on light detection and ranging (LiDAR). A method of generating an input feature map of an artificial intelligence model to detect an object based on LiDAR, which is to be performed by a computing apparatus, may include collecting LiDAR sensor data in the form of a three-dimensional point cloud for a predetermined area, generating one or more first feature maps by processing the collected LiDAR sensor data based on one or more predefined indicators, and generating the input feature map of the artificial intelligence model by combining, as an independent channel, each of the one or more generated first feature maps with one of the second feature maps previously generated based on a distance and an angle.
1 . A method of generating an input feature map of an artificial intelligence model to detect an object based on light detection and ranging (LiDAR), which is performed by a computing apparatus, the method comprising:
collecting LiDAR sensor data in a form of a three-dimensional point cloud from an autonomous driving vehicle located within a predetermined area of the computing apparatus;
generating one or more first feature maps by processing the collected LiDAR sensor data based on one or more predefined indicators;
generating the input feature map of the artificial intelligence model by combining, as an independent channel, each of the one or more generated first feature maps with one of second feature maps previously generated by pre-computing a distance and an angle for a preset object detection area;
detecting objects around the autonomous vehicle by inputting the input feature map into the artificial intelligence model, and controlling the driving of the autonomous driving vehicle based on the detected the objects;
generating a plurality of grids having a predetermined size by dividing a preset object detection area based on an XY plane;
generating a second feature map related to a distance based on distances between the plurality of generated grids and a reference point; and
generating a second feature map related to an angle based on angles between the plurality of generated grids and the reference point,
wherein the generating of the second feature map related to the distance includes:
calculating a distance value between center point coordinates of each of the plurality of generated grids and reference point coordinates;
standardizing the calculated distance value using a preset maximum distance value, which is a preset value corresponding to a maximum detection range of a LiDAR sensor; and
generating the second feature map related to the distance by inputting the standardized distance value into a matrix corresponding to the preset object detection area, and
wherein the generating of the second feature map related to the angle includes:
calculating an angle value between a line connecting the center point of each of the plurality of generated grids and the reference point and an X axis or a Y axis;
transforming the calculated angle value using Euler angle transformation; and
generating the second feature map related to the angle by inputting the transformed angle value into the matrix corresponding to the preset object detection area.
2 . The method of claim 1 , wherein the generating of the one or more first feature maps includes:
generating a plurality of grid pillars by dividing the collected LiDAR sensor data based on an XY plane;
calculating the one or more predefined indicators for each of the plurality of generated grid pillars; and
generating the one or more first feature maps by inputting the one or more calculated indicators into a matrix corresponding to the collected LiDAR sensor data.
3 . The method of claim 2 , wherein the one or more predefined indicators includes the number, maximum heights, which are heights of points having a largest Z-axis coordinate value, among LiDAR sensor points, minimum heights, which are heights of points having a smallest Z-axis coordinate value, among the LiDAR sensor points, and an average height, which is a mean value of Z-axis coordinate values of the LiDAR sensor points, and
the generating of the one or more first feature maps includes generating, as the independent channel, a first feature map corresponding to each of the number, maximum heights, minimum heights, and average height of the LiDAR sensor points.
4 . The method of claim 1 , wherein the generating of the second feature map related to the distance includes:
calculating a distance value between the center point coordinates of each of the plurality of generated grids and the reference point coordinates;
transforming the calculated distance value using a predefined logarithmic function; and
generating the second feature map related to the distance by inputting the transformed distance value into the matrix corresponding to the preset object detection area.
5 . The method of claim 1 , wherein the generating of the second feature map related to the angle includes:
calculating an angle value between a line connecting the center point of each of the plurality of generated grids and the reference point and an X axis or a Y axis;
calculating a trigonometric ratio for the calculated angle value; and
generating two second feature maps related to the angle by inputting the calculated trigonometric ratio into the matrix corresponding to the preset object detection area.
6 . An apparatus for generating an input feature map of an artificial intelligence model to detect an object based on light detection and ranging (LiDAR), the apparatus comprising:
a processor;
a network interface;
a memory; and
a computer program loaded into the memory and executed by the processor,
wherein the computer program includes:
an instruction for collecting LiDAR sensor data in a form of a three-dimensional point cloud from an autonomous driving vehicle located within a predetermined area of the apparatus;
an instruction for generating one or more first feature maps by processing the collected LiDAR sensor data based on one or more predefined indicators;
an instruction for generating the input feature map of the artificial intelligence model by combining, as an independent channel, each of the one or more generated first feature maps with one of second feature maps previously generated by pre-computing a distance and an angle for a preset object detection area;
an instruction for detecting objects around the autonomous vehicle by inputting the input feature map into the artificial intelligence model, and controlling the driving of the autonomous driving vehicle based on the detected the objects;
an instruction for generating a plurality of grids having a predetermined size by dividing a preset object detection area based on an XY plane;
an instruction for generating a second feature map related to a distance based on distances between the plurality of generated grids and a reference point; and
an instruction for generating a second feature map related to an angle based on angles between the plurality of generated grids and the reference point,
wherein the instruction for generating of the second feature map related to the distance includes:
an instruction for calculating a distance value between center point coordinates of each of the plurality of generated grids and reference point coordinates;
an instruction for standardizing the calculated distance value using a preset maximum distance value, which is a preset value corresponding to a maximum detection range of a LiDAR sensor; and
an instruction for generating the second feature map related to the distance by inputting the standardized distance value into a matrix corresponding to the preset object detection area, and
wherein the instruction for generating of the second feature map related to the angle includes:
an instruction for calculating an angle value between a line connecting the center point of each of the plurality of generated grids and the reference point and an X axis or a Y axis;
an instruction for transforming the calculated angle value using Euler angle transformation; and
an instruction for generating the second feature map related to the angle by inputting the transformed angle value into the matrix corresponding to the preset object detection area.
7 . A non-transitory computer-readable medium having stored therein a computer program for causing a computing apparatus to execute the following operations of:
collecting LiDAR sensor data in a form of a three-dimensional point cloud from an autonomous driving vehicle located within a predetermined area of the computing apparatus;
generating one or more first feature maps by processing the collected LiDAR sensor data based on one or more predefined indicators;
generating an input feature map of an artificial intelligence model by combining, as an independent channel, each of the one or more generated first feature maps with one of second feature maps previously generated by pre-computing a distance and an angle for a preset object detection area;
detecting objects around the autonomous vehicle by inputting the input feature map into the artificial intelligence model, and controlling the driving of the autonomous driving vehicle based on the detected the objects;
generating a plurality of grids having a predetermined size by dividing a preset object detection area based on an XY plane;
generating a second feature map related to a distance based on distances between the plurality of generated grids and a reference point; and
generating a second feature map related to an angle based on angles between the plurality of generated grids and the reference point,
wherein the generating of the second feature map related to the distance includes:
calculating a distance value between center point coordinates of each of the plurality of generated grids and reference point coordinates;
standardizing the calculated distance value using a preset maximum distance value, which is a preset value corresponding to a maximum detection range of a LiDAR sensor; and
generating the second feature map related to the distance by inputting the standardized distance value into a matrix corresponding to the preset object detection area, and
wherein the generating of the second feature map related to the angle includes:
calculating an angle value between a line connecting the center point of each of the plurality of generated grids and the reference point and an X axis or a Y axis;
transforming the calculated angle value using Euler angle transformation; and
generating the second feature map related to the angle by inputting the transformed angle value into the matrix corresponding to the preset object detection area.