Method and System for Classification of an Object in a Point Cloud Data Set
A method for classifying an object in a point cloud includes computing first and second classification statistics for one or more points in the point cloud. Closest matches are determined between the first and second classification statistics and a respective one of a set of first and second classification statistics corresponding to a set of N classes of a respective first and second classifier, to estimate the object is in a respective first and second class. If the first class does not correspond to the second class, a closest fit is performed between the point cloud and model point clouds for only the first and second classes of a third classifier. The object is assigned to the first or second class, based on the closest fit within near real time of receiving the 3D point cloud. A device is operated based on the assigned object class.
1 . A light detection and ranging (LIDAR) sensor system for a vehicle, comprising:
a sensor configured to:
transmit a plurality of transmit beams at a plurality of angles relative to the sensor;
receive a plurality of return beams from reflection by an object of the plurality of transmit beams; and
output a point cloud to represent the object based on the plurality of return beams; and
one or more processors configured to:
determine a plurality of classification statistics regarding the object based on the point cloud, wherein a first classification statistic of the plurality of classification statistics is different from a second classification statistic of the plurality of classification statistics; and
output a class of the object based on the plurality of classification statistics.
2 . The LIDAR sensor system of claim 1 , wherein the one or more processors are configured to determine the class of the object based on a comparison of the point cloud with a model point cloud corresponding to the class.
3 . The LIDAR sensor system of claim 1 , wherein the one or more processors are configured to:
determine a first distance in a first plane defined by a first point of the point cloud that corresponds to a first return beam of the plurality of beams and a second point of the point cloud that corresponds to a second return beam of the plurality of beams;
determine a second distance in a second plane defined by the first point and the second point; and
determine the first classification statistic as a histogram based on the first distance and the second distance.
4 . The LIDAR sensor system of claim 1 , wherein the one or more processors are configured to control the vehicle to avoid collision with the object based on the class of the object.
5 . The LIDAR sensor system of claim 1 , wherein the sensor is configured to generate the point cloud to include a first data point representing a first range to the object determined from a first return beam of the plurality of return beams and to include a second data point representing a second range to the object determined from a second return beam of the plurality of return beams.
6 . The LIDAR sensor system of claim 1 , wherein the sensor comprises:
a laser source configured to output a carrier wave;
a modulator configured to modulate the carrier wave to provide the carrier wave as the plurality of transmit beams; and
one or more scanning optics configured to scan the plurality of transmit beams over the plurality of angles.
7 . The LIDAR sensor system of claim 1 , wherein the transmit beam is a chirp signal.
8 . The LIDAR sensor system of claim 1 , wherein the sensor is configured to output the point cloud for use as training data.
9 . The LIDAR sensor system of claim 1 , wherein the one or more processors are configured to determine the class from a predetermined number of classes.
10 . The LIDAR sensor system of claim 1 , wherein the one or more processors are configured to determine the object class from a vehicle class and a roadside structure class.
11 . An autonomous vehicle control system, comprising:
one or more processors configured to:
receive a data signal comprising a three-dimensional (3D) point cloud representing an object;
determine a plurality of classification statistics regarding the object based on the point cloud, wherein a first classification statistic of the plurality of classification statistics is different from a second classification statistic of the plurality of classification statistics;
determine a class of the object based on the plurality of classification statistics; and
generate a control signal to control operation of an autonomous vehicle based on the class of the object.
12 . The autonomous vehicle control system of claim 11 , wherein the 3D point cloud comprises a first data point corresponding to a first range to the object and a second data point corresponding to a second range to the object.
13 . The autonomous vehicle control system of claim 11 , wherein the one or more processors are configured to generate the control signal to avoid collision with the object.
14 . The autonomous vehicle control system of claim 11 , wherein the class of the object comprises at least one of a vehicle class or a roadside structure class.
15 . The autonomous vehicle control system of claim 11 , wherein the one or more processors are configured to:
determine one or more distances defined relative to a first data point of the 3D point cloud and a second data point of the 3D point cloud;
determine one or more angles defined relative to the first data point and the second data point; and
determine the plurality of classification statistics based on the one or more distances and the one or more angles.
16 . A LIDAR sensor system for a vehicle, comprising:
a laser source configured to generate a carrier wave;
an optic configured to output the carrier wave as a plurality of transmit signals;
one or more detectors configured to detect a plurality of return signals from reflection of the plurality of transmit signals by an object; and
one or more processors configured to:
determine a point cloud to represent the object based on the plurality of return signals;
determine a plurality of classification statistics regarding the object based on the point cloud, wherein a first classification statistic of the plurality of classification statistics is different from a second classification statistic of the plurality of classification statistics; and
output a class of the object based on the plurality of classification statistics.
17 . The LIDAR sensor system of claim 16 , further comprising a modulator configured to modulate at least one of a phase or a frequency of the carrier wave.
18 . The LIDAR sensor system of claim 16 , wherein the plurality of classification statistics include a spin image determined from the point cloud and a covariance matrix determined from the point cloud.
19 . The LIDAR sensor system of claim 16 , wherein the one or more processors are configured to select the class based on a match between the plurality of classification statistics and the class.
20 . The LIDAR sensor system of claim 16 , wherein the class represents a plurality of object types.